GEOCHEMISTRY, GROUNDWATER AND POLLUTION

SECOND EDITION

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

GEOCHEMISTRY,
GROUNDWATER AND
POLLUTION,
2ND EDITION

C.A.J. APPELO
Hydrochemical Consultant, Amsterdam, the Netherlands

D. POSTMA
Environment & Resources DTU, Technical University of Denmark, Kgs. Lyngby, Denmark

A.A. BALKEMA PUBLISHERS Leiden/London/New York/Philadelphia/Singapore

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

Library of Congress Cataloging-in-Publication Data
Applied for

Cover Illustration: Ben Akkerman – Landscape

© Ben Akkerman collection Stedelijk Museum, Amsterdam

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

All rights reserved. No part of this publication or the information contained herein may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying,
recording or otherwise, without written prior permission from the publisher.

Although all care is taken to ensure the integrity and quality of this publication and the information herein, no
responsibility is assumed by the publishers nor the author for any damage to property or persons as a result of
operation or use of this publication and/or the information contained herein.

Published by: A.A. Balkema Publishers, Leiden, The Netherlands a member of

Taylor & Francis Group plc
www.balkema.nl and www.tandf.co.uk

ISBN 04 1536 421 3 Hardback

ISBN 04 1536 428 0 Student paper edition

Printed in Great Britain

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

Preface

The first edition of Geochemistry, groundwater and pollution has become a popular textbook that
is used throughout the world in university courses and as a reference text. The new developments in
our science during the last decade have inspired us to update the book. There are now more data
available, more experiments have been done, new tools were developed, and of course, new problems
emerged. An important catalyst for rewriting the book was the advancement of the computer
code PHREEQC. From the start in 1980, this model was developed to simulate field data and
laboratory experiments and at present it encompasses over 20 years of experience in modeling
water quality. The code is used all over the world in research and to predict the long-term effects of
pollution and radioactive waste storage. Modeling with PHREEQC is an integrated part of the
presentation in the second edition. It is used to elucidate chemical concepts when they become
complicated and in particular to analyze the interactions between chemical processes and
transport. Numerous examples in this book show PHREEQC applications with input files listed in
the text. The more comprehensive ones can be downloaded from the net and are useful templates for
similar problems.
The basic structure of the first edition is maintained. However, flow and transport in aquifers is
introduced already in Chapter 3 since it is important to become familiar with the principles that gov-
ern water flow and residence times in aquifers at an early stage. The transport equations developed
in this chapter are applied throughout the text when transport and chemical reactions interplay, for
example to calculate retardation and chromatographic patterns. The behavior of heavy metals is
treated more extensively in a separate chapter with emphasis on surface complexation theory.
A chapter on organic pollutants has also been added, discussing the evaporation and leaching of
these chemicals as well as inorganic and microbial degradation. The last chapter introduces numeri-
cal transport models and provides application examples of organic pollution, acid mine water, in-situ
aquifer remediation, arsenic in groundwater, and more. Similar to the previous edition, examples
serve to illustrate the application of theory. In addition to the problems at the end of each chapter,
questions have been interspersed in the text to focus the reader and to provide diversion with simple
calculations that elucidate the concepts.
We like to thank colleagues and friends for their contributions to the second edition. First
and foremost, we owe much to David Parkhurst who initiated PHREEQC, solved many of our prob-
lems, and reviewed Chapter 11. Vincent Post developed the graphical interface for PHREEQC
that is used throughout this book. Individual chapters were read and commented by Martin S.
Andersen, Philip Binning, Peter Engesgaard, Rien van Genuchten, Pierre Glynn, Jasper Griffioen,
Rasmus Jakobsen, David Kinniburgh, Claus Kjøller, Niel Plummer, Ken Stollenwerk and Art White.
We also appreciate the contributions and comments of Boris van Breukelen, Liselotte Clausen,
Flemming Larsen, Uli Mayer, Hanne Dahl Pedersen and Judith Wood. We are indebted and
immensely thankful for the time spent by all. Projects and studies with Paul Eckert, Laurent Charlet,
V
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
VI Preface

Christoph König and Alain Dimier have helped to practize ideas and provided financial support
that is gratefully acknowledged (TA). Dieke thanks Philip Binning and the University of
Newcastle N.S.W., Australia, for hospitality and financial support during a sojourn in the winter
of 2002. Finally, Tony could not have imagined to have started the work without having Dorothée,
Els and Maria close-by for inspiration, wisdom and love, and Tobias for surviving.

January 2005.
Amsterdam, Tony Appelo
Hanoi, Dieke Postma

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

Contents

LIST OF EXAMPLES XII

NOTATION XV

1 INTRODUCTION TO GROUNDWATER GEOCHEMISTRY 1

1.1 Groundwater as drinking water 1
1.1.1 Standards for drinking water 1
1.2 Units of analysis 3
1.3 Groundwater quality 7
1.4 Sampling of groundwater 10
1.4.1 Depth integrated or depth specific sampling 10
1.4.2 Procedures for sampling of groundwater 12
1.5 Chemical analysis of groundwater 15
1.5.1 Field analyses and sample conservation 15
1.5.2 Accuracy of chemical analysis 17
Problems 20
References 21

2 FROM RAINWATER TO GROUNDWATER 23

2.1 The hydrological cycle 23
2.2 The composition of rainwater 26
2.2.1 Sources and transport of atmospheric pollutants 31
2.3 Stable isotopes in rain 31
2.3.1 Isotopic ratios and the
notation 32
2.3.2 The Rayleigh process 33
2.3.3 The isotopic composition of rain 37
2.4 Dry deposition and evapotranspiration 41
2.5 Mass balances and ecosystem dynamics 46
2.5.1 Water quality profiles in the unsaturated soil 49
2.6 Overall controls on water quality 51
Problems 58
References 59

3 FLOW AND TRANSPORT 63

3.1 Flow in the unsaturated zone 63
3.2 Flow in the saturated zone 64
VII
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
VIII Contents

3.2.1 Darcy’s law 64

3.2.2 Flowlines in the subsoil 67
3.2.3 Effects of non-homogeneity 70
3.2.4 The aquifer as a chemical reactor 71
3.3 Dating of groundwater 72
3.4 Retardation 75
3.4.1 The retardation equation 76
3.4.2 Indifferent and broadening fronts 79
3.4.3 Sharpening fronts 82
3.4.4 Solid and solute concentrations 84
3.5 Diffusion 86
3.5.1 Diffusion coefficients 87
3.5.2 Diffusion as a random process 89
3.5.3 Diffusive transport 93
3.5.4 Isotope diffusion 96
3.6 Dispersion 99
3.6.1 Column breakthrough curves 102
3.6.2 Dispersion coefficients and dispersivity 105
3.6.3 Macrodispersivity 107
Problems 113
References 115

4 MINERALS AND WATER 119

4.1 Equilibria and the solubility of minerals 119
4.2 Corrections for solubility calculations 123
4.2.1 Concentration and activity 123
4.2.2 Aqueous complexes 127
4.2.3 Combined complexes and activity corrections 128
4.2.4 Calculation of saturation states 131
4.3 Mass action constants and thermodynamics 132
4.3.1 The calculation of mass action constants 132
4.3.2 Calculation of mass action constants at different temperature 133
4.4 Equilibrium calculations with PHREEQC 135
4.4.1 Speciation calculations using PHREEQC 135
4.4.2 The PHREEQC database 137
4.4.3 Mineral equilibration 141
4.5 Solid solutions 142
4.5.1 Basic theory 142
4.5.2 The fractionation factor for solid solutions 148
4.5.3 Kinetic effects on the fractionation factor 149
4.6 Kinetics of geochemical processes 152
4.6.1 Kinetics and equilibrium 152
4.6.2 Chemical reactions and rates 153
4.6.3 Temperature dependency of reaction rates 159
4.6.4 Mechanisms of dissolution and crystallization 160
4.6.5 Rate laws for mineral dissolution and precipitation 162
Problems 169
References 171

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Contents IX

5 CARBONATES AND CARBON DIOXIDE 175

5.1 Carbonate minerals 176
5.2 Dissolved carbonate equilibria 178
5.2.1 The carbonic acid system 179
5.2.2 Determining the carbonate speciation in groundwater 183
5.3 Carbon dioxide in soils 186
5.4 Calcite solubility and PCO2 191
5.4.1 Calcite dissolution in systems open and closed for CO2 gas 193
5.4.2 Two field examples 195
5.5 Carbonate rock aquifers 197
5.5.1 Dolomite and dedolomitization 201
5.5.2 Pleistocene carbonate aquifers 205
5.6 Kinetics of carbonate reactions 210
5.6.1 Dissolution 210
5.6.2 Precipitation 217
5.7 Carbon isotopes 218
5.7.1 Carbon-13 trends in aquifers 221
5.7.2 14C and groundwater age 226
5.7.3 Retardation by sorption and stagnant zone diffusion 228
Problems 232
References 236
6 ION EXCHANGE 241
6.1 Cation exchange at the salt/fresh water interface 242
6.2 Adsorbents in soils and aquifers 247
6.2.1 Clay minerals 248
6.3 Exchange equations 251
6.3.1 Values for exchange coefficients 254
6.3.2 Calculation of exchanger composition 255
6.3.3 Calculation of exchanger composition with PHREEQC 257
6.3.4 Determination of exchangeable cations 260
6.4 Chromatography of cation exchange 262
6.4.1 Field examples of freshening 263
6.4.2 Salinity effects on cation exchange 268
6.4.3 Quality patterns with salinization 271
6.4.4 Fronts and chromatographic sequences 272
6.4.5 Modeling chromatographic sequences with PHREEQC 275
6.5 Physical non-equilibrium 283
6.5.1 Modeling stagnant zones 285
6.6 The Gouy-Chapman theory of the double layer 288
6.6.1 Numerical integration of the double layer equations 293
6.6.2 Practical aspects of double layer theory 296
6.7 Irrigation water quality 299
Problems 303
References 306
7 SORPTION OF TRACE METALS 311
7.1 The origin and occurrence of heavy metals in groundwater 311
7.2 Sorption isotherms and distribution coefficients 315

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

X Contents

7.2.1 Distribution coefficients from ion exchange 318

7.3 Variable charge surfaces 322
7.3.1 Titration curves with suspended oxide particles 322
7.3.2 Surface charge and point of zero charge, PZC 324
7.3.3 Sorption edges 328
7.3.4 Sorption, absorption, and coprecipitation 333
7.4 Surface complexation 334
7.4.1 Surface complexation models 338
7.4.2 The ferrihydrite (Fe(OH)3) database 340
7.4.3 Diffuse double layer concentrations in surface complexation models 343
7.5 Complexation to humic acids 344
7.5.1 The ion association model 346
7.5.2 Tipping and Hurley’s discrete site model “WHAM” 348
7.5.3 Distribution models 354
7.5.4 Humic acids as carriers of trace elements 356
7.6 Kinetics of surface complexation 358
7.6.1 Extrapolation of adsorption kinetics for other metal ions 363
7.7 Field applications 363
Problems 367
References 369
8 SILICATE WEATHERING 375
8.1 Weathering processes 375
8.2 The stability of weathering products 380
8.3 Incongruent dissolution of primary silicates 383
8.4 The mass balance approach to weathering 389
8.5 Kinetics of silicate weathering 395
8.6 Field weathering rates 400
8.7 Acid groundwater 404
8.7.1 Buffering processes in aquifers 405
Problems 410
References 412
9 REDOX PROCESSES 415
9.1 Basic theory 415
9.1.1 The significance of redox measurements 420
9.1.2 Redox reactions and the pe concept 422
9.2 Redox diagrams 423
9.2.1 Stability of water 424
9.2.2 The stability of dissolved species and gases: Arsenic 425
9.2.3 The stability of minerals in redox diagrams 432
9.3 Sequences of redox reactions and redox zoning 438
9.3.1 Decomposition of organic matter 442
9.4 Oxygen consumption 446
9.4.1 Pyrite oxidation 450
9.4.2 Kinetics of pyrite oxidation 450
9.4.3 Oxygen transport and pyrite oxidation 453
9.5 Nitrate reduction 458
9.5.1 Nitrate reduction by organic matter oxidation 459
9.5.2 Nitrate reduction by pyrite and ferrous iron 462

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Contents XI

9.6 Iron reduction and sources of iron in groundwater 465

9.6.1 Iron in aquifer sediments 465
9.6.2 Reductive dissolution of iron oxides 466
9.7 Sulfate reduction and iron sulfide formation 472
9.7.1 The formation of iron sulfides 476
9.8 The formation of methane 477
Problems 479
References 480
10 POLLUTION BY ORGANIC CHEMICALS 489
10.1 Gas-water exchange 489
10.1.1 Evaporation of a pure organic liquid 494
10.2 Transport of pure organic liquids through soil 496
10.3 Sorption of organic chemicals 499
10.3.1 Sorption of charged organic molecules 504
10.3.2 Sorption in stagnant zones 506
10.3.3 Release from stagnant zones and blobs 510
10.4 Transformation reactions of organic chemicals 516
10.4.1 Monod biotransformation kinetics 518
10.5 Kinetic complexation of heavy metals on organics 529
Problems 535
References 537
11 NUMERICAL MODELING 541
11.1 Numerical modeling of transport 543
11.1.1 Only diffusion 543
11.1.2 Advection and diffusion/dispersion 550
11.1.3 Non-linear reactions 558
11.2 Examples of hydrogeochemical transport modeling 560
11.2.1 Tritium-Helium age dating 561
11.2.2 Toluene degradation in an aquifer 565
11.2.3 Remediation of a BTEX polluted site 568
11.2.4 Acid drainage from a Uranium mine 570
11.2.5 In-situ iron removal from groundwater 579
11.2.6 Arsenic in Bangladesh groundwater 585
11.2.7 Fractionation of isotopes 590
References 595

APPENDIX A: HYDROGEOCHEMICAL MODELING WITH PHREEQC 599

APPENDIX B: ANSWERS TO PROBLEMS 617

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

List of Examples

1.1 Recalculation of concentration units 5

1.2 Effect of iron oxidation on analytical results 15
1.3 Estimating the reliability of water analyses 17
1.4 Analytical errors due to precipitation in the sampling bottle 20

2.1 Calculate
18O of rain condensing from vapor 35

2.2 Calculate recharge using the Cl mass balance 44
2.3 Estimate the Cl and NO 3 concentration in groundwater below agricultural land 46

3.1 Calculate the travel time from midway in the Vejen waste site to 125 m 69
downstream
3.2 Calculate the water level in the Vejen river, 1 km downstream from the waste site 69
3.3 Flushing of NO 3 from an aquifer 72
3.4 Retardation and isotherm slope 80
3.5 Analytical modeling of column elution 81
3.6 Retardation of a sharp front 83
3.7 Recalculate 10 ppm As in sand to solute concentration in mg/L pore water 84
3.8 Travel time of diuron in a soil 84
3.9 Calculation of a diffusion profile 95
3.10 Diffusive flux through a clay barrier (after Johnson et al., 1989) 96
3.11 Chloride isotope fractionation during diffusion 99
3.12 Dispersion coefficient from a single shot input 101
3.13 Front dispersion in a column 103
3.14 Pollutant spreading during transport in an aquifer 109
3.15 Longitudinal dispersivity in the Borden aquifer 112

4.1 Gypsum addition to high fluor groundwater 121

4.2 Calculate ionic strength and ion activity coefficients 126
4.3 Solubility of gypsum 129
4.4 Calculation of solubility products from Gibbs free energy data 133
4.5 Temperature dependency of the solubility product 134
4.6 Calculate the speciation of a water analysis using PHREEQC 135
4.7 Solubility of quartz at 150°C 140
4.8 Equilibrate a water sample with minerals 141
4.9 Ideal solid solution of Cd2 in calcite 144
4.10 Non-ideal solid solution of Cd2 in calcite 146
4.11 Miscibility gap in the solid solution 147
4.12 Oxidation of Fe(2) 156
4.13 Dissolution of hydroxyapatite; transport or surface reaction controlled? 161

XII
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 List of examples XIII

4.14 Dissolution rate of quartz 163

4.15 Kinetic dissolution of quartz with PHREEQC 164

5.1 Calculate TIC in water for a fixed CO2 pressure 182

5.2 Calculation of the aqueous carbonate system with PHREEQC 183
5.3 Manual calculation of carbonate speciation and SIcalcite 185
5.4 PHREEQC calculation of carbonate speciation and SIcalcite 186
5.5 Calculate mixing effects on calcite saturation state with PHREEQC 193
5.6 PHREEQC calculation of open and closed system calcite dissolution 194
5.7 Propagation of the calcite dissolution front in the Rømø aquifer 197
5.8 Dissolution of dolomite in the Italian Dolomites 201
5.9 Kinetic dissolution of calcite calculated with PHREEQC, comparing simplified
and parent “PWP” rates 213
5.10 PHREEQC calculation of
13C during calcite dissolution 219
5.11 PHREEQC calculation of 13C evolution during dedolomitization 223
5.12 Groundwater age from 14C 227
5.13 Estimate exchangeable carbonate on Chalk and the retardation of 14C 229
5.14 Estimate the retardation of 14C by matrix diffusion 231

6.1 Recalculate CEC (meq/kg soil) to concentration (meq/L pore water) 248
6.2 Structural charge of smectite 251
6.3 Exchange coefficients as a function of solution normality 253
6.4 Calculate the cation exchange complex in equilibrium with groundwater 256
6.5 Calculate the exchanger composition in contact with groundwater, using PHREEQC 258
6.6 Calculate the exchanger composition, using the Rothmund-Kornfeld equation 259
6.7 Simulate the analytical measurement of exchangeable cations 261
6.8 Flushing of an exchange complex 263
6.9 The water composition after passage of a salinity front during freshening 268
6.10 Flushing of K from a column 274
6.11 PHREEQC model for Valocchi’s field injection experiment 279
6.12 Calculate the surface potential on montmorillonite 292
6.13 Calculation of the Exchangeable Sodium Ratio (ESR) 300
6.14 Calculation of SAR adjusted for calcite precipitation 302

7.1 Freundlich sorption isotherm for Cd2 on loamy sand 316

7.2 Estimate the distribution coefficient of Cd2 318
7.3 Distribution coefficient for Sr2 319
7.4 Exchange coefficient of Cd2 vs Na on montmorillonite 320
7.5 Calculate the specific capacitance of a -Al2O3 surface 336
7.6 Calculate the surface potential on -Al2O3 337
7.7 Calculate the equivalent ionic strength for a constant capacity model 339
7.8 Recalculate sites/nm2 to mol/L 340
7.9 Sorption of Cd2 to iron oxyhydroxide in loamy sand 342
7.10 Estimate the spherical surface area of a fulvic acid 345
7.11 Sorption of Cd2 to organic matter in loamy sand 351
7.12 Calculate heavy metal removal from groundwater with aeration and filtration 364

8.1 Incongruent dissolution of K-feldspar 387

8.2 Mass balance and the water chemistry of the Sierra Nevada (USA) 390
8.3 Mass balance calculation of mineral weathering using PHREEQC 392
8.4 Dissolution kinetics of K-feldspar as calculated by PHREEQC 398
8.5 Acid groundwater formation and gibbsite buffering 406
8.6 Modeling acidification with PHREEQC 408

9.1 Calculation of redox speciation with the Nernst equation 419

9.2 Calculate E0 from G 0r 420

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

XIV List of examples

9.3 Calculation of redox speciation using the pe concept 422

9.4 Calculation of K from thermodynamic data 423
9.5 Oxidation of the atmosphere’s N2 content to nitrate 430
9.6 Calculation of redox zonation with PHREEQC 440
9.7 Applying Henry’s law to oxygen dissolution 446
9.8 Compare oxygen flux and carbon productivity in a soil 449
9.9 Modeling gas loss during pyrite oxidation with oxygen 455
9.10 Construct a redox balance for nitrate reduction by organic matter oxidation 460
9.11 Reaction of benzene with iron-oxide and methanogenesis 470
9.12 pH buffering by pyrite and kinetically dissolving iron oxides 471

10.1 Estimate the flux of freon-11 (CCl3F) into the sea 493
10.2 Retardation of Lindane and PCB 502
10.3 Calculate sorption and ion exchange of Quinoline with PHREEQC 505
10.4 Estimate the composition of a DNAPL pool in an aquifer 512
10.5 Model the extraction of a DNAPL pool with PHREEQC 512
10.6 Estimate the hazard of groundwater pollution by methyl bromide 517
10.7 PHREEQC model of phenol degradation 519
10.8 Xylene degradation with biomass growth 522
10.9 Speciation of EDTA in Glatt river water 530
10.10 Kinetic exchange of Fe(3)EDTA 532

11.1 Calculation of aquifer pollution by waste site leachate 542

11.2 A Pascal code to model Cl diffusion from seawater into fresh water sediment 546
11.3 Implicit calculation of diffusion 548
11.4 Model the linear retardation of -HCH in a laboratory column 553
11.5 Effect of the Freundlich exponent on breakthrough curves from a column 558

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

Notation

Greek characters
 Speciation factor, fraction of total concentration (), or
isotope fractionation factor, see below (), or
solid solution fractionation factor, the activity ratio of two elements in a solid, divided by
the activity ratio in solution (), or
exchange factor between mobile and immobile regions (1/s), or
spatial weighting factor in numerical scheme ()
18Ol/g isotope fractionation factor (), here for 18O in liquid/gas equilibrium written in the form
H218O(g)  H216O(l) ↔ H218O(l)  H216O(g)
L longitudinal dispersivity (m); subscript T for transversal dispersivity
i equivalent fraction of ion i in water ()
i equivalent fraction of i on exchanger ()
iM molar fraction of i on exchanger ()

difference of isotopic ratio with respect to a standard (‰)
i activity-coefficient of species i in water ()
 porosity, a fraction of total volume (), or
isotopic enrichment factor (), notation follows example 18Ol/g, or
dielectric constant (F/m)
w water filled porosity, a fraction of total volume (); subscript g for gas filled porosity
m mobile water filled porosity, a fraction of total volume (); subscript im for immobile water
viscosity (g/m/s)

tortuosity, the ratio of actual path over straight line path (), or
water content (g/g)
1/ Debye length (m)
m molar electrical conductivity (S m2/mol)
 radioactive decay constant (1/s), or
activity coefficient in solid solution (), or
correlation length (m)
 reduced mass (g/mol)
max specific degradation rate (1/s)
v stoichiometric coefficient in reaction ()
 charge density (C/m3), or
solid or liquid density (g/cm3)
b bulk density of solid (with air filled pores) (g/cm3)
 charge density at solid surface (eq/kg solid); subscripts 0 for structural charge, H for protons,
M for metals, A for anions, D for double layer charge
DL double layer charge (C/m2 solid)
2 variance in a set of parameters; of the diffusion curve (units depend on parameter)
 residence time (s)
m mobile fraction of pore water (), m/w

XV
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
XVI Notation

 mole fraction in solid or liquid solution ()

 potential in a double layer (V)
0 potential at a solid surface (V)
 saturation state or ratio, IAP/K ()
 weighting factor for time in numerical scheme ()

Capital letters
A surface area (m2), or
temperature dependent coefficient in Debye-Hückel theory ()
A0 initial surface area (m2), or
total cation concentration (meq/L)
A11 longitudinal macrodispersivity (m)
Alk Alkalinity (meq/L)
CEC cation exchange capacity (meq/kg)
B number of bacterial cells as dry biomass (mg/L, mol C/L), or
temperature dependent coefficient in Debye-Hückel theory (1/m)
D diffusion coefficient (m2/s), or
dry (atmospheric) deposition (g/m2/yr), or
aquifer thickness (m)
De effective diffusion coefficient in a porous medium (m2/s)
Df diffusion coefficient in free water (m2/s); subscript a in free air
DL longitudinal dispersion coefficient (m2/s); subscript T for transversal dispersion coefficient
DDL diffuse double layer
E evapotranspiration (m), or
electrical field strength (V/m), or
kinetic energy (kg m2/s2), or
efficiency of in situ iron removal ()
E.B. electrical balance of water analysis (%)
EC electrical conductivity of water sample (S/cm)
Eh redox potential relative to the H2 / H half reaction (V)
E0 standard (redox) potential (V)
ESR exchangeable sodium ratio ()
F flux (mass/m2/s)
FNa fractionation factor with respect to Na, used when comparing concentrations in rainwater
and seawater ()
Gr0 Gibbs free energy of a reaction (J/mol)
Gf0(i) Gibbs free energy of formation of i from elementary i (J/mol)
H 0r reaction enthalpy (heat) (J/mol)
I ionic strength ()
IAP ion activity product ()
K mass action constant () or solubility product ()
Ka acid dissociation constant ()
Ki\j exchange constant, the subscript ions indicate solute ions in the order of appearance in the
mass-action equation: i  j-X ↔ i-X  j, Gaines and Thomas convention ()
KGi\j idem, Gapon convention ()
Kd distribution coefficient (mg sorbed/kg solid)/(mg solute/L pore water) (L/kg)
KF constant in Freundlich sorption isotherm (units depend on concentration units and on exponent)
KH Henry’s constant for liquid ↔ gas equilibrium (atm L/mol)
KH Henry’s constant with the gaseous concentration in mol/L gas (L water/L gas); KH  KH/RT
Kd distribution coefficient (mg sorbed/L porewater)/(mg solute/L pore water) ()
KL constant in Langmuir sorption isotherm (mol/L)
Koc distribution coefficient among 100% organic carbon and water (L/kg)
Kow idem, among octanol and water (L/kg)
L length of column or flowtube (m)
M molality (mol/kg H2O)
M mass (mol, kg)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Notation XVII

N normality (eq/L), or
N a number of molecules ()
P gas pressure (atm), or
net rate of precipitation (groundwater infiltration) m/yr
Pe Peclet number ()
PV pore volumes injected/displaced, V/V0 ()
Q discharge (m3/s)
R gas constant (8.314 J/K/mol), or
retardation (), or
rate of dissolution/precipitation (mol/L/s), or
resistance (, V/A), or
isotopic ratio ()
S substrate concentration (mg/L, mol/L)
SR saturation state or ratio, IAP/K ()
SAR sodium adsorption ratio (mmol/L)0.5
SI Saturation Index, log(IAP/K) ()
U potential energy in Boltzmann equation (J/atom)
V volume (m3)
V0 pore volume of a column (m3)
X soil exchanger with charge number 1
Y yield factor, transforming substrate (S) into biomass (B)

Lower case letters

a ion radius (m), or
ion size in Truesdell-Jones Debye-Hückel equation (m), or
stagnant body size (m)
å ion size in Debye-Hückel equation (m)
b coefficient in Truesdell-Jones Debye-Hückel equation ()
c concentration (mol/L, mol/kg, mg/L, etc)
c0 concentration at column inlet (x  0) (mol/L)
ci initial concentration (mol/L)
d differential symbol
d depth (m), or
deuterium excess in rain (‰)
f a fraction (), or
dilution factor ()
foc fraction of organic carbon in soil ()
h hydraulic head or potential for water flow (m)
k hydraulic conductivity (m/day), or
rate constant in kinetic rate law (units depend on mechanism)
k 1⁄2 substrate concentration giving half of the maximal Monod rate (mol/L)
kmax maximal Monod rate (mol/L/s)
m molality (mol/kg H2O), or
temporal or spatial moment (units depend on number)
m/m0 mass fraction ()
meqI-X exchangeable cation I (eq/kg)
p in pH, pK, etc. indicates log, the negative, decimal logarithm
q sorbed concentration (on solid), expressed per L porewater (mol/L, mg/L, etc.)
r specific kinetic rate (mol/m2/s for solids, mol/s otherwise), or
radial distance (m)
s sorbed concentration (on solid) (mol/kg, mg/g, etc.); q  (b /)s
u ionic mobility (m2/s/V)
v velocity (m/s)
vD specific discharge, or Darcy velocity, vD  vH2O /w (m/s)
vH2O pore water flow velocity (m/s)
vc velocity of a given concentration (m/s)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

XVIII Notation

za boundary film thickness of air (m); subscript w for water

zi charge number of ion i ()

General data
Na Avogadro’s number (6.0221023/mol)
F Faraday constant (96485 C/mol or J/Volt/g eq)
R gas constant (8.314 J/K/mol)
T temperature in K ( °C  273.15)
qe charge of the electron (1.6021019 C)
k Boltzmann constant (1.381023 J/K)
1 cal  4.184 J
1 atm  1.013105 Pa
at T  298.15 K:
RT  2.479 kJ/mol
RT/F  25.69 mV; 2.303 RT/F  59.16 mV
Vgas  RT/P  24.79 L/mol at 1 atm
H2O 0.891 g/m/s

Global data
Earth surface: 5.11014 m2
Ocean surface: 3.61014 m2
Land surface: 1.51014 m2

Water in
Oceans: 137001020 g
Ice caps and glaciers: 2901020 g
All groundwater: 951020 g
Shallow groundwater  100 m: 5.61020 g
Atmosphere: 0.131020 g
Global precipitation 4.71020 g H2O/yr

Prefixes

p n  m c d k M G
pico nano micro milli centi deci kilo mega giga
1012 109 106 103 102 101 103 106 109

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

1

Introduction to Groundwater Geochemistry

Groundwater geochemistry is the science that explores the processes controlling the chemical com-
position of groundwater, the groundwater quality. The groundwater quality influences the use of this
resource. Groundwater may contain hazardous substances that affect health when consumed or
which deteriorate the environment when the water is thoughtlessly spilled at the surface. The
groundwater quality may change during the exploitation or it may be affected by human activities of
which the impact is not always immediately evident.
The interest of society in groundwater geochemistry is mainly to ensure good quality drinking
water. Although drinking water can be manufactured, for example by desalinization, this still is a
costly affair, and to surrender to this option is in conflict with our desire to utilize groundwater as
a sustainable resource, refreshingly and cleanly flowing from a well. Preservation of good ground-
water therefore has a high priority for environmental authorities.
This chapter introduces basic subjects such as the concentration units, water sampling tech-
niques, and how to examine the accuracy of a chemical analysis. We begin with the concentration
limits for drinking water.

1.1 GROUNDWATER AS DRINKING WATER

Access to clean freshwater will be one of the biggest global resource problems of the coming
decades. One billion people had no access to clean drinking water from public supply in the year
2003. Probably, between 2 and 7 billion people will live in water scarce countries in the middle of
this century. Recent estimates suggest that climate change will account for about 20 percent of the
increase in global water scarcity (www.unesco.org). Lack of fresh water presents a global problem
of huge dimensions and a major effort is required to ensure good quality drinking water for the world
population. Much of the drinking water is derived from surface waters but, particularly in develop-
ing countries, groundwater is often preferred because it needs less treatment and has a better bacte-
riological quality which helps to minimize the spread of water-born diseases like cholera.
The general public is well aware of the value of good quality drinking water and consequently,
the sale of bottled waters is world-wide a big business. Figure 1.1 shows the labels of various min-
eral waters, and they seem to suggest that the composition of the bottled water has a favorable health
aspect.

1.1.1 Standards for drinking water

Table 1.1 shows the maximal admissible concentration in drinking water for a number of
common constituents imposed by most national authorities. Generally, the concentration
limits in Table 1.1 comply with the guidelines of the World Health Organization (WHO).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

2 Introduction to groundwater geochemistry

Figure 1.1. Labels of mineral water bottles, advertizing beneficial effects of drinking mineral water because
of their composition.

Interestingly, bottled mineral waters are not subject to these drinking water limits since, from a leg-
islation point of view, they are considered a medicine rather than drinking water and the comparison of
the analyses printed on labels of mineral water bottles with the values in Table 1.1 may be an interesting
exercise. Table 1.1 contains both elements with a natural origin and constituents mainly derived from
pollution (like nitrate and nitrite) or resulting from water treatment and distribution systems (Cu, Zn).
For pesticides the concentration of each pesticide should be less than 0.0001 mg/L while the sum of
the pesticides should not exceed 0.0005 mg/L.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Units of analysis 3

Table 1.1. Standards for the composition of drinking water and contribution of drinking water to the intake of
elements in nutrition. The European Union drinking water limits can be found on www.europa.eu.int in the
Official Journal L330 of December 5, 1998. For the USA look at www.epa.gov/safewater/. For WHO guidelines
see www.who.int/entity/dwq/en.

Constituent Contribution to Maximal admissible Comment

mineral nutrition (%) concentration (mg/L)

Mg2 3–10 50 Mg/SO4 diarrhea

Na 1–4 200
Cl 2–15 250 taste; safe  600 mg/L
SO42 250 diarrhea
NO3 50 blue baby disease
NO 2 0.5
F 10–50 1.5 lower at high water consumption
As
30 0.01 black-foot disease, skin cancer
Se 0.01
Al .. 0.2 acidification/Al-flocculation
Mn 0.05
Fe 0.2
Ni 0.02 allergy
Cu 6–10 2 3 mg/L in new piping systems
Zn negligible 0.1 5 mg/L in new piping systems
Cd .. 0.005
Pb .. 0.01
Cr 20–30 0.05
Hg 0.001

The average contribution of drinking water to the daily elemental intake with nutrition is also listed
(Safe Drinking Water Comm., 1980). For example for Na, the intake through drinking water is
insignificant as the Na intake completely is dominated by the addition of salt (NaCl) to our food.
However, for elements like F and As, the intake through drinking water is highly significant, and a
too high concentration of these two elements is a severe health threat affecting millions of people in
many places around the world. The intake of excess fluoride leads to painful skeleton-deformations
termed fluorosis; it is a common disease in African Rift Valley countries such as Kenya and Ethiopia,
where volcanic sources of F are important, and in India and West Africa where salts and sedimen-
tary F-bearing minerals are the primary sources. In India alone about 67 million people are at risk of
developing fluorosis (Jacks et al., 2000).
Groundwater high in arsenic may be found in inland basins in arid and semi-arid climate (Welch
et al., 2000). Another typical setting for high arsenic groundwater is in alluvial plain sediments
where strongly reduced groundwater, often with a high Fe2 concentration, is present. In both cases
the aquifers are situated in geologically young sediments and in rather flat, low-lying areas were the
extent of flushing with groundwater is low (Smedley and Kinniburgh, 2002). Arsenic is also associ-
ated with sulfide minerals, and may therefore be high in mine waste waters. Chronic arsenic poi-
soning can result in skin lesions, hyperkeratosis (thickening, hardening and cracking of palms and
soles), skin cancer and liver disease (Karim, 2000; Figure 1.2). In Bangladesh an estimated 20–70
million people are at risk from drinking high arsenic groundwater and it probably presents the biggest
case of mass poisoning in history (Halim, 2000).

1.2 UNITS OF ANALYSIS

The concentration of dissolved substances in water can be reported in different units, depending on
the purpose of the presentation and also on tradition. Some common units are listed in Table 1.2.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
4 Introduction to groundwater geochemistry

Figure 1.2. Skin lesions (black foot disease) as result of drinking water with high concentrations of arsenic.

What is the best unit to use? The labels of bottled mineral waters in Figure 1.1 suggest that
milligrams per liter is the obvious unit to express the analysis and mg/L is used by most laboratories
to report their analytical results. Other common units are parts per million (ppm  1 mg/kg) and
parts per billion (ppb  1 g/kg). These are numerically equal to mg/L and g/L, respectively, when
the density of the water sample is 1 kg/L, as is the case for dilute fresh waters. However, the density
of evaporation brines is much higher and a correction is required when converting the results
reported per volume to concentrations by weight.

Table 1.2. Concentration units for dissolved substances.

mg/L milligrams per liter sample

g/L micrograms per liter sample
ppm parts per million by weight of sample
ppb parts per billion by weight of sample
mmol/L millimoles per liter sample
mol/L micromoles per liter sample
meq/L milliequivalents per liter of sample
mmolc/L milliequivalents per liter of sample
epm equivalents per million, by weight of sample
M molality, moles per kg of H2O
mM millimoles per kg of H2O
N normality, equivalents per liter

The unit molarity (mol/L) reflects the number of molecules of a substance (Avogadro’s number:
1 mol  6.0221023 molecules), rather than their weight, and has the advantage that changes in
concentration can be related directly to the coefficients in reaction equations. For example, if 1 mmol
gypsum/L dissolves (the chemical formula of gypsum is CaSO4 2H2O), the Ca2 concentration

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Units of analysis 5

increases by 40 mg/L and the SO42 concentration by 96 mg/L, while the increase for both is simply
1 when expressed in mmol/L.
The official SI-unit is moles/m3, but this unit is not commonly applied. Numerically, moles/m3
are equal to mmol/L, the unit we often use in this book. To recalculate an analysis from mg/L to
mmol/L, the numbers are divided by the gram formula weight for each species. The gram formula
weight is the weight in grams of 1 mol of atoms or molecules (Example 1.1). The unit milliequivalent
per liter (meq/L or mmolcharge/L, abbreviated as mmolc /L) is the molar concentration multiplied by
the charge of the ions (1 mmol Na/L  1 meq Na/L; 1 mmol Ca2/L  2 meq Ca2/L).
Milliequivalents per liter are useful for checking the charge balance of chemical analyses.

Table 1.3. Recalculation of analysis units.

mmol/L  mg/L / (gram formula weight)

mmol/L  ppm (density of sample) / (gram formula weight)
mmol/L  meq/L / (charge of ion)
(weight solution  weight solutes)
mmol/L  molality  density   1000
(weight solution)

The unit molality gives the number of moles per kg of H2O and is generally used in physical chemistry.
The advantage is that this concentration unit is independent of a change in density with temperature,
or a change in concentration of other constituents. In normal fresh to slightly brackish water, molality
is identical with molarity (mol/L). The density of sea water at 25°C is 1.023 kg/L as compared to
0.997 kg/L for air-free pure water so that the difference between molality and molarity amounts to
approximately 2.5%. Table 1.3 presents equations for the conversion between different units and
some examples are given in Example 1.1.

EXAMPLE 1.1. Recalculation of concentration units

1. Gram formula weights are calculated from the periodic system as reproduced in Table 1.4 from the
Handbook of chemistry and physics.
The mass of 1 mol Ca is 40.08 grams.
1 mol SO42 weighs: 32.06 grams from sulfur  4  15.9994 grams from oxygen, in total: 96.06 grams.
2. Conversion of mg/L to mmol/L is obtained by dividing by the weight of the element or molecule.
Thus, a river water contains 1.2 mg Na/L;
This corresponds to 1.2 / 22.99  0.052 mmol Na/L;
The sample also contains 0.6 mg SO24 /L;
This equals 0.6 / 96.06  0.006 mmol SO24 /L.

3. The term mmol/L indicates the number of ions or molecules in the water when multiplied by Avogadro’s
number. For Na in the river water sample it amounts to 0.052103  6.0221023  3.11019 ions of
Na in 1 liter of water. (Quite a lot really!)
4. Ions are electrically charged, and the sums of positive and negative charges in a given water sample must
balance. This condition is termed the electroneutrality or electrical balance of the solution. Since mmol/L
represents the number of molecules, it should be multiplied by the charge of the ions to yield their total
charge in meq/L. Thus:
0.052 mmol Na/L  1  0.052 meq/L;
1.8 mmol Ca2/L  2  3.6 meq/L;
0.41 mmol SO2 4 /L  2  0.82 meq/L.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

Table 1.4. Periodic table of the elements. Reprinted with permission from Chem. Eng. News, February 4, 1985, 63(5), p. 26–27. Published in 1985 by the American

6
Chemical Society.

Introduction to groundwater geochemistry

1 2 New notation 13 14 15 16 17 18
Group Previous IUPAC form IIIB IVB VB VIB VIIB
IA IIA CAS version IIIA IVA VA VIA VIIA VIIIA Orbit

1 1 2 0
H 1 He
(Numbers in parentheses are mass numbers of most stable isotope of that element)
1.00794
4.0020602
1 2 K
3 1 4 2 KEY TO CHART 5 3 6 2 7 1 8 2 9 1 10 0
4 N 2 F Ne
Li Be B C 3 O
Atomic Number 50 2 Oxidation States 4
4
Symbol Sn 4 5
1987 Atomic Weight 118.71 1
6.941 9.012182 10.811 12.011 14.00674 2 15.9994 18.9984032 20.1797
2-1 2-2 18 18 4 Electron Configuration 2-3 2-4 2-5 3 2-8 K-L
2-6 2-7
11 1 12 2 13 3 14 2 15 3 16 4 17 1 18 0
Na Mg AI 4 P 5 S 6 CI 5 Ar
Si 3 7
4 2
3 4 5 6 7 8 9 10 11 12 1
22.989768 24.3050 IIIA IVA VA VIA VIIA VIIIA 26.981539 28.0855 30.97362 32.066 35.4527 39.948
2-8-1 2-8-2 IIIB IVB VB VIB VIIB VIII IB IIB 2-8-3 2-8-4 2-8-5 2-8-6 2-8-7 2-8-8 K-L-M
19 1 20 2 21 3 22 2 23 2 24 2 25 2 26 2 27 2 28 2 29 1 30 2 31 3 32 2 33 3 34 4 35 1 36 0
K Ca Sc Ti 3 V 3 Cr 3 Mn 3 Fe 3 Co 3 Ni 3 Cu 2 Zn Ga Ge 4 As 5 Se 6 Br 5 Kr
4 4 6 4 3 2 1
5 7
39.0983 40.078 44.955910 47.88 50.9415 51.9961 54.93085 55.847 58.93320 58.69 63.546 65.39 69.723 72.61 74.92159 78.96 79.904 83.80
-8-8-1 -8-8-2 -8-9-2 -8-10-2 -8-11-2 -8-13-1 -8-13-2 -8-14-2 -8-15-2 -8-16-2 -8-18-1 -8-18-2 -8-18-3 -8-18-4 -8-18-5 -8-18-6 -8-18-7 -8-18-8 -L-M-N
37 1 38 2 39 3 40 4 41 3 42 6 43 4 44 3 45 3 46 2 47 1 48 2 49 3 50 2 51 3 52 4 53 1 54 0
Rb Sr Y Zr Nb 5 Mo Tc 6 Ru Rh Pd 4 Ag Cd In Sn 4 Sb 5 Te 6 I 5 Xe
7 3 2 7
1
85.4678 87.62 88.90585 91.224 92.90638 95.94 (98) 101.07 102.90550 106.42 107.8682 112.411 114.82 118.710 121.75 127.60 126.90447 131.29
-18-8-1 -18-8-2 -18-9-2 -18-10-2 -18-12-1 -18-13-1 -18-13-2 -18-15-1 -18-16-1 -18-18-0 -18-18-1 -18-18-2 -18-18-3 -18-18-4 -18-18-5 -18-18-6 -18-18-7 -18-18-8 -M-N-O
55 1 56 2 57* 3 72 4 73 5 74 6 75 4 76 3 77 3 78 2 79 1 80 1 81 1 82 2 83 3 84 2 85 86 0
Cs Ba La Hf Ta W Re 6 Os 4 Ir 4 Pt 4 Au 3 Hg 2 Ti 3 Pb 4 Bi 5 Po 4 At Rn
7
132.90543 137.327 138.9055 178.49 180.9479 183.85 186.207 190.2 192.22 195.08 196.96654 200.59 204.3833 207.2 208.98037 (209) (210) (222)
-18-8-1 -18-8-2 -18-9-2 -32-10-2 -32-11-2 -32-12-2 -32-13-2 -32-14-2 -32-15-2 -32-16-2 -32-18-1 -32-18-2 -32-18-3 -32-18-4 -32-18-5 -32-18-6 -32-18-7 -32-18-8 -N-O-P
87 1 88 2 89** 104 105 106 107
Fr Ra Ac 3 Unq 4 Unp Unh Uns
(223) 226.025 227.028 (261) (262) (263) (262)
-18-8-1 -18-8-2 -18-9-2 -32-10-2 -32-11-2 -32-12-2 -32-13-2 OPQ

58 3 59 3 60 3 61 3 62 2 63 2 64 3 65 3 66 3 67 3 68 3 69 3 70 2 71 3
Ce 4 Pr Nd Pm Sm 3 Eu 3 Gd Tb Dy Ho Er Tm Yb 3 Lu
*Lanthanides 140.115 140.90765 144.24 (145) 150.36 151.965 157.25 158.92534 162.50 164.93032 167.26 168.93421 173.04 174.967
-20-8-2 -21-8-2 -22-8-2 -23-8-2 -24-8-2 -25-8-2 -25-9-2 -27-8-2 -28-8-2 -29-8-2 -30-8-2 -31-8-2 -32-8-2 -32-9-2 NOP
90 4 91 5 92 3 93 3 94 3 95 3 96 3 97 3 98 3 99 3 100 3 101 2 102 2 103 3
Th Pa 4 U 4 Np 4 Pu 4 Am 4 Cm Bk 4 Cf Es Fm Md 3 No 3 Lr
5 5 5 5
**Actinides 6 6 6 6
232.0381 231.03588 238.0289 237.048 (244) (243) (247) (247) (251) (252) (257) (258) (259) (260)
-18-10-2 -20-9-2 -21-9-2 -22-9-2 -24-8-2 -25-8-2 -25-9-2 -27-8-2 -28-8-2 -29-8-2 -30-8-2 -31-8-2 -32-8-2 -32-9-2 OPQ

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Groundwater quality 7

Some terms often used in relation to groundwater chemistry are explained in Table 1.5. Apart from
standard chemical terminology, like pH and Eh, it also contains specific items from the water world.
For example, hardness of water is among other things used to dose the amount of soap needed in a
washing machine.

Table 1.5. Common parameters used in the water world (cf. Standard Methods, 1985; Hem, 1985).

Hardness Sum of the ions which can precipitate as “hard particles” from water. Sum of
Ca2 and Mg2, and sometimes Fe2. Expressed in meq/L or mg CaCO3/L or
in hardness degrees. 100 mg CaCO3/L  1 mmol CaCO3/L  2 meq Ca2/L
Hardness degrees 1 german degree  17.8 mg CaCO3/L
1 french degree  10 mg CaCO3/L
Temporary hardness Part of Ca2 and Mg2 concentrations which are balanced by HCO3
(all expressed in meq/L) and can thus precipitate as carbonate
Permanent hardness Part of Ca2 and Mg2 in excess of HCO 3 (all expressed in meq/L)
Color Measured by comparison with a solution of cobalt and platinum
EC Electrical Conductivity, in S/cm ( mho/cm), EC
100  meq (anions or
cations)/L
pH log[H], the log of H activity (dimensionless).
Eh Redox potential, expressed in Volt. measured with platinum/reference electrode
pe Redox potential expressed as log[e]. [e] is “activity” of electrons.
pe  Eh/0.059 at 25°C.
Alkalinity (Alk) Acid neutralizing capacity. Determined by titrating with acid down to a pH of
about 4.5. Equal to the concentrations of mHCO3  2 mCO 23(mmol/L) in most
samples.
Acidity Base neutralising capacity. Determined by titrating up to a pH of about 8.3.
Equal to H2CO3 concentration in most samples except when Al3 or Fe3 are
present
TIC Total inorganic carbon
TOC Total organic carbon
COD Chemical oxygen demand. Measured as chemical reduction of permanganate or
dichromate solution, and expressed in oxygen equivalents
BOD Biological oxygen demand.

1.3 GROUNDWATER QUALITY

The classical use of water analyses in hydrology is to show the regional distribution of the water com-
positions in a map. Such maps serve environmental authorities, water resource managers, drilling
operators and other practitioners to identify aquifers with good quality groundwater, but they are also
useful for a first assessment of the relation between the aquifer mineralogy and groundwater com-
position. For example, groundwater from a limestone aquifer is likely to contain enhanced calcium
and bicarbonate concentrations.
A standard groundwater chemical analysis will as a minimum comprise values for temperature, EC,
pH, the four major cations (Na, K, Mg2, Ca2) and four major anions (Cl, HCO 2 
3 , SO4 , NO3 ),
in other words eleven variables. When the number of available water analyses in an area accumulates,
it becomes increasingly more difficult to overview all the numbers. Therefore graphical methods have
been developed to display the chemical composition of the main components in groundwater at a
glance. Figure 1.3 shows examples of two common diagram types, the bar diagram and the circle dia-
gram. The bar diagram consists of two adjacent columns; the left column for cations and the right col-
umn for anions. Each column shows the stacked contribution of the various ions with concentrations
expressed in milliequivalents per liter. Since the sum of positive charges should be balanced by the sum
of negative charges, the height of the two columns should be the same for a good water analysis.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

8 Introduction to groundwater geochemistry

Sample 1 Sample 1 Sample 3

15 Na  K
Ca

Mg
Cl SO4

0 1000 mg/L
10
meq/L

HCO3 Scale of diameters

Sample 3

Alkali – Na  K

Magnesium  Mg2
5
Calcium  Ca2

Chloride  Cl
2
Sulfate  SO4
 2
Carbonates  HCO3  CO3

Figure 1.3. Bar and circle diagrams, two graphical methods to display the bulk chemical composition of
groundwater (Domenico and Schwartz, 1997).

In the circle diagram the size of the circle is proportional to total dissolved solids. The circle is then
subdivided in an upper half showing the relative composition of the cations, and a lower half with the
anions. Again, concentrations are expressed in milliequivalents per liter. In both bar and circle
diagrams, waters with a similar chemical composition are quickly recognized.
A further step in visualizing is the Stiff diagram. It consists of three to four horizontal axes dis-
playing selected components. On each axis a cation is plotted to the left and an anion to the right,
again in milliequivalents per liter. The uppermost axis has Na to the left and Cl to the right and
reflects a possible marine influence since in seawater NaCl is the dominant salt. The second axis has
Ca2 to the left and HCO 3 to the right and this axis is meant to display the dissolution of CaCO3.
The third axis has Mg2 to the left and SO42 to the right, presenting the remaining two major com-
ponents in most waters. The fourth axis is optional and its variables may change from study to study.
The values on each axis are connected by lines and a typical shape emerges for a given water com-
position. In Figure 1.4, sample 1 is rainwater with a low concentration of all the components. Sample 4
is from a limestone aquifer where CaCO3 dissolution is the predominant process yielding a typical
bird-like shape. Sample 3 is from an aquifer where dolomite, CaMg(CO3)2, is the dissolving mineral
and therefore the concentrations of Ca2 and Mg2 are similar, the bird seems to have lost half of its
left wing.
A large number of chemical analyses can be compiled in the so-called Piper diagram (Figure 1.5).
The Piper diagram contains two triangular charts for depicting the proportions of cations and anions,
expressed in meq/L. The triangle for cations has 100% Ca2 in the left corner, 100% Na  K
towards the right and 100% Mg2 upwards (traditionally, Na and K are combined because these
ions were analyzed jointly in a precipitate with uranium). The sum of the concentrations of the three
ions, in milliequivalents per liter, is recalculated to 100% and the relative composition is plotted in
the triangle. For anions the triangle has 100% carbonate to the left, Cl to the right and SO42 on top.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Groundwater quality 9

1 2 3

6 10 5 5 10 meq/L
K  Na CI
Ca2 HCO
3
2
Mg2 SO4
Fe NO3

Figure 1.4. Stiff-diagrams reflecting the chemical composition of groundwaters. For explanation see text.
Free software to construct Stiff diagrams can be downloaded from http://www. twdb.state.tx.us/publications/
reports/GroundWaterReports/Open-File/Open-File_01-001.htm

The two data points in the triangles are joined by drawing lines parallel to the outer boundary until
they unite in the central diamond shape (Figure 1.5). The relative chemical composition of the water
sample is now indicated by a single point. The diamond diagram often forms the background for
a descriptive terminology of the chemical composition of groundwater. For example a calcium-
bicarbonate water type must plot near the lefthand corner of the diamond while a sodium-bicarbonate
type will be located closer to the bottom corner.

80 80
Scale of diameters
0 mg/L
SO4  Cl Ca  Mg
500
1000

5000 20 3 20
Sample 2
10000

20 20
3
O

80 80
HC
Na

1



3
K

Mg SO4
CO

3 2
80 80
20 20
1 % meq/L 3
2 1
80 20 20 80
Ca Cl

Figure 1.5. The Piper diagram, used for graphically displaying the bulk chemical composition of groundwaters.
The two triangular plots give the relative composition of cations and anions, expressed in percentage of total cation
or anion meq/L. From the triangular diagrams the points are conducted to the diamond shaped diagram parallel to
the outer side of the diagram until they intersect (dotted lines for sample 1). The size of the circle in the diamond
indicates the total dissolved solid concentration (Domenico and Schwartz, 1997). Free software for constructing
Piper diagrams can be downloaded from http://water.usgs.gov/nrp/gwsoftware/GW_Chart/GW_Chart.html

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

10 Introduction to groundwater geochemistry

(Ca, Mg)SO4
EC(µS/cm)
(Ca, Mg)CI2
Shale, sandstone, greywacke, quartzite, ......... 150–350
schist, gneiss
80 80
Marl  CaSO4/MSO4 (magmatic rock) ............ 800–1300

60 60 Limestone, marl (magmatic rock) ..................... 1200–2000

CI

Ca



Magmatic rock .................................................. 1500–1700

Mg
SO 40 40
Schist, quartzite, granite, sandstone (FeS2) .. 1300–2100
20 20
(Ca, Mg)(HCO3)2 (Na, K)2SO4
(Na, K)CI
Mg SO4

80 80
Na

K

60 3
O 60
Mg

HC

SO
40 40

4
20 20

Ca 80 60 40 20 NaK 20 40 60 80 CI
Ca HCO3 Cl

Figure 1.6. A Piper plot of European bottled mineral waters and their relation to the rock type from which the
water was extracted (modified from Zuurdeeg and Van der Weiden, 1985).

Figure 1.6 shows a Piper diagram with the composition of mineral waters from Europe and their rela-
tion to the rock type from which they have been extracted (Zuurdeeg and Van der Weiden, 1985). Note
that mineral waters are retrieved from almost any rock type and accordingly there is also a large vari-
ation in their chemical composition. The diagram shows that limestones and marls produce a Ca, Mg-
HCO3 type of water, whereas groundwater in metamorphic rocks (schists, sandstones, etc.) or igneous
rocks (granites and magmatic rocks) also contain high concentrations of elements such as Na, K,
and Cl. The electrical conductivity (EC) of the mineral waters indicate that appreciable amounts of
salts are present that may exceed the limit for a regular drinking water supply (Table 1.1).

QUESTION:
The composition of sample 1 in Figure 1.5 is (mg/L): Na 245; Mg2 21.9; Ca2 61.7; Cl 11.3; HCO
3
906; SO42 21.2. Calculate the cation and anion percentages, and compare with Figure 1.5.

1.4 SAMPLING OF GROUNDWATER

Both the sampling and the chemical analysis of groundwater are expensive and particularly the cost
of drilling new boreholes is high. It is therefore important to evaluate beforehand which data are
required to solve a specific problem. Furthermore, in the interpretation of the chemical analyses,
both the well construction and the sampling procedures should be considered when judging the rep-
resentativeness of the samples for the actual groundwater composition.

1.4.1 Depth integrated or depth specific sampling

Figure 1.7 shows two different approaches towards sampling of groundwater. The borehole on the
left has a long screen and the sample obtained is integrated over the length of the screen as well as
over the permeability of the formation. The screen interval with fine grained sand will contribute
relatively less to the total yield than the coarse grained part with a much higher permeability.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sampling of groundwater 11

Depth integrated Depth specific

Q Q

Long screen

Short screen

Figure 1.7. Depth integrated versus depth specific groundwater sampling. Coarse dotted areas represent
coarse grained sand and densely dotted areas fine grained sand. The size of the arrows reflects the flow rate
(modified from Cherry, 1983).

The borehole to the right has a short screen and will provide depth specific information on the chem-
ical composition of the groundwater in a distinct sediment layer. Depth specific sampling is usually
required for detailed studies of the chemical processes in the aquifer.
In many cases groundwater compositions show major variations with depth even on a small scale.
Integrated samples from a screen interval of several meters may accordingly represent mixtures of
waters with different concentrations and the mixing process may even induce chemical reactions dur-
ing sampling. Figure 1.8 displays a groundwater chemistry profile, obtained by depth specific sam-
pling in a sandy aquifer with a well defined oxic zone on top of an anoxic zone containing Fe2.

Concentration (mg/L)
0 2 4 6 8 10 12
10
20/10-88
15/11-88

O2
20
Depth (m)

30
Fe2ⴙ

40

Figure 1.8. The distribution of O2 and Fe2 in a sandy aquifer.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

12 Introduction to groundwater geochemistry

Suppose the borehole has a screen installed that draws water from both the oxic and anoxic zones.
The water with O2 will mix and react with Fe2 containing water and the concentrations of O2, Fe2
and pH or alkalinity will alter by the reaction:
2Fe2  1⁄2O2  3H2O → 2FeOOH  4H (1.1)
The composition of the resulting sample depends on the extent of mixing and reaction. In principle
even a sample containing both Fe2 and O2 might be captured although such a water quality is not
present in the subsoil. In any case, the resulting sample will not reflect the in situ conditions in the
aquifer very well. Similar reactions occur in shallow, open wells with a large body of stagnant water
in contact with air. Production wells, usually equipped with screens of tens of meters, may be
pumped sectionwise in an attempt to obtain more detailed, depth specific information (Lerner and
Teutsch, 1995).

1.4.2 Procedures for sampling of groundwater

The first concern in sampling is the contamination and disturbance of natural conditions caused by
drilling operations. It may take a long time before the influence of new materials brought into the
aquifer, including drilling fluids, gravel pack or casing materials, is sufficiently diminished to allow
for representative sampling. Water from boreholes where drilling mud has been used or where clay
layers are sealed with bentonite (a swelling clay mineral) often displays cation exchange with the
clay. If trace metals or trace organics are going to be analyzed, special care is required to avoid
contamination by or sorption onto the sampling materials. The time period that should pass before
representative samples can be obtained depends on the groundwater flow rate, the ion exchange
capacity etc., and may be in the order of two to three months when pumping on the screen is limited
(as is often the case with depth specific samplers).
Wells which have been out of production for some time may yield a water chemistry that is dif-
ferent from the composition during pumping. The main reason is the presence of stagnant water
above the screen in the well and it is therefore necessary to empty the well for a number of volumes.
On the other hand, excessive pumping may draw waters with a different composition towards the
screen and cause mixing of waters. Thus, a balance has to be found between these two aspects.

12
Temp.(°C)

11
EC (µS/cm)

1900
EC
10
Temp.
9 1800

50
9 Eh
Eh (mV)

150
pH

8
pH 250
7
350
0 1 2 3 4
Well volumes

Figure 1.9. The change in chemical composition of discharge during well flushing (modified from Lloyd and
Heathcote, 1985).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sampling of groundwater 13

Estimates for the number of well volumes to be emptied vary between 2 and 10 times and depend on
the local hydrological conditions (Barber and Davis, 1987; Robin and Gillham, 1987; Stuyfzand,
1983). In most cases 2 to 4 times seems sufficient.
To accomplish effective flushing, the well should be pumped from just below the air-water interface
(Robin and Gillham, 1987). The best way to evaluate the degree of flushing needed is to monitor easily
measured field parameters such as the electrical conductivity (EC) or pH over time. The example
shown in Figure 1.9 demonstrates that in this case stationary conditions are obtained after emptying
about two well volumes.
Problems with sampling from existing wells are often related to faulty completion of wells and
are not always easy to detect. Some examples of commonly encountered problems are illustrated in
Figure 1.10. In Figure 1.10A, the screen length is too large causing drainage of water from an upper
sandy layer, with a higher pressure, into a lower sandy layer with a lower hydraulic potential. Case B
displays an improperly placed bentonite seal which results in leakage from the upper part of the
aquifer into the lower part. In C, the gravel pack functions as a drain which also may cause short cir-
cuiting. There are examples of the application of herbicide to well fields, intended to make things
look pleasant at the surface, but resulting in downward seepage of the chemical along the outer side
of the well casing and polluting the well. Finally, case D displays the effect of leaky coupling of the
casing.

A B C D

Gravel
pack

Clay

Figure 1.10. Different forms of leakage and short circuiting during groundwater sampling from screened
wells (modified from Stuyfzand, 1989). For explanation see text.

For proper depth specific sampling, dedicated systems are required (Figure 1.11). In shallow
aquifers a drive-point piezometer, hammered or pushed down, may provide high-resolution profiles
of the groundwater chemical composition (Jakobsen and Postma, 1999). Towards greater depth, a
hollow stem auger equipped with a water sampling system may be useful (Figure 1.11), especially
when combined with geophysical sensors to delineate the geological variations (Sørensen and
Larsen, 1999). It is also possible to take cores, which can be centrifuged or pressurized to procure
water samples from clay layers. The disadvantage of the drive-point piezometer is that resampling
the groundwater needs renewed augering, and therefore permanent installations might be preferred
for monitoring purposes. Sampling points can be installed individually, or in a multiple-level, single
borehole piezometer (Obermann, 1982; Lerner and Teutsch, 1995).
Single hole multiple piezometers can be introduced as bundles (Cherry et al., 1983; Appelo et al.,
1982; Stuyfzand, 1983; Leuchs, 1988) or enclosed in a casing (Pickens et al., 1978; Postma et al.,
1991). While the first approach is by far the cheapest, it requires cohesionless aquifer material to
ensure sealing between sampling points, and the risk of short circuiting between different sample
levels through the bundles of tubing always remains even when bentonite is applied in-between.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

14 Introduction to groundwater geochemistry

Conventional piezometers Multiple piezometers Hollow stem

in a single casing auger

Air tube
Water tube

Inflatable
Standpipe
bladder
piezometer
Casing
Water inlets Check valve
Screen or
Narrow dia. Cleaning tube
slotted tip
piezometer

Figure 1.11. Three approaches for depth specific groundwater sampling; left, multiple-borehole piezometer
nest; middle, multiple piezometers enclosed in a casing (Cherry, 1983); right, hollow stem auger with a water
sampling device (Sørensen and Larsen, 1999).

This risk increases with the number of sample levels used and the difference in hydraulic head in the
layers which have been penetrated.
Sample retrieval is done most easily by suction if the water table is less than 9 m below the sur-
face. However, the application of vacuum may result in degassing of the sample (Suarez, 1987;
Stuyfzand, 1983) and produce erratic results for dissolved gases, pH and volatile organics.
Alternative approaches are downhole pumps (Gillham and Johnson, 1981), or double line gas driven
sampling devices (Appelo et al., 1982). In the latter approach (Figure 1.12), a downhole reservoir
being filled with groundwater, is emptied by pressurizing the sampler with an inert gas like N2 or Ar.

N2
Sample

Casing

Sample reservoir

Valve
Sample intake
Filter

Figure 1.12. Sampling of groundwater using a downhole reservoir with a check valve. Pressurizing the reser-
voir with N2 or Ar, closes the check valve and the sample is transported towards the surface.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chemical analysis of groundwater 15

A check valve at the bottom of the reservoir is then closed and the sample transported to the surface.
Upon (gentle) release of the pressure, the reservoir is filled again. There is, however, still a risk for
gas exchange at the interfaces between the sample and the pressurizing gas. Samples should there-
fore preferably be taken in the middle part of the sampling volume.

1.5 CHEMICAL ANALYSIS OF GROUNDWATER

There is probably no such thing as a perfect chemical water analysis carried out by routine proce-
dures. For specific purposes both the sampling and analytical procedures have to be adjusted to
obtain the best possible results for the critical parameters in the given study.
The analytical chemistry of groundwater in general is beyond the scope of this book. In most
cases water analyses are carried out by standard procedures such as described in handbooks like
“Standard Methods for the examination of water and wastewater” (1985). However, some aspects
concerning collection and conservation of water samples, and the evaluation of the quality of chem-
ical analysis deserve special attention.

1.5.1 Field analyses and sample conservation

Water samples may contain variable concentrations of suspended solids or colloids that need to be
removed before the analysis of the “real” solute ions can begin. The size of the colloids ranges from
109 to about 104 m, and actually, the scale from solutes to colloids is a sliding one (Stumm and
Morgan, 1996). The standard procedure is to filtrate through 0.45 m membrane filters. However,
finely dispersed Fe- and Al-oxyhydroxides may pass through 0.45 m filters, and 0.1 m is a better
choice and able to remove viruses as well (Kennedy and Zellweger, 1974; Laxen and Chandler,
1982). Pressure filtration, using an inert gas is preferred since vacuum application may degas the
sample. Disposable syringes and filter cartridges are often handy for sample manipulation and to
keep the sample out of contact with the atmosphere. Access of atmospheric oxygen may result in the
oxidation of components like Fe2, H2S, etc. which are often present in anoxic groundwater.
Furthermore, degassing of CO2 may occur, causing changes in pH, alkalinity and total inorganic car-
bon, which may also induce carbonate precipitation (Suarez, 1987). Also degassing of methane may
occur.
Precautions to avoid changes in the chemical composition of the sample before analysis consist
of either measuring sensitive components in the field or the conservation of samples for later analy-
sis in the laboratory. An overview of the required treatment is presented in Table 1.6. Conservation
is in most cases done by adding acid to the sample until the pH is 2 (0.7 mL of 65% HNO3 is usu-
ally sufficient to neutralize alkalinity and to acidify 100 mL sample). Acidification stops most bac-
terial growth, blocks oxidation reactions, and prevents adsorption or precipitation of cations. The
effect of improper handling on the analytical results is illustrated in Example 1.2.

EXAMPLE 1.2. Effect of iron oxidation on analytical results

A groundwater sample contains 20 mg Fe2/L. How does the alkalinity change when this iron oxidizes and
precipitates as Fe(OH)3 in the sample bottle?
The reaction in the bottle is:

2Fe2  4HCO
3  ⁄2O2  5H2O → 2Fe(OH)3  4H2CO3
1

When 20 mg Fe/L precipitates, this corresponds to 20 / 55.8  0.36 mmol/L. The reaction consumes twice
this amount of HCO
3 , and the alkalinity is expected to decrease with 2  0.36  0.72 meq/L.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

16 Introduction to groundwater geochemistry

Table 1.6. Conservation of chemical parameters of water samples.

Parameter Conservation/field analysis

Na, K, Mg2, Ca2 Acidify to pH 2 in polyethylene container (preferably HNO3 for AAS or ICP-analysis).
NH 3
4 , Si, PO4
Heavy metals Acidify to pH 2 in glass or acid rinsed polypropylene container.
SO42, Cl Cool to 4°C.
NO3 , NO2

Store cool at 4°C and analyze within 24 hours or add bactericide like thymol.
(Note that NO  
3 may form from NH4 in reduced samples. NO2 may
self-decompose even when a bactericide is added.)
H2S To avoid degassing, collect sample in a Zn-acetate solution, precipitating ZnS.
Spectrophotometry in the field or later in the laboratory.
TIC Dilute sample to TIC  0.4 mmol/L. (This effectively reduces CO2 pressure,
and prevents the escape of CO2).
Alkalinity Field titration with the GRAN method (Stumm and Morgan, 1996)
Fe2 Spectrophotometry in the field. Alternatively determined as Fe-total in an
acidified sample.
pH, Temp., EC, O2 Field measurement in a flow cell.
CH4 Unfiltered sample collected avoiding degassing, then acidified.

Field analyses are usually carried out for parameters like pH, EC, Eh, O2, which are measured by elec-
trode, and sometimes also for alkalinity, Fe2 and H2S. Electrode measurements should preferably be
carried out in a flow cell mounted directly on the well head in order to prevent air admission. EC
measurements are particularly useful as a control on analysis and conservation of samples. Eh
measurements only give a qualitative indication of the redox conditions and should be made as sloppy
as possible, so you will not be tempted to relate them to anything quantitative afterwards (Chapter 9).
Figure 1.13 compares pH measurements carried out immediately in the field with later labora-
tory measurements and shows substantial differences. These differences are even more significant
when it is remembered that pH is a logarithmic unit (pH  log[H]). Since the pH is of major
importance in geochemical calculations, care should be taken to obtain reliable measurements.

7
pH field

4
4 5 6 7 8
pH laboratory

Figure 1.13. Comparison of field measurements of pH, in a carbonate-free sandy aquifer, with those per-
formed in the laboratory (Postma, unpublished results).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chemical analysis of groundwater 17

While a pH measurement in principle is a simple operation, there are a number of potential sources of
error. Some problems are related to the removal of groundwater from its position in the subsoil, usually
anoxic and possibly at a high hydrostatic pressure. Most problems can be overcome by careful sampling
and field measurement procedures. In particular, the use of in-line flow cells in a pressurized sampling
system helps to minimize the problems of pH measurement related to the change of environment.
Another source of error in low ionic strength groundwaters is the liquid junction between the refer-
ence electrode and the solution to be measured. The fundamental problem is that the liquid junction
potential across the porous ceramic plug of the reference electrode will vary with the solution com-
position (Bates, 1973). Liquid junction problems in low ionic strength waters have been recognized
(Illingworth, 1981; Brezinski, 1983; Davison and Woof, 1985) and may amount to several tens of a
pH unit. A tricky aspect of this error is that it is not revealed by a standard two buffer calibration
(Illingworth, 1981; Davison and Woof, 1985), since commercial buffers are all high ionic strength
solutions. Electrode performance should therefore be checked in dilute solutions, for example a
solution of 104 M HCl or 5105 M H2SO4, which should yield a pH of 4.00  0.02, cf. Davison
(1987) and Busenberg and Plummer (1987).

1.5.2 Accuracy of chemical analysis

In general, two types of errors are discerned in chemical analyses:
Precision or statistical errors which reflect random fluctuations in the analytical procedure.
Accuracy or systematic errors displaying systematic deviations due to faulty procedures or
interferences during analysis.
The precision can be calculated by repeated analysis of the same sample. It is always a good idea to
collect a number of duplicate samples in the field as a check on the overall procedure. Systematic
errors can be tested only by analyzing reference samples and by interlaboratory comparison of the
results. At low concentrations, duplicate analyses may show large variations when the sensitivity of
the method is insufficient. The accuracy of the analysis can for major ions be estimated from the
electrical balance (E.B.) since the sum of positive and negative charges in the water should be equal:

(Sum cations  Sum anions)

Electrical Balance (E. B., %)   100 (1.2)
(Sum cations  Sum anions)

where cations and anions are expressed as meq/L and inserted with their charge sign. The sums are
taken over the cations Na, K, Mg2 and Ca2, and anions Cl, HCO 2 
3 , SO4 and NO3 . Sometimes
2 
other elements contribute significantly, for example ferrous iron (Fe ) or NH4 in reduced ground-
water, or H and Al3 in acid water. The presence of the last two substances in significant amounts
requires more accurate calculations using a computer speciation program like PHREEQC which will
be presented in Chapter 4. Differences in E.B. of up to 2% are inevitable in almost all laboratories.
Sometimes an even larger error must be accepted, but with deviations in excess of 5% the sampling
and analytical procedures should be examined.

EXAMPLE 1.3. Estimating the reliability of water analyses

Your laboratory returns the following water analysis:
pH  8.22 EC  290 S/cm
Na K Mg2 Ca2 Cl HCO
3 SO42 NO
3

13.7 1.18 3.2 42.5 31.2 79.9 39 1.3 mg/L

Is this a reliable analysis?

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

18 Introduction to groundwater geochemistry

A first requirement for a water analysis is a reasonable charge balance of cations and anions: the solution
should be electrically neutral. The analysis is recalculated from mg/L to mmol/L and then to meq/L.
Formula wt.
mg/L (gram/mol) mmol/L charge meq/L

Na 13.7 / 22.99  0.60  1  0.60

K 1.18 / 39.1  0.03  1  0.03
Mg2 3.21 / 24.31  0.13  2  0.26
Ca2 42.5 / 40.08  1.06  2  2.12
 3.01
Cl 31.2 / 35.45  0.88  1  0.88
HCO 3 79.9 / 61.02  1.31  1  1.31
SO42 39 / 96.06  0.41  2  0.82
NO3 1.3 / 62.0  0.02  1  0.02
 3.03

The difference between cations and anions is only 0.02 meq/L. The electrical balance is:

E.B.  (3.01  (3.03)) / (3.01  (3.03))  100  0.02 / 6.04  0.3%.

which is very good indeed.

Another useful technique is to compare the calculated and measured electrical conductivity.
Different ions have a different molar conductivity m which is defined as:
 m  ECi / mi (1.3)
where ECi is the electrical conductivity (Siemens/m, S/m), contributed by ion i with a concen-
tration of mi mol/L. The m has accordingly the units of S m2/mol. The conductivity is related to
the mobility of the ion and its charge, and varies for different electrolyte ions as noted in Table 1.7.

Table 1.7. Mobility and conductivity for ions at trace concentration in water at 25°C. Conductivities from
Landolt and Bornstein (1960).

Na K Mg2 Ca2 Cl HCO

3 SO42

Conductivity 50.1 73.5 106 119 76.35 44.5 160 S cm2/mol

8 2
Mobility 5.19 7.62 5.46 6.17 7.91 4.61 8.2910 m /s/V

The unit for the molar conductivity in the table is S cm2/mol which is equal to S/cm L/mmol. The
conductivity of a solution is, at least for trace concentrations, obtained by multiplying the conductivity
listed in Table 1.7 with the concentrations of the ions in mmol/L and summing up. For example, the EC
of a solution of 0.1 mM NaCl is 50.1  0.1  76.35  0.1  12.6 S/cm, and of 0.3 mM Ca(HCO3)2
is 119  0.3  44.5  2  0.3  62.4 S/cm, both at 25°C. The ion mobility is obtained by dividing
the conductivity by the charge number |z| of the ion and the Faraday constant, F  96485 C/mol:
u   m /(| z | F ) (1.4)
where u is the mobility in 108 m2/s/V. The ion mobilities are related to the diffusion coefficients, as
we will see in Chapter 3.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chemical analysis of groundwater 19

150 KCI

CaCl2
NaCl
S/cm
meq/L

Ca(HCO3)2
100
NaHCO3
Equivalent conductance

Ca Mg
SO SO
4 4

50

Range in natural
fresh water

0
1 0 1 2
Concentration (log (meq/L))

Figure 1.14. The relation between equivalent electrical conductivity and concentration for different salts.

The molar conductivity generally decreases with increasing concentration as is illustrated in Figure
1.14 for various salts. Here, the equivalent conductivity (or conductance) is plotted, i.e. the molar
conductivity is divided by the sum of the charge numbers of the ions in the salt. The decrease of the
conductivity is due to ion-association and electrostatic shielding effects which increase with con-
centration. Figure 1.14 shows that a Ca(HCO3)2 fresh water will have an equivalent conductivity of
close to 100 S/cm per meq/L. Therefore, at 25°C, the EC divided by 100, will in most cases give
a good estimate of the sum of anions or cations (both in meq/L):
 anions   cations (meq/L)  EC /100 (S/cm ) (1.5)
This relation is valid for EC-values up to around 1500 S/cm. More complicated expressions have
been derived by Rossum (1975) and Stuyfzand (1987). Also relationships between the EC and total
dissolved solids (TDS in mg/L) have been suggested (Hem, 1985), but these are less reliable since
they contain an additional assumption about average molecular weight of the conducting ions.
The calculation of the E.B., and comparison with the EC, as a check on the chemical analyses,
is only applicable for major elements. The accuracy for minor elements is much more difficult
to estimate. Sometimes the incompatibility of different elements found together in one sample can
be a warning that something went wrong. Thus, it is unlikely that O2 or NO 3 , which indicate oxi-
dizing conditions, are found together with appreciable concentrations of Fe2 in natural water (see
Figure 1.8). Ferrous iron (Fe2) only occurs in appreciable concentrations (more than 1 mol/L)
in reduced environments. The solubility of ferric iron (Fe3) is low in water with a pH between 3
and 11, and it precipitates rapidly as Fe(OH)3. The total iron concentration should therefore also be
low in water where O2 or NO 3 is present. Also, the solubility of aluminum is low in the pH range
5 to 10, and it is unlikely to find an Al concentration of more than 1 mol/L in water having a near
neutral pH.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

20 Introduction to groundwater geochemistry

EXAMPLE 1.4. Analytical errors due to precipitation in the sampling bottle

In samples brought from Portugal to the laboratory in Amsterdam a large charge imbalance was found. A
sample consisted of 100 mL water acidified to pH 2 for the analysis of cations, and 200 mL water for the
analysis of HCO  2 
3 , Cl , SO4 , NO3 as well as pH and EC. The samples were transported by car, and were
subjected to significant temperature variations. A typical example is given below:

temp.  18°C
pHfield  7.00 pHlab.  7.62
ECfield  720 EClab.  558 S/cm

Na K Mg2 Ca2 Cl HCO

3 SO42 NO3

0.55 0.01 0.35 3.00 0.64 4.31 0.16 0.12 mmol/L

Sum cations  7.36; Sum anions  5.38 meq/L

The sum of anions and cations in this sample differs by 16%, and could not be improved by duplicate analy-
ses. It was realised, however, that cations and anions were analyzed in different bottles. When the cations were
analyzed in the unacidified sample bottle, lower concentrations of Ca2 were found. The reason for the imbal-
ance was clearly that calcite had precipitated in the bottle without acid, thus decreasing the HCO
3 concentration.
The reactions in the bottle can also be discovered when field measurements are compared with laboratory
measurements. EC of the samples measured in the field was consistently higher by about 100 S/cm than
EC measured in the laboratory. Note the good agreement between ECfield / 100 and the sum of cations.

PROBLEMS
1.1. An analysis of groundwater gave as results:

pH Na K Mg2 Ca2 Cl Alk SO42 mg/L

8.5 62 0.1 9 81 168 183 

Alkalinity is expressed as HCO

3

a. Is the analysis correct?

b. Calculate total hardness, in degrees german hardness, and non-carbonate hardness.

1.2. Analyses of three water samples are listed below. Check the analyses and indicate where they might be
wrong (concentrations in mmol/L).

1 2 3

pH () 8.01 7.2 7.5

EC (S/cm) 750 150 450
Na 2.3 0.3 0.5
K 0.1 0.08 0.001
Mg2 1.1 0.2 0.2
Ca2 1.4 1.4 1.5
Cl 2.0 0.3 0.5
HCO3 0.3 0.7 1.5
SO42 0.9 0.2 0.5
NO3 1.8 0.1 1.5
Al 0.001 0.2 0.001
Fe 0.001 0.001 0.3

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 21

1.3. The analyses listed below reflect a range of water compositions in Denmark (concentrations in mg/L).

A  St. Heddinge waterworks, C  Harald’s Mineralwater, B  Maarum waterworks, D  Rainwater Borris

Sample A B C D

pH 7.37 7.8 5.5 4.4

EC (S/cm) 750.0 1900.0 88.0
Na 18.0 440.0 9.0 7.2
K 6.3 14.0 1.0 0.63
Mg2 14.0 29.0 0.78
Ca2 121.0 42.0 1.4
NH4 0.012 3.5 1.23
Cl 29.0 201.0 17.0 12.0
Alk (mg HCO
3 /L) 297.0 1083.0 0.0
SO42 61.0 0 4.48
NO3 35.0 2.0 5.0 4.52
PO4 0 0.02
F 0.32 1.7 0

a. Compare the water analyses with the drinking water limits in Table 1.1. Are these waters suitable as
drinking water?
b. Check the quality of each analysis by setting up a charge balance. Compare the total amount of equivalents
of ions with the electrical conductivity (EC) and identify possible errors in the analysis. Analysis C
(Harald’s Mineralwater) is incomplete. However, the hardness in german degrees is known to be 1°dH.
Use this information to calculate the electrical conductivity. Note that this also sets limits to the missing
anions.
c. Analysis D is rainwater from Borris (C. Jutland). Calculate the contribution of rainwater ions to the
groundwater composition, assuming that chloride behaves like a conservative element. When would this
assumption become dubious?
d. Draw a Stiff diagram for each analysis and discuss which processes have influenced the water
composition. Consider what kind of geological deposit the analysis has been extracted from.

REFERENCES
Appelo, C.A.J., Krajenbrink, G.J.W., Van Ree, C.C.D.F. and Vasak, L., 1982. Controls on groundwater quality
in the NW Veluwe cacthment (in Dutch). Soil Protection Series 11, Staatsuitgeverij, Den Haag, 140 pp.
Barber, C. and Davis, G.B., 1987. Representative sampling of ground water from short-screened boreholes.
Ground Water 25, 581–587.
Bates, R.G., 1973. Determination of pH; theory and practice, 2nd ed. Wiley and Sons, New York, 479 pp.
Brezinski, D.P., 1983. Kinetic, static and stirring errors of liquid junction reference electrodes. Analyst 108,
425–442.
Busenberg, E. and Plummer, L.N., 1987. pH measurement of low-conductivity waters. U.S. Geol. Surv. Water
Resour. Inv. Rep. 87–4060.
Cherry, J.A., 1983. Piezometers and other permanently-installed devices for groundwater quality monitoring.
Proc. conf. on groundwater and petroleum hydrocarbons: protection, detection and restoration, IV-1 IV-39.
Cherry, J.A., Gillham, R.W., Anderson, E.G. and Johnson, P.E., 1983. Migration of contaminants in ground-
water at a landfill: a case study. 2. Groundwater monitoring devices. J. Hydrol. 63, 31–49.
Davison, W., 1987. Measuring pH of fresh waters. In A.P. Rowland (ed.), Chemical analysis in environmental
research. ITE symp. no 18, 32–37, Abbots Ripton.
Davison, W. and Woof, C., 1985. Performance tests for the measurement of pH with glass electrodes in low
ionic strength solutions including natural waters. Anal. Chem. 57, 2567–2570.
Domenico, P.A. and Schwartz, F.W., 1997. Physical and chemical hydrogeology, 2nd ed. Wiley and Sons,
New York, 506 pp.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

22 Introduction to groundwater geochemistry

Gillham, R.W. and Johnson, P.E., 1981. A positive displacement ground-water sampling device. Ground Water
Monitor. Rev. 1, 33–35.
Halim, N.S., 2000. Arsenic mitigation in Bangladesh, The Scientist, March 6, 14–15.
Hem, J.D., 1985. Study and interpretation of the chemical characteristics of natural water, 3rd ed. U.S. Geol.
Survey Water Supply Paper 2254, 264 pp.
Handbook of Chemistry and Physics. Am. Rubber Cy. (CRC Press)
Illingworth, J.A., 1981. A common source of error in pH measurement. Biochem. Jour. 195, 259–262.
Jacks, G., Bhattacharya, P. and Singh, K.P., 2000. High-fluoride groundwaters in India. In P.L. Bjerg,
P. Engesgaard and Th.D. Krom (eds), Groundwater 2000, 193–194, Balkema, Rotterdam.
Jakobsen, R. and Postma, D., 1999. Redox zoning, rates of sulfate reduction and interactions with Fe-reduction
and methanogenesis in a shallow sandy aquifer, Rømø, Denmark. Geochim. Cosmochim. Acta 63, 137–151.
Karim, M. M., 2000. Arsenic in groundwater and health problems in Bangladesh. Water Res. 34, 304–310.
Kennedy, V.C. and Zellweger, G.W., 1974. Filter pore-size effects on the analysis of Al, Fe, Mn, and Ti in water.
Water Resour. Res. 10, 785–790.
Landolt and Bornstein, 1960. Zahlenwerte und Funktione, 7th Teil. Springer, Berlin.
Laxen, D.P.H. and Chandler, I.M., 1982. Comparison of filtration techniques for size distribution in freshwa-
ters. Anal. Chem. 54, 1350–1355.
Leuchs, W., 1988. Vorkommen, Abfolge und Auswirkungen anoxischer Redoxreaktionen in einem pleistozänen
Porengrundwasserleiter. Bes. Mitt. dt. Gewässerk. Jb. 52, 106 pp.
Lerner, D. and Teutsch, G. 1995. Recommendations for level-determined sampling in wells. J. Hydrol. 171,
355–377.
Lloyd, J.W. and Heathcote, J.A., 1985. Natural inorganic hydrochemistry in relation to groundwater. Clarendon
Press, Oxford, 296 pp.
Obermann, P., 1982. Hydrochemische/hydromechanische Untersuchungen zum Stoffgehalt von Grundwasser
bei landwirtschaftlicher Nutzung. Bes. Mitt. dt. Gewässerk. Jb. 42.
Pickens, J.F., Cherry, J.A., Grisak, G.E., Merrit, W.F. and Risto, B.A., 1978. A multilevel device for ground-
water sampling and piezometric monitoring. Ground Water 16, 322–323.
Postma, D., Boesen, C., Kristiansen, H. and Larsen, F., 1991. Nitrate reduction in an unconfined sandy aquifer:
Water chemistry, reduction processes, and geochemical modeling. Water Resour. Res. 27, 2027–2045.
Robin, M.J.L. and Gillham, R.W., 1987. Field evaluation of well purging procedures. Ground Water Monitor.
Rev. 7, 85–93.
Rossum, J.R., 1975. Checking the accuracy of water analyses through the use of conductivity. J. Am. Water
Works Ass. 67, 204–205.
Safe Drinking Water Comm., 1980. Drinking water and health, Vol. 3. Nat. Acad. Press, Washington, 415 pp.
Smedley, P.L. and Kinniburgh, D.G., 2002. A review of the source, behavior and distribution of arsenic in natu-
ral waters. Appl. Geochem. 17, 517–568.
Sørensen, K. and Larsen, F., 1999. Ellog auger drilling: Three-in-one method for hydrogeological data collec-
tion. Ground Water Monit. Rev. 17, 97–101.
Standard Methods for the examination of water and wastewater, 1985. Joint publication of APHA, AWWA,
WPCF, 16th ed. Am. Publ. Health Ass., Washington, 1268 pp.
Stumm, W. and Morgan, J.J., 1996. Aquatic chemistry, 3rd ed. Wiley and Sons, New York, 1022 pp.
Stuyfzand, P.J., 1983. Important sources of errors in sampling groundwater from multilevel samplers (in Dutch).
H2O 16, 87–95.
Stuyfzand, P.J., 1987. An accurate calculation of the electrical conductivity from water analyses (in Dutch).
KIWA SWE-87.006, Rijswijk, 31 pp.
Stuyfzand, P.J., 1989. Hydrochemical methods for the analysis of groundwater flow (in Dutch). H2O 22,
141–146.
Suarez, D.L., 1987. Prediction of pH errors in soil-water extractors due to degassing. Soil Sci. Soc. Am. J. 51,
64–67.
Welch, A.H., Westjohn, D.B., Helsel, D.R. and Wanty, R.B., 2000. Arsenic in ground water of the United States:
occurrence and geochemistry. Ground Water 38, 589–604.
Zuurdeeg, B.W. and Van der Weiden, M.J.J., 1985. Geochemical aspects of European bottled waters. In
Geothermics, Thermal-mineral waters and hydrogeology, 235–264. Theophrastus Publ., Athens.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

2

From Rainwater to Groundwater

Rainwater is the source of most groundwater and a logical starting point for the study of groundwater
geochemistry. Continental rainwater is dominated by oceanic vapor and it does indeed resemble
strongly diluted seawater. However, natural and anthropogenic dusts and gases modify the composi-
tion. Before the rain turns into groundwater, various processes in the soil may affect the concentra-
tions. Dry-deposited dust particles and gases will dissolve. Evapotranspiration concentrates the
solutes, and vegetation selects essential elements to store them temporarily in the biomass.
Particularly the weathering of minerals in the soil is of importance in changing concentrations, and all
the soil processes together often generate already much of the groundwater chemistry. In this chapter,
we will follow the evolution in water chemistry from rain, via soil, to the aquifer.
Mass balance calculations are useful to obtain first insight in the water chemistry. The mass bal-
ance approach is fundamental and universally applicable, and we will use it to decipher the hydro-
logical cycle, the sources of the constituents in rainwater, the isotopic evolution of rainwater, cycling
of elements in the biosphere, and to derive the groundwater quality below farmland.

2.1 THE HYDROLOGICAL CYCLE

The inventory of the global water cycle in Figure 2.1 shows the amounts of water in the various
reservoirs. The oceans contain 13,700  1020 g H2O, or 97% of the total amount of water on earth.
The water in the ice caps and glaciers amounts to 2% of the total. If all the ice would melt, the depth
of the oceans would increase by 2% and the sea level would rise by 2% of the average depth, or about
80 m. Groundwater is placed third with about 1% of the total amount of water on earth, and it is
many times larger than water in rivers and lakes (together about 0.009%). Finally the amount of
water present in the atmosphere constitutes only 0.001% of the global total.

Atmosphere:
0.13
Pt E Pt E
R 0.34
3.6 3.95 1.1 0.75 Ice: 290

G 0.01
Oceans: 13,700
Groundwater: 95
3800 m 100 m: 5.6

Figure 2.1. The global hydrologic cycle showing amounts in 1020 g H2O (encircled entities), and mass transfers
among the reservoirs in 1020 g H2O/yr (numbers associated with arrows). Based on data in Berner and Berner,
1996; Holland and Petersen, 1996; Budyko, 1982.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

24 From rainwater to groundwater

Figure 2.1 also indicates the mass transfers among the reservoirs. The numbers are based on meas-
urements and calculations from mass and energy balances and contain a variable uncertainty. Well
measured are the continental precipitation (Pt) and the continental surface runoff (R) into the oceans.
The amount of groundwater draining directly into the oceans (G) is by comparison rather small.
From these numbers (cf. Figure 2.1), we can estimate that over land the evapotranspiration
(E  Pt  R  G  1.1  0.34  0.01  0.751020 g H2O/yr) amounts to 68% of the precipitation.
The oceanic evaporation is estimated with the energy budget method, and the precipitation over
the oceans is subsequently obtained from mass balance (Budyko, 1982). Over the oceans, evapora-
tion exceeds precipitation and a vapor transport from the ocean to the continent balances the conti-
nental surface runoff and the groundwater outflow into the oceans. Water evaporated over the oceans
contributes with 32% to the continental precipitation on average, but the contribution varies depend-
ing on the distance from the coast and the meteorological conditions. The proportion of land derived
vapor or recycled precipitation (Eltahir and Bras, 1996) can be estimated from a mass balance,
which along a transect accounts for precipitation and evapotranspiration (Budyko, 1982). The
atmospheric vapor transfer at any point x (Figure 2.2) is the sum of the inflow at the border (x  0),
the loss by precipitation, and the gain from evapotranspiration:

I x  I 0  Pt x  E x (2.1)

Here Ix is the vapor mass transfer (m H2O  m/yr) at distance x from the coast, I0 is the atmospheric
input at the coast, Pt is precipitation and E is evapotranspiration along the transect (both in m H2O/yr).
For the global data in Figure 2.1, Pt / E  1.1 / 0.75  1.47, or simplified Pt  1.5E. Substituted in
Equation (2.1) this gives Ix  I0  0.5E x. The ratio of the land derived vapor (E x) to the total
vapor transfer (Ix) then is:
E x E x
 (2.2)
Ix I 0  0.5 E x

I0 is the product of atmospheric vapor content (u) and the air flow velocity (v) across the sea-land
boundary. To calculate Figure 2.2 we used u  (atmospheric vapor content) / (density of water 
earth’s surface)  0.131020 g H2O / (106 g/m3  5.11014 m2)  0.025 m H2O. Furthermore, a wind
velocity v  4 m/s directed eastward, the average for Europe at 50°N (Peixóto and Oort, 1983), gives
I0  0.025 m  4 m/s  3.15106 m2/yr. And lastly, E  (g H2O/yr) / (density of water  earth’s
land surface)  0.751020 / (106  1.51014)  0.5 m/yr.

I0 Ix

E Pt

0 x

1.5
E⋅x
1 Recycled precipitation
Ix
0.5

0
0 2500 5000 km

Figure 2.2. Recycled precipitation, the ratio of land derived vapor (E x) over the total vapor transfer (Ix) in
the atmosphere as a function of the distance to the coast.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The hydrological cycle 25

Figure 2.2 shows that the importance of land-derived vapor increases as a function of distance to the
coast. However, first at 2500 km from the coast does the land-derived vapor exceed 50% of the
atmospheric water vapor. The integrated amount over this distance gives an even smaller percentage
(Problem 2.2), illustrating that the chemistry of rainwater on the continents, including its isotopes, is
very strongly influenced by oceanic water vapor.
Assuming constant volumes of water in each reservoir and constant mass transfers among them
(steady state conditions) we can calculate the residence time, which is the volume divided by the mass
transfer:
V
 (2.3)
Q

where  is the residence time (s), V is volume of the reservoir (g H2O) and Q is discharge (g H2O/yr)
into or out from the reservoir.
For the oceans, the residence time based on surface runoff and groundwater outflow is 13,700 /
(0.34  0.01)  39,000 years. For the atmosphere, based on precipitation, the residence time is
0.13 / (3.6  1.1)  10 days. For groundwater, the residence time based on direct groundwater discharge
into the ocean would be 95 / 0.01  9,500 years. However, most groundwater discharges via rivers
and a minimum estimate of the residence time would rather be 95 / (0.34  0.01)  271 years (all
the basic data are in Figure 2.1). The residence time indicates the timescale that the subsystem oper-
ates on and marks its susceptibility towards changes. Clearly, the oceans are large, slowly responding
systems, while the atmosphere is highly dynamic. Groundwater reservoirs are intermediate and
operate on timescales of at least hundreds of years.
The global water cycle averages the existing variations with lattitude and ignores the large differ-
ences among the continents. Figure 2.3 shows global net recharge variations as a function of lattitude.
Precipitation is highest in the tropics and in the temperate climate. Evaporation is highest in arid
regions with a strong heat source (sun radiation) and low air moisture. In the subtropics the potential
evaporation exceeds precipitation. Over the continents this results in the formation of deserts like the
Sahara and the Great Desert of Australia.

South Latitude (deg) North

80 60 40 20 0 20 40 60 80
80
Precipitation minus evaporation (cm/yr)

60

40

20

20

40

60

80

Figure 2.3. Net precipitation (precipitation minus evaporation) as a function of lattitude. Positive values indi-
cate net infiltration and negative values potential evaporation (Peixóto and Keitani, 1973).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

26 From rainwater to groundwater

QUESTIONS:
Estimate the surface area of the oceans, using Figure 2.1.
ANSWER: 1.371018 m3/3800 m  3.61014 m2
About 6% of groundwater resides in shallow aquifers 100 m with 20% porosity. Estimate the percentage
land on earth with shallow aquifers.
ANSWER: 5.61014 m3/0.2/100 m  2.81013 m2 / (earth’s land surface  1.51014 m2)
 100  18.7%
Estimate the residence time of groundwater in shallow aquifers.
ANSWER: Qin  0.351020 g H2O/yr  0.187  0.0651020 g H2O/yr.
  5.61020 g H2O/0.0651020 g H2O/yr  86 yr.

2.2 THE COMPOSITION OF RAINWATER

The composition of rainwater is determined by the source of the water vapor and by the ions which are
acquired (or lost) by water on its journey through the atmosphere. Over the oceans, rainwater resembles
strongly diluted seawater with a chloride concentration of 10–15 mg/L. Given the dominance of oceanic
vapor in the atmosphere (Figure 2.2), this composition will form the basis of inland rainwater quality thou-
sands of kilometers away from the sea. The sea-salt content originates from marine aerosols that form from
the liquid droplets, produced when gas bubbles burst at the sea surface (Monahan, 1986; Blanchard and
Syzdek, 1988). Evaporation of the water from the droplets produces aerosols with different size-ranges.

1.0 2.0
0.6 4.0
8.0
0.4 8.85
0.3
0.2

2.0 1.00.6
0.3
22.58 0.4 0.2 0.26
2.0 0.31 0.12
1.0 0.20 0.17 0.25
0.14
0.68 0.13
0.1 0.14 0.23 1.19
0.6 0.08 0.10
0.16 0.27 5.89
0.2 0.15
0.21 0.19 0.40
0.11
0.22 0.27 0.14 0.15 0.17
0.18 0.35
0.6 0.10
0.31 12.41
0.35 0.28 0.12 0.23 6.90
0.12
0.16
0.67 0.16 0.27
1.0 0.1 0.13
2.0 0.09 0.14 8.0
0.44 0.62
3.31
0.17
CIⴚ
0.43 4.0
0.6 0.35 1.02
2.0 1.0 0.19 0.21
0.24 0.76 0.66
0.4 0.3
0.3 0.37
0.4 0.80
0.6 3.81 2.44
1.0 1.0 1.0
2.0
2.0 4.0 2.0
4.0 21.96
8.0 8.0

Figure 2.4. Average chloride concentrations (mg/L) in precipitation over the United States (Junge and
Werby, 1958).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The composition of rainwater 27

The coarsest particles have diameters exceeding 10 m and retain a composition that is closely akin
to seawater. From smaller particles, chloride evaporates presumably in the form of HCl, which
results in Cl depletions relative to Na of up to 40% (Raemdonck et al., 1986; Möller, 1990). Rainout
of Cl depleted particles produces a Cl depleted rain, but since the evaporated HCl also contributes
to the total rainwater chemistry, the net effect is that the Na/Cl ratios remain similar to the ratio in
seawater, as found in coastal areas of Europe (Möller, 1990), the United States (Wagner and Steele,
1989), and in the Amazonas (Andreae et al., 1990). The importance of the sea-salt contribution is
easily traced in the chloride concentration of rainwater. Figure 2.4 shows the distribution of chloride
in the rain over the USA. The highest concentrations are found near the coast as a result of rapid
washout of the largest sea salt particles, and the concentrations decrease exponentially inland to a
background value of 0.1–0.2 mg/L at about 600 km from the coast. Over the continent, there are
additional Cl sources in industrialized areas, like burning of waste plastics (Lightowlers and Cape,
1988), coal fired power plants (Wagner and Steele, 1989), or even the evaporation of salt water used
as a coolant in the steel industry (Vermeulen, 1977).
The relative composition of rainwater and seawater can be compared by using a fractionation
factor. For example, for Cl relative to Na the fractionation factor FNa is:
(Cl/Na) rain
FNa  (2.4)
(Cl/Na )seaw.

Since we are comparing ratios, the Cl and Na concentrations in rain or seawater can be expressed
in any concentration-units. Table 2.1 compiles fractionation factors for rainwater on the oceanic island
of Hawaii, which for major cations will be similar to other near-coastal stations.
Table 2.1 indicates that the major elements are not fractionated very much. For Cl, the fraction-
ation factor ranges from 0.93 to 1.0, and perhaps the lower value is the result of some volatilization.
For Ca2, the fractionation factor ranges from 0.97 to 1.22, indicating some enrichment of Ca2
with respect to Na, which suggests the presence of a land-based Ca2 source. Bromide and organic-
N have fractionation factors much higher than 1. These elements are part of lipid containing organic
matter that is present in the surface water layer. When gas bubbles from below the water surface burst
into the air, the interfacial foam is dragged along and the associated elements are enriched in the atmos-
phere and in rainwater. For major components the absence of extensive fractionation leads to the con-
clusion that rainwater of marine origin is basically strongly diluted seawater.
During transport over land, the air masses and clouds pick up continental dust and gases from
natural or industrial origin, which modifies the composition of rainwater. Figure 2.5 shows the dis-
tribution of Ca2 in rain over the USA. The concentration pattern is the reverse of that for Cl. The
lowest Ca2 concentrations are found near the coast and the concentrations are particularly high in
the arid southwestern part of the US. Soil dust, containing CaCO3 particles, is the source of most
Ca2 in rain, together with dust from cement industry, and fuel and waste burning. Na in rain-
water may also originate from soil dust particles in arid climates (Nativ and Issar, 1983), and
roadspray used for de-icing may increase the Na/Cl ratio with respect to seawater (Baisden
et al., 1995).

Table 2.1. Fractionation factors, given as a range, of seawater elements in marine aerosols. Fractionation
factors expressed with respect to Na (FNa). (Data from Duce and Hoffman, 1976.)

Mg2 Ca2 K Sr2 Cl SO2

4 Br Org.N

FNa upper limit 1.07 1.22 1.05 0.89 1.0 ? 12 1  106

lower limit 0.98 0.97 0.97 0.84 0.93 1 1 2  104

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

28 From rainwater to groundwater

0.78

0.5
1.0 1.5 1.0
0.5 0.73 1.5

3
1.0

0.6
0.3
1.72
0.47 0.38 0.91
0.19 1.44 0.58
0.85 0.99
0.29
1.02
0.52 0.93 0.43
0.50 0.30
2.40 1.24
1.11 1.01 0.96
0.77 0.3
0.30
1.38 3.12 2.82 0.69
0.25 0.5
3.28 1.19
3.41 2.91
0.37 1.20 3.0 0.44
3.79 2.12 0.32
0.27 1.90 0.76
0.31
4.74 2.0
0.3 2.26
2.17
0.83 0.35
0.5 0.32
1.22 3.31 1.14 0.48
2.18 0.89
1.0 0.28
0.3 0.43 Ca2ⴙ
1.5 1.42
2.0
0.54 0.56
3.0
4.63 1.0

6.50 1.5
3.0 2.0

Figure 2.5. Average calcium concentrations (mg/L) in precipitation over the United States (Junge and Werby, 1958).

Table 2.2 contains a number of rainwater analyses from different parts of the world that illustrate the
compositional variations. The older analyses are all from bulk rain samplers that are permanently
open. However, for the stations Beek (Netherlands) and Delhi (India), more recent analyses have
been included of ‘wet-only’ samples collected with samplers that open only when it rains, which
excludes dry deposition (Section 2.4) and generally results in lower concentrations.

Table 2.2. Rainwater analyses (values in mol/L).

a b c c d e e f
Kiruna Hubbard De Kooy Beek Beek Thumba Delhi Delhi
Sweden Brook,US Neth. Neth. wet-only India India wet-only
Period ’55–’57 ’63–’74 ’78–’83 ’78–’83 ’97–’98 1975 1975 1997

pH 5.6 4.1 4.40 4.75 5.5 6.3

Na 13 5 302 43 26 200 30 9
K 5 2 9 7 3 6 7 9
Mg2 5 2 35 9 4 19 4 2
Ca2 16 4 19 47 10 23 29 13
NH4 6 12 78 128 74 21
Cl 11 7 349 54 29 229 28 13
SO2
4 21 30 66 87 29 7 4 10
NO3 5 12 63 63 35 9
HCO 3 21 – – 9 25

(a) Granat, 1972; (b) Likens et al., 1995; (c) KNMI / RIV, 1985; (d) Boschloo and Stolk, 1999; (e) Sequeira
and Kelkar, 1978; (f) Jain et al., 2000.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 The composition of rainwater 29

To facilitate the interpretation of these analyses and to illuminate supplementary sources of ions, the
sea salt contribution can be subtracted from the rainwater analyses. Chloride is chosen as an inert
reference component in the calculation because, in an overall sense, it shows little fractionation. For
Na
the calculation proceeds as follows (Table 2.3). In seawater the Na
concentration is 485 mmol/L
and the Cl concentration is 566 mmol/L. The ratio of Na / Cl in seawater is therefore 0.86. In Kiruna
rain, the Cl concentration is 11 mol/L and the part of Na
that is derived from seawater amounts
therefore to 0.86  11  9 mol/L. The observed Na
concentration in Kiruna rainwater is
13 mol/L, and 13  9  4 mol Na
/L must have a non-marine origin. This difference (which
indicates either a continental source of Na
, or a depletion of Cl) is, however, barely significant
compared to the analytical accuracy. Similar calculations were performed for all the rainwater analy-
ses in Table 2.2, and the results are given in Table 2.3. After subtracting the seawater contribution, the
precipitation of the near coastal stations, De Kooy or Thumba, shows low residuals of Na
, K
and
Mg2
ions, which indicates that the content of these ions in rain primarily is derived from seawater.
On the other hand, precipitation in Delhi is enriched in Ca2
derived from soil dust and if 1975 and
1997 numbers are comparable, more than half comes down as dry deposition.
A conspicuous enrichment of the anions NO3 and SO42 is found in the rain of the United States
and Europe (Table 2.3, Hubbard Brook and Beek). These ions originate mainly from industrial and
traffic fumes containing gaseous NOx and SO2. The gases become oxidized in the atmosphere, pro-
ducing the strong acids HNO3 and H2SO4. Dissociation of these acids lowers the pH of rainwater
down to values of 4 and are the cause of acid rain (Berner and Berner, 1996). In Europe the emis-
sion of SO2 has decreased substantially during the last decade and the increase in the pH of rainwater
in Beek from 4.75 during the period 1978–1983 to 5.5 in 1998 looks promising.
The pH of rainwater is also affected by base producing processes which vary over time. The intensi-
fication of animal husbandry increasingly produces the neutralizing base NH4OH, which evaporates
from manure that is spread over agricultural land. Its influence can be traced again in the analysis of Beek,
located in an area of the Netherlands where factory farming has grown rapidly since 1970. Not much
NH
4 is present in the rain at Hubbard Brook and the pH is therefore lower at this site. Cement industry,
and waste burning, also emit particles that contribute with base. The analysis of Beek rainwater shows a
large excess of Ca2
when the sea salt contribution has been subtracted (Table 2.3) due to the presence of
a local cement industry. Ironically, the increased implementation of filters on chimneys has reduced the
emission of base producing dust, which has led to a sharp decline in base cations in rain and counteracted
the effect of pH increase that would result from the reduction in SO2 emission (Hedin et al., 1994).

Table 2.3. Concentrations in rainwater from sources other than seawater. For the Kiruna station both the
contribution from seawater and from additional sources is shown while for the other stations only the additional
sources are tabulated. (Units are mol/L, except for seawater).

seaw. Kiruna Hub. De K. Beek Beek Thum Delhi Delhi

(mmol/L) Br. wet-only wet-only
rain seaw. other
contr. src.

Na
485 13 9.4 4 1 3 3 1 4 6 2

K
10.6 5 0.2 5 2 3 6 2 2 6 9
Mg2
55.1 5 1 4 1 1 4 1 3 1 0
Ca2
10.7 16 0.2 16 4 12 46 9 19 28 13
NH
4 2106 6 0 6 12 78 128 74 21
Cl 566 11 11 – – – – – – – –
SO2
4 29.3 21 0.6 20 30 48 84 27 5 3 9
NO3 5106 5 0 5 12 63 63 35 9
HCO 3 2.4 21 21 0 1 9 24

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

30 From rainwater to groundwater

Free acids
100 
H

H

H

Microequivalents/liter
80 H


Neutralized strong acids

NO3
SO4
Ammonia

NO3
SO4

NO3
SO4
 

NO3
NH4

SO4
NH4

60 NH4


NH4

Soil dust
Ca2 Ca2
40 Ca 2
Ca2
2
Mg
Mg2 Mg2 Mg2

20 Sea salt Na

CI CI

 
Na Na 
CI Na CI

0 West wood Pasadena Riverside

Figure 2.6. Composition of rain in southern California, 1978–79, interpreted in terms of input components and

source type (Morgan, 1982). Hatched areas in the SO2
4 and NO3 fields represent immobile sources, unhatched
is the contribution from traffic.

The net effect of different acid and base producing processes on the rainwater pH is
illustrated in a bar diagram by Morgan (1982) showing rainwater analyses from California (Figure 2.6).
The bottom part shows the sea salt component with the anions Cl and part of the SO42, and the
cations Na and Mg2. The remainder of SO42 and NO 3 come from industrial sources and were
originally present as sulfuric and nitric acid but have become partly neutralized by CaCO3 from con-
tinental dust and NH4OH from manure evaporation. The unneutralized part of the acids is present as
free H, and determines the pH of rainwater.

QUESTION:
Calculate the concentration of free acids in Hubbard Brook and Delhi rain (eq/L) from Table 2.2.
ANSWER: Hubbard Brook: 104.1  106  79 eq/L. Delhi: 106.3  106  0.5 eq/L.

180°

150° 17 150°

18

19 120°
120°
21
20

22 24
23
90° N.P. 90°
23

16 22
24
21
17 60°
60° 20
18
19

30° 30°

0°

Figure 2.7. Tracks of air parcels around the world. The numbers represent successive days in April, 1964.
(Modified from Newell, 1971.)
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Stable isotopes in rain 31

2.2.1 Sources and transport of atmospheric pollutants

Atmospheric pollution operates on an international scale due to the rapid dispersion and world-wide
transport of local emissions through the atmosphere. The transport velocity in the atmosphere has
been demonstrated in trajectory studies in which atmospheric particles (or air parcels) are traced
back in time, based on meteorological observations. Figure 2.7 presents the results of an early study
(Newell, 1971) and shows that particles or gases travel the distance from the United States to Europe
in only about 4 days. The industrial pollution of the atmosphere is clearly not limited by national
frontiers.
Mass balances for the acidifying gases NOx and SO2 have been constructed for individual
countries, and an example for the Netherlands is shown in Figure 2.8 for the year 1980. The
figure shows that of 550,000 tons of NOx emitted (expressed as NO2), only 90,000 tons were
deposited within the Netherlands. The remainder was dispersed, in part to the sea (60,000 tons), and
in part to other countries (400,000 tons). The exported amounts exceeded the 150,000 tons deposited
as import from foreign sources. Similar numbers applied for SO2 before 1980 but the emission of
this acidifying oxide has been forcefully reduced in later years and amounted to about 70,000 tons
in 1999.

310 400
d d

270 150
e e
100 490 80 60 550 90
c a b c a b

SO2 NOx

Figure 2.8. Mass balances for SO2 and NOx in the Netherlands in 1980. The values are given in 1000 tons/year.
(a) is total industrial emission from the Netherlands, of which (b) is deposited in the Netherlands, (c) in sea, (d) in
other countries and (e) is deposited in the Netherlands from foreign sources.

2.3 STABLE ISOTOPES IN RAIN

Many elements, including hydrogen and oxygen, exist in nature with different mass numbers, also
called isotopes. For example, hydrogen occurs in nature as a mixture of the isotopes 1H and 2H (deu-
terium) while oxygen is found as isotopes with the atomic masses 18O, 17O and 16O. Ocean water
contains two 18O atoms for every thousand 16O atoms (Table 2.4), but the ratio is different in fresh
water. 1H, 2H, 18O and 16O do not engage in nuclear transformations and are therefore termed stable
isotopes. In contrast, radioactive isotopes like tritium (3H) or radiocarbon (14C) will decay over time
and can be used for dating (Section 3.3).
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
32 From rainwater to groundwater

Due to the difference in mass, stable isotopes behave slightly differently in physical, chemical and bio-
logical processes. For example, during evaporation of water and again during rainout from the atmos-
phere, the stable isotopes 2H / 1H and 18O / 16O become fractionated. The resulting small variations in
isotopic concentrations may yield information on the climate at the point of infiltration or the prove-
nance of the water. Also for other elements, isotope fractionation may help to elucidate the geochemical
processes and reactions which have been creating a given water composition (Clark and Fritz, 1997).

Table 2.4. The stable environmental isotopes.

Isotope Standard Ratio R Measured phases/process
-range (‰)

2
H VSMOW1 2
H / 1H  1.5576104 H2O, CH2O, CH4, H2, 450 (CH4), 200
OH minerals (arctic ice), 50
(evap. lakes)
3
He Air 3
He / 4He  1.3106 Gas-sources, dating
6
Li NBS2 L-SVEC 6
Li / 7Li  0.0832 Water, rocks
11
B NBS 951 11
B / 10B  4.0436 Water, pollution from waste sites
13
C VPDB3 13
C / 12C  0.011237 TIC, CH4, carbonates, 14C dating 120 (bact. CH4),
20 (non-marine
carb.)
15
N Air 15
N / 14N  3.677103 NO 
3 , N2, NH4 , N-organics
18
O VSMOW 18
O / 16O  2.005103 H2O, CH2O, HCO 2
3 , SO4 , 60 (arctic ice)
NO3 , O minerals 40 (CO2(g))
34
S CDT4 34
S / 32S  0.045 SO2
4 , S 2
37
Cl SMOC5 37
Cl / 35Cl  0.32398 Water, salts, diffusion 1.5 to 1.0
87
Sr – 87
Sr / 86Sr
0.710 Water, minerals, weathering
1
Vienna Standard Mean Ocean water; 2 National Bureau of Standards; 3 Vienna Pee Dee Belemnite; 4 Cañon
Diablo Troilite; 5 Standard Mean Ocean Chloride, ratio from natural abundances (Wedepohl, 1969).

2.3.1 Isotopic ratios and the
notation

The concentration of stable isotopes is normally given as the ratio of the least abundant
isotope over the most abundant isotope and expressed relative to a standard. For water the
internationally agreed standard is called “Vienna Standard Mean Ocean Water” (VSMOW).
According to Table 2.4, in VSMOW the isotopes of oxygen and hydrogen are present in the ratios
18
O / 16O  2.005103 and 2H / 1H  1.56104. Standards for stable isotopes of other elements
that are commonly used in geochemical studies are also listed in Table 2.4.
The variations of isotopic ratios in nature are small and are studied more easily by using the

notation. The
notation expresses the deviation of the isotopic ratio R in the sample with respect to
the ratio in the standard:
Rsample  Rstandard

sample  1000 (2.5a)
Rstandard

or:
Rsample
sample
 1 (2.5b)
Rstandard 1000

The least abundant isotope is specified with the
symbol, for example the ratio 18O / 16O in rain is
described as
18Orain.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Stable isotopes in rain 33

It follows from Equation (2.5) that the
value of a mixture of n waters can be calculated by adding
the
values of the contributing waters multiplied by the mass fraction of the reference isotope, and
dividing by the concentration of the reference isotope in the mixture:
n

1 (x1 ref m1 ) 

n ( xn ref mn )

∑
i ( xi ref mi )
i1

mixture   (2.6)
(x1 ref m1 ) 
 (xn ref mn ) ref mmixture

where xi is the fraction of water i, which has concentration refmi of the reference isotope (e.g. 16O, or 12C).
Usually to a good approximation, the total concentration of the element mi ( refmi  jmi, where
j is the isotope of interest) can be used instead of the reference isotope:
n

1 (x1 m1 ) 

n ( xn mn )
∑
i ( xi mi )
i1

mixture   (2.6a)
(x1 m1 ) 
 (xn mn ) mmixture

For example, when a water (1) with TIC1  10 mM and
13C1  0‰ is mixed 1:1 with a water (2)
with TIC2  2 mM and
13C2  18‰, Equation (2.6a) gives
13Cmixture  3‰, while the exact
answer is 3.0005‰ according to Equation (2.6).

QUESTIONS:
Calculate the ratio of 13C and 12C in water (1) with
13C1  0‰.
ANSWER: 13C / 12C  0.011237
Calculate the concentrations of 13C and 12C in water (1) with TIC  10 mM and
13C  0‰.
ANSWER: 12m1  10 / (1  0.011237)  9.888879 mM, 13m1  0.111121 mM.
Calculate the concentrations of 13C and 12C in water (2) with TIC  2 mM and
13C  18‰.
ANSWER: 12m2  2 / (1  0.011035)  1.978171 mM, 13m2  0.021829 mM.
Calculate
C when waters (1) and (2) are mixed 1:9.
13

ANSWER:
13C  11.57‰

Calculate
O of a 1:1 mixture of river Rhine water with
18O  9‰ and seawater.
18

ANSWER: 4.5‰

2.3.2 The Rayleigh process

Rayleigh originally derived a formula to calculate the products of distillation for a mixture of
liquids with different boiling points. The principles are, however, applicable to any separation
process with a constant fractionation factor, as is the case for stable isotopes. For example
during condensation from water vapor, the heavier molecule H218O condenses more readily
than H216O, producing liquid water that is isotopically heavier than the vapor. The remaining
water vapor must become increasingly enriched in the light isotope 16O and the change in composi-
tion can be calculated with the Rayleigh formula. In the Rayleigh process, the product is
removed continuously from the reactant, and further interactions are eliminated. This condition
applies to the situation where a raindrop falls out of a cloud, or where isotopes fractionate
between water and a solid phase along a flowline, for example in partitioning of 13C and 12C during
calcite precipitation.
Consider a small number, dN, molecules of light H216O that condense and are accompanied by dNi
molecules of the heavy H218O. The ratio dNi / dN in the condensate will be higher than the ratio Ni / N

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

34 From rainwater to groundwater

in the vapor:

dN i N
 i (2.7)
dN N

where  is the (isotope) fractionation factor (no dimension, the fractionation process is similar to
element fractionation among rain and seawater, Equation 2.4). Going from vapor (g) to liquid (l) the
fractionaton factor is 18Ol/g  1.0098 at 20°C.
The derivative of the equation y  (ln x) is d(y) / dx  1 / x, or d(ln x) / d x  1 / x. Hence d(ln
x)  d(x) / x. Thus, in Equation (2.7) we have d(ln Ni)  dNi / Ni and d(ln N)  dN / N, which we
substitute:

d ln N i
 (2.8)
d ln N

Substracting 1 from the left and right hand sides and recalling that ln a  ln b  ln(a / b), leads to:

d ln N i d ln N i d ln N d ln (N i / N )
1      1
d ln N d ln N d ln N d ln N

and inserting the isotopic ratio R  Ni / N yields:

d ln R (2.9)
  1
d ln N

Integration of Equation (2.9) from the initial values R0 and N0:

R N
∫ d ln R  (  1) ∫ d ln N
R0 N0

solves to:

R  N  ( 1)
ln R  ln R0  (  1) ( ln N  ln N 0 ) or    (2.10)
R0  N0 

Since Ni is small compared to N, N
N  Ni, and the ratio N / N0 is almost equal to the remaining
fraction f of the vapor. Therefore:

Rvapor
 f (1) (2.11)
Rvapor, 0

With Equation (2.11) we can then calculate how R in atmospheric vapor changes during a rain-out.
The isotopic ratio in the rain follows straightforwardly from Equation (2.7):
Rrain  dN i / dN  Rvapor (2.12)

and substitution in Equation (2.11) gives:

Rrain  Rvapor, 0 f ( 1)  Rrain, 0 f ( 1) (2.13)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Stable isotopes in rain 35

We can rewrite Equation (2.13) for 18O, inserting
instead of R according to Equation (2.5b):


18O  
18O rain, 0 

 rain
 1    1 f ( 1) (2.14)
 
 1000   1000 

The term
/ 1000 will normally be small (0.01), and therefore ln(
/ 1000  1)

/ 1000 (check this
for
 36 and
 24). Taking logarithms of Equation (2.14) and substituting ln(
/ 1000  1) 

/ 1000 yields:

18O rain 
18O rain, 0  1000 (18O l/g  1) lnn f (2.15)

For shorthand notation the enrichment factor,   1000 (  1), is used:

18O rain 
18O rain, 0  e18O l/g ln f (2.16)

We can also rewrite Equation (2.12) in
notation:

rain 
vapor  1000 ln  (2.17)

Since  is very close to 1, ln 
(  1), and therefore:

rain

vapor  1000 (  1) 
vapor  el/g (2.18)

The enrichment factor is very helpful since it immediately gives the isotopic composition
of two different phases at isotopic equilibrium. For example, at 20°C the equilibrium enrich-
ment 18Ol/g  9.8
(
18Ol 
18Og). For
18Ovapor  15.8‰ the isotopic composition of the
rain is
18O  6‰. Note that fractionation can be large for the hydrogen isotopes, in which
case the fractionation should be determined for the ratio 2H / 1H, followed by calculation of the
2H
value.

QUESTION:
Derive the equation to describe
vapor as function of f ?
ANSWER: Combine Equations (2.16) and (2.18):
18Ovapor 
18Ovapor, 0 
18Ol/g ln f

EXAMPLE 2.1. Calculate
18O of rain condensing from vapor with, initially,
18O  12.2‰, as a function
of the remaining vapor, temperature is fixed at 20°C. 18Ol/g  9.8‰ at 20°C.
We consider three possibilities,

1. the rain is separated continuously from the vapor (Rayleigh condensation),

2. the rain is separated from the vapor but collected in a vessel,
3. the rain is collected and remains in contact with the vapor for all fractions.

Ad 1. For the rain in a Rayleigh process, we use Equation (2.16) and combine with (2.18):

18O rain 
18O rain, 0  e18O l/g ln f  (
18O vapor, 0  e18O l/g )  e18O l/g ln f
 2.4  9.8  ln f

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

36 From rainwater to groundwater

Ad 2. We calculate the vapor composition according to a Rayleigh process,

vapor 
vapor, 0  el/g ln f

and combine with the mass balance (Equation 2.6a),

f
vapor  (1  f )
rain  1
vapor, 0

to obtain:

18O rain  {
18O vapor, 0  f
18O vapor, 0  f el/g ln f } / (1  f )

 12.2  9.8  f  ln f /(1  f )

Ad 3. We combine the mass balance (Equation 2.6a),

f
vapor  (1  f )
rain  1
vapor, 0

and the equilibrium relation (Equation 2.18),

rain 
vapor  el/g

which gives:

rain 
vapor, 0  el/g f  12.2  9.8  f

The resulting
18O’s in rain for the three options are shown in Figure 2.9, as a function of the fraction of
remaining vapor, f. Also shown is the composition of the vapor which loses water in the Rayleigh process
(options 1 and 2).

5 2: Collected

10
3: Equilibrium

18O (‰)

15
1: Rayleigh rain

20 1 & 2: Rayleigh vapor

25

30
0 0.2 0.4 0.6 0.8 1
Fraction of vapor, f

Figure 2.9. Isotopic composition of rain for three options of separating and equilibrating rain from vapor.

QUESTION:
Draw
18O of vapor for option 3?
ANSWER: a line parallel to case 3 but shifted 9.8‰

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Stable isotopes in rain 37

It is difficult to predict the value of a fractionation factor (cf. Chacko et al., 2001; Thorstenson and
Parkhurst, 2004), but in general, fractionation is related to the movement of individual molecules
and it decreases with increasing temperature. At high temperature,   0 for any isotope pair. Also,
qualitative reasoning suggests that the denser phase (liquid compared to gas) or the denser com-
pound (CaCO3 vs CO2) will contain more of the heavy isotope. Furthermore, heavier isotopes are
discriminated in kinetic and diffusion processes, and therefore in biological activities as well.

2.3.3 The isotopic composition of rain

The rainout of atmospheric water vapor is driven by the cooling of air during transport to higher lat-
itude or altitude. The change in isotopic composition of rain and snow, as a function of the remain-
ing fraction f, and indirectly of temperature, was calculated according to option 1 of Example 2.1
and is shown in Figure 2.10 (Clark and Fritz, 1997). Note that the enrichment factor increases
when temperature decreases, and that Clark and Fritz used a slightly larger value at 20°C than in
Example 2.1.
Figure 2.11 shows the isotopic composition of rainwater from various locations on earth in a plot
of
2H vs
18O. It is found that the stable isotopes for hydrogen and oxygen in rainwater obey the
empirical relation:

2H  8
18O  10 (2.19)

This equation is known as the meteoric water line. The slope of the line d(
2H) / d(
18O)  8,
and has its origin in the way most precipitation is generated, viz. derived from evaporated
oceanic water in the tropics, then transported to higher latitudes where it cools and rains out.

30
Moisture co
ntent (C/C 20
Temperature (°C)
0)

10
0
10
20
0 30
10 ε  10.2‰ Rain
Water vapor
Sn
20 ow
δ18O (‰)

30
40 ε  11.4‰

50
60 ε  14.7‰
70
1 0.8 0.6 0.4 0.2 0
Residual vapor fraction, f

Figure 2.10. Isotopic composition of rain and snow during progressive cooling and rainout. Initial temperature
of 25°C and
18Ovapor  11‰. Dashed lines link
18O of precipitation with temperature of condensation.
(From Clark and Fritz, 1997, or calculated with PHREEQC, Chapter 11).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

38 From rainwater to groundwater

18O (‰) Bombay

30 25 20 15 10 5
Cape Hatteras
Djakarta
Karachi
Tokyo Rangoon
Hong Kong
Manila New Delhi
Reykavik Valentia
Santa Maria Genova
Copenhagen Hollandia
Luang Prabang
Stuttgart
Isfjord Lista
Pohang
Scoresbysund Chicago

2H (‰)
Wien
Flagstaff 100
Gronnedal
Holsteinsborg
Bethel 120
Goose Bay

Umanak
Edmonton 140
Fort Smith
160
Whitehorse
Sdr. Stromfjord 180
2
Stat. Nord
200
220
Iceberg
1 Iceberg

Figure 2.11. The relation of
2H and
18O in rain for various locations (Dansgaard, 1964). Line 2 is the meteoric
water line, line 1 has the same slope but an intercept of zero.

Comparison with the pattern expected for Rayleigh distillation proceeds as follows. The Rayleigh
equation for
18Orain is:

18O rain 
18O rain, 0  e18O l/g ln f (2.20)

Differentiation towards f, we find d(
18Orain) / df  (18Ol/g) / f in the rain. (The slope of the equation
y  a  b ln x is equal to dy / dx  b/x). A similar relation applies for
2H. The slope among
2H and

18O now becomes:

d
2H / df e2H l/g

  8.6 (18C) (2.21)
d
18O / df e18O l/g

The slope calculated for the Rayleigh process (Equation 2.21) is close to the meteoric water line
(Equation 2.19) and indicates that rainout is an equilibrium process for isotope fractionation. The
slope for the Rayleigh proces is slightly higher because we have neglected that the initial values for

2Hrain, 0 and
18Orain, 0 are non-zero when substituting the
notation for isotopic ratios in the
Rayleigh equation (Problem 2.4).
In contrast, evaporation is a kinetic fractionation process. Kinetic fractionation produces a
vapor that is lighter than predicted by equilibrium fractionation. Moreover, kinetic fractionation is
stronger for 18O than for 2H, and the vapor becomes, relative to the equilibrium process, more
depleted in 18O than in 2H. For example, oceanic vapor at 25°C would have
18O  9.4‰ and

2H  85‰ at equilibrium, but the observed values are close to 13‰ and 95‰, respectively.
As the result, rain obtains its isotopes by equilibrium fractionation during condensation from a vapor

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Stable isotopes in rain 39

that is depleted in 18O compared to the original liquid from which the vapor was generated. Since the
depletion is stronger for 18O than for 2H, the rain has a deuterium excess, which is calculated from
analyzed values by:

d 
2H rain  8
18O rain (2.22)

The meteoric water line (Equation 2.19) gives d  10‰ for average precipitation. A higher deuterium
excess may indicate a contribution from evaporated water of an arid climate with low relative humidity
in the atmosphere (Gonfiantini, 1986; Clark and Fritz, 1997), or a contribution from evaporated inland
water to the rain.
The temperature is the driving force in cooling and condensing atmospheric vapor and results in
the global variations of the isotopic composition of rain (Figures 2.12 and 2.13). In coastal areas the
relation between
18O and temperature is (Dansgaard, 1964):

18O  0.695 t c  13.6‰ (2.23a)

and for deuterium:

2H  5.6 t c  100‰ (2.23b)

where tc is the average yearly temperature in °C. For continental stations the slope in Equation
(2.23a) is less, for example 0.58‰/°C instead of 0.695‰/°C (Rozanski et al., 1993). There is also
a latitudinal effect, indicating that rain is generally more depleted at higher and colder lati-
tudes (Figure 2.13). Altitude is another temperature-related effect that depletes
18O by 0.1
to 0.5‰/100 m.
The temperature dependence also leads to seasonal fluctuations of the isotopic composition of
rain, which are shown for various stations in Figure 2.12. As expected, the rain turns isotopically
lighter in the colder season. The seasonal effect introduces a greater range of isotope compo-
sitions for continental stations with winter-snow (The Pas) than for near-coastal stations (Stanley).

The Pas Addis Ababa Stanley

54.0N 101.1W 9.0N 38.7E 51.7S 57.9W
20 20
T
10 T 10
T
δ18O (‰) or T(°C)

δ 18O
0 0
δ18O
10 δ18O 10

20 20

30 30
JFMAMJJASOND JFMAMJJASOND JFMAMJJASOND

Figure 2.12. Seasonal variations in
18O and temperature in The Pas (Canada), Addis Ababa (Ethiopia) and
Stanley (Falkland Islands) (Rozanski et al., 1993).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 40
From rainwater to groundwater
20 26
10

20 8 24
22 60
6 20 10
18
10 16 8
12 14
8 10
6
6
4 4
68 30
4 2 4 4

8 2
4 6 0
4
6 6
6 6
8
8 6
10 10 4
12 2 30
14 8
16 4 10
4
18 20
6 12
20 14
8 30 16
10 60
40 18
30 12
14 20
16 50
18
22

180 150 120 90 60 30 0 30 60 90 120 150 180 150

Figure 2.13. The global picture of
O in precipitation (Clark and Fritz, 1997).
18

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dry deposition and evapotranspiration 41

Lastly, a continental effect makes the isotopic composition of rain lighter, first because the atmos-
pheric vapor needs progressively lower temperatures for rainout, second because evaporation of
inland waters is a kinetic process which lightens the atmospheric vapor as was discussed before.
During transpiration, on the other hand, plants take up water indiscriminately, and although some
fractionation may occur in the leaves, steady state requires that the water that evaporates from the
leaves has the same isotopic composition as soil water extracted by the roots.

QUESTIONS:
Estimate
18O and
2H of rain in your home town (e.g. Amsterdam, avg. temp. 10°C).
ANSWER: Amsterdam:
18O  6.65‰,
2H  44‰.
Find the deuterium excess in your home town’s rain.
ANSWER: (Amsterdam) d  9.2‰.
What is the meteoric water line according to Equations (2.23a  b)?
ANSWER:
2H  8.06
18O  9.58‰.
What will be the slope of
2H with temperature in continental rain according to Rozanski?
ANSWER: 8  0.58  4.6‰/°C.

2.4 DRY DEPOSITION AND EVAPOTRANSPIRATION

Dry deposition comprises both the deposition of particulate aerosols and of atmospheric gases by
adsorption. Since the Cl concentration in groundwater is being used as a conservative parameter for
estimating the recharge (Schoeller, 1960; Eriksson, 1960; Lerner et al., 1990; Edmunds and Gaye,
1994; Wood and Sanford, 1995; Wood, 1999), it is important to estimate the contribution of dry depo-
sition to the total atmospheric input of Cl. The importance of dry deposition shows up as the difference
between the composition of rainwater collected by wet-only rain gauges, open only during rain, and of
rainwater collected by bulk samplers which remain open all the time (Galloway and Likens, 1978; Table
2.2). During dry periods, bulk samplers collect aerosols as well as anomalies such as bird droppings on
the funnel that wash into the sampling bottle with the next rain. In the temperate climate of the
Netherlands where 800 mm rain is distributed regularly over the year, bulk samplers overestimate the
wet input of the major ions by a factor of about 1.3 (Ridder et al., 1984). The difference between wet-
only and bulk sampling is also fundamental when estimating the atmospheric input of different ele-
ments to the aquifer, since dry deposition depends on the surface characteristics, vegetation, wind
direction etc., all factors that cannot be easily extrapolated from the rain gauge to the natural surface.
The large variability in dry deposition is illustrated in Figure 2.14. It shows the sum of anions,
basically chloride, in soil water below a forest in Denmark. At the windward margin of the forest, a
high rate of dry deposition produces high ion concentrations in the unsaturated zone. Further towards
the center of the forest, the air is filtered and the concentration decreases by up to a factor of 10.
Dry deposition is an important source of elements, as shown already by Garrels and Mackenzie
(1972). Aluminum in dry deposition is mostly derived from soil dust and its dry deposition flux
ranges from 13–380 mg/m2/yr at various locations (Shahin et al., 2000). Associated with aluminum
are trace elements in the same concentration ratio as in the soil and in many cases there are further
increases as the result of anthropogenic emissions.
Over the North Sea, the dry deposition of trace elements such as Cd, Cu, Pb and Zn is about equal
to the wet deposition. The total atmospheric input of Pb into the North Sea is about equal to the total
riverine input, while the atmospheric input of Cd and Zn amounts to about 1⁄3 of the riverine input
(Van Aalst et al., 1983). Dry deposition also contributes with significant amounts of NOx and SO2,
and since gaseous adsorption is relatively important for these components, dry deposition is nor-
mally greater for conifer than for deciduous canopies (Fowler, 1980).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

42 From rainwater to groundwater

Wind

50 m

2 3
4
0

2
Depth (m)

0 2 4 0 2 4 6 0 2 4 6 8 10 12 14
Sum anions (meq/L)

Figure 2.14. Dry deposition and total anion concentrations in the unsaturated zone of a sandy aquifer. At the
windward margin, boring 4, the concentrations are high and decrease then further inward into the forest
(Hansen and Postma, 1995).

Unfortunately, dry deposition on vegetation is notoriously difficult to measure. Samples of through-

fall under the canopy consist of wet input affected by canopy-exchange. Ions may have been lost by
adsorption to the canopy, or concentrations may have increased by admixture of flushing of salts
deposited during the previous dry period and leachates from the vegetation (e.g. Ulrich et al., 1979;
Miller and Miller, 1980). Some idea of the importance of dry Cl deposition can be obtained from
chemical balances of watersheds or lysimeters in which both input and output have been measured
over a period of several years. Figure 2.15 shows the ratio of input and output for a number of ecosys-
tems with different vegetations. The input is the product of the Cl concentration in precipitation mul-
tiplied by the amount of precipitation over the study area (g/year); the output is estimated from either
the concentrations in streamflow water and discharge measurements, from groundwater analyses and
precipitation-surplus, or from lysimeter studies (also in units of g/year). The ratio of output over input
is plotted against Cl concentration in precipitation since a higher Cl concentration in rain should be
associated with a higher abundancy of Cl in aerosols, leading to increased dry deposition.
Figure 2.15 shows a ratio close to 1 in areas covered with heather or low shrubs, so that the output
is balanced by the input measured with bulk rain samplers, without much contribution from additional
dry deposition. In forests with very low Cl concentrations in precipitation, ratios less than 1 are
found, indicating an actual loss of Cl in the watershed. This could possibly be due to biological uptake
(Oberg, 1998) or alternatively to anion adsorption on clay minerals (Feth et al., 1964), although this has
not been documented in the field studies compiled in Figure 2.15. When the average Cl concentra-
tion in wet deposition of forested areas exceeds 50 mol/L, a ratio of around 2 is found. This suggests
a total Cl deposition that is about twice the amount supplied with rain and underlines the importance
of land use for dry deposition. In Figure 2.15 there is no clear difference between deciduous and coni-
ferous forests, except for the 3 lysimeters at a coastal site in the Netherlands with a 40 year long record
(Stuyfzand, 1984) that plot at the extreme right of Figure 2.15. Here, the pine forest has a ratio exceed-
ing 5, which shows that its perennial canopy effectively collects aerosols during winter storms.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dry deposition and evapotranspiration 43

Pine
5 Deciduous/mix.
Bare/heather

Cl-input (prec.) 4
Cl-export

0
0 100 200 300
Cl concentration in precipitation (µmol/L)

Figure 2.15. Ratio of Cl output/input in watersheds and lysimeters against Cl concentration in rain. The
Cl input was measured with bulk rain samplers (Appelo, 1988).

The dry deposition derived from a chemical balance can be compared with a physical model
(Eriksson, 1959). Dry deposition can be calculated from:

D  3.15105  d c (2.24)

where D is dry deposition (g/m2/yr), vd is the dry deposition velocity (cm/s), c is concentration in
(g/m3), and 3.15105 recalculates cm/s into m/yr. The value of vd depends, according to Stokes’ law,
on the aerosol particle diameter, and the various size fractions can be summed to obtain the operational
dry deposition (Holsen and Noll, 1992). The size of aerosol particles often shows a bimodal character,
with one fraction between 0.1 and 0.8 m, and a second fraction between 10 and 100 m. The larger par-
ticles result from an equilibrium of production from seasalt, soil dust, soot, etc. and sedimentation, while
the fine particles are produced by coagulation and condensation of particles, which become strongly
hygroscopic when finer than 0.1 m (Whitley et al., 1972). The small-size aerosols, which are the most
important for transport over more than 10–100 kms, have a considerably higher deposition velocity than
calculated from Stokes’ law, probably because the particles are filtered out by adsorption on fixed
surfaces. Their velocity of deposition is estimated to range from 0.2 to 2 cm/s (Mészáros, 1981; Möller,
1990). If we take vd  1 cm/s, and cCl  5 g/m3, which is the value observed at 600 m high above the
sea (Blanchard and Syzdek, 1984), then D  1.58103 mg/m2/yr is the continental dry deposition rate
of Cl in areas where vegetation scavenges the small aerosol particles. This value translates to 2.0 mg/L
in 800 mm rain per year and dry deposition then approximately doubles the wet-input in forested areas
situated at more than about 10 km from the coast. This is in reasonable agreement with estimates derived
from chemical balances for forests in temperate climate (Figure 2.15; Matzner and Hesch, 1981; Appelo,
1988; Mulder, 1988; Rustad et al., 1994). The model assumes a linear relation among dry deposition and
atmospheric concentration that may be correct for stagnant air in forests, but becomes somewhat ques-
tionable for fields with low shrubs or grass vegetations where air turbulence apparently reduces dry
deposition. For such areas, the total atmospheric input equals the input measured with bulk rain samplers.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

44 From rainwater to groundwater

mm
180 Bulk prec.

Throughfall

120

60

60

Infiltration

120
1984 1986 1987 1988 1989 1990

Figure 2.16. Hydraulic budgets in the temperate climate, conifer forest at Klosterhede, Denmark. (Modified
from Rasmussen, 1988).

Only part of the precipitation will ultimately infiltrate into the aquifer. Figure 2.16 shows the annual
water balance in a conifer forest in a temperate climate with an annual precipitation of 800 mm/yr
and an infiltration of 400 mm/yr. The difference is lost by evapotransporation from vegetation and
soil. Throughfall, water dripping from the canopy, is already seen to be significantly less than the
bulk precipitation, while infiltration in most years starts during the fall and stops again in the spring
so that there is no net infiltration during the summer. Any salt deposition by rain or dry deposition
will accumulate during summer and be flushed down with the autumn rain.
The evapotranspiration in an area can be calculated by comparing the Cl concentrations in soil-
or groundwater and in rainwater:
cCl, rain  Pt  D
E  Pt  (2.25)
cCl, soil

where E is evapotranspiration (mm/yr), Pt is total precipitation (mm/yr), cCl is the Cl concentration
in precipitation (mg/L) and D is 1575 mg/m2/yr for deciduous or mixed forests and 0 for bare or
grass-covered land. When applying Equation (2.25), the precipitation chemistry from bulk samplers
should be used, and the soil- or groundwater concentrations must be averaged. Also, the relation
becomes invalid when other sources of Cl are present, such as manure or waste disposal, or when
the chloride mass transfer has changed over time, for example by changes in land use (Wood, 1999).

EXAMPLE 2.2. Calculate recharge using the Cl mass balance and estimate the flow velocity in the unsatu-
rated zone. Gehrels (1999) sampled and analyzed soil water Cl in the Netherlands’ Veluwe area below grass-
land (Molinia sp.) and Picea/Prunis forest. He obtained bulk rain and dry deposition rates of Cl for the two
vegetation types in the period 1993–1994, when total precipitation was Pt  975 mm/yr.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dry deposition and evapotranspiration 45

Precipitation grass forest

bulk 0.099 0.099 mmol Cl/L
dry 0 99 mmol Cl/m2/yr
total 0.099 0.201 mmol Cl/L
soil, below root 0.19 0.45 mmol Cl/L
w 0.089 0.072 m3/m3
Result
P 508 435 mm net recharge/yr
v 5.7 6.0 m/yr
The dry deposition of 99 mmol Cl/m2/yr is equivalent to 99/975  0.102 mmol Cl/L, which is added to the
concentration in bulk rain. The net recharge P  Pt  mCl, total rain / mCl, soil and the flow velocity v  P/w
are given in the table.

Finally, the importance of evaporation for the chemical composition of surface waters is illustrated
in Figure 2.17. It shows the distribution of total dissolved solids (TDS) in surface waters of
the USA with the highest concentrations occurring in the central parts with a high evaporation,
and the lowest concentrations in the marginal parts of the continent. This pattern is the inverse of the
distribution of Cl in rainwater (Figure 2.4) and underpins the importance of evaporation.

QUESTION:
The dry deposition of Cd is 1 mg/m2/yr. Calculate the concentration in 1000 mm/yr of rain.
ANSWER: 1 g/L

0 1000

km

100 100–350 350–700 700–1150 !1150 parts per million

Figure 2.17. Dissolved solids concentrations in stream waters of the United States (Rainwater, 1962).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

46 From rainwater to groundwater

2.5 MASS BALANCES AND ECOSYSTEM DYNAMICS

Mass balances can be used to estimate the water quality in an area. They consist of an inventory of
all the sources and sinks for the elements of interest, including water, and the calculation of the water
composition from the resulting net output. For agricultural areas the most important terms in the
balance are related to harvesting and the application of manure and fertilizers. The composition of
crops and the manure from various animals is continuously analyzed by various institutions and
some uptake and release rates are listed in Table 2.5. The numbers allow to assess the groundwater
quality below agricultural land as demonstrated in Example 2.3.

Table 2.5. Elemental uptake rates by crops and release rates from animal manure. (1 ha  104 m2)

Water N P K Na Mg Ca Cl S

Crop
Grass, 4 cuttings 4000 (m3/ha/yr) 15.4 0.9 5.4 0.1 0.5 1.0 2 0.7 (kmol/ha/yr)
Green maize 3000 13.4 1.1 4.9 0.1 0.8 0.9 1.7 0.43
Barley 3000 6.1 0.6 1.8 0.05 0.3 0.3 0.5 0.2
Animal manure
Dairy cow 18.1 (m3/yr) 6.2 0.56 2.1 0.64 0.50 0.72 1.7 0.45 (kmol/yr)
Fatting calf 2.2 0.46 0.04 0.11 0.04 0.07 0.06 0.16 0.01
Porker 1.7 0.90 0.12 0.15 0.06 0.05 0.11 0.08 0.06
Breeding pig 3.7 2.0 0.26 0.34 0.13 0.10 0.25 0.17 0.12
1000 chickens 3.7 16 3.0 4.0 1.1 1.3 3.2 1.6 0.99
1000 laying-hens 27 36 11 7.6 2.6 2.5 17 4.0 2.0

EXAMPLE 2.3 Estimate the Cl and NO 3 concentration in groundwater below agricultural land
We consider a farm run by a family of four, with 10 ha grassland and 40 cows. The input consists of precipi-
tation, waste water, manure and fertilizer. The export is by vegetation uptake, harvesting and groundwater
outflow. Concentrations in the groundwater are obtained dividing the net output of an element by the net
output of water.

H2O N Cl

Yearly input
Rain 840 mm 0.18 0.08 mmol/L
1 person 36.5 m3 6 3 mmol/L
1 cow 18.1 m3 340 95 mmol/L
Yearly output
Grass (4 cuttings) 400 mm 15.4 2 kmol/ha
Mass transfer for 10 ha
Rain 8.40104 m3/yr 15.1 6.7 kmol/yr
4 persons 1.46102 0.88 0.44
40 cows 7.24102 246 68
Total 8.49104 262 75.9
Grass 4.00104 154 20
Net 4.49104 m3/yr 108 55.9 kmol/yr
Concentration 2.4 1.2 mmol/L

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Mass balances and ecosystem dynamics 47

The concentrations calculated in Example 2.3 can be compared with drinking water limits, which are
0.4 and 5.5 mmol/L for NO 
3 and Cl (Table 1.1), respectively. The estimated N concentration by far
exceeds the drinking water limit, but fortunately not all the nitrogen will be in the form of NO 3.
Most of the N is introduced as ammonium and amines associated with the organic waste and will
only partly be transformed into nitrate by microbes under aerobic conditions. A part will evaporate
already as ammonia during field application, and some of the nitrate in the soil will be denitrified to
N2 gas and escape to the atmosphere. The various processes which affect nitrogen conversions can
be calculated for agricultural soils with computer programs (Rijtema and Kroes, 1991; Wu and
McGechan, 1998). A rough estimate for an aerobic soil with a neutral pH is, that about half of N will
persist in the form of nitrate and thus will be flushed. However, note that the resulting 1.2 mmol
NO 3 /L in Example 2.3 is still three times higher than the drinking water limit.
The comparison of observed and calculated concentrations will rapidly reveal whether processes
have been neglected in the mass balance. For example, the influence of agricultural practices on
water quality is illustrated in Figure 2.18. It shows the differences in groundwater composition with
depth in a borehole equipped with multilevel samplers. The high NO 3 concentrations in the upper
part of the boring are due to heavy application of manures and fertilizers to fields upstream, and are
close to what was calculated in Example 2.3.
The mass balance calculation is an overall summary in which concentration fluctuations in
time and space are averaged and internal biogeochemical cycles are neglected. However, uptake
and cycling of elements by vegetation and storage in accreting biomass, or the release from decay-
ing organic matter can profoundly influence the concentrations of elements in water leached from
the soil.

Boring V1 (meq/L)
Depth
m.b.s 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8 9 10
0 Legend

NH4 Fe Mg Ca K Na Cl HCO3 SO4 NO3

2
Chemical constituents

Medium sand
4
Gravels
Clay
6
Measured tritium
content
8

10

12

14

16

0 10 20 30 40 50 60 70 80
TU

Figure 2.18. The groundwater chemistry below agricultural land. Concentrations of cations are plotted cumu-
latively to the left, and anions cumulatively to the right (Appelo, 1988).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

48 From rainwater to groundwater

Figure 2.19 illustrates the transfers of Ca2 in the forested ecosystem developed on gneiss and till
deposits at Hubbard Brook (Likens and Bormann, 1995). The input of Ca2 into the ecosystem
comes mainly from the weathering of minerals and for only 10% from precipitation. The outflow of
Ca2 with the runoff is about half of the total input. The difference is due to the uptake of Ca2 in
the biomass of growing trees. Furthermore, the Ca2 uptake by the biomass is 3 times the Ca2
weathering flux, implying that Ca2 is extensively recycling in the soil-biomass system.

Precipitation 2.2

Biomass
increment
8.1
Litter plus
throughfall
Forest floor 54.1
increment 1.4

Uptake 62.2

Weathering 21.1 Soil solution

Outflow to stream
runoff 13.7
Units are kg/ha/yr

Figure 2.19. Cycling of Ca2 in a forested ecosystem at Hubbard Brook (Drever, 1997).

Table 2.6 displays the transfers for other elements in the Hubbard Brook ecosystem and shows signifi-
cant difference among the elements. For example, Mg2 is recycled in the biomass to a much smaller
extent than Ca2. On the other hand K is very strongly conserved within the soil and biomass. The veg-
etation uptake of K is 64 kg/ha/yr, whereas the streamwater output is only 1.9 kg/ha/yr. A nutrient like
P is even more strongly recycled and withhold in the biomass. Natural vegetations are, as may be
expected from an evolutionary standpoint, masterful in storing essential elements with a limited avail-
ability. However, this also implies that a small disturbance in the flux through living matter, for
example by logging trees or bush fires, may greatly disturb the much smaller flux of dissolved con-
stituents leaving the biosphere.

Table 2.6. Elemental fluxes in the biomass compared to fluxes with rain, weathering and streamwater output
in the Hubbard Brook ecosystem (fluxes in kg/ha/yr from Likens and Bormann, 1995).

Ca Mg Na K N S P Cl

Bulk precipitation input 2.2 0.6 1.6 0.9 6.5 12.7 0.04 6.2
Streamwater output 13.7 3.1 7.2 1.9 3.9 17.6 0.01 4.6
Vegetation uptake 62 9 35 64 80 25 9 small
Root exudates 4 0.2 34 8 1 2 0.2 1.8
Weathering release 21 4 6 7 0 1 ? small

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Mass balances and ecosystem dynamics 49

Table 2.6 shows for N and S that precipitation constitutes the main input into the ecosystem. This is
partly due to adsorption of gases like SO2, NH3 and NO2 on the vegetation. Some of these gases are
taken up directly in the plant, while the remainder is flushed away with the rain.
Vegetation uptake is usually determined by considering above-ground biomass only, neglecting
the growth and production of roots. However, the proportion of below-ground production of the total
net primary production varies from about 0.2–0.3 for forests to 0.6 for grasses (Gower et al., 1999)
and is therefore highly significant in the chemical balance, but also quite difficult to analyze.

QUESTIONS:
Estimate the Cl concentration in the urine of a dairy cow and its calf.
ANSWER: From Table 2.5, cow: 1700 mol/yr / 18.1 m3/yr  94 mM; calf: 73 mM.
A dairy cow produces about 50 L milk/day. Estimate the daily water consumption of the animal.
ANSWER: 50 L/day (milk)  50 L/day (urine, Table 2.5)  100 L/day.
Calculate the E.B. of a dairy cow’s faeces, using Table 2.5
ANSWER:   5.18 keq/yr.   2.6 keq/yr (Cl and SO42 only); estimated organic
acids 2.58 keq/yr; N is mainly present in amines.

2.5.1 Water quality profiles in the unsaturated soil

The effects of variations in rainwater composition and dry deposition, and of seasonal fluctuations
in infiltration and evapotranspiration on the groundwater chemistry are seen most clearly in the
unsaturated zone, and particularly when the rocks contain no highly reactive minerals such as car-
bonates. This is illustrated by a water chemistry profile through the unsaturated zone of a sandy
deposit near the coast in Denmark in Figure 2.20.
Very large variations in the total dissolved ion concentrations are mostly due to variations in
the concentrations of Na and Cl which both have a marine origin. At this site the average
water transport rate through the unsaturated zone is about 5 m/yr, so that the whole profile repre-
sents roughly one year of infiltration, although net infiltration occurs only from October to May.

meq/L meq/L
0 1 2 3 4 5 0 1 2 3 4 5
0 0

1 1

K Ca
2 Mg 2
Na AI CI
Depth (m)

3 3
SO4

4 4

5 5

6 6

Figure 2.20. The groundwater chemistry in the unsaturated zone of a sandy, carbonate free sediment under
a conifer forest in Denmark. Concentrations are plotted cumulatively. The annual precipitation is 860 mm/yr,
infiltration 400 mm/yr, and unsaturated zone pore water velocity 4–5 m/yr (Hansen and Postma, 1995).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

50 From rainwater to groundwater

During the fall, the salts accumulated in the soil over the previous summer period start to be washed
down into the unsaturated zone. During the same season westerly storms are frequent and may
deposit large amounts of sea salt. The effects of these two processes are not easily separated and in
addition, transpiration may also contribute to the maximal concentrations observed in Figure 2.20 in the
upper meters.
The influence of dry deposition, and differences in evapotranspiration on the Cl content of
water in the unsaturated zone of a sandstone in England are illustrated in Figure 2.21. Below heath-
land, Cl concentrations are low and show little variation. Below the forest, chloride concentrations
are much higher and vary widely, although generally lower concentrations are found below the root-
ing depth. The high Cl concentrations below the forest can partly be explained by the effect of dry
deposition (Figure 2.15). Moreover, evapotranspiration varies distinctly between heath/grassland
and forested areas. In particular interception (evaporation from wetted surfaces) is much higher in
forested areas and may amount to about 40% of the annual precipitation (Calder, 1990, 2003). The
concentration of dissolved ions becomes correspondingly higher. Moss and Edmunds (1989) found
that the recharge rate below heathland was roughly 3 times higher than below the forest.
Unsaturated profiles often display the characteristic shape of Figures 2.20 and 2.21 with higher con-
centrations present in the rooted, upper part than at greater depth. In most cases these profiles are stable
throughout the year in both arid and temperate climate (Sharma and Hughes, 1985; Gehrels, 1999).
This indicates that the rain is short-circuiting to greater depths via root channels and along water-
repellent, humic surfaces of soil cutans, and only partly exchanges with the more concentrated soil
solution in the stagnant structure. Short-circuiting implies that the estimates of net recharge with the
Cl mass balance method must be based on water quality from below the root zone. The bypass flow
also explains why seasonal signals in isotope and Cl concentrations in rain are quickly lost in the
soil solution.

EF1

HGC10 HGC4-7
Depth (m)

12
0.0 4.0 4.0 0.5
CI (mmol/L)

Figure 2.21. Chloride profiles in the unsaturated zone of the Sherwood sandstone below birch woodland, Cyprus
Pine and heathland (Moss and Edmunds, 1989).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Overall controls on water quality 51

Depth (m) 2

5
0 0.02 0.04 0 4 8 12 16 20

g (g/g) Cl (g/L)

Figure 2.22. Profiles of gravimetric water content (left, grams per gram) and Cl concentration in the soil
solution below an Eucalyptus stand in a semi-arid climate, Murbko, S. Australia. Annual precipitation is
260 mm/yr, potential evaporation 1800 mm/yr and the recharge rate 0.1 mm/yr (Cook et al., 1994).

In arid or semi-arid areas where evapotranspiration is high, very high concentrations of salts may
accumulate in the soil solution. Figure 2.22 shows the Cl distribution in a soil solution profile
below a stand of Eucalyptus in S.Australia. As the water moves downward through the soil, roots
selectively remove water molecules and the Cl concentration increases. First at a depth of about 3 m
the effect of evaporation ceases and the Cl concentration, at 14 g/L, becomes constant over depth.
If the chloride mass balance method is used to calculate the recharge rate, clearly the Cl concen-
tration at depths !3 m should be used in this case. The calculated recharge rate was 0.1 mm/yr which
can be compared with the mean annual precipitation of 260 mm/yr. The low recharge rates will also
give long reponse time of the Cl profiles upon changes in landuse or climate before a new steady
state situation is attained (Allison et al., 1994; Phillips, 1994).

2.6 OVERALL CONTROLS ON WATER QUALITY

The chemical composition of fresh groundwaters of the USA is shown in a frequency plot in Figure 2.23.
The most abundant cations are Ca2, Na and Mg2. The main anions are HCO 2 
3 , SO4 and Cl . The

data in Figure 2.23 are older than 1960, before groundwater became polluted with NO3 . More recent
data would place NO3 together with the other major anions. Minor components are Fe2, Mn2, and
many trace metals, as well as anions like F and boron, are also plotted. For some elements, such as
Na, SO42, and Cl the distributions approach the sigmoidal shape of a normal distribution. This
suggests that the concentration of these species depends on their availability in rocks, very slowly
dissolving minerals, or control by biological processes, which have a random character on the con-
tinental scale. For other components, like Ca2, HCO  
3 , SiO2, K and F , the gradient steepens at
higher concentrations which could indicate that the solubility of a mineral places an upper limit on
the maximal concentration of the species in natural waters.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

52 From rainwater to groundwater

100
ium
an
90 Ur n 2 2


ro Fe

SO 2
2 Bo Sr

4
Mn 

Cl 
80 Li

Ca 2 

lids
2


70 Sr

3
NO
3

so
Br 
Al

lved
HCO 
disso 3
K

60
Fe 2

2
SiO

Na 
% 50
F

Total

g 2
40

Ca 2
M
30 
3
NO
20
K
2

2
SiO
10 SO 4

0.01 0.1 1 10 100 1000

Parts per million

Figure 2.23. Concentrations of different species in groundwaters in the USA displayed in a frequency plot
(Davis and De Wiest, 1966). Reprinted by permission of John Wiley & Sons, Inc.

Zooming in from the continental scale to a single watershed, the complexity increases and a myriad
of processes may affect the groundwater chemistry, as is outlined schematically in Figure 2.24. We
have already discussed the importance of the external factors like rainwater chemistry, evapotran-
spiration and uptake by vegetation in some detail. What follows is a concise introduction to the
processes that affect water quality in the soil and groundwater zone.

10

1
10
2
3
4 5 9
6
7
1
1
7 8 1
11
clay
11
8
sh
fre lt
sa
5

1 Evaporation 7 Precipitation of secondary minerals

2 Transpiration 8 Mixing of water
3 Selective uptake by vegetation 9 Leaching of fertilizers, manure
4 Oxidation/Reduction 10 Pollution
5 Cation exchange 11 Lake/sea biological processes
6 Dissolution of minerals

Figure 2.24. An overview of processes that affect the water quality in a watershed.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Overall controls on water quality 53

Oxidation/reduction processes are of importance and often related to the decay of organic matter. It
is the reverse process of uptake and storage of elements in growing plants. Decay of organic matter
is an oxidation reaction. The process uses oxygen and produces carbonic acid:
CH 2O  O 2 → H 2O  CO 2 (2.26)

In this reaction a carbohydrate, CH2O, is used as a simple representative of the, in reality, highly
complex organic compounds in organic material.
The oxidation of organic matter may occur in the soil but also within aquifers where fossil
organic matter is present as peat, lignite, etc. Once oxygen becomes exhausted other electron accep-
tors, like nitrate, iron oxides or sulfate may mediate the oxidation of organic matter. The reduction
of iron oxides results in high iron concentrations in groundwater, the reduction of sulfate in badly
smelling hydrogen sulfide and finally, fermentation of organic matter may result in groundwaters
containing methane. The production of CO2 as result of organic matter oxidation (Equation 2.26)
may also enhance the weathering of carbonate and silicate minerals.
Weathering and dissolution of minerals releases elements to the water. The relation between rock
type and water composition is illustrated in the composition of the spring waters of the Ahrntal
(Zillertal Alps) in N. Italy. The geology displays a typical central alpine sequence of micaschists, lime-
stones, dolomites, gypsiferous layers, and serpentinite rock. The rocks dip vertically and are incised
by alpine trough valleys (Figure 2.25). Table 2.7 lists the characteristic chemical composition for each
of the four rock types; Group 1 consists of calcareous schist, a metamorphic rock containing CaCO3
as its most reactive component and Ca2 and HCO 3 are the predominant ions in the spring water.

WEST GERMANY
N Zurich
Munich Wien
AUSTRIA
SWITZERLAND
FRANCE

1 Bolzano
Milano
Zagreb
ITALY
PUR

iss
BU

ne
SCHB.

St. Jacob
IN

og
th
LA

1 1
or 1
ND

ern
au
B.

T HR
BÄR

A 1 1 1
1 1
ENB.

Steinhaus 2
GR

2 2
OS S

3 2
KL
EIN

3
KL

2 2 3 3
K

AU
LA

2 3
4
SE
US

32 2 Weisse wand
NB.
EN

2
2 1
JA

2
13 3
B.
R

4 14
LB

2 1 11 4
.

3 Katzenkofel
2 2
3 1
1 3 SPRINGS
4 water-types: Calcite-saturation
1: Calcareous schist log(IAP/K ):
2: Dolomite ! 0.2 (Supersat.)
Hirbernock 3: Serpentinite/ofiolite  0.2
4: Micaschist  0.2 (Undersat.)
LITHOLOGY (after geological map of Italy)
Durreck Unconsolidated
(Quatenary)
Rauchkofel Prasinite/ofiolite
Pennide nappe
0 1 2 3 km Calcareous schists
Serpentinite/ofiolite
Quartzite “Matrei zone”
Dolomite
Micaschist Austride nappe

Figure 2.25. The geology of the Ahrntal (N. Italy) with location and classification of springs.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

54 From rainwater to groundwater

Table 2.7. Typical composition of spring-waters. “mcs” micaschist; “serp”  serpentinite; “pras” 
prasinite/ofiolite; “dol”  dolomite; “cc”  calcareous schist (Appelo et al., 1983).

Group code 4 4 3 3 2 2 1
Geol. unit “mcs” “mcs” “serp” “pras” “dol” “dol” “cc”

pH 7.63 7.70 8.47 7.41 8.00 7.93 8.12

EC (S/cm) 52 100 152 610 325 505 250
Temp (°C) 3.7 4.0 4.0 4.8 5.5 7.5 2.0
Na (mmol/L) 0.04 0.03 0.02 0.06 0.04 0.04 0.04
K 0.016 0.026 0.011 0.015 0.011 0.008 0.011
Mg2 0.11 0.12 0.44 2.30 0.67 0.93 0.31
Ca2 0.17 0.35 0.33 1.20 0.77 1.95 0.94
Cl 0.05 0.05 0.06 0.05 0.04 0.06 0.05
HCO 3 0.37 0.86 1.35 5.54 2.10 3.34 2.23
SO2
4 0.08 0.04 0.05 0.72 0.14 1.19 0.15
Si 0.05 0.05 0.06 0.13 0.06 0.09 0.05

Group 2 comprises rock with the mineral dolomite (CaMg(CO3)2) and the water contains, apart from
Ca2 and HCO 3 , significant amounts of Mg . At some places the mineral gypsum (CaSO4 2H2O) is
2

present and the water becomes enriched in SO42. Group 3 contains magmatic rocks with Mg-minerals
like serpentine and talc. These weather easily and produce waters enriched in Mg2, HCO 3 and silica.
Finally group 4 consists of metamorphic micaschists. The minerals in this rock have a low reactivity and
the solute concentrations are low. Table 2.8 lists the major elements and their main mineral source (as
well as some additional sources), and the range of concentrations which may be expected in fresh waters.

Table 2.8. Normal ranges of concentrations in unpolluted fresh water and the sources of elements.

Element Concentrations (mmol/L) Source

Na 0.1–2 Feldspar, Rock-salt, Zeolite, Atmosphere, Cation exchange

K 0.01–0.2 Feldspar, Mica
Mg2 0.05–2 Dolomite, Serpentine, Pyroxene, Amfibole, Olivine, Mica
Ca2 0.05–5 Carbonate, Gypsum, Feldspar, Pyroxene, Amfibole
Cl 0.05–2 Rock-salt, Atmosphere
HCO 3 0–5 Carbonates, Organic matter
SO42 0.01–5 Atmosphere, Gypsum, Sulfides
NO3 0.001–0.2 Atmosphere, Organic matter
Si 0.02–1 Silicates
Fe2 0–0.5 Silicates, Siderite, Hydroxides, Sulfides
PO4 0–0.02 Organic matter, Phosphates

Different mechanisms of rock weathering can be deducted from Figure 2.26. It shows the maximal
dissolved solids concentration in streams from small catchments, located in different rock types, as
a function of the mean annual runoff (Walling, 1980). It is instructive to interpret the mean annual
runoff in terms of the residence time of water. The average catchment will have a soil that is 1 m
thick with a water filled porosity of 10%. The soil profile then contains 0.1 m water. With a precipi-
tation surplus (rain – evapotranspiration) of 0.1 m/yr, it will take 1 year for the water to travel
through the soil. If the precipitation surplus were 0.2 m/yr, the water travel time would be half a year,
etc. The annual runoff is therefore inversely related to the residence time of water in the soil; the
higher the annual runoff, the shorter the contact time between water and rock.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Overall controls on water quality 55

5000

3000

Maximum dissolved solids concentration (mg/L)

Sh
ale
&
sa
1000

nd
sto
n e
500

Limestone

G
ra

San
n ite
100

dsto
Volcanic
s

ne
50

Sc
his
Sand & gravel

t, g
ne
iss
10

5
30 50 100 500 1000
Mean annual runoff (mm)

Figure 2.26. Differences in total dissolved solid concentration in surface runoff as a function of rock type and
annual runoff (Walling, 1980). Reprinted by permission of John Wiley & Sons, Ltd.

Two trends can be identified in Figure 2.26. First, the content of total dissolved solids in the water
derived from limestone, volcanics, sand and gravel deposits is almost independent of the amount of
runoff, which is not the case for the other rock types. Second, for a specific value of the annual
runoff, the total dissolved concentration varies for different rocks. The differences are related to the
solubilities of the minerals present in the parent rock, and the rate of dissolution of these minerals.
Gravel consists mainly of insoluble quartz pebbles which are highly resistant to dissolution.
Accordingly, the concentration level remains low, and the composition of stream water is mainly
controlled by rainwater chemistry and evapotranspiration. On the other hand, volcanics and lime-
stones contain more soluble minerals so that the solute concentration in runoff increases. The dis-
solved solute concentration here is nearly constant and independent of runoff. This indicates that the
dissolution rate of the solids is fast compared to the residence time of the water in the drainage area,
and the concentration level may well represent the solubility of the minerals.
On the other hand, streamflow from granites and sandstones show concentration levels
that depend strongly on the annual runoff. This suggests that the dissolution rate of the minerals
is slow compared to the residence time and that the concentration level is determined by the latter.
The dominant minerals in granite and sandstone are feldspars, micas and quartz and, as is shown in
Chapter 8, these minerals exhibit indeed very slow dissolution kinetics. In other words, Figure 2.26
demonstrates that in some cases, as for limestones, an equilibrium approach is appropriate, while in
other situations, as with most silicate rocks, dissolution kinetics must also be considered.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

56 From rainwater to groundwater

10

pH pH
9 0.10
HCO
3


8

3
O
C

0.01 a 0.01
Molality of dissolved species

Molality of dissolved species

N

SiO2
HCO
3
K
0.001 Silicagel 
0.001
4
SO
pptes
2
O
Si


l
C

Ca2
Calcite pptes
Sepiolite pptes  precipitates
pptes
2
Mg
0.0001 0.0001

CO
3
0.00001 0.00001
Ca2
Mg2

1 2 5 10 30 100 1000
Concentration factor

Figure 2.27. Calculated results of evaporation of Sierra Nevada (USA) spring water (modified from Garrels and
Mackenzie, 1967).

Minerals may also precipitate in soils or aquifers. Weathering of silicate minerals leads generally to the
formation of secondary clay minerals. The type of clay mineral that forms depends on the composi-
tion of the parent rock and the weathering stage. Basically, clay minerals consist of Al-silicates with
or without other cations. As weathering proceeds, the clay minerals are stripped first of cations and
then of silicon, until the sparingly soluble Al-hydroxide remains. If iron is present in the parent rock,
the formation of Fe-oxides can also be expected.
In arid climates, evaporation may lead to the precipitation of a suite of minerals in the soil. These
include calcite, gypsum, and chloride-salts. The changes in solute concentration and the sequence of
minerals that precipitate during evaporation can be calculated for a given initial water composition, as
shown for the Sierra Nevada spring waters in Figure 2.27. In this case, the initial water stems from
springs draining a granitic terrain, and the minerals which precipitate are sepiolite (a Mg-silicate), cal-
cite, and silica-gel. The final water is a Na-Cl  HCO3 brine which is commonly found in soda lakes.
Ion exchange reactions, between dissolved cations and those sorbed to mineral surfaces may pro-
foundly change the water chemistry when the infiltrating water differs from what is already present.
Examples are found in fresh and salt water displacements in coastal aquifers, pollution plumes

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Overall controls on water quality 57

spreading out from waste sites, and acid rain moving downward through soil. Ion exchange tends to
smoothen concentration gradients which are the result of changing conditions in the groundwater
reservoir, but it can also produce concentration peaks in a chromatographic sequence. The sediment
exchange complex in a soil consists of clay minerals, iron oxides, weathered primary minerals and
organic material. Per volume of wet sediment, the exchange complex can easily contain an amount
of in essence mobile cations that is 300 times larger than is present in soil moisture!
Mixing of different water qualities occurs in coastal zones between fresh water and seawater, but
also in groundwater seepage zones, and near springs. An upland area with locally different characteris-
tics (geology, vegetation) can discharge different water qualities to the same spring. Mixing of different
waters by dispersion is also important for transport of pollutants in aquifers. Mixing of waters can lead
to subsaturation with respect to calcite, even if the original waters were at saturation before mixing.
Anthropogenic activities can thoroughly change water quality. These contributions may be of any
kind, from NaCl used in households to heavy metals or poisonous organic constituents leached from
tip heaps. Air pollution and acid rain already have been mentioned. The large amounts of NO 3 and
pesticides leaching from agricultural fields constitute a serious problem in many countries.

Major oceans Major oceans

Black

Caspian
10,000 10,000
Jordan Baltic
Pecos Sacramento

os
Pec os
Pec
Total dissolved salts (ppm)

Total dissolved salts (ppm)

Jordan
1000 1000
Colorado Colorado
Don
Rio Grande Rio Grande
Volga

Yukon Rhine Indus

Mississippi
Nile
Ganges
Columbia Lena Columbia Parana
Ob Mobile
100 Congo Niger 100 Niger
Lake Superior
Orinoco Pechora

10 Negro 10
Negro

.1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1.0
Na/(Na  Ca) Cl/(Cl  HCO3)

Figure 2.28. The chemistry of world surface waters expressed as function of TDS and the relative contribution
of Na and Ca2, and of Cl and HCO 3 (Gibbs, 1970).

This chapter started with the hydrological cycle, noting that groundwaters mostly discharge
to rivers and surface waters. Thus, surface runoff will present an overall average of groundwater
compositions in the drainage basin (although quality changes are retarded by the long groundwater
residence times). Figure 2.28 assembles surface water compositions from all over the world,

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

58 From rainwater to groundwater

graphing total dissolved solids as a function of the dominance of Na or Ca2 and of Cl or HCO 3.
The figure illustrates the processes in groundwater quality in a nutshell. At low dissolved concen-
trations, the dominant ions are Na and Cl, contributed by rainwater without much geochemical
reaction. Contact with rapidly dissolving calcite and Ca-silicates increases the relative content of
Ca2 and HCO 3 . If subsequent evaporation concentrates the solution, Ca
2
and HCO 3 are lost by
precipitation of CaCO3 and the water composition moves upwards in the figures and returns to the
seawater composition dominated by Na and Cl.
Thus, the overall picture of groundwater quality can be understood with simple, qualitative rea-
soning. However, for the details a variety of chemical, physical and biological processes must be
considered, which are discussed in the remainder of this book.

PROBLEMS
2.1. Estimate the distance from the coast where the global averaged proportion of recycled precipitation equals 1.
2.2. Estimate the integrated proportion of recycled precipitation from 0 to 2500 km from the coast, using
Equation 2.2.
2.3. The composition of rain at 3 locations in the Netherlands is listed in the table (Summer-
averages, 1979, concentrations in mol/L, except pH).

pH NH4 Na K Ca2 Mg2 Cl NO

3 SO42

Vlissingen 4.00 49 265 8 39 34 297 92 85

Deelen (Veluwe) 3.95 119 30 3 16 5 37 90 ..
Beek (Limburg) .... 143 23 11 50 5 34 87 96
Sea-water (mmol/l) – – 485 10.6 10 55 566 – 29.5

a. Check the quality of the Vlissingen-analysis. Estimate missing parameters in the two other analyses.
b. Calculate seawater-influence in the three samples.
c. Which other influences can be discerned?
2.4. Calculate the MWL slope by differentiating Equation (2.11) with respect to f after substituting
for R. Use

02H  95‰ and
018O  12.2‰ (values for oceanic vapor), f  0.9, l/g2H  1.0876 and
l/g18O  1.0099 (values for 18°C). Also calculate the deuterium excess.
2.5. A sample contains 16O  55.45 mol/L. Calculate
18O when 18O  0.11117, 0.11118 and 0.11 mol/l. Also
find the number of digits accuracy for 18O, to obtain 0.05‰ accuracy in the
value.
2.6. In unsatured profiles, Gehrels (1999) measured Cl concentrations of 0.7 and 0.19 mmol/L in the root-
zone and below, respectively. The water filled porosity was rz  0.25 in the root zone and w  0.089
below. Estimate the concentration in the stagnant part of the root zone.
2.7. How does the ratio of Na- over Ca-concentration change in rain from the coast to further inland locations?
2.8. How does the Na / Ca-ratio change when surface water evaporates?
2.9. Which elements are enriched in vegetation when compared with soil moisture?
2.10. What is the effect of a larger discharge on runoff composition from:
a. gravel; b. calcareous rocks; c. granite.
2.11. Find the change of
2H in water condensing in a closed bucket cooling from 20°C to 0°C. Initially there
is only vapor in the bucket,
2Hvapor, 0  80‰. 2Hl/g  91.3‰ (at 15°C).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 59

REFERENCES
Allison, G.B., Gee, G.W. and Tyler, S.W., 1994. Vadose-zone techniques for estimating groundwater recharge in
arid and semiarid regions. Soil Sci. Soc. Am. J. 58, 6–14.
Andreae, M.O., Talbot, R.W., Berresheim, H. and Beecher, K.M., 1990. Precipitation chemistry in Central
Amazonia. J. Geophys. Res. 95, 16987–16999.
Appelo, C.A.J., 1988. Water quality in the Hierdensche Beek watershed (in Dutch). Free University,
Amsterdam, 100 pp.
Appelo, C.A.J., Groen, M.M.A., Heidweiller, V.M.L. and Smit, P.M.H., 1983. Hydrochemistry of springs in an
alpine carbonate/serpentinite terrain. Int. Conf. Water-Rock Interaction, Misasa, Japan, Extend. Abstr.
26–31.
Baisden, W.T., Blum, J.D., Miller, E.K. and Friedland, A.J., 1995. Elemental concentrations in fresh snowfall
across a regional transect in the northeastern US – apparent sources and contribution to acidity. Water Air
Soil Poll. 84, 269–286.
Berner, E.K. and Berner, R.A., 1996. Global environment: water, air, and geochemical cycles. Prentice-Hall,
Englewood Cliffs, 376 pp.
Blanchard, D.C. and Syzdek, L.D., 1988. Film drop production as a function of bubble size. J. Geophys. Res. 93,
3649–3654.
Boschloo, D.J. and Stolk, A.P., 1999. National network for analysis of rainwater chemistry (in Dutch). RIVM
Rep. 723101049/054.
Budyko, M.I., 1982. The earth’s climate: past and future. Academic Press, New York, 307 pp.
Calder, I.R., 1990. Evapotranspiration in the uplands. Wiley and Sons, New York, 148 pp.
Calder, I.R., Reid, I., Nisbet, T.R. and Green, J.C., 2003. Impact of lowland forests in England on water
resources: Application of the hydrological land use change (HYLUC) model. Water Resour. Res. 39 (11),
art. 1319.
Chacko, T., Cole, D.R. and Horita, J., 2001. Equilibrium oxygen, hydrogen and carbon isotope fractionation fac-
tors applicable to geologic systems. Rev. Mineral. 43, 1–81.
Clark, I.D. and Fritz, P., 1997. Environmental isotopes in hydrogeology. CRC Press, Boca Raton, 328 pp.
Cook, P.G., Jolly, I.D., Leaney, F.W., Walker, G.R., Allan, G.L., Fifield, L.K. and Allison, G.B., 1994.
Unsaturated zone tritium and chlorine 36 profiles from southern Australia: Their use as tracers of soil water
movement. Water Resour. Res. 30, 1709–1719.
Dansgaard, W., 1964. Stable isotopes in precipitation. Tellus 16, 436–468.
Davis, S.N. and DeWiest, R.C.M., 1966. Hydrogeology. Wiley and Sons, New York, 463 pp.
Drever, J.I., 1997. The geochemistry of natural waters, 3rd ed. Prentice-Hall, Englewood Cliffs, 436 pp.
Duce, R.A. and Hoffman, E.J., 1976. Chemical fractionation at the air/sea interface. Ann. Rev. Earth Planet. Sci.
4, 187–228.
Edmunds, W.M. and Gaye, C.B. 1994. Estimating the spatial variability of groundwater recharge in the Sahel
using chloride. J. Hydrol. 156, 47–59.
Eltahir, E.A.B. and Bras, R.L., 1996. Precipitation recycling. Rev. Geophys. 34, 367–378.
Eriksson, E., 1959. The yearly circulation of chloride and sulfur in nature; meteorological, geochemical and
pedological implcations. Part I. Tellus 11, 375–403.
Eriksson, E., 1960, The yearly circulation of chloride and sulfur in nature; meteorological, geochemical and
pedological implications, Part II. Tellus 12, 63–109.
Feth, J.H., Roberson, C.E. and Polzer, W.L., 1964. Sources of mineral constituents in water from granitic rocks,
Sierra Nevada, California and Nevada. US Geol. Surv. Water Supply Pap. 1535 I, 70 pp.
Fowler, D., 1980. Removal of sulphur and nitrogen compounds from the atmosphere in rain and by dry deposi-
tion. In D. Drabløs and A. Tollan (eds), Ecological impact of acid precipitation, 22–32, SNSF Project, Oslo.
Galloway, J.N. and Likens, G.E., 1978. The collection of precipitation for chemical analysis. Tellus 30, 71–82.
Garrels, R.M. and Mackenzie, F.T., 1967. Origin of the chemical compositions of some springs and lakes. In
W. Stumm (ed.), Equilibrium concepts in natural water systems. Adv. Chem. Ser. 67, 222–242.
Garrels, R.M. and Mackenzie, F.T., 1972. Evolution of sedimentary rocks. Norton, New York, 397 pp.
Gehrels, J.C., 1999. Groundwater level fluctuations. Ph.D. thesis, Free University, Amsterdam, 269 pp.
Gibbs, R., 1970. Mechanisms controlling world water chemistry. Science 170, 1088–1090.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

60 From rainwater to groundwater

Gonfiantini, R., 1986. Environmental isotopes in lake studies. In P. Fritz and J.Ch. Fontes (eds), Handbook of
environmental isotope geochemistry, 113–168, Elsevier, Amsterdam.
Gower, S.T., Kucharik, C.J. and Norman, J.M., 1999. Direct and indirect estimation of leaf area index, f_apar,
and net primary production of terrestrial ecosystems. Remote Sens. Env. 70, 29–51.
Granat, L., 1972. On the relation between pH and the chemical composition in atmospheric precipitation. Tellus
24, 550–560.
Hansen, B.K. and Postma, D., 1995. Acidifcation, buffering, and salt effects in the unsaturated zone of a sandy
aquifer, Klosterhede, Denmark. Water Resour. Res. 31, 2795–2809.
Hedin, L.O., Granat, L., Likens, G.E., Buishand, T.A., Galloway, J.N., Butler, T.J. and Rodhe, H., 1994. Steep
declines in atmospheric base cations in regions of Europe and North America. Nature 367, 351–354.
Holland, H.D. and Petersen, U., 1996. Living dangerously. Princeton Univ. Press, 490 pp.
Holsen, T.M. and Noll, K.E., 1992. Dry deposition of atmospheric particles – application of current models to
ambient data. Env. Sci. Technol. 26, 1807–1815.
Jain, M., Kulshrestha, U.C., Sarkar, A.K. and Parashar, D.C., 2000. Influence of crustal aerosols on wet deposi-
tion at urban and rural sites in India. Atmos. Env. 34, 5129–5137.
Junge, C.E. and Werby, R.T., 1958. The concentration of chloride, sodium, potassium, calcium and sulfate in
rain water over the United States. J. Meteorol. 15, 417–425.
KNMI/RIVM, 1985. Chemical composition of precipitation over the Netherlands. Ann. Rep. 1983. KNMI, De
Bilt, the Netherlands.
Lerner, D.N., Issar, A.S. and Simmers, I., 1990. Groundwater recharge. Int. Contrib. Hydrogeol., vol 8, IAH,
345 pp.
Lightowlers, P.J. and Cape, J.N., 1988. Sources and fate of atmospheric HCl in the UK and Western Europe.
Atmos. Env. 22, 7–15.
Likens, G.E. and Bormann, F.H., 1995. Biogeochemistry of a forested ecosystem, 2nd ed. Springer, New York, 159 pp.
Matzner, E. and Hesch, W., 1981. Beitrag zum Elementaustrag mit dem Sickerwasser unter verschieden
Ökosystemen im Nordwestdeutschen Flachland. Z. Pflanzenern. Bodenkd. 144, 64–73.
Mészáros, E., 1981. Atmospheric chemistry, fundamental aspects. Elsevier, Amsterdam, 201 pp.
Miller, H.G. and Miller, J.D., 1980. Collection and retention of atmospheric pollutants by vegetation. In
D. Drabløs and A. Tollan (eds), Ecological impact of acid precipitation, 33–40, SNSF Project, Oslo.
Möller, D., 1990. The Na/Cl ratio in rainwater and the seasalt chloride cycle. Tellus 42B, 254–262.
Monahan, E.C., 1986. The ocean as a source for atmospheric particles. In P. Buat-Menard (ed.), The role of air-
sea exchange in geochemical cycling, 129–163, D. Reidel, Dordrecht.
Morgan, J.J., 1982. Factors governing the pH, availability of H+, and oxidation capacity of rain. In E.D.
Goldberg (ed.), Atmospheric chemistry, 17–40. Springer, Berlin.
Moss, P.D. and Edmunds, W.M., 1989. Interstitial water-rock interaction in the unsaturated zone of a Permo-tri-
assic sandstone aquifer. In D.L. Miles (ed.), Proc. 6th water-rock bnteraction symp., 495–499, Balkema,
Rotterdam.
Mulder, J., 1988. Impact of acid atmospheric deposition on soils. Ph.D. thesis, Wageningen.
Nativ, R. and Issar, A., 1983. Chemical composition of rainwater and floodwaters in the Negev desert, Israel.
J. Hydrol. 62, 201–223.
Newell, R.E., 1971. The global circulation of atmospheric pollutants. Scientific American 224, 32–43.
Oberg, G., 1998. Chloride and organic chlorine in soil. Acta Hydrochim. Hydrobiol. 26, 37–44.
Peixóto, J.P. and Kettani, M.A., 1973. The control of the water cycle. Scientific American 228, 46–61.
Peixóto, J.P. and Oort, A.H., 1983. The atmospheric branch of the hydrological cycle and climate. In A. Street-
Perrott et al. (eds), Variations in the global water budget, 5–65. D. Reidel, Dordrecht.
Phillips, F.M., 1994. Environmental tracers for water movement in desert soils of the American Southwest. Soil
Sci. Soc. Am. J. 58, 15–24.
Raemdonck, H., Maenhout, W. and Andreae, M.O., 1986. Chemistry of marine aerosol over the tropical and
equatorial Pacific. J. Geophys. Res. 91, 8623–8636.
Rasmussen, L., 1988. Sur nedbørs effekt på ionbalancen og udvaskningen af metaller og anioner i danske
nåleskovøkosystemer i perioden 1983–1987. Report DTH, Lyngby.
Rainwater, F.H., 1962. Stream composition of the conterminous United States. US Geol. Surv. Hydrol. Inv. Atlas
HA 61.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 61

Ridder, T.B., Baard, J.H. and Buishand, T.A., 1984. The influence of sampling methods on the quality of preci-
pitation (in Dutch). KNMI Techn. Rep. 55, 40 pp.
Rijtema, P.E. and Kroes, J.G., 1991. Some results of nitrogen simulations with the model ANIMO. Fert. Res. 27,
189–198.
Rozanski, K., Araguás-Araguás, L. and Gonfiantini, R., 1993. Isotopic patterns in modern global precipitation.
In P.K. Swart et al. (eds), Continental isotope indicators of climate, AGU Monogr. 78, 1–36.
Rustad, L.E., Kahl, J.S., Norton, S.D. and Fernandez, I.J., 1994. Underestimation of dry deposition by through-
fall in mixed northern hardwood forests. J. Hydrol. 162, 319–336.
Schoeller, H., 1960. Salinity of groundwater, evapotranspiration and recharge of aquifers (in French). IASH
Pub. 52, 488–494.
Sequeira, R. and Kelkar, D., 1978. Geochemical implications of summer monsoonal rainwater composition
over India. J. Appl. Meteor. 17, 1390–1396.
Shahin, U, Yi, S.M., Paode, R.D. and Holsen, T.M., 2000. Long-term elemental dry deposition fluxes measured
around Lake Michigan with an automated dry deposition sampler. Env. Sci. Technol. 34, 1887–1892.
Sharma, M.L. and Hughes, M.W., 1985. Groundwater recharge estimation using chloride, deuterium and
oxygen-18 profiles in the deep coastal sands of Western Australia. J. Hydrol. 81, 93–109.
Stuyfzand, P.J., 1984. Effects of vegetation and air pollution on groundwater quality in calcareous dunes near
Castricum: measurements from lysimeters (in Dutch). H2O 17, 152–159.
Thorstenson, D.C. and Parkhurst, D.L., 2004. Calculation of individual isotope equilibrium constants for geo-
chemical reactions. Geochim. Cosmochim. Acta 68, 2449–2465.
Ulrich, B., Mayer, R. and Khanna, P.K., 1979. Deposition of air pollutants and their effects on forest ecosystems
in Sölling (in German). Sauerländer, Frankfurt a.M., 291 pp.
Van Aalst, R.M., Van Aardenne, R.A.M., De Kreuk, J.F. and Lems, Th., 1983. Pollution of the North Sea from
the atmosphere. TNO-Report CL 82/152., TNO, Delft, 124 pp.
Vermeulen, A.J., 1977. Immision measurements with rain gauges (in Dutch). Prov. North Holland, Haarlem.
Wagner, G.H. and Steele, K.F., 1989. Na+/Cl—ratios in rain across the USA, 1982–1986. Tellus 41B, 444–451.
Walling, D.E., 1980. Water in the catchment ecosystem. In A.M. Gower (ed.), Water quality in catchment
ecosystems. 1–48. Wiley and Sons, New York.
Wedepohl, K.H. (ed.), 1969–1976. Handbook of geochemistry. Springer, Berlin.
Whitley, K.T., Husar, R.B. and Liu, B.Y.H., 1972. Aerosol size distribution of Los Angeles smog. J. Coll. Interf.
Sci. 39, 177–204.
Wood, W.W., 1999. Use and misuse of the chloride-mass balance method in estimating groundwater recharge.
Ground Water 37, 2–3.
Wood, W.W. and Sanford, W.E., 1995. Chemical and Isotopic methods for quantifying ground-water recharge in
a regional, semiarid environment. Ground Water 33, 458–468.
Wu, L. and McGechan, M.B., 1998. A review of carbon and nitrogen processes in four soil nitrogen dynamics
models. J. Agr. Eng. Res. 69, 279–305.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

3

Flow and Transport

In order to understand the spatial and temporal variability of groundwater quality, it is imperative to
have a perception of the groundwater flowpath and the travel time from the site of infiltration to the
sampling point. Along the flowline, chemicals may become sorbed which retards their transport
compared to the velocity of water. Furthermore, physical processes like diffusion and dispersion
cause mixing and smoothen concentration changes. To achieve a conceptual understanding of the
processes, we will in this chapter focus on hand calculations that provide insight in flowlines and res-
idence times, in retardation, and in diffusion and dispersion.

3.1 FLOW IN THE UNSATURATED ZONE

Water in the unsaturated zone percolates vertically downward along the maximal gradient of the soil
moisture potential when the relief is moderate. The rate of percolation in an unsaturated profile can
be derived from a mass balance, dividing the precipitation surplus by the water filled porosity of the
soil (Figure 3.1; Allison et al., 1994; Phillips, 1994):
H  P / ew (3.1)
2O

where vH2O is the velocity of water (m/yr), P is the precipitation surplus (m/yr) and w the water-filled
porosity (m3/m3). This relationship has been verified in the field by Andersen and Sevel (1974) and
Engesgaard et al. (1996) by analyzing tritium transport in a 20 m thick unsaturated zone in sandy
sediments.
A water velocity that is simply determined by mass balance, means that infiltrating water pushes
the old water ahead, a type of flow adequately termed piston flow. Figure 3.2 shows how the tritium
peak in the rain from 1963 is translated by piston flow in the unsaturated zone of chalk at 3 locations in
England. The velocity of water in the profile measure in 1968 is vH2O  4.4 m / 5 yr  0.9 m/yr, and
similar at the other sites. The tritium peak decreases with time due to radioactive decay (Section 3.3).

H2O
Grains  (ε  εw)

εw

Figure 3.1. A precipitation surplus of 100 mm/yr translates to vH O  500 mm/yr in a soil with w  0.2.
2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

64 Flow and transport

Tritium concentration (TU)

0 200 400 600

1968
5
1970
Depth (m)
10
1977

15

Figure 3.2. Tritium variation in rain and piston flow displacement at three sites in the English chalk, measured
in 3 different years (Foster and Smith-Carrington, 1980).

However, the observed displacement of about 1 m/yr was too low compared with the precipitation sur-
plus and the water filled porosity of the chalk. Part of the infiltration was therefore believed to proceed
along preferential flowpaths through macro-fissures (Foster and Smith-Carrington, 1980). This rapid
bypass flow is also observed in soil moisture in the root zone (Section 2.5.1).

QUESTION:
Estimate the age of water at 3 m depth in Figure 2.22: average water content
g  0.012 g/g, P 
0.1 mm/yr. (Recalculate
g from g/g into water filled porosity w using a bulk density of 1.8 g/cm3).
ANSWER: Average
g  0.012 g/g, or 0.022 g H2O/cm3. Hence vH O  0.1 / 0.022 
2
46 mm/yr.
Age: 3000 mm / 46 mm/yr  65 yr.

3.2 FLOW IN THE SATURATED ZONE

Consider the landfill site shown in cross section in Figure 3.3. Leachate from the waste is percolat-
ing into the aquifer and the question is, how fast the various chemicals travel through the subsoil, and
where and when they pollute drinking water wells further downstream. For an answer, we must
define the flowlines of groundwater in the aquifer and derive the travel time. The detailed calcula-
tion of flow patterns is usually done with numerical models (Zheng and Bennett, 2002; Chiang and
Kinzelbach, 2001). However, a first guess can be readily obtained by hand calculations.

3.2.1 Darcy’s law

The groundwater levels measured in wells (the piezometric level) can be compiled in a map of
equipotentials (isohypses), the lines that connect points with the same groundwater elevation or
potential (Figure 3.4). Water flows from a high to a low potential and the groundwater flow direction
is at right angles to the isohypses if the aquifer is isotropic, which means that the hydraulic proper-
ties are equal in all directions. Hydraulic gradients follow primarily the local relief, but are also
influenced by recharge and discharge rates, and groundwater withdrawal (Freeze and Cherry, 1979;
Domenico and Schwartz, 1997; Fetter, 1994).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Flow in the saturated zone 65

Chloride
Distance from divide (m)
0 100 200 300 400
!1000 mg/L
Landfill
B1 B2 B3 B4 B5 B6 B7 B8 B9 500–1000
38
Distance above sea level (m)
250–500
Water level
100–250
34
50–100
30–50
30
35
Clay/silt
26

22 Clay

Figure 3.3. Cl concentration contours in the aquifer below the Vejen waste site. The small dots on the
vertical borings (B1–B9) represent sample points (Lyngkilde and Christensen, 1992).

Groundwater flow thus depends on the hydraulic gradient, but it depends also on the hydraulic con-
ductivity of the subsoil. These two parameters are combined in Darcy’s law:
vD  k dh / dx (3.2)

where D is the specific discharge or Darcy flux (m/day), k is the hydraulic conductivity (m/day), and
dh / dx is the hydraulic gradient.

South Australia NSW

82
81
83
SA 86
80
78 71
88 77 22
Study 91
76 20
75 Victoria 21
area 90 19
89 74
73
68 72
18 Loxton 16
10

65
Murray

69
64
N 14
62
63 67
61 59 20
Sunset
58 55 57 Country
ver
66
Ri 60 54 44 15
52 70
53 12
51
50 45
11
43 46 48 49 10
Murray 13 42 30
Bridge 47
40 39 38
41 37
Pinnaroo 28
9
27 26
36 25
40 24
35 Big Desert
23
22
8
20 21
34
19
7 33
60 17
6
5

Southern Bordertown
Ocean 80 Nhill
4
14 13 0
Murray Group 10
Little Desert
Renmark Group
120
12
10 9
8 5
2
0 100 km 7 8 2 140
3 1

Figure 3.4. Groundwater potential and groundwater flow direction. The contours represent the hydraulic head
(isohypses), and the arrows indicate the inferred direction of groundwater flow (Dogramaci and Herczeg, 2002).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

66 Flow and transport

The piezometric level of groundwater gives, after corrections for density differences, the potential
(h) for groundwater flow. The discharge is obtained by multiplying the specific discharge with the
surface area perpendicular to the flow:

Q  v D A  kA dh / dx (3.3)

where Q is discharge (m3/day), and A is the total surface area which includes pore space and grain-skele-
ton together.
The hydraulic conductivity of a sediment can be determined in a permeameter and is calculated
with Equation (3.3) from Q, A and h / x. The Darcy flux is defined as specific discharge for a
medium containing both grains and water. However, only the pore space contributes to the water
flow, so that the actual velocity of water through the pores must be larger. Mass balance applies as
for the unsaturated zone, and the velocity is:
"  vD / ew  k / ew dh / dx (3.4)
2O

The water velocity can also be calculated from the discharge and the cross section A w:

H  Q / (A ew ) (3.5)
2O

Some information on groundwater flow velocities can be obtained using hydraulic conductivities
and porosities of sediments presented in Table 3.1. For example, the actual distance traveled by
groundwater in a sandy aquifer with porosity w  0.3, hydraulic conductivity k  50 m/day, and
potential gradient dh / dx  0.001 (1 m per 1 km), amounts to 60 m/yr. The hydraulic conductivity
varies by many orders of magnitude for different sediments, while the porosity only varies from 0.2
for coarse, unsorted sands to 0.65 for clay.
Layers of clay, peat and loam, which are considered impermeable, are termed aquicludes or,
when almost impermeable, aquitards. Such layers can separate more permeable (sandy) aquifers.
If the watertable in the aquifer is higher than the boundary with the aquiclude, the aquifer is con-
fined, otherwise the aquifer is unconfined and possibly unsaturated in the upper part. If the upper-
most aquifer extends right to the earth’s surface it is called a phreatic aquifer.
The groundwater velocity can, in principle, be calculated with Darcy’s law (Equation 3.2), when the
hydraulic gradient and the hydraulic conductivity are known. However, the large variation in hydraulic
conductivities makes the calculated velocity rather uncertain. Often, the water balance based on
Equation (3.5) may provide a better idea of the average, or effective parameters. Effective parameters
may differ from the ones that can be measured directly. The effective porosity comprises the porosity
containing water that participates in flow, in contrast to the total porosity which includes the pores
filled with stagnant water. The effective hydraulic conductivity is similarly obtained from the quotient
of discharge and hydraulic gradient, and so averages small scale variations in hydraulic conductivity.

Table 3.1. Hydraulic conductivity and porosity of

different sediments.

k (m/day)  (fraction)

Gravel 200–2000 0.15–0.25

Sand 10–300 0.20–0.35
Loam 0.01–10 0.30–0.45
Clay 105–1 0.30–0.65
Peat 105–1 0.60–0.90

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Flow in the saturated zone 67

In numerical models, the water balance is, together with the measured groundwater potential, used to
optimize the transmissivity (the product of hydraulic conductivity and aquifer thickness). If only the
quantity of (horizontal) water flow is important, the transmissivity is an adequate parameter. However,
increased pollution has aroused interest for unraveling the actual flowlines in the aquifer. It has proven
difficult to assess flowlines exactly using numerical hydrological models because permeability variations
have a large influence on flowpaths but only small effect on hydraulic heads, and thus do not show up in
the traditional hydrological measurements.

3.2.2 Flowlines in the subsoil

The cross section through a Canadian prairie landscape in Figure 3.5 illustrates how water may flow
along curved flowlines from recharge areas to discharge areas, often near creeks. The vertical scale
is grossly exaggerated and in reality the flowpaths are more horizontal. Nevertheless, the figure
illustrates the important points that the flowpaths are at right angles to the isolines for the hydraulic
potential (dashed lines in Figure 3.5, note that scaling does affect the intersection angle in a graph),
and that the intersection of the isoline with the water table defines its potential in the terms of
hydraulic head. By specifying the shape of the water table with a mathematical formula, it is possible
to solve the potential field in the cross section and to obtain the flowlines (if the flow domain is
homogeneous, otherwise numerical models need to be used) (Tóth, 1962; Freeze and Witherspoon,
1966; Domenico and Schwartz, 1997).
It is easier to use the water balance to visualize how groundwater flows in aquifers (Vogel, 1967;
Ernst, 1973; Gelhar and Wilson, 1974; Dillon, 1989). Let us take a segment of the cross section of
Figure 3.5, from Kneehills Creek in the West to the divide at the topographic high, and redraw it as
the phreatic aquifer shown in Figure 3.6. The sediment is homogeneous so that porosity and
hydraulic conductivity are the same everywhere. In the rectangular aquifer of Figure 3.6 the veloc-
ity is (very nearly) equal at all depths along a vertical line. (Why does an aquifer not show the veloc-
ity profile of a river?). The point along the upper reach where water infiltrates, and depth in the
aquifer are then related proportionally:
x D
 (3.6)
x0 Dd
where D is the thickness of the aquifer. Water infiltrated at a point x0, upstream of x, is at a given
time, found at depth d in the aquifer. Above d flows water that infiltrated between point x0 and x,
while water at greater depths infiltrated further upstream of x0.
ek
re

Equipotential lines
k
ee

tC
Cr

Flow lines
e

en

W E
vid

itt
lls

3400
Di

rm

ll A
hi
Elevation above mean

ee

te

Hi
In
Kn

3200
sea level in feet

3000
2800
2600
2400
0 1 2 3 4 5 6 7 8 9 10
Distance in miles

Figure 3.5. Groundwater potential distribution and flow pattern, and the effect of a highly permeable body,
near 7 miles, on the flow system (Tóth, 1962).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

68 Flow and transport

x
x0

t1
d t2
D
t3
t4

Figure 3.6. A profile through a homogeneous, phreatic aquifer.

The flow velocity in the aquifer can also be calculated. If a precipitation surplus of P m/yr enters the
aquifer, Q  P x m2/yr must be discharged through the aquifer at point x. Using Equation (3.5),
the flow velocity at point x is:

dx Px
v  (3.7)
dt De w

The thickness D has replaced the surface area A, since we consider flow in a profile (for example per
m width). The groundwater velocity increases with distance from the divide, since more precipita-
tion must be discharged through the same layer. The increase of the flow velocity requires that the
hydraulic gradient increases with the square of distance.
Integration of (3.7) gives, when water infiltrates at t  0 at x  x0:

x P
ln  t (3.8)
x0 De w

or the distance x reached in time t:

 Pt 
x  x0 cxp   (3.9)
 De w 

We may also substitute the proportionality relationship (3.6) into (3.8), which yields:

D Pt
ln  (3.10)
Dd De w

or:
  Pt  
d  D 1  exp    (3.11)
 
 D e w 


This is a remarkable result. It demonstrates that in a homogeneous aquifer with uniform infiltration,
water at a specific depth is all of the same age, independent of the location. In other words, the
isochrones are horizontal, water resides in the aquifer in planes of equal age.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Flow in the saturated zone 69

EXAMPLE 3.1. Calculate the time for water to flow from midway in the Vejen waste site to 125 m down-
stream (Figure 3.3), and the thickness of the plume at that point. The waste site is located on the groundwa-
ter divide for the clay/silt layer and is 50 m long, the aquifer is 9 m thick. The precipitation surplus is
0.4 m/yr, the porosity 0.35 (Brun et al., 2002).

ANSWER:
The midpoint of the site is at x0  25 m, x  125 m from the divide. With Equation (3.8), we obtain
t  9  0.35 / 0.4  ln(125 / 25)  12.7 yr. The percolate from 50 m waste is distributed proportionally
over depth, and the thickness of the plume is 50 / 125  9  3.6 m. Note that the actual plume is much more
spread out in the vertical as a result of the hydraulic conductivity variations in the subsoil, the irregular fill-
ing of the site, and probably because density flow occurs (Christensen et al., 2001).

The depth/time relation is valid for an aquifer of infinite length, or in practice, not too close to the
point of discharge of the aquifer (in formula: 0  x  (L  D), where L is the distance from divide
to drain). Ernst (1973) and Gelhar and Wilson (1974) have derived formulas which can be used near
the discharge point, assuming radial upward flow into the spring or seepage zone.
In principle, seepage occurs where the water table intersects the topography and from where a
drainage network develops (Ernst, 1978; De Vries, 1995). The drainage distance can be estimated
from Darcy’s law for the homogeneous aquifer:

Q  P x  kD dh / dx (3.12)

Integration from hm at x  0 at the divide to h at the discharge point at x  L gives:

2kD
L (hm  h) (3.13)
P

Applied to the aquifer in Figure 3.5 between the divide and Kneehills Creek in the West, the
hydraulic conductivity k can be calculated using Equation (3.13). With L  4216 m (2.62 miles),
(hm  h)  62.8 m (206 ft), D  185 m (600 ft) and P  0.03 m/yr, we obtain k  0.06 m/day.
(Tóth, pers. comm., estimated k  0.08 m/day for the area).

EXAMPLE 3.2. Calculate the water level in the Vejen river, 1 km downstream from the Vejen waste site and
estimate the travel time of waste leachate to the river. The hydraulic conductivity of the aquifer is 35 m/day
(Brun et al., 2002), the other data are given in Example 3.1.

ANSWER:
The water level at the confinement of the waste (50 m from the divide) is 38 m above sea level. We integrate
Equation (3.12) from hx50 m  38 m to hx1000 m:

1
⁄2 (10002  502 )  (35  9 / (0.4 / 365)) (hx1000 m  38)

which solves to hx1000 m  36.3 m. The travel time from the border of the waste is 24 yrs.
Actually, the water level in the river is about 32 m. There is a greater hydraulic gradient along the flow-
line towards the river, indicating that the hydraulic conductivity decreases, or that the aquifer thins out, or
that the clay layer disappears allowing water from below to enter the aquifer. Figure 3.3, but also the regional
hydrogeology suggest the importance of the two latter factors, implying that the travel time will be less than
estimated by our calculation.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

70 Flow and transport

QUESTIONS:
Why does an aquifer not show the velocity profile of a river?
ANSWER: The flow profile in water is determined by the resistance at the boundaries: in
the river at the river bed and the air, in the aquifer by the resistance at the pore
boundaries which remain the same over depth.
Estimate the flowtime from midway in the Vejen waste site (x0  25 m) to the river?
ANSWER: less than 29 yr (note that the last 850 m are traveled in the same time as the
first 150 m).
In Example 3.2, consider the effect of the decrease of water filled thickness by the gradient of h with x?
ANSWER: Take D  7.3 m, then hx1000 m  35.9, still higher than observed; the travel time
decreases slightly: 19.4 yrs.

3.2.3 Effects of non-homogeneity

Most aquifers are not homogeneous and large differences in hydraulic conductivity are usual. The
distribution of flowlines can be calculated by proportioning the volume of flow according to the
transmissivity difference. The correction is performed in the vertical plane as shown in Figure 3.7.
For two layers of thickness D1 and D2, porosities 1 and 2, and with hydraulic conductivities k1 and
k2 (m/day), the discharge through the vertical section can be divided in two portions Q1 and Q2. The
ratio of the two is (assuming equal hydraulic gradient in both layers):
Q1 kD
 1 1 (3.14)
Q2 k 2 D2

and the velocities in the layers are:

v1 k e
 1 2 (3.15)
v2 k 2 e1

Various examples of flowpatterns affected by inhomogeneities are illustrated in Figure 3.37 and
Problem 3.7, and can further be found in Freeze and Cherry (1979). Analytical solutions for various
cases may be found in Bruggeman (1999) and more flexible computer programs for calculating and
visualizing flowlines in 2D profiles are TopoDrive&ParticleFlow by Hsieh (2001) which can be down-
loaded from water.usgs.gov/nrp/gwsoftware/tdpf/tdpf.html and Flownet by Van Elburg, Hemker and
Engelen which can be downloaded from www.microfem.nl.

k1, ε1
D1 Q1

k2, ε2

D2 Q2

Figure 3.7. Effect of hydraulic conductivity-changes on flowlines.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Flow in the saturated zone 71

3.2.4 The aquifer as a chemical reactor

Concentrations of dissolved substances were so far considered as depth related entities which can be
measured using depth specific samplers (Chapter 1). On the other hand, a large pumping well with a
long screen yields an average of the concentrations present along the length of the screen. Assume,
for example in Figure 3.6, that the concentration of NO 3 in infiltrating water increases from an ini-
tial ci at time t  0, to c0 at time t  0. When the concentration front has arrived at depth d, the aver-
age concentration is:

c0 d  ci (D  d )
c (3.16)
D

Now, assume that the initial concentration in the aquifer ci  0, and use (3.11) to obtain:

d
c  c0  c0 (1  ePt /Dew ) (3.17)
D

This shows that the concentration in a fully penetrating well (screen length equals D) increases with
time to the final value by an exponential function. This function is similar to the one used for an
“ideal mixed reactor” in chemical engineering (Levenspiel, 1972). For such a reactor the residence
time is defined as:

Total mass M tot

t  (3.18)
Mass exchange dM in / dt

where  is the residence time (s). dMin / dt can be replaced with dMout / dt, since the Mtot does not
change. Replacing mass with volume of water, the residence time becomes:

t V /Q (2.3)

which we used in Chapter 2 for calculating the residence time of water.

When a step input c0 is introduced at time t  0 in the input stream, the concentration in the exit
stream of the reactor increases as:

c  c0 (1  et /t) (3.19)

Equation (3.17) is identical with (3.19) when   Dw / P, and this is correct since in Equation (2.3)
V  Dw x, and Q  Px. The comparison is logical since a fully penetrating well samples a mixture
of waters which infiltrated from time zero in the uppermost part to successively earlier times deeper
down. With uniform flow (no velocity variations with depth) the averaging in the groundwater well
is similar to the mixing in the reactor.
A difference between a stirred chemical reactor and the phreatic aquifer is that the reactor is well
mixed to let all the reactions elapse in the same way everywhere. In the phreatic aquifer, mixing only
takes place in the well, while the chemical reactions among water and solids are variable in time and
distance along the flowlines through the aquifer. Nevertheless, the chemical reactor equations are
useful when reactions between water and sediment are unimportant. Equation (3.19) has been used
to derive characteristics of the groundwater reservoir from tritium measurements in spring water
(Yurtsever, 1983; Maloszewski and Zuber, 1985). When the residence time is known, either from
tracer measurements or from hydrogeological conditions, a first estimate of the change in concen-
tration over time is easily obtained with Equation (3.19), for example of the NO 3 concentration in
production wells after the application of excess fertiliser on the surface has been banned.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

72 Flow and transport

EXAMPLE 3.3. Flushing of NO 3 from an aquifer

Estimate the time required to reduce the NO 3 concentration from 150 mg/L to 25 mg/L (the drinking water
limit) in an aquifer after fertilizer input has been stopped. D  10 m, w  0.3, P  0.6 m/yr.

ANSWER:
The residence time in the aquifer is   D  / P  5 yr. From (3.16) and (3.17) we obtain:

c  c 0 (1  et/t )  c i et/t

Here c0  0 mg/L, ci  150 mg/L. From c  ci et/ find 25 / 150  et/, or t   ln(25 / 150)  9 yrs.

3.3 DATING OF GROUNDWATER

Calculated groundwater flow patterns and residence times can be verified by groundwater dating.
There is a range of dating methods available based on radioactive decay (Cook and Herczeg, 2000).
These comprise tritium (3H), 3H / 3He and 85Kr for young groundwaters of up to about 50 years,
radiocarbon (14C) for older groundwaters up to 30,000 years and finally isotopes like cosmogenic
81
Kr and 36Cl for even older groundwaters. Using the latter method, groundwater with an age of
400,000 years has been dated (Collon et al., 2000).
For all radioactive decay processes, the rate equals a decay constant, , times the concentration
of the reactant. For example, tritium (3H) disintegrates to 3He and  at a rate:

d 3H
  3 H (3.20)
dt
The tritium concentration is expressed in tritium units (TU). One TU equals 1018 mol 3H/mol H,
and is equivalent to 3.2 pCi/L from the emitted  particles. The decay constant   0.0558/yr. The
equation can be integrated from the initial value 3H0 at time t  0 and gives:

ln(3H / 3H0 )   t (3.20a)

The time to let tritium decay to half of its initial concentration is called the half-life,
t ⁄  ln(0.5) /   12.43 years.
1
2

Figure 3.8 shows the distribution of 3H and its daughter 3He in an aquifer. The peak in 3H and 3He
at 10 m depth corresponds to 1963, the year of maximal hydrogen bomb testing in the atmosphere
(cf. Question). Tritium/Helium offers an absolute dating method because the initial 3H0 is given by
the sum of 3H and 3He in the sample. However, the resulting age is affected by dispersion (Schlosser
et al., 1989) and by different diffusion rates of 3He and 3H (Chapter 11). Moreover, radiogenic and
atmospheric 3He contributions must be subtracted, and some tritiogenic 3He may escape into the
unsaturated zone, which can give substantial corrections in young or old (infiltrated before 1950)
groundwater (Schlosser et al., 1989; Cook and Solomon, 1997).
Krypton 85 (85Kr) is a radioactive, and chemically inert, noble gas that decays to stable 85Rb with a
half-life of 10.7 years. Almost all 85Kr in the environment is of anthropogenic origin. It is a fission prod-
uct of uranium and plutonium and released by nuclear weapon testing, nuclear reactors and nuclear
reprocessing plants. The activity of 85Kr in the atmosphere has steadily increased except for a few years
around 1970 (Figure 3.9) and may be used for groundwater dating. 85Kr is measured as specific activity,
i.e. as the ratio of 85Kr to stable Kr (in disintegrations per minute of 85Kr per volume Kr, dpm/cm3).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dating of groundwater 73

0
3
H Hydraulic age
3
3
He H / 3He age
5
Depth (m)

10

15

20

0 200 400 0 20 40 60
TU Years

Figure 3.8. Groundwater 3H and 3He concentrations and 3H / 3He age profiles at Sturgeon Falls, Ontario
(Solomon et al., 1993). The 3He concentration is expressed in TU (1018 mol/mol H).

Because the ratio of 85Kr over stable Kr is measured, the Kr concentration in the recharge water, which
depends on krypton solubility and temperature, does not have to be known (Ekwurzel et al., 1994).
Groundwater dating proceeds by fitting downward decreases in groundwater 85Kr to the historical
atmospheric record.
Carbon 14 is a natural radioactive compound that forms in the atmosphere. The decay constant
for 14C is   1.21104/yr, and the half-life is 5730 yr. It is therefore suitable for older ground-
waters. The 14C activity of a sample is expressed as percent modern carbon (pmc or 14A) given by
the ratio:

(14C / gC) sample (13C / 12C)2standard

14 A   100 (3.21)
(14C / gC) standard (13C / 12C)2sample

where the 14C / gramC in the standard equals the ratio in equilibrium with a “modern”, pre-nuclear
bomb testing atmosphere. The squared ratio (13C / 12C)2 accounts for fractionation of 14C. The radio-
carbon age of groundwater from simple radioactive decay is t  ln(14A / 14A0) / 1.21104 yr (the
“conventional” age). However, the measured 14C concentration may have been diluted by ‘dead’ car-
bon derived from carbonate dissolution or degradation of organic carbon. In that case a number of
corrections must be applied before 14C can be used for dating (Chapter 5).
Chlorofluorocarbons (CFCs) and sulfurhexafluoride (SF6) have been introduced in the atmos-
phere and hydrosphere during the last couple of decades. CFCs, also called freons, are stable syn-
thetic organic compounds produced for use as refrigerants, aerosol propellants, solvents etc.
(Busenberg and Plummer, 1992). Gaseous SF6 is a product of industrial processes (Bauer et al.,
2001). CFCs have received much attention because of the detrimental effects on the atmospheric
ozone layer. The increase in the atmospheric concentration of various CFCs is also shown in
Figure 3.9.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

74 Flow and transport

600 150 2000

CFC-12
500
1500
400 100

dpm/cm3
pptv

TU
300 CFC-11 1000
85
200 Kr 50
500
100 CFC-113

0 0 0
1940 1950 1960 1970 1980 1990 2000 1950 1960 1970 1980 1990
Year Year

Figure 3.9. Left: Atmospheric concentrations of CFC-11, CFC-12 and CFC-113 (part per trillion by volume
(pptv)) and 85Kr (disintegration per minute per cubic centimeter) (Cook and Solomon, 1997); Right: Tritium
concentrations in rain in the Netherlands.

The steady increases in both CFCs and 85Kr concentration have been used for dating of young
groundwaters. While for the solutes 36Cl and 3H the clock is set at the time of infiltration, for the
atmospheric gases like CFCs the relation is more complex. For gases, the temperature-dependent
solubility and transport through the unsaturated zone plays a role (Cook and Solomon, 1997).
Dating of young groundwater using CFCs proceeds by fitting downward increases in measured
CFC concentration to the historical atmospheric record and thereby interpreting the age of the
groundwater. Such an age interpretation over depth is for three different CFCs given in Figure 3.10.
For comparison the position of the 1963 tritium peak in 1991 is also shown, presenting in general a
good agreement with the CFC ages.
Comparison (Figure 3.10) of CFC-11 with CFC-12 and CFC-113, suggests consistently
higher ages derived from CFC-11. Probably this is an artefact due to degradation of CFC-11.

Apparent age (yr)

0 10 20 30 40 50 60
0
Hydraulic age
CFC-11
CFC-12
CFC-113
5
Depth (m)

10

3H

15

20

Figure 3.10. Apparent groundwater age profiles estimated from CFC’s and 3H. The error bar depicts the posi-
tion of the 3H peak in 1991. The shaded region depicts the hydraulic age profile calculated with Equation (3.11)
and a recharge rate derived from the position of the tritium peak (Cook et al., 1995).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Retardation 75

It appears that CFC-11 is readily degraded under anoxic conditions, while CFC-12 degrades at a ten
times slower rate (Oster et al., 1996). CFC-113 on the other hand has been reported to be slightly
adsorbed to sediment particles (Bauer et al., 2001). Thus, care should be taken in the interpretation
of CFC concentrations for groundwater dating and SF6 may in some cases be a better option (Bauer
et al., 2001).

QUESTIONS:
Calculate the maximal tritium concentration in rain for Figure 3.8?
ANSWER: The maximum is 306 (3He)  165 (3H)  471 TU.
Estimate the decay constant for 3H for this maximum?
ANSWER: est  0.055/yr (for the plotted age of 19 yr).
In Figure 3.8, find the depth where 3H and 3He have equal concentration, D  35 m, w  0.35,
P  0.16 m/yr?
ANSWER: t  12.43 yr, d  5.24 m
Two groundwater samples have 14A of 50 and 25%, respectively. Calculate the uncorrected
(conventional) ages?
ANSWER: 5730 and 11460 yrs.

3.4 RETARDATION
Again, the aquifer downstream from the Vejen waste site is depicted, showing the Na concen-
trations in Figure 3.11. Probably, most of the Na is derived from dissolution of salt that is
present in the waste, which implies that Na and Cl are released in equal (molar) quantities to
the percolate (but the rain may have the ratio of Na/Cl from seawater). When the distribu-
tion of the two ions is compared, it is evident that the Na plume has traveled less far than Cl
(Figure 3.3). Part of Na is sorbed to the sediment, while sorption of Cl is negligible in most
situations.

Sodium
Distance from divide (m)
0 100 200 300 400

Land fill
!500 mg/L
B1 B2 B3 B4 B5 B6 B7 B8 B9 250–500
38
Distance above sea level (m)

Water level 100–250

50–100
34
20–50
20
30
Clay/silt

26

22 Clay

Figure 3.11. Profile through the Vejen aquifer showing Na concentrations (Christensen et al., 2001).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

76 Flow and transport

Pollutant X enters an aquifer

Case A: no adsorption Case B: sorbed is 1 out of 2

Soil surface

H2O

adsorbed 0 adsorbed 1
Ratio   0  Kd   1  Kd
solute 1 solute 1
νX  νH O 1
2
νX  νH2O
(1  1)

Figure 3.12. Transport of chemicals is retarded by sorption. Case A: no sorption; retardation is 1  0  1,

relative to water flow. Case B: sorption of half of total mass, and the chemical migrates at half the velocity
of water; retardation is 1  1 / 1  2.

The effect of sorption on the transport velocity of solutes is illustrated in Figure 3.12. In case (A) the
solute does not sorb to the sediment grains. Consequently this solute is transported with the velocity
of water. In case (B), one out of two solute molecules is sorbed. Thus, half the mass of chemical is lost
from solution and therefore it will travel with half the velocity of water. The solute is retarded com-
pared to water migration.

3.4.1 The retardation equation

For quantifying retardation, we must consider the effects of sorption processes on transport.
First, examine the left part of Figure 3.13 which shows the concentration profile of a solute,
sampled at time t1 along a flowline. If we come back at a later time t2, the solute has migrated
downstream and the concentration profile has shifted along x, but also the shape has changed.
The other way to observe the migration of the solute is to sit down at a single well, for example
located at x1 and to measure the concentration changes over time. The corresponding result is
shown in Figure 3.13B. The slopes of the lines in Figure 3.13A, or the gradients dc / dx, are
denoted by partial signs ∂, which indicates that c not only depends on x, but also on t. The time
is fixed in (∂c / ∂x)t, and likewise, x is fixed in Figure 3.13B where the gradient (∂c / ∂t)x is given
by the slope of the c / t plot. Often, the subscripts are omitted when the fixed dimension is
obvious.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Retardation 77

A: (c/x)t B: (c/ t)x1

c2 c2

t1 t2

∂c
c1 c1
∂t

x1 x2 t1 t2

Figure 3.13. A: A concentration profile over distance, profiles are given for times t1 and t2, B: The change in
concentration over time at point x1. Dotted line in B shows the gradient (∂c / ∂t)x.

QUESTION:
Can you indicate the slope (∂c2 / ∂x)t in Figure 3.13?
1
Draw the c / t profile at x2.
ANSWER: Note that the (c / t)x profile steepens at x2 for c2 because (∂c / ∂x) is
2
steeper at t2.

Next, consider a small volume at position x1 with dimensions dx dy dz as in Figure 3.14, and cal-
culate the mass balance for solute c when water passes the volume with a fixed velocity vH2O in the
x direction.

#c, #q

Mass in: vH2O c dy dz (mol/s)

  Mass out: vH2O(c  #c / #x dx) dy dz (mol/s)
#c
vc v c+ dx Change in sorbed mass: #q / #t dx dy dz (mol/s)
 #x  The sorbed concentration q is expressed in
dz mol/L pore water.
dy
dx

Figure 3.14. Mass balance of solute transport through a cube.

The mass balance for the solute in the volume is:

∂M ∂c ∂q
 In  Out  Sorbed  H O dx dy dz  dx dy dz (3.22)
∂t 2 ∂x ∂t
Dividing the mass M by the volume dx dy dz gives the concentration c, and thus:
∂c ∂c ∂q
 H O  (3.23)
∂t 2 ∂x ∂t

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

78 Flow and transport

dq
5 dc2

Sorbed q (µmol/L)
C
3 dq
q  Kd c,  Kd
dc

2 dq
dc1
B

1
dq
q  0, 0
A dc
0
0 0.5 1 1.5 2
c1 Solute c (µmol/L) c2

Figure 3.15. Examples of sorption isotherms: A: no sorption, B: linear sorption, C: non-linear sorption.

The third step is to combine the time gradients of q and c. We expect that, when the concentration of
solute c increases, also the concentration of the sorbed species q will increase, but the exact rela-
tionship will depend on the chemistry of the system. Figure 3.15 shows three examples of sorption
isotherms which couple q and c.
In case A (Figure 3.15), there is no sorption (q  0) and dq / dc  0 for all c (the total derivative
is used here because dq / dc is independent of x and t). This case applies to Cl which conserves all
the mass in solution (and is hence called a conservative substance). In case B, q increases linearly
with c, with q  2c and slope dq / dc  2. When the slope is identical for all c, the distribution
coefficient, Kd  dq / dc  q / c is constant. Usually, sorption diminishes as solute concentration
increases, and the slope dq / dc decreases as c increases (the slope is smaller for c2 ! c1, case C in
Figure 3.15). The isotherm of case C is defined by q  3.5c0.5 and dq / dc  1.75c0.5. The isotherm
is conform with the general equation q  KF cn, also called a Freundlich isotherm.
The slope of the sorption isotherm can be introduced into Equation (3.23) by replacing ∂q / ∂t with
(dq / dc) (∂c / ∂t) which gives:

 dq   ∂c  ∂c
1      H2O (3.24)
 dc   ∂t  ∂x

Multiplying both sides with ∂x / ∂c yields:

 dq   ∂c   ∂x   ∂c   ∂x 
1        H2O     (3.25)
 dc   ∂t   ∂c   ∂x   ∂c 

The product (∂c / ∂x) (∂x / ∂c)  1 (the value of a derivative multiplied by its inverse is 1). But the
product (∂c / ∂t) (∂x / ∂c) is not simply (∂x / ∂t)! As you can see in Figure 3.13, a negative slope
(∂c / ∂x) translates into a positive slope (∂c / ∂t), and the product of the two terms must be negative.
More formally it is derived by implicit differentiation of a constant concentration:

∂c ∂c
dc  0  dt  dx (3.26)
∂t ∂x

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Retardation 79

which gives:

∂c ∂x  dx   ∂x 
      (3.27)
∂t ∂c  dt constant concentration  ∂t c

We have implied that the concentration is constant, and therefore (dx / dt) is a partial differential for a
given, constant concentration. The partial differential (∂x / ∂t)c (m/s) is equal to the velocity vc of that
specific concentration. If we combine Equations (3.27) and (3.25):
vH vH
2O 2O
vc   (3.28)
dq Rc
1
dc
Here, the retardation is defined as:
dq
Rc  1  (3.29)
dc
which varies according to isotherm slope. Equation (3.28) is called the retardation equation. The
retardation equation shows that the transport velocity is different for different concentrations when
the slope of the sorption isotherm is variable. It was first derived by DeVault (1943) and presented
by Sillén (1951) in a particularly insightful manner.

QUESTIONS:
In Figure 3.13A,
Compare the distances x traveled by c1 and c2 in time (t2  t1).
ANSWER: x is larger for c2 than for c1.
Compare the flow velocities v  x / (t2  t1) for c1 and c2.
ANSWER: v is greater for c2 than for c1.
Compare the retardation for c1 and c2.
ANSWER: the retardation is smaller for c2 than for c1.
Sketch the isotherm for the chemical transported in Figure 3.13.
ANSWER: the isotherm is convex, i.e. everywhere the slope lies above the line, and the
slope is smaller for c2 than for c1 (case c in Figure 3.15).

3.4.2 Indifferent and broadening fronts

The retardation equation yields vc for a specific concentration c and we can evaluate how retarded
solutes move compared to water. We can make time-distance graphs for a concentration using
vH
2O
xc  vc t  t (3.30)
Rc

We can also plot concentrations as a function of distance for a fixed time:

xH
2O
xc  (3.31)
Rc

or we can chart them as a function of time for a fixed position:

t c  tH Rc (3.32)
2O

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

80 Flow and transport

Let us explore the effect of the isotherm slope dq / dc on the retardation of different concentrations
in Example 3.4.

EXAMPLE 3.4. Retardation and isotherm slope

The first chemical to be considered shows linear adsorption with q  c and therefore dq / dc  1. The ini-
tial concentration along the flowline c is 0.8 mol/L which is flushed by a 0.1 mol/L solution. The sorbed con-
centration q is expressed in mol/L pore water. Calculate xc and tc for xH2O  20 m and tH2O  1 yr. Plot the
concentrations in c / x and c / t graphs.

Table 3.4.1. Linear retardation, linear isotherm q  c dq / dc  1.

c q dq / dc R xH2O xc tH2O tc

0.8 0.8 1 2 20 10 1 2
0.45 0.45 1 2 20 10 1 2
0.1 0.1 1 2 20 10 1 2

The second chemical shows a convex adsorption isotherm with q  c0.5 (a Freundlich type isotherm).
Do the same calculation as above and add the calculated curves to the previous plots.

Table 3.4.2. Broadening front, convex isotherm, q  c0.5. dq / dc  0.5c0.5.

c q dq / dc R xH2O xc tH2O tc

0.8 0.894 0.559 1.559 20 12.829 1 1.559

0.45 0.671 0.745 1.745 20 11.459 1 1.745
0.1 0.316 1.581 2.581 20 7.749 1 2.581

0.8
Sorbed q (mol/L)

0.6 q  c 0.5

0.4 qc

0.2

0
0 0.2 0.4 0.6 0.8 1
Solute c (mol/L)

1 1
Solute c (mol/L)

0.8 0.8
Solute c (mol/L)

qc qc
0.6 q  c 0.5 0.6 q  c 0.5
q0 q0
0.4 0.4

0.2 0.2

0 0
0 10 20 0 1 2
Distance (m) Time (year)

Figure 3.16. Isotherms and concentration/distance and concentration/time plots for Example 3.4.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Retardation 81

For linear adsorption (q  c in Figure 3.16) the sorbing chemical is delayed compared to a non sorb-
ing chemical (q  0) while the shape of the concentration front remains unchanged (indifferent
front). When the adsorption isotherm is convex, a high concentration gives a relatively low dq / dc,
thereby a low retardation and consequently, the concentration moves fast. When the concentration is
low, the slope dq / dc is relatively high, retardation is high, and the concentration moves slower. This
results in a broadening front when the chemical is eluted.

EXAMPLE 3.5. Analytical modeling of column elution

A 4 cm long sediment column has been spiked with 37 g -HCH/L (-hexachlorocyclohexane). The sorption
isotherm is q  7c0.7 g/L. Estimate the elution of -HCH with respect to conservative breakthrough.

ANSWER:
We relate the breakthrough time of -HCH to Cl by:

tHCH  dq 
 Rc  1    1  4.9c0.3
tCl  dc 

The breakthrough of Cl shows dispersion (see Section 3.6) and t-HCH is calculated relative to the con-
centrations of the dispersed Cl curve. The breakthrough time is expressed in ‘pore volumes’ injected in the
column: after 1 pore volume the resident water (with Cl) is displaced and Cl drops to zero (when disper-
sion is absent). Two pore volumes need twice more time than one, etc. Figure 3.17 compares measured -
HCH concentrations in a column experiment with the calculated breakthrough curve (Rc  (V *  Vc) / Vc,
where V *  dq / dc indicates the variable retardation) and also shows results from a numerical model which
accounts for dispersion (see Chapter 11) (Appelo et al., 1991). Clearly, the simple formula provides very
good first estimates, although numerical modeling is essential for fine tuning.

40

35
Analytical
30
q  7c 0.7
25
HCH (µg/L)

Numerical model

20

15

10 Vc V*

5 Cons. Measured HCH

0
0 2 4 6 8 10 12 14 16
Pore volumes

Figure 3.17. Elution of a sediment column spiked with 37 g -HCH/L. Thin lines are results from
numerical model, dotted line is calculated with the retardation equation. The Cl breakthrough curve is
scaled to the maximal concentration of -HCH.

QUESTION:
For one year, a sorbing chemical is spilled in soil at a concentration of 1 mM. The isotherm is linear,
dq / dc  4; vH2O  10 m/yr.
1) After 10 yrs, where is the pollutant forefront?
ANSWER: at 100 / 5  20 m

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

82 Flow and transport

2) After 10 yrs, where is the pollutant hindfront?

ANSWER: at 90 / 5  18 m
3) What is the solute concentration in the soil water?
ANSWER: 1 mM
4) What is the sorbed concentration in the soil (expressed per L pore water)?
ANSWER: 4 mM (Note the mass balance 1 mM  10 m  (1  4) mM  2 m)
5) When arrives the chemical at 30 m?
ANSWER: after 15 years
6) When is the chemical flushed at 30 m?
ANSWER: after 16 years
Estimate the pore volumes needed to lower the concentration of -HCH to below 1 g/L in Example 3.5.
ANSWER: 10.8 in case dispersion is absent.

3.4.3 Sharpening fronts

In Example 3.4, the solution with a low concentration flushed the solution with a high concentration.
Now let us look at the reverse situation where a high concentration (0.8 mol/L) displaces a low
concentration (0.1 mol/L) of the chemical with the convex isotherm. The slope dq / dc at 0.1 mol/L
is steeper than at 0.8 mol/L and the retardation for 0.1 mol/L is larger than for 0.8 mol/L. Thus, the
concentration 0.8 mol/L travels faster than 0.1 mol/L. We may draw a c / x diagram in Figure 3.18,
which shows that the high concentration will tend to overtake the low concentration.

0.8
Solute c (mol/L)

0.6
q  c0.5 q0
0.4

0.2

0
0 10 20
Distance (m)

Figure 3.18. A chemical with convex isotherm q  c0.5 enters the flowline.

This situation is, of course, impossible: at 10 m a choice of 3 concentrations can be made of, 0.8, or 0.1
or about 0.2 mol/L. What happens is, that any contact of c  0.8 mol/L with soil that contains sorbed
chemical in equilibrium with a lower solute concentration, will result in sorption and thus lower the
solute concentration. Accordingly, the higher migration rate of the high concentration is arrested to the
lower velocity of the smaller concentration. The result is a sharp front where the solute concentration
jumps directly from 0.1 to 0.8 mol/L, accompanied by a corresponding jump in sorbed concentrations.
Over the front, mass balance applies in which the change q of sorbed mass is connected with
the change c in solution. Obviously, retardation is stronger when the ratio q / c is greater.
In addition to the retardation equation that is based on the slope of the isotherm (Equation 3.29), we
can formulate the retardation equation for a sharp front:
q
R jump  1  (3.33)
c

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Retardation 83

1 1
c

Sorbed q (mol/L)

Solute c (mol/L)
0.8 0.8

0.6 q 0.6 q  c 0.5 q0

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 10 20
Solute c (mol/L) Distance (m)

Figure 3.19. Convex isotherm gives a sharp front when concentrations increase, here from 0.1 to 0.8 mol/L.

The sharp front equation is illustrated in Figure 3.19. The differences between the initial and the
final sorbed and solute concentrations are sufficient for calculating the retardation of the front.
To discern whether a front is sharpening, we look at the sorption isotherm. A slope dq / dc that is
smaller for the final concentration than for the initial solution, will give a jump-like concentration
change, a sharpening front, or a shock. On the other hand, a slope that is greater for the the final con-
centration will yield a broadening front, or a wave. When the slope is constant (linear isotherm) the
front is not affected by concentration dependent retardation and we have an indifferent front.
The retardation of an indifferent front is also calculated with Equation (3.33). This reasoning can be
applied piecewise along isotherms with variable shape and any slope dq / dc (Problem 3.11).

EXAMPLE 3.6. Retardation of a sharp front

In continuation of Example 3.4 we calculate the fronts for the chemical with q  c0.5 when the solute con-
centration increases from 0.1 to 0.8 mol/L. Moving along the sorption isotherm from 0.1 to 0.8 mol/L, the
slope dq / dc decreases and we are dealing with a sharp front. The retardation is then calculated from
R  1  (qi  q0) / (ci  c0), using the initial and the final concentration, and is identical for all concentra-
tions in between. The sorbed concentration q is expressed in mol/L pore water.

Table 3.6.1. Sharpening front, convex isotherm q  c0.5. dq / dc  0.5c0.5.

c q dq / dc R xH2O xc tH2O tc

0.1 0.32 1.58 1.82 20 10.95 1 1.82

0.45 0.67 0.75 1.82 20 10.95 1 1.82
0.8 0.89 0.56 1.82 20 10.95 1 1.82

0.8
Solute c (mol/L)

0.6 q0 q  c 0.5

0.4

0.2

0
0 0.5 1 1.5 2 2.5
Time (year)

Figure 3.20. A c / t plot for a sharp front, Example 3.6 (the c / x plot is shown in Figure 3.19).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

84 Flow and transport

Sharp fronts are found when chemicals with convex isotherms enter an aquifer, or when several ions
or species compete for the same sorption site, and the displacing species has a stronger affinity for
the site than the displaced one. Sharpening fronts counteract the tendency of concentration changes
to become more diffuse with time or distance. Retardation is a matter of mass transfer between the
solid and the solution only and is therefore not confined to sorption. Also mineral dissolution and
precipitation yield retardation and when the dissolution/precipitation reactions are fast enough to
reach equilibrium, the associated fronts will be sharp as well.

QUESTION:
Calculate the retardation when 36 g -HCH/L enters a pristine aquifer (cf. Example 3.5). First
consider, is the front sharp?
ANSWER: The front is sharp. R  1  (7  360.7  0) / (36  0)  3.4.

3.4.4 Solid and solute concentrations

Chemical analysts usually express the concentration of substances in solids as concentration per
mass of solid, mol/kg, g/kg, ppm by weight, etc. These numbers can be converted to equivalent solute
concentrations, by multiplying with the bulk density b and dividing by the water filled porosity w,
cf. Example 3.7. The factor b / w
6 kg/L for sandy aquifers.

EXAMPLE 3.7. Recalculate 10 ppm As in (unsaturated) quartz sand to an equivalent solute concentra-
tion in mg/L pore water. Total porosity   0.3, water filled porosity w  0.2. Density of quartz is
qu  2.65 g/cm3.

ANSWER:
The bulk density of the sediment is the density with air filled pores:

b  qu (1  e)  0(e)  2.65  0.7  1.855 g/cm 3

Take 1 kg aquifer, the volume is 1 / b  0.54 L. Hereof, w  0.2 is filled with water, or 0.108 L. The 1 kg
aquifer contains 10 mg As, which is 10 / 0.108  93 mg As/L pore water.

The concentrations expressed per liter pore water always appear surprisingly high compared
to those expressed relative to the solid phase. This is a simple consequence of the fact that most
of the mass and volume in the aquifer is located in the solid phase. The dimensionless ratio of the
concentrations, or the slope of the isotherm when the ratio varies, allows for a readily estimate of
the mobility of chemicals. In Example 3.8 a comparison is given of the calculation of the her-
bicide diuron’s travel time in an aquifer using both the dimensional and the non-dimensional
approach.

EXAMPLE 3.8. Travel time of diuron in a soil

In a specific soil, diuron (a herbicide) was found to be sorbed 6 g/g soil at a solute concentration of
1 g/mL (Nkedi-Kizza et al., 1983). Estimate the arrival time of 1 mg diuron/L at 1 m depth when irrigation
water percolates with 0.3 m/yr, b  1.8 g/cm3, w  0.1.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Retardation 85

ANSWER:
By full mass balance: We take a soil column 1 m2  1 m depth. When diuron has arrived at 1 m depth, the
column contains 0.1 m3 water with 1 g diuron/m3 and 1800 kg soil with 6 g diuron/kg. Total is
0.1  1  1800  6  10.9 g. Irrigation water is applied at 0.3 m3/yr and contains 1 g/m3. Thus, it takes
10.9 / 0.3  36.3 yr.
By pore volume and retardation: Express sorbed diuron in g/mL pore water, q  6 g/g  b / w 
108 g/mL. The retardation is R  1  q / c  1  108 / 1  109. It takes water 0.1 / 0.3  0.33 yr to
arrive at 1 m depth and to flush one pore volume. Hence, the diuron arrives after 0.33  109  36.3 yr.

The effect of the curved sorption isotherm on transport time becomes clear when transport is
calculated for a concentration pulse (Scheidegger et al., 1994). For diuron the following sorption
isotherm was determined: s (g/g soil)  6c0.67 (Nkedi-Kizza et al., 1983). If in Example 3.8, diuron
was applied for only one year, while everything else remains the same, then the front of 1 mg diuron/L
still has a retardation of 109 because it is sharp. The zone with 1 mg diuron/L compacts to the distance
traveled by water in one year, divided by the retardation (Equation 3.31) or 3 m / (Rjump  109) 
0.0275 m and arrives at 1 m depth after 0.33  109  36.3 yr. However, a concentration of 0.1 g/L
has a retardation of 2257, and it requires 0.33  2257  752 yr before the concentration at 1 m
depth is reduced to 0.1 g/L (the drinking water limit). We have neglected dispersion and decay of
the chemical here, but the essential feature still emerges that transport in the soil and subsoil is slow
indeed, and is slowed down even further by sorption and chemical reactions.
Let us return once more to the Vejen aquifer, Figures 3.3 and 3.11. Can we estimate now how
transport of Cl and Na are interrelated? Figure 3.21 shows averaged concentrations of Cl and
Na in the plume downstream of the waste site and a model line for Na that is calculated from the
trendline for Cl as follows. The background concentrations for Na and Cl are 11 and 20 mg/L
respectively. (The molar Na/Cl ratio in background water is 0.85, closely equal to the ratio in sea-
water as expected). From the waste site, NaCl is leached with a molar ratio of 1. Thus, the concen-
tration of Na can be calculated from the concentration of Cl:

23 g Na/mol
cNa  11  (c  20)
35.45 g Cl/mol Cl

500

400

300 Cl
mg/L

Na
200

100

0
0 100 200 300 400 500
Distance (m)

Figure 3.21. Concentrations of Na (crosses) and Cl (squares) in the plume of the Vejen waste site, averaged
over depth from the multilevel samplers. Dotted line for Na shows estimated concentrations based on the Cl
trendline with a retardation of 1.5.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

86 Flow and transport

However, Na is retarded by ion exchange. The retardation factor is almost constant 1.5 for the
concentrations in the aquifer (ion exchange is discussed in Chapter 6, cf. Problem 6.17). Thus,
xNa  xCl / 1.5 was used to obtain the travel distance of Na, shown in Figure 3.21, dotted line. We
observe a concentration pattern that is irregular, even when averaged over depth. This will be related
to irregularities in the filling of the site, the subsequent leaching of the waste and, most importantly,
be influenced by the hydraulic conductivity variations in the aquifer as was noted before. Finally, the
low concentrations are running ahead of the main front, and therefore we must also discuss front
spreading and dispersion.

3.5 DIFFUSION
A concentration difference between two points in a stagnant solution will be leveled out in time by
the random Brownian movement of molecules. The process is called molecular diffusion and the
movement of molecules by diffusion is described by Fick’s laws. Fick’s first law relates the flux of a
chemical to the concentration gradient:

∂c
F  D (3.34)
∂x

where F is the flux (mol/s/m2), D the diffusion coefficient (m2/s), and c the concentration (mol/m3).

∂c / ∂x

C F
∆F
0
∆x

0
x x

Figure 3.22. Diffusion develops as a result of a concentration gradient.

Figure 3.22 illustrates how the flux depends on the concentration gradient. Diffusion changes the
concentration at a given x and the flux is therefore time dependent.The concentration change can be
found by constructing a mass balance for a small volume, as shown in Figure 3.23, similar to Figure
3.14 and Equation (3.22). The mass that enters through the left side is:

∂c
F1  dzdy  D d zd y (3.35)
∂x

and the mass leaving through the right side is:

 ∂F   ∂c ∂ ∂c 
F2  dzdy   F1  dx  dzdy  D   d x  d zd y (3.36)
 ∂x   ∂x ∂x ∂x 

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Diffusion 87

F1 F2
dz

dy
dx

Figure 3.23. Concentration changes in a cube as result of diffusion.

The change in number of moles M in the cube is the difference:

∂M  ∂c ∂c ∂2 c 
 (F1  F2 )  dzdy  D D D d x  d zd y
∂t  ∂x ∂x ∂x 2  (3.37)
∂2 c
D dxdydz
∂x 2
Dividing by the volume gives the change in concentration:
c 
M / (dx dy dz). We therefore
obtain:
∂c ∂2 c
D (3.38)
∂t ∂x 2
Equation (3.38) is known as Fick’s second law of diffusion.

3.5.1 Diffusion coefficients

For solute ions the diffusion coefficient, D, will be related to their mobility, u, as derived from the
molar electrical conductivity in m/s per V/m (m2/s/V) given in Table 1.7 (Atkins and de Paula, 2002;
Lyman et al., 1990). Differences in mobility are a function of the friction ions experience when trav-
eling through water, and are related to the ion-size and the viscosity:

u  |z|qe / (6$ a) (3.39)

where |z| is the absolute charge of the ion, qe is the charge of the electron (1.6021019 C), the vis-
cosity of water (0.891103 kg/m/s at 25°C), and a the radius of the ion. The radius of the solvated
ion includes water of hydration and is found by combining the molar conductivity (Equation 1.4)
with Equation (3.39). For example, the mobility of Na is 5.19108 m2 s/V (Table 1.7), which
gives a  0.164 nm while the radius of unhydrated Na is 0.10 nm.
Similarly, the diffusion coefficient is given by the Stokes-Einstein equation:
D  kT / (6$ a) (3.40)

where k is the Boltzmann constant (1.38051023 J/K) and T the absolute temperature (°C  273.15).
The combination of Equations (3.39) and (3.40) relates the diffusion coefficient to the ionic mobility:
D  u kT / (|z|qe ) (3.41)
Diffusion coefficients, calculated using (3.41), are listed in Table 3.2 and vary by a factor of 3. There
will be some interaction between different diffusing species because charge must be conserved.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

88 Flow and transport

When NaCl is diffusing in distilled water, the more mobile Cl ions will drag along the Na ions and
accelerate Na slightly, while the movement of Cl will be impeded somewhat. Similarly for a
Ca(HCO3)2 solution, some acceleration results for HCO 3 , and some retardation for Ca
2
. For a sym-
metric salt (in which cation and anion have the same charge), the resulting diffusion coefficient can be
calculated with the Nernst formula, which emphasizes the smaller value of the pair of ions. For multi-
component solutions, an average diffusion coefficient may be used which, for the ions in Table 3.2, is
Df  1.3109 m2/s at 25°C. Values at other temperatures can be obtained using Equation (3.40):

D f ,T  D f ,298 T 298 / (298 T ) (3.42)

which yields Df,10  0.84109 m2/s at 10°C (cf. also Li and Gregory, 1974). Finally,
the ionic mobility listed in Table 3.2 is for ions at trace concentration, and decreases towards
higher concentrations (cf. Figure 1.13). Accordingly, the diffusion coefficients decrease slightly
with increasing concentration for simple salt solutions (Lasaga, 1998, p. 317), while the
relations become quite complicated for multicomponent solutions (Felmy and Weare, 1991).

Table 3.2. Diffusion coefficients (Df 109 m2/s) in “free” water as calculated from ionic mobilities at 25°C
(obtained from molar conductivities, Table 1.7).

Na K Mg2 Ca2 Cl HCO

3 SO2
4 Average

Mobility 5.19 7.62 5.46 6.17 7.91 4.61 8.29 6.46108 m2/s/V
Diffusion coefficient 1.33 1.96 0.70 0.79 2.03 1.18 1.06 1.30109 m2/s

Overall, the differences in Df remain small and in most cases it is sufficient to use a constant diffusion
coefficient in “free” water of Df
109 m2/s for simple electrolytes. For neutral organic molecules,
the diffusion coefficient can be estimated with (Schwartzenbach, et al., 1993; Lyman et al., 1990):

D f ,298  2.8 109 Vl0.71 (3.43)

where Vl is the liquid molar volume of the substance (cm3/mol; Vl  (Mol. wt, g/mol) / (l, g/cm3).
Compared to diffusion in “free” water, solutes diffusing through a sediment-water system must
travel an extra distance because they have to circumnavigate the sediment grains. The effective dif-
fusion coefficient De corrects for the additional pathway. There is considerable controversy on how
the effective diffusion coefficient can be calculated from the free water value and related to sediment
properties (Lerman, 1979; Boudreau, 1997). In the Kozeny theory (Childs, 1969), it is related to the
tortuosity which is defined as the length of the actual travel path taken by a solute in a porous
medium, divided by the straight line distance:

De  D f /
2 (3.44)

where
is the tortuosity of the porous medium. The quadratic dependence of diffusion on tortu-
osity results from the following reasoning. Imagine a column which has its ends in two different
solutions, so that a steady state flux of some ion passes through the column. The sediment in the
column has pores which are connected by tortuous capillaries of length La. Now mould the
column in such a way that the capillaries are straightened out to length L, without affecting
the volume of the pore space, the pore geometry, or the flux of the ion through the column.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Diffusion 89

The concentration gradient in the moulded column will be smaller by (La / L). The volume of the
column has not changed, and therefore the cross-sectional area will also be smaller by (La / L).
Since the total flux through the column did not change either, the apparent diffusion coefficient
in the straightened capillaries must be larger by (La / L)2 
2. A tortuous path, at an average
–
angle of 45° with the straight line, obviously gives a tortuosity
 La / L  2 / 2  1.4, and hence
De  Df / 2.
Alternatively, the effective diffusion coefficient can be related to the formation factor F, which
is the ratio of the specific electrical resistance of a sediment with a solution, and the specific resist-
ance of the free solution. The formation factor is thus defined for a unit area of porous medium, i.e.
pore volume and grain skeleton together. The formation factor has the advantage that it can be meas-
ured directly. The formation factor itself is also a function of porosity. Empirical relationships
between formation factor and porosity take the form of Archie’s law:

F  1 / ewn (3.45)

where the exponent n varies from 1.4 to 2.0 (McNeil, 1980). When the sediment grains are perfect
insulators and only the pore space conducts electricity, the tortuosity is related to the formation fac-
tor as (Childs, 1969):

2  F ew (3.46)

Substitution of F  2 1
w yields
 w , and the effective diffusion coefficient follows from:
2

De  D f /
2  D f ew (3.47)

With w  0.5, the relation gives De  0.5Df, the same value as derived from tortuosity per se. Other
values of w and n give of course a slightly different result; Boving and Grathwohl (2001) found

2  1.2
w for sandstones and limestones, which yields De  0.44Df for w  0.5. In dry soils,
the diffusion coefficient may be further reduced when water resides only in very thin films or
unconnected pores (Hu and Wang, 2003).
When applying diffusion coefficients for “free water” to sediment-water systems, one should
remember that only the water filled porosity, w, participates in the diffusive flux. Thus, the flux
must be multiplied with w to obtain the mass transfer for a surface composed of grains and water
(compare with the difference among water velocity and Darcy flux). Some authors incorporate the
correction in the value of De, which has led partly to the confusion noted by Lerman (1979) and Bear
(1972).

3.5.2 Diffusion as a random process

Fick’s second law (3.38) gives the change of c in space (x) and time (t). For comparison with experi-
mental data we are normally more interested in the actual concentration at a given time and location,
which requires integration of Equation (3.38). By differentiation with respect to t and x, we can ver-
ify that the integration of (3.38) yields a solution such as:

 x 2 
c  A t ⁄2 exp 
1
 (3.48)
 4 Dt 

Here A is a constant that must be found from the initial and boundary conditions. Consider the initial con-
dition where no chemical is present at time t  0, while at t  0, N moles are injected at the origin, x  0.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

90 Flow and transport

This is known as a single shot input or Dirac delta function; as t → 0, c  0 everywhere except at the ori-
gin where c → %. Mass conservation at any time t requires:

%
1⁄2
 x 2 
N  A t ∫ exp  4 Dt  dx
%
(3.49)

––––
Substitution of x2 / (4Dt)  s2, and the associated derivative dx  (4Dt) ds gives:

%
N  A 4D ∫ es ds
2
(3.50)
%

The integral now contains a form closely akin to the error function erf(z):

z
2
erf(z )  ∫ es
2
ds (3.51)
√$ 0

The error function is tabulated, and numerical approximations are known, as given in Table 3.3. The
error function is 0 for z  0, and 1 for z  %. Furthermore, the error function is symmetric around
0, so that erf(z)  erf(z). We can thus evaluate the integral in (3.50) as:

%
√$
∫ es ds  2 erf (% )  √ $
2
(3.52)
% 2

The constant A in (3.50) then becomes:

A  N/(4$D) ⁄2
1

and the solution of Equation (3.38) for the initial condition stated is:

N  x 2 
c( x, t )  exp   (3.53)
4$Dt  4 Dt 

where N is the input mass (moles) at time t  0 at x  0. c(x, t) is expressed in mol/m since we consider
only one dimension. Fundamentally, (3.53) is equal to the normal density function (the Gauss curve):

N  (x  x )2 
n( x )  exp  0  (3.54)
2$ 2  2  2


where x0 is the average location (in Equation (3.53), x0  0), and 2 is the variance of the distribu-
tion. Diffusion can therefore be considered as a statistical process. Figure 3.24 illustrates the concept
as spreading of particles moving in random steps. One thousand particles start at x  0 and make
steps of random size within the range 2.5 to 2.5. The curve above the swarm sums the number of
particles for each x. In theory, a Gauss curve results with a variance that increases with each round
(i) of random steps as:

 2  i 25 / 12 (3.55)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Diffusion 91

Table 3.3. Values and approximations for the error function (erf) and the error function complement
(erfc  1  erf).

z erf(z) erfc(z)

3.00 1.0000 2.0000 Mo

2.50 0.9996 1.9996
4πDt
2.00 0.9953 1.9953 68%
1.00 0.8427 1.8427 0.61 Mo
0.50 0.5205 1.5205 4πDt
0.00 0.0000 1.0000 2 2Dt
0.10 0.1125 0.8875
0.20 0.2227 0.7773
0.30 0.3286 0.6714 σ xo σ
0.40 0.4284 0.5716
Figure 3.24. A Gaussian (normal) density function.
0.50 0.5205 0.4795
0.60 0.6039 0.3961
0.70 0.6778 0.3222
–
z  1⁄2   0.7071 0.6827 0.3173 1.2
0.80 0.7421 0.2579
0.90 0.7969 0.2031 1
erf(z)
1.00 0.8427 0.1573
2 / √π exp(s 2)

0.8
1.10 0.8802 0.1198
1.20 0.9103 0.0897 0.6

1.30 0.9340 0.0660 0.4 erfc(z)  1  erf(z)

1.40 0.9523 0.0477
1.60 0.9763 0.0237 0.2

1.80 0.9891 0.0109 0

z
2.00 0.9953 0.0047 2 1 0 1 2 3
s
2.25 0.9985 0.0015
2.50 0.9996 0.0004 Figure 3.25. The error function, erf(z), is the integral of the
2.75 0.9999 0.0001 Gaussian density function from 0 to z. The error function
3.00 1.0000 0.0000 complement, erfc(z)  1  erf(z), is the integral from z to %.

Approximations for erf(z) (Abramowitz and Stegun, 1964):

erf(z)  (1  e(4z /$)) ⁄ (0.007)

2 1
2

erf(z)  1  (1  a1z  a2z2  a3z3  a4z4)4 (0.0005)

a1 a2 a3 a4
0.278393 0.230389 0.000972 0.078108

erf(z)  1  (a1b  a2b2  a3b3  a4b4  a5b5) exp(z2) (1.5107) b  1/(1  a6z)
a1 a2 a3 a4 a5 a6
0.254829592 0.284496736 1.421413741 1.453152027 1.061405429 0.3275911
Note that the error function is symmetric around 0, and that erf(z)  erf(z).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

92 Flow and transport

150

100

Particles 0* 5* 25* 75*

50

0
50 0 50 50 0 50 50 0 50 50 0 50

Figure 3.26. Computer simulation of one thousand particles that do 5, 25 or 75 steps with a random size in the
range 2.5 to 2.5.

As the enlargement after 75 steps in Figure 3.27 indicates, the thousand particles portrayed in Figure
3.27 still only roughly approximate a smooth Gauss curve. Computed random motions require many
thousands of particles before a reasonable fit is obtained within a few percent deviation. Computer
programs which use the “random walk” based on particle tracking in combination with a random
step (e.g. Uffink, 1988; Kinzelbach, 1992, Sun, 1999), are known as devourers of computer time.

50

75*
Particles

25

0
50 0 50

Figure 3.27. Enlargement of Figure 3.26 for 75 random steps and comparison with the ideal Gauss curve.

Since Equations (3.53) and (3.54) fundamentally are the same, the variance 2 is related to the dif-
fusion coefficient by:
 2  2Dt (3.56)
where  has the dimension of length. This simple formula, first derived by Einstein, provides a rapid
estimate of the mean diffusion length, not only of chemicals in aquifers but in gases and in minerals
as well (Moore, 1972). As shown in Figure 3.24, 68% of the mass that is spreading is contained in
the area from  to , after a given period of time.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Diffusion 93

For diffusion in three dimensions the variances of the extra dimensions must be added. Thus, for dif-
fusion in water where 2x  y2  2z , the sphere where 68% of a point source is located has a radius
r  xyz:

 xyz  ( 2x   2y   2z )  6D f t (3.57)

3.5.3 Diffusive transport

When is diffusive transport important as compared to advective transport? Using Equation (3.56)
diffusive spreading can be calculated straightforwardly. Table 3.4 shows results for different time
periods, and for comparison also the traveled distance L, during the same time, by an advective
flow of 10 m/year. Over short time periods diffusive transport equals advective transport, but over
longer time periods (and hence also over larger distances) advective flow becomes more and more
important.

Table 3.4. Spreading of a point source through diffusion (Df  109 m2/s),
compared with advective transport (v  10 m/year).

Diffusion, Advection,
t (s)  (cm) L (cm)

60 (1 min) 0.03 0.002

3600 (1 hr) 0.27 0.11
21600 (6 hr) 0.66 0.68
68400 (1 day) 1.3 2.7
32106 (1 year) 25 1000

Fick’s second law can be integrated for the boundary conditions:

c(x, t)  ci, for x ! 0, t  0 and for x  %, t ! 0;

c(x, t)  c0, for x  0, t ! 0

The solution for these boundary conditions can be obtained with similar methods as applied for
Equation (3.53) (Lasaga, 1998, p. 368) or via Laplace transforms (Boas, 1983, p. 676; Van
Genuchten, 1981) and is:
 
x
c(x, t )  ci  (c0  ci )erfc   (3.58)
 4D t 
 e 

where erfc[z] is the error function complement (i.e. 1  erf[z], Table 3.3). This equation has
been used to model field data of diffusion processes. In the Champlain Sea clay deposits, the
Cl profiles reflect diffusion out of the clay over a period of 10,000 years since the regression
of the sea (Desaulniers and Cherry, 1988). The model considered an initial homoge-
neous Cl distribution (1⁄3 of seawater concentration) and an upper weathered zone, where the
salt is flushed. The Cl distribution was succesfully modeled using a diffusion coefficient
De of 2106 cm2/s. This effective diffusion coefficient can be compared with a “free” water
diffusion coefficient of about 8106 cm2/s at 10°C that we found in Section 3.5.1.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

94 Flow and transport

Moisture Cl
loss (g/L) Na K Ca2& Mg2
(wt%) 0 4 8 12 14 20 24 (g/L) (mg/L) (mg/L)
0 10 20 0 1 2 3 4 5 6 7 8 9 10 11 0 100 200 300 0 200 400 600 800 1000 1200 0

100

200
Upper Chalk

marl

300

Clⴚ Kⴙ
Depth (m)

Naⴙ Ca2ⴙ Mg2ⴙ

400

marl
Middle Chalk

500 Melbourn
Lower Chalk rock

Middle Lias
Lower Lias

600

Figure 3.28. The chemical composition of pore waters in Upper Cretaceous Chalk deposits, UK. The upward
decrease indicates an upward loss of ions by mainly diffusion (Bath and Edmunds, 1981).

Similar effective diffusion coefficients have been used by Volker (1961), Sjöberg et al. (1983),
Groen et al. (2000), Hendry and Wassenaar (2000) and Beekman (1991) in clays where a geologic
change induced a salinity jump and Cl diffusion occurred over periods of about 300 to 500 years.
Figure 3.28 shows the water composition in connate waters in chalk deposits (Bath and Edmunds,
1981). Upward diffusion is thought to be the main driving force behind the concentration profiles.
However, the upper part of the chalk is fissured and here advective flow removes solutes, and the
transport mechanism may be somewhat hybrid. Equation (3.58), assuming a diffusion coefficient De
of 51010 m2/s, gives a transport time of around 3 million years in this case, since late Tertiary. Even
longer diffusion times of over 50 million years were used to model Cl and
2H profiles in clays
which are considered as repositories for the storage of nuclear waste (Patriarche et al., 2004).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Diffusion 95

Relative concentration
0.2 0.4 0.6 0.8 1.0
0 8
2  105  108 7
1  10

Toluene 6
10
6 6
20 
10

4
Depth below interface (cm)

40

Cl

60

80

100

Figure 3.29. Chloride and toluene concentrations in interstitial solutions obtained from a clay core below a
waste site. Simulation data for 1800 days and for indicated diffusion coefficients. Reprinted with permission
from Johnson, et al., 1989. Copyright American Chemical Society.

Advective flow in clays or peats (aquicludes and aquitards) is often negligible and diffusion is then
the only transport mechanism for chemicals. An example, showing diffusion of Cl and toluene in
thick clays underlying a waste deposit, is given in Figure 3.29. The diffusion coefficients used in the
simulations are indicated in the figure; the best fit for Cl and toluene is obtained with 5106 and
5108 cm2/s, respectively. For linearly retarded chemicals De is replaced by De / R, which indicates
that R
100 for toluene.

EXAMPLE 3.9. Calculation of a diffusion profile

Calculate the relative concentration profile of Cl diffusing in the clay after 1800 days (Figure 3.29).
Df  1.1109 m2/s for Cl, w  0.5.

ANSWER:
The effective diffusion coefficient is De  Df w  1.1109  0.5  5.51010 m2/s. Time is 1.56108 s.
We use Equation (3.58) with (4Det)  0.585 m, ci  0 and c0  1 mol/L. Hence:

c(x, 1800 days)  0  (1  0)  erfc(x / (4De t )  1  erf(x / 0.585)

The simple approximation for erf(z)  (1  exp(4z2 / $))0.5 provides sufficient accuracy here, and gives
c(0.4 m, 1800 days)  0.33 mol/L, etc.

QUESTION:
Calculate c at 0.6 m
ANSWER: 0.14 mol/L

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

96 Flow and transport

Figure 3.29 shows that adsorption initially can slow down the diffusive flux of a pollutant. However,
once all the sorption sites are filled, retardation stops, and the diffusive fluxes of conservative elements
and adsorbed elements become the same. When, for example, a 1 m thick clay liner is used to act as a
barrier against downward percolation of leachates from a waste site, the steady state flux may become
considerable, and cause extensive pollution of groundwater in the aquifer below the site (Example 3.10).

EXAMPLE 3.10. Diffusive flux through a clay barrier (after Johnson et al., 1989)
Estimate the steady state flux of benzene through a 1 m thick clay liner below a 104 m2 waste site. The concen-
tration of benzene in the waste site is 1 g/L, the groundwater below the liner has a concentration at 0.01 g/L.
The clay has a porosity w  0.5, the free water diffusion coefficient of benzene Df  7106 cm2/s.
Also estimate the time needed to arrive at steady state when benzene is sorbed linearly by the clay,
Kd  2.08.

ANSWER:
An effective diffusive coefficient is calculated from Equation (3.47): De  Df  w  7106 
0.5  3.6106 cm2/s. From Fick’s first law obtain a flux

F  De ew (∂c /∂x )  3.6106  0.5  (1  0.01) / 100  1.8108 mg/s/cm 2 .

This amounts to 56 kg benzene/yr, a quantity which has the potential to contaminate 11.2109 liters of
water at the EPA drinking water limit of 0.005 mg/L. Note that we multiplied the flux with porosity to obtain
the flux through the open, pore space.
The maximum time necessary to reach steady state can be calculated from amount of adsorption that
must occur. Adsorbed with 1 g/L and 0.01 g/L in solution are 2.1 and 0.02 g/L pore water, or the average for
the whole clay liner: 1.05104  0.5  5252 kg benzene. Add the amount in solution, which is 2525kg, to
find a total quantity of 7777 kg benzene in the clay liner (one may wonder whether this is available in the
waste). This amount is carried into and through the clay in about 140 years by the steady state flux. The ini-
tial diffusion profile is concave, and consequently exhibits a larger gradient, and a larger flux; the calculated
time space to reach steady state is therefore a maximal one. The time for steady state can also be
calculated with an analytical solution (Johnson et al., 1989).

Diffusion also has important consequences for transport in fractured aquifers, where the main flow
occurs along faults or fissures, but solutes exchange through diffusion with the stagnant water in the bulk
rock. The phenomenon was described in Section 3.1 in the discussion of tritium transport in the unsatu-
rated zone of chalk (Figure 3.1). A porous formation with these characteristics is termed a dual porosity
medium, or a rock with mobile–immobile zones. Aggregated soils in which transport occurs both by
advective flow along shrinkage fissures, and by diffusion into the peds, have similar transport character-
istics. It is typical to observe tailing of the concentrations when a pulse injection has passed the observa-
tion point, due to back-diffusion, again out of the matrix into the channels which contain the mobile fluid.
These processes generally require numerical models as applied in Chapters 6 and 10, and discussed in
Chapter 11.

3.5.4 Isotope diffusion

For perfect gases, the diffusion coefficient of gas 1 into gas 2 is given by kinetic theory (Bird et al.,
1960; Tabor, 1991):

v12  v22
D1,2  (3.59)
12$(c1  c2 )(a1  a2 )2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Diffusion 97

where v is the molecular velocity (m/s), c the concentration (mol/m3) and a is the radius of the mol-
ecule. With equal kinetic energy E  1⁄2mv2 for all the molecules, and equal radii for isotopes a and
b of gas 1, we find the ratio of the diffusion coefficients for isotope tracer-diffusion in gas 2:

m m 
 1a 2 
D1a,2 v12a  v22  m m
1a 2  m1b ( m1a  m2)
   (3.60)
D1b,2 v12b  v22 m m  m1a ( m1b  m2)
 1b 2 
 m m
1b 2 

For example, for diffusion of 13CO2 and 12CO2 in air, the diffusion coefficients show the proportion:
D12
CO2, air 45  (44  28.8)
  1.0044 (3.61)
D13 44  (45  28.8)
CO2, air

where 28.8 g/mol is the molecular weight of air (with 79% N2 and 21% O2). The ratio of 1.0044 indi-
cates that the light molecule 12CO2 will diffuse a certain distance quicker than 13CO2. The ratio
m1m2 / (m1  m2) is called the reduced mass 1,2 (g/mol).
The diffusion coefficients for water molecules show a similar mass effect, as illustrated in
Table 3.5 for water labeled with tritium and deuterium. Also, the self-diffusion coefficients of H216O
––––––
and H218O show a ratio of 1.054( 20 / 18) as expected for perfect gases (Tyrrell and Harris, 1984).
Apparently, water behaves here according to simple kinetic theory and we can estimate the tracer-
diffusion coefficient for isotopes of solute ions with Equation (3.60). For the solute ions it is neces-
sary to include the water molecules that adhere to the solute ion (the hydration number is
approximately given by the Debye-Hückel å values in 1010 m, cf. Chapter 4). For example for Cl,
å / 1010  3, which gives a mass m37Cl  91 g/mol and 37Cl,H O  15.0275 g/mol. Similarly for
2

35 
Cl , we have m35Cl  89 g/mol and 35Cl,H O  14.972 g/mol. The resulting ratio of the diffusion
2

coefficients is D37Cl,H O / D35Cl,H O  (35 / 37) ⁄2  0.9981. Desaulniers et al. (1986) optimized a
1
2 2

ratio of 0.9988 for a diffusion profile in clay, and Eggenkamp et al. (1994) obtained 0.9977 for a dif-
fusion profile in a marine sediment.

Table 3.5. Tracer diffusion coefficients (Df / 109 m2/s) of tritium (T) and deuterium (D)
labeled water in ordinary water (H216O) (Mills and Harris, 1976).

(HDO / HTO) ⁄
1
°C HTO HDO Ratio 2

5 1.272 1.295 0.982

(9.2432 / 9.4737) ⁄  0.988
1
25 2.236 2.272 0.984 2

45 3.474 3.532 0.984

Beekman (1991) investigated a site flooded by seawater where Cl diffused over a period of
400 years downward into a freshwater clay. In 1932 a dike was built that isolated the area from the
sea and a freshwater lake developed. From 1932 onwards, Cl therefore diffused backward into the
freshwater. Figure 3.30 shows the profiles for Cl and
37Cl as calculated by a numerical model. The
solid lines depict the situation after 400 years of downward Cl diffusion and give a downward con-
cave Cl profile. When backward Cl diffusion into freshwater is included, the result is the dotted
line marked 1987 (Figure 3.30). Chloride is diffusing out of the upper part and a Cl maximum forms.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

98 Flow and transport

Cl (mmol/L)
37Cl (‰)

0 100 200 300 400 500 600 1.0 0.5 0.0 0.5 1.0
0 0
Depth (m below Lake Bottom)

2 1987 2
1932 1932 1987
4 4
6 6
8 8
10 10
12 12
14 14
16 16
18 18

Figure 3.30. Calculated concentrations of Cl and
37Cl in the Markermeer sediment in 1932, at the termina-
tion of salt water diffusion. Solid lines are for 400 years of downward Cl diffusion until 1932. Dotted lines
include backward Cl diffusion from 1932 to 1987.

When seawater diffuses down in the sediment, 37Cl migrates slightly slower than 35Cl, therefore neg-
ative
37Cl values are expected. The model isotope ratio (Figure 3.30, solid line) has zero enrichment
at 0 and 18 m depth, the boundary locations where seawater and initial water are found. In between,
a pronounced minimum is calculated for
37Cl at 8 m. When the Cl diffuses out of the sediment,
37
Cl is again slower than 35Cl, and it becomes enriched in the remaining salt. This enrichment results
in a distinct
37Cl maximum at 2 m depth (Figure 3.30, dotted line).
Figure 3.31 shows the measured profiles for Cl and
37Cl together with model calculated
lines. Scenario A corresponds to the 1987 dotted lines in Figure 3.30. The Cl concentra-
tion approximates the observed concentrations for depths greater than 4 m reasonably well.

Cl/566 (mmol/kg H2O)

37δ (‰)

1.0 0.5 0.0 0.5 1.0

0
Cl/566
37δ ( error limit)
2
Depth (m below Lake Bottom)

Modelling results
Scenario A
4 37δ
Scenario B

6
Cl/566
8

10

12

14

Figure 3.31. Profiles for Cl and
37Cl in the Markermeer sediment measured in 1987. Scenario A corresponds
to the dotted model line in Figure 3.30. Scenario B included a sedimentation/erosion component. (Beekman,
1991; Eggenkamp, 1994).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 99

The Cl concentration in the upper 4 m has decreased due to backward Cl diffusion, and the
observed concentrations are also well matched by the computed line. For
37Cl, the field data do
show a maximum and a minimum as predicted by the model. However, the modeled minimum is
much greater than measured, while the modeled maximum is smaller. Beekman (1991) concluded
that the bottom sediments were homogenized during storms and gradually accumulated in the period
of downward salt diffusion. Thus, more 37Cl was brought into the profile than with diffusion only.
The full lines, scenario B in Figure 3.31, are for a sedimentation/erosion model which combines
spasmodic events and calm periods in-between when the concentration variations are smoothed out
again.

EXAMPLE 3.11. Chloride isotope fractionation during diffusion

Calculate
37Cl at 8 m depth in the salt water diffusion profile shown in Figure 3.30. Diffusion time is 362 yr,
D35Cl  8.41010 m2/s, D37Cl / D35Cl  0.9988, Cli  65 mmol/L and Cl0  566 mmol/L. The upper
meter sediment is homogenized with seawater due to storm events and bioturbation (Cl0 extends to 1 m bsl).
Both fresh water and seawater have
37Cl  0‰.

ANSWER:
The Cl isotope composition of seawater is 24.47% 37Cl and 75.53% 35Cl (Table 2.4). Thus,
35
Cli  0.7553  65  49.0945 mmol/L and 35Cl0  0.7553  566  427.4978 mmol/L. We calculate the
concentration after 1.141010 seconds at x  7 m (corresponding to 8 m depth) with Equation (3.58), hence
––––
for z35  x / (4Dt)  7 / (4  8.41010  1.141010) ⁄  1.13024. It gives, with the accurate approxi-
1
2

mation of the error function:

35 Cl  49.0945  (427.4978  49.0945)  erfc(1.13024)  90.702 mmol/L
–––––
Similarly for 37Cl, with z37  z35 &  0.9988  1.13092:
35 Cl  15.9055  (138.5022  15.0945)  erfc(1.13092)  29.359 mmol/L

The total Cl concentration is 35Cl  37Cl  90.7  29.4  120 mmol/L.
37Cl is the ratio in the sample nor-
malized to the ratio in the standard:

 ( 37 Cl / 35Cl)   29.359 / 90.702 

37 Cl    1  1000  
sample
 1  1000  0.883
 ( 37 Cl / 35Cl)   24.47 / 75.53 
 standard 

QUESTION:
Calculate
37Cl at x  2 m (3 m depth)
ANSWER: 35Cl  235.77, 37Cl  76.35,
37  0.404‰

3.6 DISPERSION
Groundwater flowing through a sand layer is forced to move around the sediment grains as illustrated
in Figure 3.32. The resulting spreading of a concentration front is called dispersion. There are two
types of dispersion (Figure 3.32); differences in travel time along flowlines around grains cause lon-
gitudinal dispersion (DL), whereas transverse dispersion (DT) is due to stepover onto adjacent flow-
lines by diffusion. Also indicated in Figure 3.32 is the statistical nature of dispersion; the choice just
before a grain to go left or right leads automatically to a probabilistic interpretation of the process
exactly as for diffusion.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

100 Flow and transport

dy2
Transverse DT = 1
2 dt
B


C y
A
D

E DL
10DT

Longitudinal

dx2
DL = 1
2 dt
1
A 2 B 
3

Figure 3.32. Longitudinal and transverse dispersion viewed at the microscopic scale.

While the physical process behind dispersion is different from diffusion, mathematically we can still
use Fick’s laws to quantify the spreading of concentration fronts. Combining the reactive transport
and the diffusion equation (Figures 3.14 and 3.23) we obtain:
∂c ∂c ∂q ∂2 c
 v   DL (3.62)
∂t ∂x ∂t ∂x 2
The longitudinal dispersion coefficient, DL, is also termed the hydrodynamic dispersion coefficient.
Equation (3.62) contains three terms on the right hand side; the first describes Advective flow, the sec-
ond chemical Reactions and the third Dispersion, therefore it is also called the ARD equation. This
sequence of the three terms may be the order of their relative importance in controlling groundwater
quality in most situations. The equation requires considerable mathematical background to solve
under various conditions (Bear, 1972). However, as in the case of diffusion, front spreading by dis-
persion can be calculated with the statistical approach. Figure 3.33 shows how a Gaussian density
function migrates down a flowline and is broadening due to dispersion. The dispersion can be calcu-
lated using Equation (3.53) with the difference that x0, the initial location of the point source, is not
zero, as for diffusion, but increases with the distance covered by the moving fluid, i.e. x0  vt.

δ-input Measurement point

Figure 3.33. Propagation of a Gaussian density function along a flowline (Levenspiel, 1972).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 101

Methods for determining the value of , and hence of Df or DL, from measured data, are illustrated with
the Gaussian density function in Figure 3.24. A value of 2 gives the distance comprising 68% of the
original mass. When a variance 2 can be calculated from a series of observation points, either on time-
or on distance-basis, the value of the dispersion coefficient can be calculated from Equation (3.56).
A time-based variance can be recalculated to become dimensionless, and to a distance basis, or vice
versa as illustrated in Example 3.12.

EXAMPLE 3.12. Dispersion coefficient from a single shot input

Calculate the dispersion coefficient from tracer output concentration at the end of a column measured as a
function of t. This example is from Levenspiel (1972).

Tracer output
Time t (min) Concentration (g/L)
5 Smoothed
0 0 4 curve
5 3 3
10 5
15 5 2
20 4 1
25 2
30 1 5 10 15 20 25 30
35 0 t (min)

ANSWER:
The variance of the measured output distribution can be calculated as the variance of a number of measurements,

(individual measurements)2
 (average)2   2
(no. of observations)

Hence:

 tc 2
t2 
∑ ti2 ci t
2

∑ ti2 ci 
∑ i i
∑ ci ∑ ci  ∑c 
 i 

∑ ci  3  5  5  4  2  1  20

∑ ti ci  (5  3)  (10  5)  …  (30  1)  300 min

∑ ti2 ci  (25  3)  (100  5)  …  (900  1)  5450 min 2

Therefore:

2
5450  300 
t2     47.5 min 2
20  20 

The time-based variance is normalized by dividing by the square of the average arrival time:

t2 47.5
2  2
  0.211
t (15)2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

102 Flow and transport

The variance can be expressed on distance basis: x  v t, for example when v  1 cm/min,
x  1  15  15 cm, and

 2x  t2 / t x 2  0.211  225  47.5 cm 2

2

Since x2  2 DL t, DL  x2 / (2t)  47.5 / (2  15)  1.58 cm2/min.

The transformation can only be made when dispersion coefficients are small (and observations obey, any-
how, the normal distribution, i.e. are symmetric around midpoint breakthrough).

In Example 3.12 the distribution of breakthrough concentrations was characterized by its variance.
Temporal or spatial moments of the breakthrough distribution also serve that purpose. The temporal
moments m0, m1, …, mi are defined as:
% %
m0 (L)  ∫0 t 0 cL,t dt  ∫ cL,t dt (3.63a)
0

%
m1 (L)  ∫ t1cL,t dt (3.63b)
0

or in general:
%
mi ( L)  ∫ t i cL, t d t (3.63c)
0

The zeroth moment m0(L) is the total mass of chemical, integrated over time at point L (mass s/L).
The first moment divided by the zeroth moment gives the mean arrival time –t  m1 / m0 (s). In
Example 3.12, we calculated the variance t2  m2 / m0  (m1 / m0)2 (s2). The third moment gives the
skewness of the distribution. Besides characterizing empirical distributions, moment analysis is used
as a mathematical tool to derive system properties in the Laplace space, and to compare physical
behavior such as dispersion, interaction among mobile and stagnant zones, and kinetic chemical
reactions (Valocchi, 1985; Goltz and Roberts, 1987; Harvey and Gorelick, 1995; Garabedian et al.,
1991).

3.6.1 Column breakthrough curves

When a conservative tracer is injected into a column at concentration c  c0, the front will move
with the average water flow velocity through the column. At the same time the front disperses,
as illustrated in Figure 3.34. In the figure, the concentrations are scaled from 0 to 1, by dividing all
values with c0. The concentrations are found by integrating Equation (3.62) for dq  0 which
becomes:

 ∂c   ∂c   ∂2 c 
   v    DL  2  (3.64)
 ∂t  x  ∂x t  ∂x t

To facilitate the solution, we make the column infinite and admit a non-zero initial concentration ci:

c  ci for x ! 0; t  0, and for x  %; t ! 0

c  c0 for x ' 0; t  0, and for x  %; t ! 0

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 103

Continuous supply 1
of tracer at
concentration c0 c /c 0
after time t0
0
t0 Time
(b)
v breakthrough, t2
1
First
c /c 0 appearance t1
Effect of
x dispersion
0
t0 Time
(c)

1
2
c /c0
Outflow with tracer 1
at concentration c
0
after time t1 x
(a) (d)

Figure 3.34. Longitudinal dispersion of a tracer passing through a column of porous medium. (a) Column
with steady flow and continuous supply of tracer after time t0; (b) step-function-type tracer input; (c) relative
tracer concentration in outflow from column (dashed line indicates plug flow condition and solid line illustrates
effect of mechanical dispersion and molecular diffusion); (d) concentration profile in the column at various
times. Reprinted with permission from Freeze and Cherry, 1979. Copyright Prentice Hall, Englewood Cliffs,
New Jersey.

The solution is similar to Equation (3.58), but x is taken relative to the advective front v t:

(c0  ci )  x  vt 
c( x, t )  ci  erfc   (3.65)
2  4 DL t 

Note that erfc[z]  1  erf[z], so that Equation (3.65) is symmetric around x0.5  vt. The concen-
trations from ci to (c0  ci) / 2 run ahead of the average front position at x0.5  vt, and the argument
of the error function complement is positive. The concentrations from (c0  ci) / 2 to c0, on the other
hand, lag behind x0.5, and the argument is negative.

EXAMPLE 3.13. Front dispersion in a column

Calculate the Cl concentration front in a column injected with a 1 mmol/L NaCl solution. The water veloc-
ity is 2.8104 cm/s, DL  104 cm2/s; the initial Cl concentration in the column ci  0 mmol/L.
Calculate the concentrations after 5 hr injection time at the advective front, and at x  vt  2DLt.

ANSWER:
The advective front has arrived at v t  2.8104  (5  60  60)  5.04 cm. The concentration at this
point, where x  vt, is 1⁄2 erfc[0]  0.5 mmol Cl/L (clearly just half of initial and input concentrations).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

104 Flow and transport

The concentration at x  vt  –2–D––

Lt is:

0.5 erfc  2DL t /

 ( –
)
4 DL t   0.5  1  erf  1⁄2   0.5  0.5  0.68  0.16


[ –]
where the value of erf  1⁄2 , as found from Table 3.3, is 0.68.
–––––
The concentration at x  vt   (2DLt) is similarly:

0.5 erfc  2DL t /

 ( –
)
4 DL t   0.5  1  erf 1⁄2   0.5  0.5  0.68  0.84


The spread between c  0.16 and c  0.84 is obtained over a length exactly equal to 2, and encompasses
68% of the concentrations (0.84  0.16  0.68) around the midpoint breakthrough, as illustrated in Figure
3.35. The distances from the midpoint breakthrough are 2––D––
Lt  1.9 cm.

0.84
68%

c/c0 0.5

0.16

0
0 5 10 cm

 2Dt  2Dt
vt

Figure 3.35. Dispersion at a concentration front and its relation to the variance.

QUESTION:
Calculate the concentration at x  % in the column, using Equation (3.65).
ANSWER: erfc(%)  0, hence c(%, t)  ci  0.

The solution for a dispersion front, given in Equation (3.65), is for an infinitely long col-
umn, %  x  %, and for resident concentrations, cr, within the column. If we instead consider a
semi-infinite column from x  0 to x  %, and define a constant concentration boundary at x  0
(cr(0, t ( 0)  c0), the concentrations in the column are (Lapidus and Amundson, 1952):
      
(c0  ci )  x  vt   exp  vx x  vt 
cr ( x, t )  ci  erfc  D  erfc  (3.66)
2   4D t   L   
  L   4 DL t 

This equation adds an additional term to (3.65) that describes the diffusive flux at the entrance of the
column. Resident concentrations in the column can be obtained using probes that monitor radio-
active or color tracers, or by dissecting the column when the front has penetrated a certain length
(Bond and Phillips, 1990). However, it is more common to collect the effluent at the column outlet.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 105

The concentrations obtained in this way are flux averaged and differ from the resident concentration
in that dispersion into or out of the column is discontinued. For a flux vc0 injected in a column initially
free of the chemical, the flux into the column, at the inlet, consists of an advective part vcr, x0 and a
dispersive part DL(∂cr / ∂x)x0:

 ∂c 
vc0  vcr , x0  DL  r  (3.67)
 ∂x  x0

Consequently, the resident cr is smaller than the flux-averaged concentration c0:

DL  ∂c 
cr  c0   r  (3.68)
v  ∂x 

(Note that (∂cr / ∂x) is negative because initially the concentration in the column was zero.)
Now, we can define a flux averaged concentration at the inlet of the column (cf (0, t)  c0)
and then Equation (3.66) will give the flux averaged concentration cf (x, t) inside the column
(Kreft and Zuber, 1978; Van Genuchten and Parker, 1984; it is assumed that ∂cr / ∂x  ∂cf / ∂x).
But, these flux averaged concentrations are for effluent that is sampled at the end of the column.
Therefore, although Equation (3.66) was derived for a semi-infinite column, it is valid as well for a
short laboratory column with flux boundary conditions at the inlet and outlet points.
Equation (3.66) can be used to obtain values for the dispersion coefficient from break-
through experiments. An excellent program, CXTFIT2 by Toride et al. (1995), can do the job
with a least squares fit and can be obtained from the US Soil Salinity Lab, Riverside,
www.ussl.ars.usda.gov. Sometimes it is advantageous to have a visual fit as well, when some data
points are considered less reliable. A visual fit can be obtained by programming (3.66) in a spread-
sheet program, and comparing results for various DL with observed concentrations. For retarded
chemicals that show linear retardation, Equations (3.65) and (3.66) can still be used, replacing t with
t / R. For non-linear retardation a numerical solution of Equation (3.62) is required as presented in
Chapter 11.

3.6.2 Dispersion coefficients and dispersivity

In Figure 3.36 values of dispersion coefficients are plotted as function of the non-dimensional Peclet
number, Pe (Bear, 1972). The Peclet number was defined for this figure as:

Pe  vd / D f (3.69)

where d is diameter of the particles packed in the column (m). By varying the water velocity v, the
Peclet number was varied for a given packing and by injecting a tracer the dispersion coefficient was
determined for the Peclet number. Figure 3.36 plots the ratio of the dispersion-coefficient DL over
the diffusion coefficient Df in pure water on the vertical axis. Two domains can be discerned in the
figure (Bear, 1972; Plumb and Whitaker, 1988). At low Peclet numbers (low water velocity), the
ratio of DL / Df is constant. At intermediate Peclet numbers from 1 to 105, the value of DL / Df
increases linearly with the Peclet number.
The interpretation of the dispersion domains is as follows. At a Pe  0.5 (Figure 3.36) the ratio
DL / Df  100.3, or DL  Df / 2. This value corresponds with the effective diffusion coefficient that
we found when correcting the free water diffusion coefficient for the tortuosity of a sediment.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

106 Flow and transport

106
Theoretical curves according to
105 Bear and Bachmat (1965, 1966)
δ  10 ?
δ  1.0
104
?
DL
3
Df 10

102
Theoretical curve according to
DL Saffman (1960) for δ  10
Df
II
101

I 100

1 Pe  vd/ Df 103 102 101 1 10 102 103 104 105 106

Pe  vd /Df

Figure 3.36. A plot of the ratio of dispersion coefficient/diffusion coefficient versus Peclet number in packed
laboratory columns (Bear, 1972).

At a Pe ! 1, the ratio DL / Df increases in the proportion 1 : 1 with the Peclet number and
therefore:

DL  D f Pe  dv (3.70)

For low Peclet numbers where DL is constant, it follows that 2  2DLt  constant  t, or 2 
constant/v (over a given travel length). The variance increases with time, or equivalently, the front
spreading at the column end increases with decreasing velocity. In other words, diffusion is the dom-
inant process. The upper boundary for diffusion controlled spreading is found at a Peclet number of
0.5. For d values in the range 0.4 to 2 mm (medium to coarse sands), a flow velocity lower than 37 to
8 m/yr, respectively, will result in diffusion-controlled spreading.
For higher Peclet numbers, or for flow velocities higher than about 50 m/yr, the variance 2 
2DL t  2d vt  2dx, or 2  constant (over a given travel length). Front spreading is now independ-
ent of the water velocity and depends only on the travelled length. In other words, dispersion
becomes a characteristic property of the porous medium denoted by the dispersivity:

 L  DL / v (3.71)

where L is the dispersivity (m). It can be subdivided into a longitudinal (L) and transverse (T) dis-
persivity. The variance is related to dispersivity by:

 2  2DL t  2 L vt  2 L x (3.72)

This important formula is very efficient for calculating the spread of pollutant fronts.
The contributions of diffusion and dispersion are combined in the hydrodynamic, longitudinal
dispersion coefficient:

DL  De   L v (3.73)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 107

The dispersivity in columns, as shown in Figure 3.36, is equal to the representative grain-
diameter used in packing the column. When aquifer sediments are used in the column, the grain-
diameter is taken as d10, the diameter below which 10% of particles fall, and Perkins and Johnston
(1963) suggest the relationship:
 L  3.5 d10 (3.74)

Here 3.5 is a shape factor, which increases somewhat with smaller grain size. In aquifers the disper-
sivity is generally much higher than in packed laboratory columns, and the diffusion contribution in
Equation (3.73) becomes already negligible at a water velocity higher than 1 m/yr.

QUESTIONS:
Compare the effect of dispersion in c/x plots of a pulse injection of two chemicals, one conservative, and
the other with linear retardation R, at times t1 and t1  R?
ANSWER: The front of the retarded chemical arrives R times later, both the width and the
variance of the pulse of the retarded chemical are R times smaller. For small
pulses this results in some broadening of the retarded pulse and also in some
peak reduction.
And the same question, comparing a c/t plot?
ANSWER: The width and the variance of the pulses are exactly the same, only shifted in time
by R. However, small retarded pulses will show broadening and peak reduction.
Are the answers different for diffusion, cq. dispersion dominated spreading?
ANSWER: No
Calculate the number of grains in a column packed with 0.5 kg, 2 mm diameter quartz sand, density
2.65 g/cm3?
ANSWER: 45044

3.6.3 Macrodispersivity
Looking at the Cl concentrations downstream the Vejen waste site in Figures 3.2 and 3.21, we
have noted the irregular distribution pattern, even when the concentrations are averaged over depth.
The larger spreading in aquifers, compared to laboratory columns, is due to the heterogeneous struc-
ture of natural sediments with alternating sands of different hydraulic conductivity, and with inter-
calated layers of clay, silt or gravel. It is no longer the individual grain size, but rather the variation
in flow length around low-conductivity bodies that determines the dispersivity. This effect is known
as macrodispersivity.
We have considered dispersion as a statistical process, where flow is redirected around obstacles
so many times, that a normal (or Gaussian) distribution of flowpaths is obtained. Several thousands
obstacles are required to approach a normal distribution and this is easily reached in a column with
homogeneously packed grains, but in an aquifer this is not the case.
Figure 3.37 illustrates the effect of geological structures on the Dirac delta function (“the single
shot input”), and also shows how the sampling method influences the measured field dispersivity.
When filter screens are longer than 1 m, mixing of water from different layers is inevitable and disper-
sivity is increased. Field dispersivities are therefore lowest when small filters or depth specific sam-
pling techniques are used (Schröter, 1984). It is possible to obtain dispersivity values in an aquifer that
are as small as in column experiments, when a point sampling method is employed and a single flow-
line is monitored. Pickens and Grisak (1981) found a dispersivity of 7 mm which remained constant
over distances from 0.36 to 4 m. Taylor and Howard (1987), who used gamma-ray counting of a radio-
tracer in dry access tubes, even found a constant dispersivity of 1.6 cm over a distance of up to 40 m.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

108 Flow and transport

Input Output Input Output

(δ) (δ)

Homogeneous sand

“Permeability” stratification

Cross bedding

Large clay lenses

Clay body

Small clay lenses

Figure 3.37. Effect of sedimentological structures on dispersion of a shot inpulse. At left: point specific sam-
pling, at right: depth integrated sampling in the same aquifer.

Depth integrated sampling includes the effects of the geological heterogeneity, and is thus more apt
to give an overall picture of transport properties of an aquifer. Figure 3.38 shows that depth inte-
grated dispersivity in aquifers may be orders of magnitude larger than in columns. Even more impor-
tant is that dispersivity appears to increase with distance. In other words it becomes a scale dependent
property, while before dispersivity was considered as a constant. The general trend in Figure 3.38 is
that the macrodispersivity is about 10% of the traveled distance:
A11  0.1 x (3.75)

where A11 is the longitudinal macrodispersivity. However, the increase lessens for travel distances
above 1000 m.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 109

10000

1000
Longitudinal dispersivity (m)

100

10

1.0

Tracer Contam. Envir.

test events tracers
.1 Fractured
media
Porous
media
.01
1 10 100 1000 10000 100000
Scale (m)

Figure 3.38. Values for dispersivities obtained in field experiments (Gelhar et al., 1985).

If we combine Equations (3.75) and (3.72) we obtain:

 2  2A11 x  2 0.1x x  0.2x 2 (3.76)

for the variance of a pollutant front in aquifers (cf. Example 3.14).

EXAMPLE 3.14. Pollutant spreading during transport in an aquifer

An aquifer has thickness D  50 m; P  0.3 m/yr; porosity w  0.3. Calculate the area which contains 68%
of a point source pollutant, infiltrated 1 km from the divide, after 1 km flow. Also estimate for the point at 2 km
from the divide, the time space during which 68% of the pollutant mass with highest concentrations passes.

ANSWER:
Assume that dispersivity is 10% of travel length, and obtain 2  0.2x2 (Equation 3.76); with x  1000 m,
  447 m. Hence we estimate that 68% of our pollutant is within 2447 and 1553 m (from the divide). We
next want the time period during which the highest concentrations pass, covering 68% of the pollutant mass.
This is the period between the arrival time of (  x) at 2000 m, and the time when (x  ) passes the point.
–––
Now, our estimated   x 0.2, and we find that x  2000 / (1   0.2)  1382 m. We use Equation (3.8)
–––
and find that this distance is attained in 50  ln[1382 / 1000]  16.2 years. Similarly, we find from
x    2000, that x  3618 m, and the time necessary is 64.3 years. Hence the time period for 68% of pol-
lutant mass, with highest concentrations, to pass the 2 km point is 64.3  16.2  48.1 years.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

110 Flow and transport

x0.5
x0.16 x0.84
1
1 6 k0.16
2 7
3 3
4 ki · di 2
c/c 0 k0.5
k0.5 . D 5 5 68%
6 1
7 10
8 4 k0.84
9 9
10 0 8

Figure 3.39. A hydraulic-conductivity stratified aquifer in which the layers contribute successively to tracer
breakthrough (left); ordering the layers according to hydraulic conductivity shows more clearly that travel
distance follows the conductivity distribution (right).

The larger field dispersivity is a result of the variation in arrival time at the observation well, which
means that the hydraulic conductivity distribution in the aquifer lies at the heart of the process.
Mercado (1967; 1984) developed the concept of a conductivity stratified aquifer. An aquifer, con-
sisting of stacked layers with different hydraulic conductivities, is depicted in Figure 3.39. Assume
that flow follows the layers only (transverse dispersion is absent). Tracer from the (fully penetrating)
injection well arrives at the observation well in a stepwise fashion depending on the hydraulic con-
ductivity of the layer. In Figure 3.39 layers 1, 4, 8, 9 and 10 are contributing tracer to the observation
well, and layer 5 has just started to contribute. We may reshuffle the layers, and order them from top
to bottom by increasing hydraulic conductivity (Figure 3.39, right). This orders at the same time the
distance traveled by the chemical according to the conductivity distribution. An average transmis-
sivity is obtained from:

k0.5 D  ∑ ki di (3.77)

where the individual layers have thickness di and conductivity ki, and D is the total thickness of the
aquifer, with average hydraulic conductivity k0.5. It is assumed that the porosity w is equal for all the
layers (cf. Equation 3.14). Now suppose that hydraulic conductivity variations among the layers fol-
low a normal distribution, with standard deviation:
 k  k0.84  k0.5 (3.78)

where k0.84 is the hydraulic conductivity which is larger than 84% of all values. According to Darcy’s
law, the average distance covered by tracer is:

x0.5  k0.5 (dh / dx t / e) (3.79)

while the distance for 84% of the layers is less than

x0.84  k0.84 (dh / dx t / e) (3.80)

If the potential gradient (dh / dx) and the porosity w are equal for all the layers, then

x0.84  x0.5  (k0.84  k0.5 ) x0.5 / k0.5   k x0.5 / k0.5 (3.81)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dispersion 111

The tracer front is spread according to the distances covered in the individual layers, and the spread
is related to the dispersivity:

 x  x0.84  x0.5  2A11 x0.5 (3.82)

Combining (3.81) and (3.82) yields the longitudinal dispersivity as:

2
1  
A11   k  x0.5 (3.83)
2  k0.5 

Recall that x0.5 is the distance covered by the tracer in half of the layers, which is equal to vt.
Equation (3.83) therefore suggests that dispersivity increases linearly with distance, as also was indi-
cated by Figure 3.38. If we assume that the hydraulic conductivity is log-normally distributed we
obtain in a similar way:
A11  1⁄2 (exp[ ln(k ) ]  1)2 x0.5 (3.84)

where ln(k) is the standard deviation of the log-normally distributed conductivity.

It is of interest that the value for 1⁄2(k / k0.5)2 can be around 0.1 (Pickens and Grisak, 1981;
Molz et al., 1983; random sample of Smith, 1981), conform with Equation (3.75). For log-normally
distributed conductivities, the value of ln(k) is 0.4 for the Borden aquifer (Sudicky, 1986) and
0.49 for the Cape Cod aquifer (Garabedian et al., 1991), which give A11  0.1x0.5 and 0.2x0.5,
respectively.
A remarkable point of the conductivity stratified aquifer is that dispersion diminishes again when
flow direction is inverted (a tracer would return in the same step form during pumping as it was
injected in the well). This is different from the diffusion concept, where spreading increases with the
square root of time irrespective of flow inversion. However, the Mercado concept neglects that the
layers have limited extent in the real aquifer and that mixing occurs where a high conductivity layer
ends. Consequently, the increase of the dispersivity with distance becomes smaller and will level off
to a constant value. More intricate models have been developed in which horizontal space variabil-
ity is allowed, as well as transverse dispersion (Gelhar and Axness, 1983; Dagan, 1989). The final,
asymptotic value of the macrodispersivity can be well approximated by (Gelhar, 1997):

A11  ( ln(k ) )2  (3.85)

where (ln(k))2 is the variance of the log-normally distributed conductivities, and  is the correlation
length of the conductivity. The correlation length indicates the continuity of the layers and can
be obtained from spatially distributed data (De Marsily, 1986; Domenico and Schwartz, 1997).
Correlation lengths of 2 m, observed in the Borden and Cape Cod aquifers, lead to values of A11
of 0.3 and 0.6 m, respectively. In these cases, the asymptotic dispersivity is attained already
within a flowlength of 10 m. The Borden and Cape Cod studies are for single aquifers and flow-paths
of a few 100 m. Larger units will have a larger macrodispersivity and a larger correlation length,
and so on and on, which can be introduced in flow models (Elfeki et al., 1999; Rubin and
Bellin, 1998).
Transverse dispersivity has been less studied, but it is smaller than the longitudinal dispersivity.
Measurements from tracer studies indicate that transverse dispersivity is about 10% of the longitudinal
dispersivity in the bedding plane, and about 1% at straight angles, corresponding to transverse hori-
zontal and vertical dispersivity, respectively (Gelhar, 1997). Klenk and Grathwohl (2002) found
transverse vertical dispersivities that were determined mostly by diffusion.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

112 Flow and transport

EXAMPLE 3.15. Longitudinal dispersivity in the Borden aquifer

A point injection of a number of chemicals in an 8 m thick aquifer was followed over two years with a
dense network of depth specific samplers (Mackay et al., 1986). Figure 3.40 shows the depth averaged
Cl-cloud after 462 days, with a midpoint at 41 m from the injection point. Estimate the longitudinal
dispersivity.

ANSWER:
The cross section suggests approximate Gaussian behaviour, and we may apply estimation techniques from
Figure 3.24 in addition to a specific calculation of variance as in Example 3.12. If we take the value of
(0.61 top) as border of the 68% area which encompasses 2, we obtain 2  9 m, and A11  0.25 m.

60
40
20

30 35 40 45 50 55
m. from injection point

60 “top”

40
0.61  “top”
mg/L

68%
20

0
30 35 40 45 50 55
2s ∼ 9 m.
m. from injection point

Figure 3.40. Vertically averaged chloride concentrations in the Borden experiment, 462 days after
injection, and cross section over the plume. (Plume delineation after Sudicky, 1986).

Calculating the variance as in Example 3.12 gives:

∑ xi2ci  ∑ x i ci 
2

2     25 m  2DLt
2
∑ ci  ∑ ci 

where xi is the distance (m) traveled by concentration ci.

With 2  2A11 vt  2A11 41, we obtain A11  0.3 m. This value should be compared with 0.36 m,
given by Freyberg (1986). It is of interest to show here the development of the longitudinal spread 11 2
and
the transverse spread 222
as observed when the plume was transported with groundwater flow. The dashed
curves are calibration fits by Freyberg (1986) (note that the longitudinal variance increases linearly with time,
and hence with travel distance; the asymptotic value of the macrodispersivity is rapidly reached), and the
solid lines indicate model calculations based on hydraulic conductivity variations in the aquifer by Sudicky
(1986).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 113

80 8 2
11
2
22
60 6
11 (m2)

2 (m2)
40 4

22
2

20 2

0 0
0 100 200 300 400 500 600 700
Time (days)

Figure 3.41. Development of longitudinal and transverse variance of the spread of Cl, injected in the
Borden aquifer (Sudicky, 1986).

QUESTION:
The Vejen waste site is 50 m wide and is located 1 km from the Vejen river. Estimate the dilution of waste
leachate.
ANSWER: 20 times (e.g. 1000 mg Cl/L will dilute to 50 mg/L in the seepage zone).

PROBLEMS
3.1. Calculate the downward velocity in chalk, when P  0.3 m/yr, and w  0.1. Compare results with
Figure 3.2.
3.2. An aquifer has thickness D  50 m; P  0.3 m/yr; w  0.3. At 900–1100 m from the divide infiltrates
polluted water in the aquifer, and a private well at 2000 m from the divide may be affected. Calculate the
thickness of the tongue of polluted water, its mean depth, and arrival time at the well point.
3.3. Draw a sketch of the hydrological situation described in Problem 3.2.
a. Calculate pore water velocity and Darcy velocity in the aquifer at 1000 and 2000 m.
b. Calculate the travel time of water between these two points.
c. What is the age of water at d  10, 20, 30, 40 m below the phreatic surface?
d. Compare the result of c. with the formula t  d P / w which is valid for an unsaturated zone (but gives
approximate results for the saturated zone as well).
3.4. Derive the time/distance and time/depth relationships for groundwater in a uniform phreatic aquifer with
radial outward flow (e.g. below a hill).
3.5. Derive the time/distance relationship for groundwater in a uniform phreatic aquifer towards a pumping
well with a radius of influence of r m.
3.6. A municipal well in a 100 m thick phreatic aquifer,   0.3, P  0.3 m/yr, has an influence up to
r  5 km. Calculate the traveltime for water from a factory situated at 1 km from the well.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

114 Flow and transport

3.7. Consider an impermeable clay layer that divides the aquifer in a phreatic part and a confined part.
The flow pattern in the confined part remains intact when the confining layer is strictly impermeable
(Figure 3.42). For the upper, phreatic aquifer the Equations (3.6) to (3.10) apply again. For the confined
part of the aquifer the age distribution with depth is conserved. The flowtime at point x* in the confined
part is (see Figure 3.42 for explanation of symbols):

De w  D  ( x *  x0  x )(D  D* )ew
t  ln  
P Dd D 
* Px0

Redraw the Figure on transparant paper.

a. Draw the hydraulic gradients in the upper and lower aquifer.

x * x
x0* x0
d*
D*

D
d

Figure 3.42. A clay layer separates an aquifer in two parts.

b. Indicate the direction of water flow through the confining layer, when it is leaky.
c. Draw isochrones (lines connecting water with equal age) in the confined aquifer.
3.8. The clay layer in the aquifer of Figure 3.42 starts at 1 km from the divide, and is 1 km long.
D*  2 / 3 D; what is x0? The two aquifers merge again downstream from the clay layer. Compare travel
times for water just above and below the clay layer.
3.9. An aquifer sediment is made of quartz grains, with density 2.65 g/cm3; porosity is 0.2. Calculate the bulk
density of the sediment, and express sorbed concentration of Cd, sCd  2 ppm, as mg/L pore water.
3.10. Bulk density of a limestone is 2.0 g/cm3; calcite, the only mineral in the rock, has density 2.7 g/cm3.
Calculate porosity of the rock.
3.11. The sorption isotherm has the shape shown in the figure.
a. Draw c / x and c / t diagrams when a solution with concentration c2 displaces c1.
b. And also when a solution with c1 displaces c2.

0.8

0.6
q
0.4

0.2

0
0 0.05 0.1 0.15 0.2
c1 c c2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 115

3.12. Calculate the distance covered by diffusion in 10 year in clay, for De  106 and 105 cm2/s.
3.13. A single shot of 10 g. NaCl is injected in a clay, there is no water flow. Calculate the mass of Cl between
x  x0, and x  x0  1 m after 10 years; De  109 m2/s. Hint: Integrate Equation (3.53), and substitute
the error function.
3.14. Derive the formula for the dispersivity in a permeability stratified aquifer with log-normally distributed
permeabilities (Equation 3.84). Hint: ln(k)  ln(k0.84)  ln(k0.5), use x0.5  exp[ln(k0.5)] (dh / dx t)
and similar for x0.84.
3.15. Calculate the value of the dispersion coefficient for a column with dispersivity L  5 mm, for
vH 2O  104, 105, 106 m/s while Df  109 m2/s, porosity is 0.3.
3.16. Give an estimated macro-dispersivity for an aquifer, 50 m thick, k0.5  50 m/day, k  10 m/day. Also for
an aquifer with log-normally distributed k,  2(ln/k)  0.3.
3.17. Estimate the diffusion coefficient of NaCl and Ca(HCO3)2 in water at 10°C, 298  0.891 cP,
283  1.301 cP.

REFERENCES

Abramowitz, M. and Stegun, I.A. 1964. Handbook of mathematical functions, 7th print. Dover, New York, 1064 pp.
Allison, G.B., Gee, G.W. and Tyler, S.W., 1994. Vadose-zone techniques for estimating groundwater recharge in
arid and semiarid regions. Soil Sci. Soc. Am. J. 58, 6–14.
Andersen, L.J. and Sevel, T., 1974. Six years’ environmental tritium profiles in the unsaturated and
saturated zones, Grønhøj, Denmark. In Isotope techniques in groundwater hydrology, vol. 1, 3–18. IAEA,
Vienna.
Appelo, C.A.J., Bäer, T., Smit, P.H.H. and Roelofzen, W., 1991. Flushing a column spiked with HCH (in Dutch).
Report, Free University, Amsterdam.
Atkins, P.W. and de Paula, J., 2002. Atkins’ Physical chemistry, 7th ed. Oxford Univ. Press, 1149 pp.
Bath, A.H. and Edmunds, W.M., 1981. Identification of connate water in interstitial solution of chalk sediment.
Geochim. Cosmochim. Acta 45, 1449–1461.
Bauer, S., Fulda, C. and Schäfer, W., 2001. A multi-tracer study in a shallow aquifer using age dating tracers 3H,
85
Kr, CFC-113 and SF6-indication for retarded transport of CFC-113. J. Hydrol. 248, 14–34.
Bear, J., 1972. Dynamics of fluids in porous media. Elsevier, Amsterdam, 764 pp.
Beekman, H.E., 1991. Ion chromatography of fresh- and seawater intrusion. Ph.D. thesis, Free University,
Amsterdam, 198 pp.
Bird, R.B., Stewart, W.E. and Lightfoot, E.N., 1960. Transport phenomena. Wiley and Sons, New York, 780 pp.
Boas, M.L., 1983. Mathematical methods in the physical sciences, 2nd ed. Wiley and Sons, New York, 793 pp.
Bond, W.J. and Phillips, I.R., 1990. Cation exchange isotherms obtained with batch and miscible displacement
techniques. Soil Sci. Soc. Am. J. 54, 722–728.
Boudreau, B.P., 1997. Diagenetic models and their implementation. Springer, Berlin, 414 pp.
Boving, T.B. and Grathwohl, P., 2001. Tracer diffusion coefficients in sedimentary rocks: correlation to poros-
ity and hydraulic conductivity. J. Contam. Hydrol. 53, 85–100.
Bruggeman, G.A., 1999. Analytical solutions of geohydrological problems. Elsevier, Amsterdam, 959 pp.
Brun, A., Engesgaard, P., Christensen, T.H. and Rosbjerg, D., 2002. Modelling of transport and biogeochemical
processes in pollution plumes: Vejen landfill, Denmark. J. Hydrol. 256, 228–247.
Busenberg, E. and Plummer, L.N., 1992. Use of chlorofluorocarbons (CCl3F and CCl2F2) as hydrologic tracers
and age-dating tools. Water Resour. Res. 28, 2257–2283.
Chiang, W.-H. and Kinzelbach, W., 2001. 3D-Groundwater modeling with PMWIN. Springer, Berlin, 346 pp.
Childs, E.C., 1969. The physical basis of soil water phenomena. Wiley and Sons, New York, 493 pp.
Collon, P., Kutschera, W., Loosli, H.H., Lehmann, B.E., Purtschert, R., Love, A., Sampson, L., Anthony, D.,
Cole, D., Davids, B., Morrissey, D.J., Sherrill, B.M., Steiner, M., Parso, R.C. and Paul, M., 2000. 81Kr in the
Great Artesian Basin, Australia: a new method for dating very old groundwater. Earth Planet. Sci. Lett. 182,
103–113.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

116 Flow and transport

Cook, P.G. and Herczeg, A.L. (eds), 2000. Environmental tracers in subsurface hydrology. Kluwer, Boston, 529 pp.
Cook, P.G. and Solomon, D.K., 1997. Recent advances in dating young groundwater: chlorofluorcarbons,
3 3
H/ He and 85Kr. J. Hydrol. 191, 245–265.
Cook, P.G., Solomon, D.K. and Plummer, L.N., 1995. CFC’s as tracers of groundwater transport processes.
Water Resour. Res. 31, 425–431.
Christensen, T.H., Kjeldsen, P., Bjerg, P.L., Jensen, D.L., Christensen, J.B., Baun, A., Albrechtsen, H.-J. and
Heron, G., 2001. Biogeochemistry of landfill leachate plumes. Appl. Geochem. 16, 659–718.
Dagan, G., 1989. Flow and transport in porous formations. Springer, Berlin, 465 pp.
De Marsily, G., 1986. Quantitative Hydrogeology, Academic Press, London, 440 pp.
Desaulniers, D.E. and Cherry, J.A., 1988. Origin and movement of groundwater and major ions in a thick
deposit of Champlain Sea clay near Montreal. Can. Geotechn. J. 26, 80–89.
Desaulniers, D.E., Kaufmann, R.S., Cherry, J.A. and Bentley, H.W., 1986. 37Cl-35Cl variations in a diffusion-
controlled groundwater system. Geochim. Cosmochim. Acta 50, 1757–1764.
DeVault, D., 1943. The theory of chromatography. J. Am. Chem. Soc. 65, 532–540.
De Vries, J.J., 1995. Seasonal expansion and contraction of stream networks in shallow groundwater systems.
J. Hydrol. 170, 15–26.
Dillon, P.J. 1989. An analytical model of contaminant transport from diffuse sources in saturated porous media.
Water Resour. Res. 25, 1208–1218.
Dogramaci, S.S. and Herczeg, A.L., 2002. Strontium and carbon isotope constraints on carbonate-solution
interactions and inter-aquifer mixing in groundwaters of the semi-arid Murray Basin, Australia. J. Hydrol.
262, 50–67.
Domenico, P.A. and Schwartz, F.W., 1997. Physical and chemical hydrogeology, 2nd ed. Wiley and Sons,
New York, 506 pp.
Eggenkamp, H.G.M., 1994. The geochemistry of chlorine isotopes. Ph.D. thesis, Utrecht, 150 pp.
Eggenkamp, H.G.M., Middelburg, J.J. and Kreulen, R., 1994. Preferential diffusion of 37Cl in sediments of Kau
Bay, Halmahera, Indonesia. Chem. Geol. 116, 317–325.
Ekwurzel, B., Schlosser, P., Smethie, W.M., Plummer, L.N., Busenberg, E., Michel, R.L., Weppernig, R. and
Stute, M., 1994. Dating of shallow groundwater-comparison of the transient tracers 3H/3He, chlorofluoro-
carbons and 85Kr. Water Resour. Res. 30, 1693–1708.
Elfeki, A.M.M., Uffink, G.J.M. and Barends, F.B.J., 1999. Groundwater contaminant transport. Balkema,
Rotterdam, 300 pp.
Engesgaard, P., Jensen, K.H., Molson, J., Frind, E.O. and Olsen, H., 1996. Large-scale dispersion in a sandy
aquifer: Simulation of subsurface transport of environmental tritium. Water Resour. Res. 32, 3253–3266.
Ernst, L.F. 1973. Determining the traveltime of groundwater (in Dutch). ICW-nota 755, Wageningen, 42 pp.
Ernst, L.F. 1978. Drainage of undulating sandy soils with high groundwater tables. J. Hydrol. 39, 1–50.
Felmy, A.R. and Weare, J.H., 1991. Calculation of multicomponent ionic diffusion from zero to high concen-
tration: 1. The system Na-K-Ca-Mg-Cl-SO4-H2O at 25(C. Geochim. Cosmochim. Acta 55, 113–131.
Fetter, C.W., 1994. Applied hydrogeology. Macmillan, London, 691 pp.
Foster, S.S.D. and Smith-Carrington, A., 1980. The interpretation of tritium in the Chalk unsaturated zone.
J. Hydrol. 46, 343–364.
Freeze, R.A. and Cherry, J.A., 1979. Groundwater. Prentice-Hall, Englewood Cliffs, 604 pp.
Freeze, R.A. and Witherspoon, P.A., 1966. Theoretical analysis of regional groundwater flow. I. Analytical and
numerical solutions to the mathematical model. Water Resour. Res. 2, 641–656.
Freyberg, D.L., 1986. A natural gradient experiment on solute transport in a sand aquifer. 2, Spatial moments
and the advection and dispersion of nonreactive tracers. Water Resour. Res. 22, 2031–2046.
Garabedian, S.P., LeBlanc, D.R., Gelhar, L.W., Celia, M.A., 1991. Large-scale natural gradient tracer test in
sand and gravel, Cape Cod, Massachusetts. 2. Analysis of spatial moments for a nonreactive tracer. Water
Resour. Res. 27, 911–924.
Gelhar, L.W., 1997. Perspectives on field-scale application of stochastic subsurface hydrology. In G. Dagan and
S.P. Neuman (eds), Subsurface flow and transport: a stochastic approach, 157–176. Cambridge Univ.
Press.
Gelhar, L.W. and Axness, C.L., 1983. Three-dimensional stochastic analysis of macrodispersion in aquifers.
Water Resour. Res. 19, 161–170.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 117

Gelhar, L.W. and Wilson, J.L., 1974. Ground-water quality modeling. Ground Water 12, 399–408.
Gelhar, L.W., Mantoglou, A., Welty, C. and Rehfeldt, K.R., 1985. A review of field-scale physical solute trans-
port processes in saturated and unsaturated porous media. EPRI, Palo Alto, CA 94303.
Goltz, M.N. and Roberts, P.V., 1987. Using the method of moments to analyze three-dimensional diffusion-
limited solute transport from temporal and spatial perspectives. Water Resour. Res. 23, 1575–1585.
Groen, J., Velstra, J. and Meesters, A.G.C.A., 2000. Salinization processes in paleowaters in Surinam. J. Hydrol.
234, 1–20.
Harvey, C.F. and Gorelick, S.M., 1995. Temporal moment-generating equations: modeling transport and mass-
transfer in heterogeneous aquifers. Water Resour. Res. 21, 159–169.
Hendry, M.J. and Wassenaar, L.I., 2000. Controls on the distribution of major ions in pore waters of a thick sur-
ficial aquitard. Water Resour. Res. 36, 503–513.
Hsieh, P.A., 2001. TopoDrive and ParticleFlow–Two computer models for simulation and visualization of
ground-water flow and transport of fluid particles in two dimensions. U.S. Geological Survey Open-File
Report 01–286, 30 pp.
Hu, Q. and Wang, S.Y., 2003. Aqueous-phase diffusion in unsaturated geologic media: a review. Crit. Rev. Env.
Sci. Technol. 33, 275–297.
Johnson, R.L., Cherry, J.A. and Pankow, J.F., 1989. Diffusive contaminant transport in natural clay: a field
example and implications for clay-lined waste disposal sites. Env. Sci. Technol. 23, 340–349.
Kinzelbach, W., 1992. Numerische Methoden zur Modellierung des Transports von Schadstoffen im
Grundwasser, 2th Auflage. Oldenbourg, München, 343 pp.
Klenk, I.D. and Grathwohl, P., 2002. Transverse vertical dispersion in groundwater and the capillary fringe.
J. Contam. Hydrol. 58, 111–128.
Kreft, A. and Zuber, A., 1978. On the physical meaning of the dispersion equation and its solutions for differ-
ent initial and boundary conditions. Chem. Eng. Sci. 33, 1471–1480.
Lapidus, L. and Amundson, N.R., 1952. Mathematics of adsorption in beds. VI. The effect of longitudinal dif-
fusion in ion-exchange and chromatographic columns. J. Phys. Chem. 56, 984–988.
Lasaga, A.C., 1998. Kinetic theory in the earth sciences. Princeton Univ. Press, 811 pp.
Lerman, A., 1979. Geochemical processes. Wiley and Sons, New York, 481 pp.
Levenspiel, O., 1972. Chemical reaction engineering. Wiley and Sons, New York, 857 pp.
Li, Y.-H. and Gregory, S., 1974. Diffusion of ions in seawater and in deep-sea sediments. Geochim. Cosmochim.
Acta 38, 703–714.
Lyman, W.J., Reehl, W.F. and Rosenblatt, D.H., 1990. Handbook of chemical property estimation methods. Am.
Chem. Soc., Washington.
Lyngkilde, J. and Christensen, T.H., 1992. Redox zones of a landfill leachate pollution plume (Vejen,
Denmark). J. Contam. Hydrol. 10, 273–289.
Mackay, D.M., Freyberg, D.L., Roberts, P.V. and Cherry, J.A., 1986. A natural gradient experiment on solute
transport in a sand aquifer. 1. Approach and overview of plume movement. Water Resour. Res. 22,
2017–2029.
Maloszewski, P. and Zuber, A., 1985. On the theory of tracer experiments in fissured rocks with a porous
matrix. J. Hydrol. 79, 333–358.
McNeil, J.D., 1980. Electrical conductivity of soils and rocks. Geonics Ltd, Missisauga, 22 pp.
Mercado, A., 1967. The spreading pattern of injected water in a permeability stratified aquifer. IASH Pub. 72,
23–36.
Mercado, A., 1984. A note on micro and macrodispersion. Ground Water 22, 790–791.
Molz, F.J., Güven, O. and Melville, J.G., 1983. An examination of scale-dependent dispersion coefficients.
Ground Water 21, 715–725.
Mills, R. and Harris, K.R., 1976. The effect of isotope substitution on diffusion in liquids. Chem. Soc. Rev. 5,
215–231.
Moore, W.J., 1972. Physical chemistry, 5th ed. Longman, London, 977 pp.
Nkedi-Kizza, P.S.C., Rao, P.S.C. and Johnson, J.W., 1983. Adsorption of diuron and 2,4,5-T on soil particle-size
separates. J. Env. Qual. 12, 195–197.
Oster, H., Sonntag, C. and Munnich, K.O., 1996. Groundwater age dating with chlorofluorocarbons. Water
Resour. Res. 32, 2989–3001.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

118 Flow and transport

Patriarche, D., Ledoux, E., Michelot, J.-L., Simon-Coincon, R. and Savoye, S., 2004. Diffusion as the main
process for mass transport in very low water content argillites: 2. Fluid flow and mass transport modeling.
Water Resour. Res. 40, W01517, DOI 10.1029/2003WR002700.
Perkins, T.K. and Johnston, O.C., 1963. A review of diffusion and dispersion in porous media. Soc. Petrol.
Engin. J., march 1963, 70–84.
Phillips, F.M., 1994. Environmental tracers for water movement in desert soils of the American Southwest. Soil
Sci. Soc. Am. J. 58, 15–24.
Pickens, J.F. and Grisak, G.E., 1981. Scale-dependent dispersion in a stratified granular aquifer. Water Resour.
Res. 17, 1191–1211.
Plumb, O.A. and Whitaker, S., 1988. Dispersion in heterogeneous porous media. I. Local volume averaging and
large-scale averaging. Water Resour. Res. 24, 913–926.
Rubin, Y. and Bellin, A., 1998. Conditional simulation of geologic media with evolving scales of heterogeneity.
In G. Sposito (ed.), Scale dependence and scale invariance, 398–420, Cambridge Univ. Press.
Scheidegger, A., Bürgisser, C.S., Borkovec, M., Sticher, H., Meeussen, H. and Van Riemsdijk, W.H., 1994.
Convective transport of acids and bases in porous media. Water Resour. Res. 30, 2937–2944.
Schlosser, P., Stute, M., Sonntag, C. and Münnich, K.O., 1989. Tritiogenic 3He in shallow groundwater. Earth
Planet. Sci. Lett. 94, 245–256.
Schröter, J., 1984. Micro- and macrodispersivities in porous aquifers (in German). Meyniana 36, 1–34.
Schwartzenbach, R.P., Gschwend, P.M. and Imboden, D.M., 1993. Environmental organic chemistry. Wiley and
Sons, New York, 681 pp.
Sillén, L.G., 1951. On filtration through a sorbent layer. IV. The  condition, a simple approach to the theory of
sorption columns. Arkiv Kemi 2, 477–498.
Sjöberg, E.L., Georgala, D. and Rickard, D.T., 1983. Origin of interstitial water compositions in postglacial
black clays (Northeastern Sweden). Chem. Geol. 42, 147–158.
Smith, L., 1981. Spatial variability of flow parameters in a stratified sand. Mathem. Geol. 13, 1–21.
Solomon, D.K., Schiff, S.L., Poreda, R.J. and Clarke, W.B., 1993. A validation of the 3H / 3He method for deter-
mining groundwater recharge. Water Resour. Res. 29, 2951–2962.
Sudicky, E.A., 1986. A natural gradient experiment on solute transport in a sand aquifer: spatial variability of
hydraulic conductivity and its role in the dispersion process. Water Resour. Res. 22, 2069–2082.
Sun, M.-Z., 1999. A finite cell method for simulating the mass transport process in porous media. Water Resour.
Res. 35, 3649–3662.
Tabor, D., 1991. Gases, liquids and solids, 3rd ed. Cambridge Univ. Press, 418 pp.
Taylor, S.R. and Howard, K.W.F., 1987. A field study of scale-dependent dispersion in a sandy aquifer.
J. Hydrol. 90, 11–17.
Toride, N., Leij, F.J. and Van Genuchten, M.T., 1995. The CXTFIT code for estimating transport parameters
from laboratory or field tracer experiments, version 2. US Salinity Lab. Res. Rep. 137. Riverside, Cal.
Tóth, J., 1962. A theory of groundwater motion in small drainage basins in Central Alberta, Canada. J. Geophys.
Res. 67, 4375–4387.
Tyrrell, H.J.V. and Harris, K.R., 1984. Diffusion in liquids. Butterworths, London, 448 pp.
Uffink, G.J., 1988. Modeling of solute transport with the random walk method. In E. Custodio, A. Gurgui and
J.P. Lobo Ferreira (eds), Groundwater flow and quality modelling, 247–265, D. Reidel, Dordrecht.
Valocchi, A.J., 1985. Validity of the local equilibrium assumption for modeling sorbing solute transport through
homogeneous soils. Water Resour. Res. 21, 808–820.
Van Genuchten, M.Th., 1981. Analytical solutions for chemical transport with simultaneous adsorption, zero-
order production and first-order decay. J. Hydrol. 49, 213–233.
Van Genuchten, M.Th. and Parker, J.C., 1984. Boundary conditions for displacement experiments through short
laboratory soil columns. Soil Sci. Soc. Am. J. 48, 703–708.
Vogel, J.C., 1967. Investigation of groundwater flow with radiocarbon. Proc. 1966 Symp., 355–369. IAEA, Vienna.
Volker, A., 1961. Source of brackish ground water in Pleistocene formations beneath the Dutch polderland.
Econ. Geol. 56, 1045–1057.
Yurtsever, Y., 1983. Models for tracer data analysis. In Guidebook on nuclear techniques in hydrology,
381–402, IAEA, Vienna.
Zheng, C. and Bennett, G.D., 2002. Applied contaminant transport modeling, 2nd ed. Wiley and Sons, New
York, 621 pp.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

4

Minerals and Water

The solubility of a mineral constrains the maximal concentration of its components in water. For
example, the mineral fluorite has the composition CaF2 and by doing solubility calculations we can
predict concentration ranges of Ca2 and F in groundwater. In the first part of this chapter we reca-
pitulate the basics of equilibrium chemistry and demonstrate the applicability in hydrogeochemistry.
Minerals present in aquifers are often not pure phases but rather mixtures or solid solutions of
different minerals. We will discuss the theory behind solid solutions and analyze the interrelations
between the composition of the aqueous solution and the solid solution.
Some minerals react fast upon contact with water. This is particularly the case for the more
“soluble” minerals such as gypsum (CaSO4 2H2O), halite (NaCl), fluorite as well as most carbon-
ate minerals, and equilibrium will be attained within a timespan that is short compared to the resi-
dence time of groundwater. Other minerals, typically silicates, react so sluggishly that equilibrium is
never attained at low temperatures and therefore reaction kinetics have to be considered.
The general strategy of this chapter is to explore first the basic principles in manual calculations.
Subsequently follows a demonstration of similar calculations with a geochemical model. Present-
day geochemical models are extremely powerful tools to analyze complex natural systems.
Throughout this chapter, and in fact throughout the remainder of this book the code PHREEQC
(Parkhurst and Appelo, 1999) will be used as an integrated part of the presentation.

4.1 EQUILIBRIA AND THE SOLUBILITY OF MINERALS

Fundamental to any description of equilibria in water is the law of mass action. It states that for a
reaction of the generalized type:

aA  bB ↔ cC  dD
the distribution at equilibrium of the species at the left and right side of the reaction is given by:

[C]c [D]d
K  (4.0)
[A]a [ B]b

Here K is the equilibrium constant and the bracketed quantities denote activities or “effective con-
centrations”. For the present we will equate activities with molal concentrations, but return to the dif-
ference between the two in Section 4.2. The law of mass action is applicable to any type of reaction,
the dissolution of minerals, the formation of complexes between dissolved species, the dissolution
of gases in water, etc. For dissolution of the mineral fluorite (CaF2) we may write:

CaF2 ↔ Ca 2  2F

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

120 Minerals and water

At equilibrium the aqueous concentrations obey the solubility product:

Kfluorite  [Ca 2 ][F ]2  1010.57 at 25C (4.1)

The solubility product is a direct application of the law of mass action, except that [CaF2] is omitted
in (4.1) since by definition the activity of a pure solid is equal to one. The exact value of Kfluorite is
somewhat uncertain (Nordstrom and Jenne, 1977). Here we use the value of Handa (1975) in order
to be consistent with the following example. Relation (4.1) can be rewritten in logarithmic form as:

log Kfluorite  log[Ca 2 ]  2 log[F ]  10.57 (4.2)

On a logarithmic plot, the equilibrium condition (4.2) between fluorite and the solution is a straight
line, as illustrated in Figure 4.1. All combinations of [Ca2] and [F] that plot below the line are sub-
saturated with respect to fluorite, while those that plot above the line are supersaturated for fluorite.
Together with the fluorite solubility line in Figure 4.1, a number of groundwater analyses from
Rajasthan, India are plotted and clearly, the equilibrium with fluorite places an upper limit to the F
(and Ca2) concentrations in the groundwater.
Although F at low concentration in drinking water has been considered as beneficial in candy bar
consuming countries, and in fact is added to drinking water in many water supplies where fluoride is
absent in groundwater, it constitutes a health hazard at concentrations above 3 ppm, causing dental
fluorosis (tooth mottling) and more seriously, skeletal fluorosis (bone deformation and painful
brittle joints in older people). High F concentrations in groundwater are found in many places of the
world, but notably in Asia and Africa. Fluoride is basically brought into the groundwater by leaching
from minerals in the rocks (Handa, 1975; Jacks et al., 1993; Saxena and Ahmed, 2003).
Volcanic rocks may have high natural fluor contents, but also the dissolution of fossil shark teeth, con-
taining fluorapatite, has been reported as a source (Zack, 1980). In arid areas evaporation may
strongly increase the F concentration in water.

log Ca2
3.5 3.0 2.5

10

Su 3.5
per
5 sat
A B ura
tion
F(ppm)

log F

2 C 4.0
Su
bs
atu
rati
on
1

4.5
10 20 30 100
Ca2(ppm)

Figure 4.1. The stability of fluorite and the saturation of groundwaters from Sirohi, W. Rajasthan, India (mod-
ified from Handa, 1975). The evolution in water chemistry upon addition of gypsum is described by the pathway
A, B to C as discussed in the text.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Equilibria and solubility of minerals 121

If equilibrium with fluorite imposes a constraint on the groundwater composition then water with a
high natural Ca2 concentration will generally contain a low F concentration, while groundwater
with a low Ca2 concentration will be rich in F. This seems substantiated by the field observations
of Handa (1975; Figure 4.1). The fluorite equilibrium control on the F concentration can be
exploited to remove F by water treatment. At saturation for fluorite, an increase of [Ca2] will,
according to Equation (4.1), decrease [F]. To bring about an increase in [Ca2] it has been proposed
(Schuiling, pers. comm.; Jacks et al., 2000) to add gypsum (CaSO4 2H2O). For gypsum dissolution
we may write, neglecting the crystal water for simplicity:

CaSO 4 ↔ Ca 2  SO 24

With the solubility product:

Kgypsum  [Ca 2 ][SO 24 ]  104.60 at 25C (4.3)

Note that gypsum is much more soluble than fluorite. The amount of gypsum which has to be added
to a water sample with composition A (Figure 4.1), in order to reach equilibrium with fluorite, can
be calculated following Example 4.1.

EXAMPLE 4.1. Gypsum addition to high fluor groundwater

A groundwater sample contains 10 ppm Ca2 and 5.5 ppm F. Is this water saturated with respect
to fluorite, and if not, how much gypsum should be added in order to reach saturation for
fluorite?
First we recalculate ppm to molal concentrations, which yields [Ca2]  103.60 and [F]  103.54.
A fast answer is obtained by plotting these values directly into Figure 4.1 which results in point A.
Alternatively we may calculate the product:

[Ca 2 ][F ]2  [103.60 ][103.54 ]2  1010.68

The value obtained is slightly lower than the solubility product for fluorite (Equation 4.2) and the sample is
subsaturated for fluorite.
When gypsum is added, the water composition will change from point A (Figure 4.1) parallel to the Ca-
axis until it meets the fluorite saturation line at point B. From the fluorite solubility product (Equation 4.2)
we find that:

[Ca 2 ]B  Kfluorite / [F ]2  [1010.57 ] / [103.54 ]2  103.49

Thus to reach equilibrium with fluorite, the amount of gypsum to be added is equal to
[Ca2]B  [Ca2]A  103.49  103.60  0.072 mmol/L.

Once equilibrium with fluorite is reached, the addition of more gypsum will modify the water com-
position following the fluorite solubility line in Figure 4.1. The change in water chemistry is described
quantitatively by Equation (4.2) and results in a decreasing F concentration due to the precipitation
of fluorite. However, the dissolution of gypsum cannot continue indefinitely since at some point also
saturation for gypsum is reached. In Figure 4.1, point C indicates the simultaneous saturation for
fluorite and gypsum and its location can be calculated as follows.
When the composition of the solution changes from B to C, the dissolution of gypsum and pre-
cipitation of fluorite occur simultaneously. We may define:
x  mol gypsum dissolved per liter water
y  mol fluorite precipitated per liter water

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

122 Minerals and water

Using these two variables we write the mass balance equations for the concentrations of Ca2 and
F at point C as:

[Ca 2 ]C  [Ca 2 ]B  x  y (4.4)

[F ]C  [F ]B  2y (4.5)

[SO 24 ]C  [SO 24 ]B  x (4.6)

Together with the solubility products for fluorite (Equation 4.1) and gypsum (Equation 4.3), this
yields 5 equations with 5 unknowns which can be solved. However, the resulting quadratic equation
is rather awkward and a more simple method is to identify in the mass balance equations a parame-
ter which can be neglected. From Figure 4.1 we can surmise that when the water composition
changes from B to C, the amount of Ca2 released by gypsum dissolution will be much larger than
the amount of Ca2 precipitated as fluorite. We introduce this as an simplifying assumption, and
check its validity afterwards:
y  x

which reduces Equation (4.4) to:

[Ca 2 ]C  [Ca 2 ]B  x (4.7)

Substitution of this relation, together with Equation (4.6), in the solubility product for gypsum
(Equation 4.3) results in:

Kgypsum  ([Ca 2 ]B  x ) ([SO 24 ]B  x )

Rearranging yields:

x 2  ([Ca 2 ]B  [SO 24 ]B ) x  ([Ca 2 ]B [SO 24 ]B  K gypsum )  0

And substitution of known values (from Example 4.1):

x 2  103.49 x  104.60  0

This is a neat standard equation that solves to:

x  4.84103  102.32
Accordingly, 4.8 mmol gypsum/L must dissolve to reach saturation with both gypsum and fluorite.
Substituting x in Equation (4.7) produces:

[Ca 2 ]C  102.29

Which again is substituted in the solubility product of fluorite (Equation 4.2):

[F ]C  (Kfluorite / [Ca 2 ]C )0.5  104.14

This corresponds to a fluoride concentration of 1.4 ppm, and gypsum addition has therefore lowered
the fluoride content of the water by 75%. A side effect of the addition of gypsum as water treatment
technique is, however, a high sulfate concentration and in our example it amounts to 468 ppm, which
exceeds the drinking water limit (Table 1.1) and may cause intestinal problems and diarrhea.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Corrections for solubility calculations 123

It remains to check the validity of the initial assumption by substitution in Equation (4.5):
y  1⁄2 ([F ]B  [F ]C )  1.08104

Since y is about fifty times smaller than x, our initial assumption, y  x, is quite reasonable. This type
of simplification is often very useful in chemical calculations. Any reasonable assumption can be
made, as long as it is tested afterwards. Neglection of terms is only allowed in mass balance equations
(which are summations), never in mass action equations (which are multiplications or divisions).
At point C there exists simultaneous equilibrium between fluorite and gypsum, and we can write
the reaction:

SO 24  CaF2 ↔ CaSO 4  2F (4.8)

The corresponding equilibrium constant is derived by combining the solubility products of fluorite
and gypsum:

[F ]2 [Ca 2 ][F ]2 K 1010.57

K    fluorite   105.97 (4.9)
[SO 24 ] [Ca 2 ][SO 24 ] Kgypsum 104.60

Equation (4.9) demonstrates that at simultaneous equilibrium with gypsum and fluorite, the ratio of
the squared F over SO24 concentration becomes invariant. If some other process, for example the
oxidation of pyrite (FeS2), increases the SO2
4 concentration, while equilibrium for both fluorite and
gypsum is maintained, then the F concentration also increases due to the conversion of fluorite to
gypsum (Equation 4.8).
In summary, we have seen in this section how to work with solubility products of minerals
and how they may constrain the composition of groundwater. Furthermore, we have calculated a
pathway in the evolution of water chemistry as a function of the dissolution and precipitation of
two minerals.

4.2 CORRECTIONS FOR SOLUBILITY CALCULATIONS

In the foregoing we have used the molal concentration of a solute to calculate mineral solubility.
However, the law of mass action is only valid for the activity of ions, which is the measured total con-
centration corrected for the effects of electrostatic shielding and for the presence of aqueous com-
plexes, as discussed in the next two sections.

4.2.1 Concentration and activity

The activity in the law of mass action is a measure for the effective concentration of the species. The
effective concentration can be thought of as indicating how, for example, a Ca2 ion would behave
when there are no interactions with other ions in solution, i.e. at infinite dilution. In thermodynamics
the activity of solutes, gases, components in solid solutions and adsorbed ions are all expressed as a
fraction relative to a standard state (Table 4.1) and as a fraction, the activity is always dimensionless.
For gases the standard state is a pure gas phase at 1 atm. Atmospheric air contains 21% O2 and the
activity, or partial pressure, of O2 in air is therefore 0.21 and is dimensionless! Similarly, the stan-
dard state for a solid is a pure solid, and for an ion exchanger the standard state is an exchanger filled
with a single ion only. For aqueous solutes the standard state is defined as an ideal solution with
solute concentration of 1 mol/kg H2O  1 molal. In this context “ideal” means, somewhat crypti-
cally, a 1 M solution behaving like at infinite dilution.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

124 Minerals and water

Table 4.1. Conventions for standard states and activities.

Concentration measure Symbol Standard state

Gases Gas pressure (atm)/1 atm Pi / P0 P 0  1 atm

Solid and liquid mixtures Mole-fraction  1
Aqueous solutes Molality/(1 molal) mi / m0 m0  1 mol/kg H2O
Exchangeable ions Equivalent- or Mole-fraction  or M   1, M  1

The activity is related to the molal concentration by an activity coefficient which corrects for non-
ideal behavior. For aqueous solutes, the relation is:

[i]  i mi / mi0  i mi (4.10)

where: [i] is the activity of ion i (dimensionless), i is the activity coefficient (dimensionless),
mi is the molality (mol/kg H2O), mi0 is the standard state, i.e. 1 mol/kg H2O. The factor 1/mi0 is
unity for all species and cancels in the practical enumeration of Equation (4.10) but causes the
activity to become dimensionless. Activity coefficients may vary, but if ion i is present at trace
concentration, and there are no other ions present, then i → 1. In this book the activity of
an ion is denoted by square brackets, but in other texts activity is indicated by braces {i}, or simply
as ai.
Activity coefficients for solutes are calculated using the Debye-Hückel theory. In this theory,
first the ionic strength, I, is defined which describes the number of electrical charges in the solution
(Atkins and de Paula, 2002):

I  1⁄2 ∑( mi /mi0 zi2 )  1⁄2 ∑ mi zi2 (4.11)

where zi is the charge number of ion i, and mi is the molality of i. Similarly to the definition of
activity, the ionic strength becomes dimensionless by division with the standard state mi0 i.e.
1 mol/kg H2O. The ionic strength of freshwater is normally less then 0.02 while seawater has an
ionic strength of about 0.7. A highly saline environment like the Dead Sea has an ionic strength
of 9.4.
For dilute electrolyte solutions, I  0.1, the Debye-Hückel equation describes the electrostatic
interactions as:

Azi2 I
log i   (4.12)
1  Ba˚i I

where A and B are temperature dependent constants; at 25°C A  0.5085 and B  0.32851010/m.
Langmuir (1997) lists the dependency of A and B on the temperature. However, in the temperature
range of most groundwater, 5–35°C, the variation in A and B remains small.
The empirical ion-size parameter åi (Table 4.2) is a measure of the effective diameter of
the hydrated ion. The value of å, multiplied by 1010/m, (roughly) indicates the number of water
molecules surrounding the ion. The smaller å is, the closer can oppositely charged ions approach
the ion and shield it, thereby lowering its activity coefficient. The symbols in Figure 4.2 depict
activity coefficients of various ions as a function of ionic strength as calculated with the
Debye-Hückel equation. Activity coefficients are seen to decrease with I, strongest for
the divalent ions. At an ionic strength of 0.1,  is about 0.8 for monovalent ions and 0.4 for
divalent ions.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Corrections for solubility calculations 125

º in the Debye-Hückel Equation (4.12). Numbers are given in

Table 4.2. The ion-size parameter, a,
Ångstrom (1010 m) (Garrels and Christ, 1965).

å/(1010 m) Ion

2.5 Rb, Cs, NH  

4 , Tl , Ag

    
3.0 K , Cl , Br , I , NO 3
3.5 OH, F, HS, BrO   
3 , IO4 , MnO4
   
4.0–4.5 Na , HCO 3 , H2PO 4 , HSO3 , Hg 2 , SO42, SeO 2
2 2 2 3
4 , CrO 4 , HPO 4 , PO 4
4.5 Pb2, CO 32, SO 32, MoO42
5.0 Sr 2, Ba2, Ra2, Cd2, Hg2, S2, WO 42
6 Li, Ca2, Cu2, Zn2, Sn2, Mn2, Fe2, Ni2, Co2
8 Mg 2, Be2
9 H, Al3, Cr3, trivalent rare earths
11 Th4, Zr 4, Ce4, Sn4

Various equations have been proposed to derive activity coefficients at ionic strength values higher
than 0.1 (Nordstrom and Munoz, 1994; Langmuir, 1997). Truesdell and Jones (1973), and subse-
quently Parkhurst (1990), fitted a modified version of the Debye-Hückel equation to activity coeffi-
cient data for chloride solutions:

Azi2 I
log i    bi I (4.13)
1  Bai I

where A and B are the temperature dependent coefficients from the Debye-Hückel Equation
(4.12), and ai and bi are ion-specific fit parameters (Table 4.3). Note that ai differs
from the ion-size parameter å; for Mg2 ai  5.51010 m while å  81010 m.

1.3
H H
1.2 Na
1.1 Na K, CI, NO
3
2
1.0 SO4
Ca2
0.9
0.8
0.7 Ca2
i

0.6
K, CI
0.5
0.4
0.3
0.2
0.1 NO
3
2
0 SO4

0.001 0.002 0.005 0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10

I

Figure 4.2. Activity coefficients for some common ions as a function of ionic strength. Symbols represent
calculations with the Debye-Hückel Equation (4.12). Lines correspond to the Truesdell and Jones Equation
(4.13). From Garrels and Christ (1965).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

126 Minerals and water

Table 4.3. The individual ion activity coefficient parameters in the Truesdell and
Jones (1973) Equation (4.13). Values for other ions can be found in Parkhurst (1990).

Ion a / (1010 m) b

Na 4.0 0.075

K 3.5 0.015
Mg2 5.5 0.20
Ca2 5.0 0.165
Cl 3.5 0.015
HCO  3 5.4 0.0
CO 2
3 5.4 0.0
SO 42 5.0 0.04

The lines in Figure 4.2 correspond to the Truesdell and Jones equation. At low ionic strength they agree
well with the values obtained with the Debye-Hückel equation. At high ionic strength the activity coef-
ficients of cations are seen to increase by the action of the term (bi I). The Truesdell and Jones equa-
tion is a reasonable approximation up to I values of about 2 in dominantly chloride solutions (Parkhurst,
1990). The Davies equation is another relation often used to calculate activity coefficients and is appli-
cable up to an ionic strength of about 0.5.

 I 
log i  Azi2   0.3I  (4.14)
1 I 

where A is the same temperature dependent coefficient as in Equation (4.12).

The net effect of the aqueous activity coefficient is that the solubility of minerals increases with ionic
strength up to I  0.7. In more saline solutions the solubility of minerals decreases again as activity
coefficients increase, an effect known as salting out. Salting out already starts at low ionic strength for
neutral species, including gases, which have bi  0.1 (at least in PHREEQC). Another point to note is
that the mole fraction of H2O decreases markedly in a saline solution, which results in a decreased vapor
pressure. Garrels and Christ (1965) have given formulas to calculate the activity of H2O in saline water.

EXAMPLE 4.2. Calculate ionic strength and ion activity coefficients

Let’s calculate the solubility of fluorite in a solution with 10 mmol NaCl/L. First calculate the ionic strength:

I  1⁄2 [mNa  (1)2  mCl  (1)2  mCa 2  ( 2)2  mF  (1)2 ]

We expect that only small amounts of fluorite dissolve and neglect mCa2 and mF in the equation.
Thus I  0.01.
Now calculate the activity coefficient for Ca2 using (4.14):

 0.01 
log Ca 2  0.5085( 2)2   0.3(0.01)   0.179
 1  0.01 

Ca2  0.66
So that:
Likewise for F
 0.01 
log F  0.5085(1)2   0.3(0.01)   0.0447
 1  0.01 

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Corrections for solubility calculations 127

which gives:
F  0.90

Note that the activity coefficient of the monovalent ion, F, is much higher than of the divalent ion, Ca2.
The solubility product of fluorite is given by Equation (4.1):

Kfluorite  [Ca 2][F]2  1010.57 at 25C (4.15)

Substituting (4.10) in (4.15) produces:

Ca 2 mCa 2 2 m 2  Kfluorite (4.16)

F F

Rearranging yields:

1
mCa 2 m 2  K (4.17)
F Ca 2 2 fluorite
F

And substituting activity coefficients

1
mCa 2 m 2  K  1.87 Kfluorite (4.18)
F (0.66)(0.90)2 fluorite

In distilled water the concentrations of calcium and fluoride in equilibrium with fluorite are so low
that the ionic strength approaches zero and the activity coefficients unity. However, in a 10 mmol NaCl/L
solution our calculation shows that the solubility of fluorite has increased by almost a factor two. In these
calculations, the contribution of Ca2 and F concentrations to I were neglected; their inclusion would give
an additional (very) slight increase of the solubility of fluorite.

QUESTIONS:
Obtain the concentrations of Ca2 and F in 10 mmol NaCl/L? Hint: use the mass balance on solutes that
dissolve from CaF2: mF  2 mCa2.
ANSWER: (0.5 mF)(mF)2  1.871010.57 solves to mF  0.465 mmol/L, mCa2
 0.233 mmol/L
Find the contribution of Ca2 and F to the ionic strength in 0.01 M NaCl?
ANSWER: I  1⁄2[mNa  (1)2  mCl  (1)2  mCa2  (2)2  mF  (1)2]
 0.0107
And calculate now the correct ’s and concentrations of Ca2 and F?
ANSWER: Ca2  0.65, F  0.90, mF  0.467 mmol/L, mCa2  0.233 mmol/L
How much does the solubility of fluorite increase in 0.1 M NaCl (cf. Figure 4.2)?
ANSWER: roughly 1 / (0.40.82)  4 times

4.2.2 Aqueous complexes

Ions in aqueous solution may become attached to each other as ion pairs or aqueous complexes. Examples
are major cation complexes like CaSO04, CaF or CaOH, but also heavy metals complexes exist like
CdCl, HgCl 
3 or PbOH and metal complexes with an organic ligand like oxalate, for example CuC2O4.
The complexes may be outer sphere, with water molecules present in between the constituent ions of the
complex, or inner sphere when a covalent bond is formed and water molecules that surrounded the ions
have been expelled. The inner/outer sphere state is gradual and fluctuates in time (Burgess, 1988).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

128 Minerals and water

Complexation will lower the activity of the “free” ion in water, thereby increasing the solubility of
minerals and also the mobility of trace metals. Grassi and Netti (2000) found that the Hg concentra-
tion in groundwater in a coastal aquifer increased with the Cl concentration, apparently due to
complexing of Hg2 and Cl resulting in the solubilization of Hg from the rock. Complexes may
have detoxifying effects, and heavy metal poisoning can be abated by ingesting a strong complexer
such as EDTA (ethylene-diamine-tetra-acetate). Complexation finds important application in ana-
lytical chemistry and separation technology and in the production of food and medicines.
The majority of complexes consist of a metal cation, surrounded by a number of ligands
(Burgess, 1988; Stumm and Morgan, 1996). For example, Al3 in water is surrounded by 6 water
molecules which together form the octahedrally shaped complex Al(OH2) 63. Depending on the pH,
protons are released from the complex and its charge increases:

Al(OH 2 )36 ↔ AlOH(OH 2 )52  H

This complex is usually written simplified as AlOH2.

The ligand may act as a bridge between two metal ions to give a binuclear complex.
For example, the chromate anion, CrO2 2
4 , is present in dilute acid as dichromate, Cr2O7 , in the
form of two tetrahedra which share a corner oxygen. Polynuclear complexes are formed by
Si(OH)04 or Al(OH)3(OH2)3 in polymers that can be a precursor for a solid. If a ligand has various
positions to bind the same cation, then it forms so-called multidentate complexes. Thus, F is a
monodentate ligand, while oxalate, C2O2 4 , usually is a bidentate ligand with two oxygens where
the cation is bound. EDTA is a hexadentate ligand, complexing both on N and O. Complexes
with multidentate ligands are called chelates. The chelates can form rings which encapsulate the
metal cations completely and these are important carriers of trace elements in biological systems
(Buffle, 1990).
The formation of aqueous complexes can be described by equilibria of the type:

Ca 2  SO 24 ↔ CaSO04 (4.19)

The distribution of the species is obtained by applying the law of mass action (4.0):

[CaSO04 ]
K   102.5 (4.20)
[Ca 2 ][SO 24 ]

The reaction is written as an ion-association reaction and the corresponding mass action constant
is therefore termed a stability constant. Sometimes equilibria for aqueous complexes are written as
dissociation reactions and the dissociation constant for the CaSO04 complex is obviously the inverse
of Equation (4.20). Clearly, these two formulations should not be confused.

4.2.3 Combined complexes and activity corrections

Both aqueous complexing and activity coefficients will lower the activity of the free ions, and their
combined effect must be calculated. The total amount of Ca2 in solution is given by a mass balance
equation comprising the various aqueous complexes of Ca2, given in molal units:
 Ca  mCa 2  mCaF  mCaSO0  mCaOH 
(4.21)
4

Similar mass balance equations can be written for all other substances in solution. In order to calcu-
late the activities of, for example [Ca2] and [SO2
4 ], we have to solve simultaneously a set of mass
balance equations, with concentrations in molal units, and a set of mass action equations like (4.20),
with concentrations expressed as activities. The calculation proceeds through an iterative procedure.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Corrections for solubility calculations 129

100
Uncomplexed
90 Activity

80
Percentage of total concentration
70

60

50

40

30

20

10

0
Na Mg2 Ca2 AI3 CI SO2
4 HCO
3
Species

Figure 4.3. The importance of complexing and activity corrections as percentage of total concentrations for
35‰ seawater with a pH of 8.22.

From the analytical data, first the activity coefficients and ion activities are calculated, which are
used to calculate activities of complexes. Then the molal concentrations of complexes are obtained
which enables us to calculate improved mass balances and thereby a better estimate of the ionic
strength. This procedure continues until no further significant improvement is obtained.
The importance of complexing and activity corrections, expressed as percentage of the total con-
centration, is illustrated for seawater in Figure 4.3. For Na and Cl the effect of complexing is small
while activity coefficients make the ultimate activity 30–40% lower than the total concentration. For
the divalent ions Ca2 and Mg2, about 10% of the total concentration is complexed, while the activ-
ity of Ca2 and Mg2 with the activity coefficient correction goes down to 20–30% of the total con-

centration. For SO24 and HCO3 , aqueous complexes constitute 40–50% of the total concentration, and
2
the activity of the free SO4 ion is about 10% of the total concentration.
The methods described here work well for electrolyte solutions of up to about seawater ionic
strength (I  0.7), or up to I  2 mol/L when HCO 2
3 and SO4 concentrations are low (Parkhurst,
1990). For more concentrated solutions, as are found in association with evaporites, other methods
are available. Particularly so-called Pitzer equations (Pitzer, 1981; Harvie et al., 1982; Monnin and
Schott, 1984; Plummer et al., 1988; Millero and Pierrot, 1998) have been successful in describing
mineral equilibria in highly concentrated solutions.

EXAMPLE 4.3. Solubility of gypsum

The solubility of gypsum in water, considering both aqueous complexes and activity corrections, can be cal-
culated as follows. For equilibrium between gypsum and water we write:

CaSO 4 2H 2O ↔ Ca 2  SO 24  2H 2O

and the mass action expression is:

[Ca 2][SO 24][H 2O]2

K   104.6
[CaSO 4 2H 2O]

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

130 Minerals and Water

Since the activity of a pure solid like gypsum is unity and also for dilute electrolyte solutions [H2O]  1,
this simplifies to the solubility product:

Kgypsum  [Ca 2][SO 24]  104.60 at 25C

Substituting molal concentrations for activities yields:

Kgypsum  ( 2 m 2 )(SO2 mSO2 )  104.60

Ca Ca 4 4

Since equal amounts of Ca and SO42 are released during gypsum dissolution, mCa2  mSO42 and accord-
2

ing to Equation (4.14), also Ca2  SO 2, the solubility of gypsum is simplified to
4

( mCa 2 )2  104.60 / ( Ca 2 )2

or

mCa 2  (5.01  103 ) / Ca 2

Ca can be estimated with the Davies Equation (4.14) and the ionic strength, I:
2

I  1⁄2 ∑ mi zi2  1⁄2 (mCa 2  4  mSO2  4)  4mCa 2

4

The problem is now that mCa2 depends on Ca2 which depends on I and again on mCa2. It may be solved in
an iterative procedure. In first approximation Ca2 is set equal to 1, which enables a first estimate of I and
so forth. In tabulated form the calculation proceeds as:

Iteration mCa2 I Ca2

0 1
1 →5.01103 →0.0201 →0.58
2 →8.71103 →0.0348 →0.50
3 →9.97103 →0.0399 →0.49
4 →10.3103 →0.0413 →0.48
5 →10.4103 →0.0417 →0.48
6 →10.4103

Note that the application of activity corrections has doubled the calculated solubility of gypsum. However,
we also know that the aqueous complex CaSO04 is of importance. The stability constant of this complex
(Equation 4.20) can be rewritten as:

[CaSO04 ]  102.5  [Ca 2][SO 24]

Substitution of the solubility product of gypsum gives:

[CaSO04 ]  102.5104.60  102.10  7.94103

Since for uncharged species the activity coefficient is close to unity, this is also the molal concentration of
the complex. Thus, the total solubility of gypsum is 10.4103  7.94103  18.3103 mol/L. Note
that 40% of the solubility is accounted for by the complex CaSO04, while 30% is due to activity corrections.
These results are easily transcribed in grams of gypsum dissolved per liter by multiplication with
the molecular weight of gypsum (172.1). The solubility product alone predicts that (5.01103)
(172.1)  0.86 gram gypsum can dissolve per liter water, while including complexes and activity correc-
tions leads to a solubility of (18.3103)(172.1)  3.15 g/L.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Corrections for solubility calculations 131

4.2.4 Calculation of saturation states

Once we are able to calculate the activities of free ions in solution, we can also calculate the state of
saturation of a groundwater sample with respect to minerals. One way to do this has already been
shown in Figure 4.1 where the analytical data, or more correctly the ion activities, are plotted in a
stability diagram for fluorite. Another approach is to compare the solubility product K with the
analogue product of the activities derived from water analyses. The latter is often termed the Ion
Activity Product (IAP). For example for gypsum:
Kgypsum  [Ca 2 ][SO 24 ] (activities at equilibrium)

and

IAPgypsum  [Ca 2 ][SO 24 ] (activities in the water sample)

An example is shown in Figure 4.4. Here the oxidation of pyrite (FeS2) in the upper part of the pro-
file produces large amounts of sulfate and equilibrium with gypsum appears to place an upper limit
to dissolved calcium and sulfate concentrations.

log IAP
7.5 6.5 5.5 4.5
0 Ox. zone
Red. zone

2
Depth (m)

Gypsum

Figure 4.4. The IAPgypsum compared with the solubility product at the site of pyrite oxidation in pore waters of
swamp sediments (Postma, 1983).

Saturation conditions may also be expressed as the ratio between IAP and K, or the saturation state :
  IAP /K (4.22)
Thus for   1 there is equilibrium,  ! 1 indicates supersaturation and   1 subsaturation. For larger
deviations from equilibrium, a logarithmic scale can be useful given by the saturation index SI:
SI  log (IAP /K ) (4.23)

For SI  0, there is equilibrium between the mineral and the solution; SI  0 reflects subsaturation,
and SI ! 0 supersaturation. For comparison, it can be useful to normalize  and SI to the number of
solutes  in the ion activity product. Thus,   2 for CaCO3 (calcite) and   4 for CaMg(CO3)2
(dolomite). The saturation state then becomes 1/ and the saturation index is SI/.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

132 Minerals and water

Different types of information can be obtained from saturation data. In few cases an actual control of
solutes by equilibrium with a mineral can be demonstrated as clearly as in Figures 4.1 and 4.4. Often
there is no equilibrium and then the saturation state merely indicates in which direction the processes
may go; for subsaturation dissolution is expected, and supersaturation suggests precipitation.

4.3 MASS ACTION CONSTANTS AND THERMODYNAMICS

Thermodynamics is the science concerned with energy distributions among substances in a system.
It offers an impressive framework of formulas which can be derived from a few basic laws. An intro-
duction to thermodynamics can be found elsewhere (Lewis and Randall, 1961; Denbigh, 1971;
Nordstrom and Munoz, 1994; Anderson and Crerar, 1993). Here we confine ourselves to some prac-
tical aspects and in particular to the calculation of mass action constants and their dependency on
temperature from thermodynamic tabulations.

4.3.1 The calculation of mass action constants

For the general reaction:
aA  bB ↔ cC  dD
we may write
[C]c [D]d
Gr  Gr0  RT ln (4.24)
[A]a [B]b

where Gr is the change in Gibbs free energy (kJ/mol) of the reaction, Gr0 is the standard Gibbs
free energy of the reaction and equal to Gr when each product or reactant is present at unit activity
(so that the log term becomes zero) at a specified standard state (25°C and 1 atm), [i] denotes the
activity of i, R is the gas constant (8.314103 kJ/mol/deg), T is the absolute temperature, Kelvin
(Kelvin  °C  273.15).
In older texts, energy is often expressed in kcal/mol which can be converted to kJ/mol by multiply-
ing with 4.184 J/cal (hence R  1.987103 kcal/mol/deg). The prefix  is used because energy can be
measured only as relative amounts. The direction in which the reaction will proceed is indicated by Gr:
Gr ! 0 the reaction proceeds to the left;
Gr  0 the reaction is at equilibrium;
Gr  0 the reaction proceeds to the right.
Therefore, in the case of equilibrium, Equation (4.24) reduces to:

[C]c [D]d
Gr0  RT ln (4.25)
[A]a [B]b

Note that the activity product in the last term is equal to the mass action constant K (Equation 4.0):

Gr0  RT ln K (4.26)

Back substitution of Equation (4.26) in (4.24) results in

[C]c [D]d
Gr  RT ln K  RT ln (4.27)
[A]a [B]b

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Mass action constants and thermodynamics 133

In Equation (4.27) the distance from equilibrium is expressed in terms of the mass action
constant and the solution composition, a formulation analoguous to the saturation index SI
(Equation 4.23).
Equation (4.26) has the practical application that it allows us to calculate the mass action
constant for any reaction from tabulated data of Gf0 for dissolved substances, minerals, and
gases. Gf0 is the free energy of formation, i.e. the energy needed to produce one mole of a
substance from pure elements in their most stable form. The latter (and the H ion) have by defini-
tion zero values. Tabulations are normally given for 25°C and 1 atm pressure. Gr0 is calculated
from:

Gr0  ∑ G 0f products  ∑ G 0f reactants

A number of compilations of thermodynamic data concerning mineral-water systems are available

(Helgeson et al., 1978; Robie et al., 1978; Wagman et al., 1982; Cox et al., 1989; Woods and Garrels,
1987, Nordstrom and Munoz, 1994, etc.). Preferably, consistent sets of data, as presented in the first
four references, should be used as otherwise erratic values can be obtained. For example, the listed
G 0f calcite could have been calculated from solubility measurements using values of G 0fCO32 and
G 0f Ca2 by the reverse procedure of Example 4.4. When Kcalcite subsequently is calculated from
thermodynamic tables with a different G f0 CO32 value, the result will clearly be erroneous. Even
when these precautions are taken you will discover that different compilations may give slightly dif-
ferent constants. For example, the value of Kcalcite, calculated in Example 4.4, is slightly higher than
the currently accepted value of 108.48 (Plummer and Busenberg, 1982) which includes corrections
for aqueous complexes.

EXAMPLE 4.4. Calculation of solubility products from Gibbs free energy data
Calculate the solubility product of calcite from the Gibbs free energies of formation at 25°C (Wagman et al.,
1982):

Gf0CaCO  1128.8 kJ/mol

3

Gf0Ca 2  553.6 kJ/m

mol
Gf0CO2  527.8 kJ/mol
3

ANSWER:
For the reaction CaCO3 ↔ Ca2  CO32
we obtain
Gr0  Gf0Ca 2  Gf0CO2  Gf0CaCO
3 3

Gr0  553.6  527.8  1128.8  47.4 kJ/mol

Gr0  RT ln K
47.4  8.314103  298.15  2.303 log K  5.708 log K
loog K  47.4/ 5.708  8.30

4.3.2 Calculation of mass action constants at different temperature

In nature, groundwater is generally not found at the standard conditions of 25°C and 1 atm pressure.
While variations in pressure have little effect on the values of the mass action constants, the temperature
variations are important. Variations of mass action constants with temperature are usually calculated

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

134 Minerals and water

with the Van’t Hoff equation:

d ln K H r (4.28)

dT RT 2
where Hr is the reaction enthalpy, or the heat lost or gained by the chemical system. For exother-
mal reactions Hr is negative (the system loses energy and heats up) and for endothermal reactions
Hr is positive (the system cools). The calculation of Hr and the definition of the standard state is
analogous to that of Gr0. At 25°C, the value of the reaction enthalphy, H r0, is calculated from the
formation enthalphies, H f0, which are listed in thermodynamic tables. Equation (4.28) shows that
K increases with temperature for positive H r0 (the slope of ln K with temperature is positive) and
decreases with temperature for negative H r0. Usually, H r0 is constant within the range of a few
tenths of degrees and therefore (4.28) can be integrated to give for two temperatures:

H r0  1 1
log KT  log KT     (4.29)
1 2 2.303R  T1 T2 

This equation is generally applicable in groundwater environments.

EXAMPLE 4.5. Temperature dependency of the solubility product

Calculate the solubility product of calcite at 10°C from the following enthalphies of formation (Wagman
et al., 1982):
H f0CaCO   1206.9 kJ/mol
3

H f0Ca 2  542.8 kJ/moll

H f0CO2  677.1 kJ/mol
3

ANSWER:
For the reaction CaCO3 ↔ Ca2  CO32
we may write,

H r0  542.8  (  677.1)  (  1206.9)  13.0 kJ/mol

which means that the reaction is exothermal: the system heats up when calcite dissolves. The difference in
log K between 25°C and 10°C, according to Equation (4.29) is:
H r0  1 1 
log K 25  log K10    
2.303 R  298.15 283.15 
13.0  1 1 
    
2.303  8.31410  298.15
3 283.15 
 0.12

In EXAMPLE 4.4 it was calculated that log Kcalcite at 25°C is 8.30, so that at 10°C Kcalcite 
8.30  0.12  8.18. Thus for an exothermal reaction the solubility increases with decreasing tempera-
ture and vice versa for an endothermal reaction.

QUESTION:
Calculate the temperature change when 3 mmol calcite dissolve in 1 L water? (Hint: to heat 1 L water from
15 to 16°C needs 1 kcal)
ANSWER: temperature increases by 0.01°C

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Equilibrium calculations with PHREEQC 135

4.4 EQUILIBRIUM CALCULATIONS WITH PHREEQC

Manual calculations are useful to gain a basic understanding of the underlying principles. Once this
basic understanding is reached the calculation of ion activities and saturation states becomes tedious
and can be left to the computer. Here we use the code PHREEQC (Parkhurst and Appelo, 1999)
which can be downloaded as freeware and has become the standard for doing a variety of hydrogeo-
chemical calculations. Appendix A contains a quick reference guide in the form of a number of get-
going sheets which explain the basic features of the code. A full description of many alternatives for
input and the mathematical backgrounds can be found in the manual of the program by Parkhurst
and Appelo (1999). Acquisition and installation of the program is explained in the get-going sheet
Getting started. PHREEQC was developed for calculating “real world” hydrogeochemistry and is a
powerful tool for modeling data. It is also instructive to try things out and to gain an understanding
how concentrations are affected by chemical processes and transport. One thing is important, how-
ever. The results of geochemical computer programs should always be inspected critically since the
reactions are often intricate and the results may depend on factors like the quality of the database or
the sequence of the calculations.

4.4.1 Speciation calculations using PHREEQC

The input file of PHREEQC is organized in KEYWORDS and associated data blocks. To do a spe-
ciation calculation, we use the keyword SOLUTION followed by the water composition, and
PHREEQC then calculates the ion activities and the saturation states for relevant minerals,
cf. Example 4.6.

EXAMPLE 4.6. Calculate the speciation of a water analysis using PHREEQC

We construct an input file for the Windows version of PHREEQC. The keyword SOLUTION
m-n is used to define the composition of the solution and analysis A, used in Example 4.1, is entered (m and
n are integers to indicate that the composition is for a range of solutions). Text following a “#” in the input
file is explanatory and is not read by the program. The keyword END is the signal for PHREEQC to start
the calculations.

SOLUTION 1 Analysis A # keyword, solution number and name

-units ppm # default units are mmol/kg H2O
Ca 10
F 5.5
END # keyword

Run the file by pressing the calculator icon, or AltC, S. Quickly, the program reports “Done” and if
Enter is pressed, the Output tab appears where the results of the calculations are recorded. The solution
composition is listed as molality and as number of moles of Ca and F. In this case the two are equal
since PHREEQC by default assumes that the solution consist of 1 kg H2O (but it may be defined otherwise
in the input file). Then follows the “Description of solution” which gives default values for pH, pe and
temperature (they were not entered under keyword SOLUTION), and calculated values for ionic
strength, alkalinity and electrical balance. Note that the solution is not well balanced. Next follows
the “Distribution of species”, first for the water species and then for the elements in alphabetical order.
The last section gives the saturation indices. The solution is subsaturated with respect to fluorite, as was
calculated in Example 4.1, but the SI values are not the same. This is because PHREEQC considers
corrections for activity and complexing and also because the log K value in the PHREEQC database is
slightly different.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

136 Minerals and water

----------------------------------Solution composition--------------------------------

Elements Molality Moles

Ca 2.495e-04 2.495e-04
F 2.895e-04 2.895e-04

-----------------------------------Description of solution---------------------------

pH = 7.000
pe = 4.000
Activity of water = 1.000
Ionic strength = 6.427e-04
Mass of water (kg) = 1.000e+00
Total alkalinity (eq/kg) = -4.154e-08
Total carbon (mol/kg) = 0.000e+00
Total CO2 (mol/kg) = 0.000e+00
Temperature (deg C) = 25.000
Electrical balance (eq) = 2.095e-04
Percent error, 100*(Cat-|An|)/(Cat+|An|) = 26.61
Iterations = 3
Total H = 1.110124e+02
Total O = 5.550622e+01

----------------------------------Distribution of species----------------------------

Log Log Log

Species Molality Activity Molality Activity Gamma

OH- 1.030e-07 1.001e-07 -6.987 -7.000 -0.012

H+ 1.028e-07 1.000e-07 -6.988 -7.000 -0.012
H2O 5.551e+01 1.000e+00 -0.000 -0.000 0.000
Ca 2.495e-04
Ca+2 2.489e-04 2.221e-04 -3.604 -3.653 -0.049
CaF+ 5.590e-07 5.431e-07 -6.253 -6.265 -0.012
CaOH+ 3.794e-10 3.687e-10 -9.421 -9.433 -0.012
F 2.895e-04
F- 2.889e-04 2.807e-04 -3.539 -3.552 -0.012
CaF+ 5.590e-07 5.431e-07 -6.253 -6.265 -0.012
HF 4.209e-08 4.210e-08 -7.376 -7.376 0.000
HF2- 4.667e-11 4.534e-11 -10.331 -10.343 -0.012
H2F2 4.618e-15 4.619e-15 -14.336 -14.335 0.000
H(0) 1.416e-25
H2 7.078e-26 7.079e-26 -25.150 -25.150 0.000
O(0) 0.000e+00
O2 0.000e+00 0.000e+00 -42.080 -42.080 0.000

----------------------------------Saturation indices---------------------------------

Phase SI log IAP log KT

Fluorite -0.16 -10.76 -10.60 CaF2

H2(g) -21.96 -22.00 -0.04 H2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Equilibrium calculations with PHREEQC 137

H2O(g) -1.59 -0.00 1.59 H2O

O2(g) -39.19 44.00 83.19 O2
Portlandite -12.45 10.35 22.80 Ca(OH)2

4.4.2 The PHREEQC database

The database of PHREEQC contains the definitions of chemical species, complexes, mineral solu-
bilities etc. and can be viewed by clicking on the Database tab. The database is structured by a
number of keywords, each followed by chemical definitions and constants needed to do the calcu-
lations. The first keyword is SOLUTION_MASTER_SPECIES and defines the elements in solution
(Table 4.4).
The first column gives the element names, like Ca, Na and Alkalinity, which are used to enter a
concentration under keyword SOLUTION. These names must start with a capital letter. Elements
with several redox levels may be repeated with the number of electrons lost, for example Fe, Fe(2)
and Fe(3). When the redox level of an element is specified in the input file, PHREEQC will main-
tain this during the initial speciation calculation regardless of the selected pe. The pe  log[e]
and indicates the general redox level of a solution. However, when “total” Fe is input, the distribu-
tion over Fe2 and Fe3 will be calculated from the pe. Always keep a watchful eye on the redox
species and see if they remain at the expected redox levels during the calculations and consult
Chapter 9 for more details on redox calculations.
The second column lists the aqueous species that are used primarily in the speciation calcula-
tions. The third column gives the contribution of that species to the alkalinity and requires
explanation. Operationally, the alkalinity is defined as the amount of acid that is needed to
bring the pH down to about 4.5 (cf. Chapter 5). For most waters, the alkalinity equals
Alk  mOH  mHCO3  2mCO32, but PHREEQC considers also all the other species that consume
protons. Thus, in the output of Example 4.6, the Total Alkalinity is negative, 4.154108 (eq/kg),
due to the presence of 4.209108 M HF. The contribution of F (the aqueous master species) to the
alkalinity is 0, and therefore HF is considered to be dissociated at pH  4.5 (the approximate end-
point of an alkalinity titration), producing an equivalent amount of protons which contribute nega-
tively to the alkalinity.
The next two columns give the chemical formula that is used to convert grams into moles when
concentrations are entered in mg/L or ppm, and the atomic weight of the element.

Table 4.4. PHREEQC elements in the database.

SOLUTION_MASTER_SPECIES
#
#element species alk gfw_formula element_gfw
#
Ca Ca+2 0.0 Ca 40.08
Mg Mg+2 0.0 Mg 24.312
Na Na+ 0.0 Na 22.9898
Fe Fe+2 0.0 Fe 55.847
Fe(+2) Fe+2 0.0 Fe
Fe(+3) Fe+3 -2.0 Fe
S SO4-2 0.0 SO4 32.064
S(6) SO4-2 0.0 SO4
S(-2) HS- 1.0 S
# etc...

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

138 Minerals and water

QUESTIONS:
Find the dominant species of F and Fe(3) at pH 4.5?
ANSWER: The input file SOLUTION 1; pH 4.5; F 1e-3; Fe(3) 1e-3;
END shows: 95% F and 90% Fe(OH) 2
Why has Fe(3) an alkalinity contribution of 2, cf. Table 4.4?
ANSWER: Fe(3) is dominantly Fe(OH) 2 at pH 4.5
If you would duly consider the hydroxy complexes of Fe3, what should be the alkalinity contribution?
ANSWER: 0.9  2 (for Fe(OH) 2 )  0.09  1 (for FeOH )  1.89
2

Under keyword SOLUTION_SPECIES (Table 4.5) the various aqueous complexes are tabulated in
the form of an association reaction, with log_k, the association constant. The definition holds for the
species right after the equal sign. The first reactions in Table 4.5 are identity reactions that define the
primary master species (as entered in the second column of SOLUTION_MASTER_SPECIES) and
with log_k  0. Changes of log_k with temperature are calculated from the reaction enthalphy
delta_h (Section 4.3.2) or with an analytical expression. The different options to calculate the activ-
ity coefficient from the ionic strength are controlled by the parameter gamma. For major ions the
database PHREEQC.DAT employs the Truesdell-Jones Equation (4.13), the two parameters a and b
follow gamma. For minor ions the Debye-Hückel Equation (4.12) is used. The ion-size parameter
å is given, in Ångstrom, as the first parameter following gamma, while the second parameter is set
to zero. If the line gamma is absent, then the Davies Equation (4.14) is used. The options are illus-
trated in Table 4.5. For neutral species, like uncharged complexes and gases, PHREEQC uses log
  0.1 I.

Table 4.5. Definition of species and complexes in solution, and parameters for activity corrections
in the PHREEQC.DAT database.

SOLUTION_SPECIES
H+ = H+
log_k 0.0
-gamma 9.0 0.0 # Debye-Hückel Equation (4.12)
Ca+2 = Ca+2
log_k 0.0
-gamma 5.0 0.1650 # Truesdell-Jones Equation (4.13)
Pb+2 = Pb+2
log_k 0.0
# Davies Equation (4.14)
Ca+2 + SO4-2 = CaSO4
log_k 2.3 # uncharged complex
delta_h 1.650 kcal
# etc...

QUESTIONS:
Change the activity coefficient for Pb2 from the Davies to the Debye Hückel equation (find å in
Table 4.2).
Find SI for cerrusite (PbCO3) using the two –gamma’s for a solution with C 1, Pb 1e3, Na 500, Cl
500 mM, pH 7?
ANSWER: SI  1.00 and 1.02 for Davies and Debye-Hückel, respectively
(a tiny difference!)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Equilibrium calculations with PHREEQC 139

Repeat the calculation for strontianite (SrCO3) for 1 M Sr2 in 1.5 M NaCl, and compare with the
Truesdell-Jones equation?
ANSWER: SI  3.61 (Davies), 4.23 (Debye-Hückel), 4.02 (Truesdell-Jones)

In the PHREEQC database, the keyword PHASES (Table 4.6) lists minerals and gases for which
saturation indices are calculated. Note that the equations for minerals are written as dissociation
reactions, therefore log_k is for the dissociation reaction. The variation of log_k with temperature is
again calculated from the reaction enthalphy (delta_h, section 4.3.2), or from an analytical expres-
sion of the type:

log K  A1  A2 T  A3 / T  A4 log T  A5 / T 2 (4.30)

where T is temperature in Kelvin and the numbers A1…5 are fit parameters. In the input file (Table 4.6),
the numbers A1…5 follow the indentifier – analytic. If an analytical expression is present, it overrides
the reaction enthalpy method.

Table 4.6. Definition of minerals and their solubility in the PHREEQC.DAT database.

PHASES
Fluorite
CaF2 = Ca+2 + 2 F-
log_k -10.6
delta_h 4.69 kcal
-analytic 66.348 0.0 -4298.2 -25.271
# etc...

It is important to emphasize that the results obtained by the speciation calculation never are better
than the quality of the analytical input and the constants used. Analytical sources of error have
already been discussed in Chapter 1 and particularly errors in pH measurements may affect the
results of saturation calculations significantly. For example, the reaction for equilibrium between
gibbsite (Al(OH)3) and water is :

3H  Al(OH)3 ↔ Al3  3H 2O

log(IAP )  log[Al3 ]  3pH

An error in the pH measurement of about 0.33 unit, which is not uncommon, would affect the
log(IAP) by a whole unit! Also for carbonate equilibria (see Chapter 5), errors in the pH may affect
the results of saturation calculations seriously and the uncertainty introduced by such errors should
be evaluated carefully in each individual case.
The mass action constants which are used in the programs should be internally consistent. For
example, the solubility product of gypsum listed in the database is probably derived from experiments
during which gypsum has been equilibrated with water. In the calculations to derive Kgypsum, a stability
constant for the aqueous complex CaSO40 has been used. Obviously, the same stability constant should be
used in all subsequent calculations involving Kgypsum. For major species, good quality databases, like
PHREEQC.DAT, are available. However, for trace components the user should maintain a healthy scep-
ticism concerning the quality and internal consistency of the data. In addition to PHREEQC.DAT there
are several other databases available in the PHREEQC package and they are listed in Table 4.7.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

140 Minerals and water

Table 4.7. Databases distributed with the PHREEQC program.

Database Features

PHREEQC.DAT Limited but most consistent database

WATEQ4F.DAT The PHREEQC.DAT database extended with many heavy metals
MINTEQ.DAT The database developed for the US EPA MINTEQ program (Allison et al., 1991).
Includes some organic compounds
LLNL.DAT A huge database containing many elements and with a large temperature range.
Developed for the program EQ3/6 (Wolery, 1992).

You can change the database with keyword DATABASE (this must be the first keyword in the
input file).
The critical use of the available databases and their effect on the calculations is illustrated in
Example 4.7.

EXAMPLE 4.7. Solubility of quartz at 150°C

Calculate the solubility of Quartz at 150°C, using the three databases PHREEQC.DAT, MINTEQ.DAT and
LLNL.DAT in PHREEQC. The log K’s and reaction enthalpies are for the reaction:

SiO 2(qu)  2H 2O ↔ H 4SiO04

log K25°C Reaction enthalpy Database

3.98 5.99 kcal/mol PHREEQC.DAT

4.006 6.22 kcal/mol MINTEQ.DAT
3.9993 32.949 kJ/mol LLNL.DAT

The reaction enthalpy in LLNL.DAT of 32.949 kJ/mol corresponds to 32.949/4.184  7.88 kcal/mol and is
quite different from the two other databases. We manually calculate the log K150°C using Equation (4.29),
and compare the results with those from PHREEQC using the following input file, run separately with each
of the three databases.
DATABASE phreeqc.dat # also with llnl.dat or minteq.dat
SOLUTION 1
temp 150
Si 1 Quartz # the Si concentration is adapted to equilibrium
# with quartz
END

The answer is:

Van ’t Hoff log K150°C PHREEQC log K150 Database
2.68 2.66 (Polynomial) PHREEQC.DAT
2.66 2.66 (Van ’t Hoff) MINTEQ.DAT
2.29 2.71 (Polynomial) LLNL.DAT
Databases LLNL.DAT and PHREEQC.DAT contain a polynomial expression (4.30) for temperature
dependency which overrides the reaction enthalphy approach. The log K has increased at 150°C, corre-
sponding to the endothermal reaction. The polynomial from PHREEQC.DAT and the Van ’t Hoff formula
with constant enthalpy used in MINTEQ.DAT yield the same log K over this temperature range and are close
to the value calculated manually. The polynomial expression from the LLNL.DAT also gives nearly the
same value. Apparently the reaction enthalpy listed in LLNL.DAT is incorrect.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Equilibrium calculations with PHREEQC 141

4.4.3 Mineral equilibration

Besides for calculating the saturation state of a solution with respect to different minerals,
PHREEQC can also be used to bring the solution in equilibrium with a specified mineral. This fea-
ture is illustrated for the fluorite/gypsum experiment of Figure 4.1, by calculating the water compo-
sitions for point B, equilibrium for fluorite by addition of gypsum, and point C, equilibrium with
both fluorite and gypsum (Example 4.8).

EXAMPLE 4.8. Equilibrate a water sample with minerals

Modify the input file from Example 4.6 and use keyword EQUILIBRIUM_PHASES to obtain equilibrium,
first with fluorite by adding gypsum (point B), then with gypsum and fluorite to arrive at point C. Keyword
END separates the simulations. The results of a simulation may be stored in temporary memory with key-
word SAVE solution no, and can be recalled again with USE solution no in a later simulation (no is a num-
ber or a range of numbers m–n).

SOLUTION 1 Analysis A
-units ppm
Ca 10
F 5.5
END # end first simulation

# Calculate point B
USE solution 1 # use already stored solution

EQUILIBRIUM_PHASES 1 # equilibrate with minerals

Fluorite 0 Gypsum # Mineral name, SI, reactant
SAVE solution 2 # save solution (Point B)
END # end second simulation

# Calculate point C
USE solution 2

EQUILIBRIUM_PHASES 2
Fluorite 0 0 # SI = 0, initially 0 moles
Gypsum # default SI = 0, 10 moles
END

Note that equilibrium can be reached by adding a reactant. The reactant can be a phase (defined with key-
word PHASES) or a chemical formula (for example Fluorite 0 CaSO4 instead of Fluorite 0 Gypsum in the
second simulation). If the reactant is omitted, the mineral itself reacts until the requested SI is attained, or
until it is exhausted.
An excerpt from the output of the last simulation is copied below, showing that 0.1 mmoles of fluorite
have precipitated, while 15.4 mmoles of gypsum have dissolved (‘Delta’ is positive for fluorite, negative for
gypsum). The fluoride concentration has decreased markedly to 0.087 mmol/L  19 g/mol  1.6 mg/L,
which is almost the same as we calculated by hand. However, the concentration of 15.6 mmol
SO42/L  96 g/mol  1500 mg SO42/L is much higher than was obtained before and will lead to serious
intestinal problems that make the water unfit for consumption.

----------------------------------Phase assemblage-----------------------------------

Moles in assemblage
Phase SI log IAP log KT Initial Final Delta

Fluorite 0.00 -10.60 -10.60 0.000e+00 1.011e-04 1.011e-04

Gypsum 0.00 -4.58 -4.58 1.000e+01 9.985e+00 -1.544e-02

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

142 Minerals and water

----------------------------------Solution composition-------------------------------

Elements Molality Moles

Ca 1.572e-02 1.573e-02
F 8.731e-05 8.736e-05
S 1.557e-02 1.558e-02

QUESTION:
Use EQUILIBRIUM_PHASES to find the log K’s for quartz at 150°C in Example 4.7?
ANSWER: SOLUTION 1; temp 150; EQUILIBRIUM_PHASES 1; Quarts; END

4.5 SOLID SOLUTIONS

Up to now the minerals have been considered to be pure phases. However, the analysis of
naturally occurring minerals by microanalytical devices, like the microprobe, shows that pure
minerals are the exception rather than the rule. Minerals of variable composition may be considered
as solid solutions of pure end-member minerals. Detrital minerals, originally formed at high tem-
peratures, such as plagioclases, amphiboles, pyroxenes, etc. are solid solutions. Also low tempera-
ture phases like carbonates form extensively solid solutions, the best studied example being
Mg-calcite which we treat in more detail in Chapter 5. Numerous other cases are known such
as replacement of Ca2 by Cd2, Zn2, Cu2, Co2, Fe2 and Mn2 in calcite (Rimstidt et al., 1998),
Ca2 in rhodochrosite (MnCO3), Mn2 in siderite (FeCO3) and Sr 2 in aragonite (CaCO3).
Additional examples are substitution of F for OH in apatite (Ca5OH(PO4)3), as happens in tooth
enamel, and substitution of Al3 in goethite (FeOOH). An example of the latter is illustrated in
Figure 4.5.

4.5.1 Basic theory

Substitution of Cd 2 in CaCO3 can be used as an example of the binary solid solution
between otavite (CdCO3) and calcite (CaCO3). Equilibrium between the aqueous solution
and the solid solution is defined by two mass action equations which must be fulfilled
simultaneously:
[Ca 2 ][CO32 ]
CaCO3 ↔ Ca 2  CO32 K cc   108.48 (4.31)
[CaCO3 ]
calcite

and
[Cd 2 ][CO32 ]
CdCO3 ↔ Cd 2  CO32 K ot   1012.1 (4.32)
[CdCO3 ]
otavite

Here [CaCO3] and [CdCO3] are the activities of the two components in the solid phase. For a pure
solid phase the activity is equal to one, and usually omitted in the mass action expression.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Solid solutions 143

Strongly acid soils

Moderately acid soils

Ferruginous bauxites

Saprolites
Scale
10

Number of samples
Mottles

Geodes

0
Ferricretes

0 5 10 15 20 25 30
AI – Substitution in goethite (Mole %)
Figure 4.5. The aluminum content of goethites in various soil environments (Fitzpatrick and Schwertmann, 1982).

However for a solid solution the solid phase activity is related to the mole fraction (Table 4.1) by a
relation like:
[CdCO3 ]  CdCO CdCO (4.33)
3 3

where  is the activity coefficient that corrects for non-ideal behavior and  is the mole fraction. In
the case of ideal solid solution,  is equal to one and:
 molCd 
[CdCO3 ]     
 molCa  molCd ss

and also:
[CaCO3 ]  (1  )

Subtraction of reaction (4.32) from (4.31) results in:

Cd 2  CaCO3 ↔ CdCO3  Ca 2 (4.34)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

144 Minerals and water

with the mass action expression:

[CdCO3 ] [Ca 2 ] K
K   cc  103.62 (4.35)
[CaCO3 ] [Cd 2 ] K ot

What is the effect of calcite and otavite being present as a solid solution instead of separate miner-
als? Consider the following two situations. In the first situation a solution is in equilibrium with cal-
cite and otavite as separate minerals. Then in Equation (4.35) [CaCO3]  1 and [CdCO3]  1 and
therefore the ratio [Ca2] / [Cd2] in solution is fixed to 103.62  4169. The concentration of Cd2
is much smaller than of Ca2 because the solubility of otavite is much smaller than of calcite.
In the second situation Cd2 is incorporated as an ideal solid solution in the calcite structure,
which can be written as CdCa(1)CO3. Let us take Cd  [CdCO3]  0.01 and [CaCO3]  0.99 so
that [CdCO3] / [CaCO3]  0.01 / 0.99
1 / 100. Equation (4.35) predicts then that the [Ca2] / [Cd2]
ratio in solution must be 100 times higher than for separate minerals. If the Ca2 concentration
remains the same as before, then the Cd2 concentration must now be a 100 fold lower, cf. Example
4.9. The formation of a solid solution may thus lower the aqueous concentration of the minor com-
ponent. Therefore, solid solutions are important for scavenging heavy metals from water and for lim-
iting the mobility of heavy metals in the environment (Tesoriero and Pankow, 1996).

EXAMPLE 4.9. Ideal solid solution of Cd 2 in calcite

Use PHREEQC to calculate the concentrations of Ca2 and Cd2 in equilibrium with the two minerals
calcite and otavite, and compare with a solid solution of the two minerals with Cd  0.01. Also use keyword
USER_PRINT to print the ratio of the activity of Ca2 and Cd2 in the output file.

ANSWER:
SOLUTION 1
pH 7.5 Calcite # pH is adjusted to equilibrium with calcite
C(4) 4.0
Ca 1.0
Cd 1 ug/kgw Otavite # Cd conc. is adjusted to equilibrium with otavite

SOLID_SOLUTION 1
CaCdCO3 # Name of solid solution
-comp calcite 99 # Ideal solid solution, calcite = 99 mol
-comp Otavite 1 # otavite 1 mol, x_Cd = 1 / (1+99) = 0.01

USER_PRINT
-start
10 print "aCa =", act("Ca+2"), " aCd =", act("Cd+2"), " aCa/aCd =", act("Ca+2")/act ("Cd+2")
-end
END

gives in the output file:

-------------------------------------User print--------------------------------------

aCa = 7.3946e-04 aCd = 1.7732e-07 aCa/aCd = 4.1703e+03 (initial solution)

aCa = 7.3837e-04 aCd = 1.7884e-09 aCa/aCd = 4.1286e+05 (equilibrated solution)

First, SOLUTION 1 is defined with a fixed concentration of 1 mmol Ca2/kg H2O (the default units)
and equilibrated with calcite by adjusting the pH. The initial Cd2 concentration of 1 g/kg is adapted to equi-
librium with otavite. Subsequently, the solution is equilibrated with SOLID_SOLUTION 1 with Cd  0.01.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Solid solutions 145

In PHREEQC, an ideal solid solution can be multicomponent, and have as many components “-comp …” as
relevant. The equilibration will exchange mass among solid and water, but in this example it will not affect the
composition of the solid solution, because the number of moles present in the solid is much larger than in the
solution. The ratio [Ca2] / [Cd2]  4.17103 in the initial solution equilibrated with the separate minerals
(first line in User print), and it increases 100 fold to 4.13105 when equilibrated with the solid solution.
Keyword USER_PRINT calls the BASIC interpreter that has been programmed in PHREEQC and per-
mits to customize printout of model variables. BASIC is a computer language that executes instructions
from numbered lines in the order of the numbers. In line 10, act(“Ca2”), gives the activity of the Ca2
species. The BASIC commands are listed in the PHREEQC manual under keyword RATES, and can be used
in the various USER_… keywords and for defining rates of kinetic reactions.

QUESTIONS:
If Cd  0.001 in the solid solution, what will [Ca2]/[Cd2] become?
ANSWER: 4.13106
Find the ratio mCa/mCd in Example 4.9?
ANSWER: 3.95103 and 3.91105, smaller because CdHCO 3 is a stronger complex than
CaHCO 3 : total Cd increases.
Find the ratio mCa/mCd in Example 4.9, with 0.1 M Cl present.
ANSWER: 6.29102 and 6.22104, even smaller because of strong CdCl complexes

Commonly, solid solutions show deviations from ideal behavior and the activity coefficients  become
a function of the excess free-energy of mixing. The free energy of mixing can be obtained from solu-
bility measurements (Busenberg and Plummer, 1989), from miscibility gaps, variations in fractionation
factors or other, non-ideal thermodynamic properties (Glynn, 2000). For a binary solid solution the
excess free-energy of mixing, G E, can be modeled with the Guggenheim series expansion (Glynn,
1991; Guggenheim, 1967):

G E  (12 ) RT (a0  a1 (1  2 )  a2 (1  2 )2  … ) (4.36)

where a0, a1, … are empirical coefficients (dimensionless), and 1 and 2 are the fractions of the two
components in the mixture (2  1  1). When all the coefficients are zero, G E  0 and the solid
solution is ideal. When a0 is non-zero, and the other coefficients are zero, the solid solution is clas-
sified as regular. When a0 and a1 are non-zero, the solid solution is termed subregular. Glynn and
Reardon (1990, 1992) observed that a0 and a1 are mostly sufficient to relate the activity coefficients
to the mole fraction (cf. Glynn, 2000).
The activity coefficients are found from the excess energy with the Gibbs-Duhem relation, and
are in the subregular model given by:

ln 1  22 (a0  a1 (41  1)) (4.37a)

and

ln 2  12 (a0  a1 (42  1)) (4.37b)

For the solid solution (Cd, Ca)CO3 Tesoriero and Pankow (1996) found that a0  1.21 and a1
0.
If Cd is small, like 0.01, Cd  exp((Ca)2  1.21)
exp(0.98  1.21)  3.27 and Ca 
exp((Cd)2  1.21)
exp(0.0001  1.21)  1. This illustrates that the main component in
the solid solution behaves close to ideal as long as  ! 0.9. Since Cd/Ca  3.27 it follows from

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

146 Minerals and water

Equation (4.33) that the ratio [CdCO3] / [CaCO3] increases compared to ideal solid solution.
Equation (4.35) then requires that the ratio [Ca2] / [Cd2] decreases (Example 4.10).

EXAMPLE 4.10. Non-ideal solid solution of Cd 2 in calcite

PHREEQC can model regular and subregular binary (two-component) solid solutions. Non-ideal behavior
can be defined with the Guggenheim parameters a0 and a1 (Equation 4.36), or from various thermodynamic
properties. We extend Example 4.9 with a regular solid solution. Simply add the following lines after the
END in Example 4.9:
USE solution 1
SOLID_SOLUTION 2
CaCdCO3
-comp1 Calcite 99 # Non-Ideal solid solution, component 1
-comp2 Otavite 1 # component 2
-Gugg_nondim 1.21 0 # a0 = 1.21, a1 = 0, Tesoriero and Pankow, 1996.
END

ANSWER:
-------------------------------------User print--------------------------------------
aCa = 7.3838e-04 aCd = 5.8541e-09aCa/aCd = 1.2613e+05

The ratio [Ca2]/[Cd2] is now 1.26105, which is exp(1.21  0.992)  3.27 times smaller than was
calculated in Example 4.9.

For solid solution of Sr2 in calcite, Tesoriero and Pankow (1996) obtained a0  5.7. The larger
value for Sr2 is related to the larger ionic radius of Sr2 (1.18 Å) than of Ca2 (1.0 Å), while Ca2
and Cd2 (0.95 Å) have a nearly similar radius. Although the solubility of SrCO3 is smaller than of
CaCO3 and therefore a fractionation factor greater than 1 is expected, Sr2 is rejected from the cal-
cite structure, as is indicated by the very large value of a0, resulting in a high activity coefficient.
However, Tesoriero and Pankow used the solubility constant of strontianite for estimating SrCO3
which has an orthorhombic crystal structure, while calcite is rhombohedral and this does influence
the thermodynamics of the system (Böttcher, 1997). For most salts, a0 ranges from 1 to 5 (Glynn,
2000), resulting in activity coefficients in solid solutions that are greater than 1. The activity coeffi-
cients can become so high that the solid solution becomes unstable with respect to the end-members
and falls apart. Figure 4.6 compares the energy levels of an ideal solid solution, a regular solid solu-
tion with a0  3, and a mechanical mixture of the two end-members.
For a mechanical mixture, where component 2 is the minor component with mole fraction , the
energy (kJ/mol) is:
Gmm  G f , comp2  (1  )G f ,comp1 (4.38)

This is the reference level indicated by the thin dotted line in Figure 4.6. Compared to the reference
level, the solid solution has the energy:
Gss  Gmm  RT  ln[2 ]  RT (1  ) ln[
1 (1  )] (4.39)

In an ideal solid solution 1  2  1, the two log terms in Equation (4.39) are both negative
(0    1), and consequently, the energy of an ideal solid solution is always less than of a mechan-
ical mixture (Figure 4.6). Accordingly, an ideal solid solution is more stable than the two separate
phases, an effect due to the entropy of mixing.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Solid solutions 147

0.20
Spinodal point Gmm
Regular
0.00
Miscibility gap

Energy (kJ/mol)
0.20

0.40
Ideal
0.60

0.80
0 0.2 0.4 0.6 0.8 1
x

Figure 4.6. Energy of an ideal solid solution and a regular solid solution (a0  3), compared with a
mechanical mixture (thin dotted line).

The regular solid solution with a0  3 has two minima in Figure 4.6 (at   0.07 and   0.93).
Thermodynamically, the compositions in between these minima are unstable and result in a
miscibility gap. The minima are found from:
dGss d
 (ln[2 ]  (1  ) ln[1 (1  )])  0 (4.40)
d d
and are also calculated by PHREEQC (Example 4.11). The occurrence of a miscibility gap can be
tested with the condition for an inflection point on the solid solution curve (d2Gss / d2  0). The
inflection point is called a spinodal point in thermodynamics.

EXAMPLE 4.11. Miscibility gap in the solid solution

Increasing a0 of the solid solution increases also , and PHREEQC may calculate a hypothetical miscibility
gap for the otavite/calcite solid solution when a0 is increased to 3. Extend the input file from Example 4.9
with the following lines after the END:

USE solution 1
SOLID_SOLUTION 3
CaCdCO3
-comp1 Calcite 0
-comp2 Otavite 0
-Gugg_nondim 3 0 # hypothetical a0
END

The output file gives:

-------------------------------------
Description of Solid Solution CaCdCO3
-------------------------------------

Temperature: 298.15 kelvin

A0 (dimensionless): 3
A1 (dimensionless): 0
A0 (kJ/mol): 7.43708
A1 (kJ/mol): 0

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

148 Minerals and water

Critical mole-fraction of component 2: 0.5

Critical temperature: 447.225 kelvin
Spinodal-gap mole fractions, component 2: 0.211324 0.788675
Miscibility-gap fractions, component 2: 0.070720 0.92928
-------------------------------------

and...
-------------------------------User print-----------------------------------

aCa = 7.3641e-04 aCd = 8.4148e-08 aCa/aCd = 8.7514e+03

The output file repeats the values of a0 and a1, and writes them in dimensionalized form by multiplying with
RT ( 2.479 kJ/mol at 298 K). The miscibility gap disappears at the critical temperature and mole-fraction
in the alyotropic or eutectic point. PHREEQC gives various definitions of this point in T- space following
Glynn (1991) and Lippmann (1980). We also note that, compared to ideal solid solution a0  3 has
decreased the ratio [Ca2] / [Cd2] by a factor of 47.

4.5.2 The fractionation factor for solid solutions

An alternative way to relate the concentration ratios in a solid solution with the ratios in aqueous solu-
tion is the fractionation factor  which already was introduced for stable isotopes (Equation 2.7).
The fractionation factor is often called the distribution coefficient, but in this book we reserve the lat-
ter term for the partitioning of a single substance between solid and solution. For the otavite/calcite
example, the fractionation factor is:

(molCdCO / molCaCO ) ss
′  3 3
(4.41)
(mCd / mCa )aq

When   1 both components are taken up in the same ratio as present in the solution. If  ! 1,
the trace metal is preferentially taken up in the solid, and if   1, the trace metal is rejected
by the solid. A theoretical fractionation factor can be derived by combining Equations (4.33)
and (4.35):
(Cd Cd / Ca Ca ) ss K cc
  (4.42)
([Cd 2 ] / [Ca 2 ]) aq
K ot

Hence,
)Cd / )Ca
′   (4.43)
Cd / Ca

where )i is the correction for aqueous ions, calculated by subtracting complexes from mi, followed
by multiplication with the activity coefficient, for example:

[Ca 2 ]  )Ca mCa  (mCa   mCa-cplxs ) Ca .

A difficulty in defining  for solutions and minerals in rocks is the often irregular and zoned com-
position of the solid solutions (Figure 4.7). The zonation reflects changes in the water composition
and may be due to a Rayleigh process as was discussed for
18O variations in atmospheric vapor dur-
ing rainout (Chapter 2). Thus, if some calcite/otavite solid solution precipitates, the Cd/Ca ratio in

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Solid solutions 149

the aqueous solution will change because the two elements precipitate at different rates. We write the
differential equation for a Rayleigh process (Equation 2.7):

 d mol    m 
 CdCO3    d mCd   ′  Cd  (4.44)
 d mol   d m
 CaCO 3 ss  Ca aq  mCa aq

which we integrate from composition b to e:

 me   me 
ln  Cd   ′ ln  Ca  (4.45)
b b
 mCd aq  mCa aq

From the change of the aqueous molar ratios in time, the fractionation factor can be calculated. Since
(d molCd /d molCa)ss  (d mCd /d mCa)aq, the aqueous molar ratios are conserved in the solid solu-
tion, which can be analyzed in a mineral with the electron microprobe. If we replace the ratio of the
aqueous concentrations with those in the solid solution at two points in space, Equation (4.45)
describes a solid solution with an onion shell structure as depicted in Figure 4.7. The equation is
known as the Doerner-Hoskins relation.

Figure 4.7. Solid solutions showing chemical zoning of Mg and Ca in calcite (Beets and Immerhauser,
pers. comm.).

4.5.3 Kinetic effects on the fractionation factor

The incorporation of elements in different proportions in a solid solution can easily lead to kinetic
effects (Rimstidt et al., 1998). During precipitation, the solutes have to diffuse through a stagnant
water layer adjacent to the mineral surface and then further, on the surface, towards the crystal
growth site. This situation is illustrated in Figure 4.8 and the otavite/calcite solid solution is again
used as example. The diffusional flux of solute i through the stagnant water layer is:
Fi  Di (miaq  mi0 ) /
(4.46)

where Di is the diffusion coefficient for i, miaq is the concentration in the free solution, mi0 the con-
centration at the surface, and
is the thickness of the stagnant layer. The diffusion coefficient will
be nearly the same for HCO 2
3 , Ca , Cd
2
and H (Section 3.5.1). The flux to the surface will then

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

150 Minerals and Water

depend on the concentration difference (m iaq  mi0 ) and reflect the rate of precipitation of the
elements. The concentration gradient will be equal but opposite in sign for HCO 3 and H ,

 2 2
and for HCO3 equal to the sum of Ca and Cd . The apparent molar fractionation factor is now
given by:
FCd / FCa (m aq  mCd
0 ) / (m aq  m0 )
′a   Cd aq
Ca
aq
Ca
(4.47)
mCd / mCa mCd / mCa

At high supersaturation, mi0 is small compared to miaq, and Equation (4.47) predicts that the frac-
tionation factor a will approach 1. Furthermore, the minimal surface concentration of the trace
element limits the degree of supersaturation that gives the molar fractionation factor. If m0Cd  0,
and   4169 (the equilibrium fractionation factor for the otavite/calcite solid solution) then
Equation (4.47) can be reworked to give mCa aq 0
/ mCa  1.0002. Any larger ratio of mCa aq 0
/ mCa increases
the flux of Ca relative to Cd and thus reduces  to an apparent  . Since the ratio mCa
2 2 a aq
/ m 0Ca is
indicative for the degree of supersaturation, it is evident that the equilibrium fractionation factor only
can be measured in a system that is very close to equilibrium, or when the precipitation takes place
very slowly. In consequence, most experimentally measured fractionation factors are kinetically
controlled (Rimstidt et al., 1998).

m0
1.2
ⴚ
HCO3
1
Concentration

0.8

Ca2ⴙ
0.6

maq
0.4
Cd 0.3Ca0.7CO3

0.2 Cd2ⴙ
Hⴙ
0
0 ␦
Distance

Figure 4.8. Schematic concentration gradients at the surface of an otavite/calcite solid solution.

The effect of the degree of supersaturation on the apparent fractionation factor can be calculated by
assuming that the concentrations in the stagnant water layer adjacent to the mineral surface (mi0 ) are
controlled by the molar (equilibrium) fractionation factor (Riehl et al., 1960; McIntyre, 1963). In
that case,
0  m0 (F
mCd Ca Cd / FCa ) / ′ (4.48)

Furthermore, let’s assume that the saturation ratio is given by   mCa

aq
/ m 0Ca. This equation, in com-
0 0
bination with (4.48), can be used to eliminate m Cd and mCa from
aq 0 ) / (m aq  m0 )
FCd / FCa  (mCd  mCd Ca Ca (4.49)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Solid solutions 151

10
9   10
8
7
6
5

Apparent a
4
3 3
2
1
0.8
0.6
0.4
0.2   0.1
0
3 2 1 0 1
log(Ω  1)

Figure 4.9. The apparent fractionation factors for solid solutions as function of a modified saturation index
(log   1) for molar fractionation factors of 10, 3 and 0.1.

and to rewrite it as:

FCd / FCa ′ 
  ′a (4.50)
aq
mCd aq
/ mCa (  1) ′  1

Figure 4.9 shows the apparent fractionation factor as a function of the modified saturation index
(log(  1)) for molar fractionation factors   10, 3 and 0.1, calculated according to Equation
(4.50). Clearly, if   1 (and log(  1)  %), then a  , and the apparent and the molar frac-
tionation factors are the same. On the other hand, if the solution is highly supersaturated,  !! 1, then
a → 1, as we also noted before.
The lines in Figure 4.9 can be compared with the experimental data from Tesoriero
and Pankow (1996) for strontium and cadmium solid solutions in calcite (Figure 4.10).

2000 0.15
Strontium
1500
0.1
a Cd

a Sr

1000
0.05
500

Cadmium
0 0
5 4 3 2 1 0 1 2 3 1 0 1 2 3
log R log R

Figure 4.10. Apparent fractionation factor (a) for solid solutions of strontium (a  1) and cadmium
(a ! 1) in calcite as function of the precipitation rate. R is the precipitation rate in nmol/mg/min.
Approximately 900 mg of seed calcite was added to 400 mL water in the reaction vessel; a rate of 1 nmol/mg/min
(log R  0) corresponds to 0.135 mmol CaCO3/L/hr, or 5.1108 mol/s/m2 seed crystal surface area, which is
low, also for natural conditions (cf. Chapter 5). Lines are calculated with Equation (4.50), using R  104.5(  1)
for Cd2, and R  101.8(  1) for Sr 2; experimental points are from Tesoriero and Pankow (1996).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

152 Minerals and water

The apparent fractionation factors are plotted as function of log R, the rate of carbonate precipitation
in the experiment. The experimental data for Cd show the expected trend with a high apparent frac-
tionation factor at low precipitation rates and a decline at higher rates. For Sr, the apparent fraction-
ation factor is less than 1 and nearly constant when the precipitation rate is smaller than
10 nmol/mg/min, but it increases towards 1 when the rate increases.
The important conclusion to be drawn is that at high precipitation rates the ions will be incorpo-
rated in the solid in the ratio they are present in the solution, while at lower precipitation rates the
thermodynamic effects may become increasingly more important. The effect of precipitation rate
will be stronger when the fractionation factor is larger. Overall, the effects render the derivation of
thermodynamic properties of solid solutions from laboratory or field data rather doubtful (Curti,
1999; Rimstidt et al., 1998; Schwartz and Ploethner, 1999).

QUESTION:
Find the miscibility gap for calcite/strontianite solid solution with PHREEQC, -comp 2 Strontianite;
-Gugg_non 5.7 0.
ANSWER: Miscibility-gap fractions, component 2: 0.0035 0.9965, i.e. barely miscible

4.6 KINETICS OF GEOCHEMICAL PROCESSES

4.6.1 Kinetics and equilibrium
If the mineral halite (NaCl) dissolves in distilled water, and the concentrations are followed as a
function of time, you may find a curve as shown in Figure 4.11. The [Na] increases with time until
equilibrium between water and mineral is attained at time t2. From t2 onwards, [Na] becomes inde-
pendent of time and is determined by the equilibrium constraint:

K halite  [Na ][Cl ]

Khalite can be measured directly in the laboratory or calculated from thermodynamic tables.
As we have seen, also the dependency of Khalite on temperature can be calculated. In many
cases, the equilibrium concept gives a satisfactory explanation for the observed groundwater
chemistry. Due to its firm fundament and ease in calculations, it should always be the first approach
to any problem.

Kinetics Equilibrium

K halite
[Na]

t1 t2
Time

Figure 4.11. The realms of kinetics and equilibrium, illustrated in a dissolution experiment with halite.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 153

However, some chemical phenomena in aquifers are insufficiently explained by equilibrium chem-
istry. We already noted that the composition of precipitating solid solutions becomes kinetically con-
trolled when the fractionation factor for the minor component is high. Many silicate minerals, like
feldspars, amphiboles, pyroxenes, etc. are thermodynamically unstable at the earth surface (ground-
waters/surface waters are strongly subsaturated for these minerals). Still they persist for thousands
or tens of thousands of years. Other examples are the observed subsaturation for calcite in some
aquifers as well as the survival of aragonite in aquifers for thousands of years. There are two funda-
mental causes for the lack of equilibrium. The first is the tectonic activity that transports minerals
from the environments where they were formed and are stable, often deep down in the earth at high
pressures and temperatures, to environments at the earth’s surface where they are unstable. The sec-
ond is photosynthesis which uses solar energy to convert CO2 into thermodynamically unstable
organic matter; its subsequent kinetic decomposition affects the equilibria of redox sensitive
elements such as iron, oxygen and nitrogen. Biological activity also disturbs carbonate
equilibria through the production or consumption of CO2 and through biomineralization. To under-
stand, and to describe quantitatively such deviation from equilibrium requires the use of reaction
kinetics.
In the example of halite dissolution (Figure 4.11), the first part of the reaction, before equilib-
rium with halite is attained, should clearly be studied from a kinetic point of view. Qualitatively, one
can expect that the rate of dissolution of halite depends on factors such as grain size (small crystals
have a larger surface area than large ones per unit weight), the amount of stirring, temperature, and
the distance from equilibrium. The link to equilibrium chemistry is to predict in which direction the
reaction will go. At time t1, equilibrium chemistry predicts that dissolution will take place rather than
precipitation. It does not tell us, however, whether it will take 10 seconds or 10 million years before
equilibrium is attained.
Our understanding of reaction kinetics is at a much lower level than of equilibrium chemistry
although empirical information is rapidly accumulating (e.g. Sparks, 1989; Hochella and White,
1990; Schwartzenbach et al., 1993; White and Brantley, 1995; Lasaga, 1998). There is no unifying
theory, and different notations are used to describe even the same problem. The approach to a kinetic
problem is usually divided into two steps:

1. to describe quantitatively the rate data measured in the laboratory or in the field.
2. to interpret the quantitative description of the rate data in terms of mechanisms.

It is often found that different mechanisms fit equally well to the same rate data. Therefore, kinetic
data derived from solution chemistry may at best be demonstrated as consistent with a given mech-
anism. Surface chemical techniques like AFM, SIMS, XPS, on the other hand, may elucidate the
mechanisms at the atomic scale (Stipp, 1994).

4.6.2 Chemical reactions and rates

Consider a simple reaction where compound A is converted to compound B by the reaction:
A → B
This reaction can be followed by recording the concentration of A as a function of time as shown
in Figure 4.12. The reaction rate is the change of A with time. Thus at time x1, the rate can be
determined from the slope of the tangent, cx1 / tx1, at this point. For the whole curve, this is
equal to:
rate  dcA / dt

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

154 Minerals and water

Concentration
Cx1

0 x1 tx1
Time

Figure 4.12. Derivation of rates from concentration/time data.

The units of the rate depend on the definition of the reaction and could for example be mol/L/s. For a
decrease in concentration of a reactant, the rate is given a negative sign (the slope of the tangent is neg-
ative), while for the increase of a product, the rate is positive (the slope of the tangent is positive). Thus:
rate  dcA / dt  dcB / dt

d[A]
Zeroth order Rate  ———  k log rate  log k
dt
[A]

Rate

log rate

k
log k
Time [A] log[A]

d[A]
First order Rate  ———  k [A] log rate  log k 
dt
log[A]
[A]

Rate

log rate

n1

k log k
Time [A] log[A]

d[A]
Second order Rate  ———  k [A]2 log rate  log k 
dt
2 log[A]
[A]

Rate

log rate

n2

log k
Time [A] log[A]

Figure 4.13. Simple rate laws and the differential method for the reaction A → B.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 155

For the more complicated reaction

A  2B → 3C
we find
rate  dcC / dt  3 dcA / dt  3/2 dcB / dt

This shows the importance of defining the reaction rate explicitly. The manner in which the reaction
rate varies with the concentrations of reacting substances is referred to as the reaction order. For
example, if it is found experimentally that the reaction rate is proportional to the th power of the
concentration of reactant A and to the th power of reactant B etc., then

rate  k cA cB … (4.51)

Such a reaction is said to be th order with respect to [A] and th order for [B]. The overall order n
of this reaction is simply
n    …

The coefficient k in (4.51) is the rate constant, also known as the specific rate, and is numerically
equal to the reaction rate when all reactants are present at unit concentrations. Since the rate has fixed
units (for example mol/L/s), the units of the rate constant depend on the overall order of the reaction.
In general for an n-order reaction, the units of k are (mol/L)1n/s. The effect of the reaction orders on
relations between concentration and time, and concentration and rate is illustrated in Figure 4.13.
For zeroth order kinetics, the rate is independent of the reactant concentration and this implies that
other factors or reactants control the rate. An example is shown in Figure 4.14 for bacterial reduction of
sulfate by organic matter, which shows zeroth order behavior as long as mSO42 is higher than 2 mmol/L.
To find the concentration of SO24 at a given time, for example t  15 days, we integrate the rate:

dmSO2 / dt  k  0.92 mmol/L/day (4.52)

4

20

15
Sulfate (mmol/L)

10

0
0 10 20 30
Time (days)

Figure 4.14. Dissolved sulfate versus time for marine sediment incubated in the laboratory (Berner, 1981).
The straight line indicates a zeroth order dependence on the sulfate concentration.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

156 Minerals and water

from mSO42  20 mmol/L at t  0 to the concentration mSO42 at t  15 days. Thus:

m
SO2 15
4

∫ dmSO2  0.92 ∫ dt
4
20 0

which gives mSO42  20  0.92  (15  0), or mSO42  20  13.8  6.2 mmol/L (after 15 days).
However, for times longer than 21.74 days, Equation (4.52) predicts a negative concentration of
SO2 2
4 , which is impossible of course. Actually, for concentrations smaller than 2 mM SO 4 the rate
2
slows down and becomes first order with respect to SO4 (Boudreau and Westrich, 1984; Section
10.4.1). As shown in Figure 4.13, a first order rate decreases proportionally as the concentration goes
down and reaches zero when the concentration becomes zero.
First order rate laws are common and comprise, for example, all radioactive decay reactions
(Equation 3.20). Example 4.12 illustrates the effect of first and second order rate laws for the oxida-
tion of Fe2. The rate is first order with respect to both Fe2 and PO2, implicating that doubling the
concentrations increases the rate with a factor of two. The rate is second order for [OH] and the rate
increases fourfold when the OH activity is doubled. It is thus clear that increasing the pH is a much
more effective way to accelerate Fe(2) oxidation than increasing the PO2.

EXAMPLE 4.12. Oxidation of Fe(2)

For oxidation of Fe2 in aqueous solutions at 20°C and 1 atm., the following rate law has been reported by
Stumm and Morgan (1996):
rate  d mFe2 / dt  k mFe2 [OH]2 PO
2

where k  81013 /min/atm.

Thus the rate equation is formulated as a decrease in mFe2 and is first order for mFe2 and PO2, and sec-
ond order for [OH]. What is the rate of Fe2 oxidation if mFe2  1 mM and PO2  0.2 atm, for pH values
5 and 7?

ANSWER:

pH  5; K w  [H][OH]  1014 → [OH]  109

d mFe2 / dt  81013  103  (109 )2  0.2  1.6108 mmol/L/min

 .001 mmol/L/hour

pH  7:
d mFe2 / dt  81013  103  (107 )2  0.2  2.0104 mol/L/min
 9.4 mmol/L/hour

Note the very large difference in rates at pH 5 and 7, due to the second order rate dependence on [OH].

The time needed to reach a given concentration is found by integrating the rate equation, which is pos-
sible for the rate of Fe(2) oxidation (Example 4.12), when pH and PO2 are kept constant. For example,
the time needed to halve the Fe2 concentration from 1 to 0.5 mmol/L, at pH  5 and PO2  0.2 atm:

d mFe2 / dt  81013  (109 )2  0.2  mFe2  1.6105 mFe2 (4.53)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 157

which we integrate from mFe2  1 mM at t  0, to mFe2  0.5 mM at t  t ⁄ : 1

2

0.5e3 t1⁄2
dmFe2
∫ mFe2
 1.6105 ∫ dt
le3 0

It gives ln(0.5103/103)  1.6105  (t ⁄  0). Thus, t ⁄  ln(1/2)/1.6105  43321 s or

1
2
1
2

12 hr.
In first order rate laws, the time needed to halve the reactant concentration is known as the half-
life of the reaction, which is independent of the initial concentration. In a more loose sense, the half-
life concept is also used for other types of reaction rates, but then the value will depend on the initial
reactant concentration.

QUESTIONS:
Find the time to lower the Fe2 concentration from 0.5 to 0.25 mM, pH  5, PO2  0.2 atm.
ANSWER: 12 hr
Find the half-life of Fe2 oxidation at pH  7 and PO2  0.2 atm?
ANSWER: 4.3 seconds

Time (min)
0 20 40 60 80 100 120 140 160 180 200 220 240 260
0
 Vermiculite
0.2  Montmorillonite
 Kaolinite
0.4
log(1  (K-X)t / (K-X)∞)

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Figure 4.15. Kinetic exchange of K on three clay minerals, plotted according to a first order rate (Sparks and
Jardine, 1984; Sparks, 1989).

Another example of a first order reaction is shown in Figure 4.15 for sorption of K on three clay
minerals. It depicts the result of the exchange reaction:
K  1⁄2 Ca-X2 ↔ K-X  1⁄2 Ca2 (4.54)
where X is the exchange site on the mineral, with a charge of X. Thus, K-X is the exchangeable K,
which is expressed as a fraction of the total amount of exchange sites. K-X% is the fraction at infinite

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

158 Minerals and water

time and equals 1 (exchange theory is discussed in Chapter 6). Sparks and Jardine (1984) postulated
the rate equation:
d (K-X) / dt  k {(K-X)  (K-X)∞} (4.55)
which is first order with respect to {(K-X)  (K-X)%}. Again, to find the concentration (K-X) at a
given time t  te, we integrate the equation from (K-X)  0 at t  0. First collect the terms:
( K-X)t te
d(K-* )
e

∫ (K-* )  (K-* )∞
 k ∫ dt
0 0

To solve the integral, we subsitute z  {(K-X)  (K-X)%}. Then, dz / d (K-X)  1, and d(K-X)  dz.
The integration boundaries change to zt0  {(K-X)  (K-X)%}  (K-X)% and to ztte 
{(K-X)  (K-X)%}  {(K-X)te  (K-X)%}. Thus:
{( K-X)t (K-X)∞ } te
e
dz
∫ z
 k ∫ dt
(K-X)∞ 0

which solves to
 {(K-X)t  (K-X)% } 
ln    k (t e  0)
e

 (K-X)%  (4.56)

or log(1  (K-X)t / (K-X)% )  k t e /2.303

e

The form of this equation is typical for an integrated first order rate in having the logarithm of the
concentration as linear function of time.
Experimental results were plotted according to Equation (4.56) in Figure 4.15. The slope
for the clay mineral kaolinite is larger than for montmorillonite and vermiculite, indicating
that k is largest for kaolinite, and therefore the reaction is quickest for this mineral. At longer
times, the slope increases for montmorillonite and vermiculite, which suggests that the first
order reaction mechanism invoked by Equation (4.55) is only applicable for the initial part of the
reaction.

QUESTION:
Find k for K exchange on kaolinite and montmorillonite in the first 100 min
ANSWER: k  0.115/min for kaol; k  0.0242/min for montm

In the examples presented so far, the reaction orders have been whole numbers. However, there is no
a priori reason why they could not be fractional numbers and these are frequently reported. As the
first guess for the rate equation one often takes the product of reactants, just as in the law of mass
action. For K-Ca2 exchange (Reaction 4.54) the forward rate could be written as:

r fw  k fw [K] [Ca-X 2 ] ⁄2
1
(4.57)

and the backward rate as:

rbw  kbw [Ca2] ⁄2 [K-X]
1
(4.58)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 159

Both rate equations are half-order with respect to the activity of Ca2 as solute and exchangeable
species. But in contrast to the easy integration of the first order rate law that led to Equation (4.56),
the chemically more attractive half-order rate formulation needs numerical integration.
If the overall reaction is split-up in a forward and a backward rate equation, the rate constants are
interconnected since at equilibrium the overall rate is zero and the two rates must cancel:
R  0  rfw  rbw  kfw [K][Ca-X2] ⁄2  kbw [Ca2] ⁄2 [K-X]
1 1
(4.59)

or:
1⁄
k fw [Ca 2 ] 2 [K-X]
  K (4.60)
kbw [K ] [Ca-X 2 ]

where K is the equilibrium constant of the reaction. This links the forward and the backward rate
constants as:
k fw  kbw K (4.61)

In deriving relation (4.61) we used the principle of detailed balancing.

4.6.3 Temperature dependency of reaction rates

The rate of most reactions is highly influenced by the temperature. According to the Arrhenius equa-
tion the rate constant changes with temperature as:

 Ea 
k  A exp   (4.62)
 RT 

where A is the pre-exponential factor (same units as k) and Ea is the activation energy (kJ/mol), R
the gas constant and T absolute temperature. Taking the logarithm of Equation (4.62) gives:
ln k  ln A  Ea / RT (4.63)

A plot of log k versus 1/T therefore gives a straight line with a slope of
d log k Ea
 (4.64)
d(1/T ) 2.303 R

Table 4.8. Typical values (and ranges) of activation energies for chemical and transport
processes, and relative increase of the rate coefficient when temperature increases from
10 to 20°C according to the Arrhenius equation.

Process Ea (kJ/mol) k20°C/k10°C

Photochemical reactions 15 (10–30)a 1.24

Diffusion in water 20 (Equation 3.42) 1.34
Dissolution/Precipitation 60 (20–80)b 2.4
Biochemical reactions 63 (50–75) 2.5
Solid state diffusion 500 (200–650)b 1400
a
Schwarzenbach et al., 1993. bLasaga, 1998.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

160 Minerals and water

Since d(T 1) / dT  T 2, we can rewrite Equation (4.64) as:

d log k Ea (4.65)


dT 2.303 RT 2

Equation (4.65) has the same form as the Van ’t Hoff equation (4.28), which related the temperature
dependency of equilibrium constants to the reaction enthalpy. In transition state theory, the activa-
tion energy is linked to the energy absorbed by the reactants when an activated complex is formed in
an intermediate step from which the products of the reaction are generated (Lasaga, 1998).
Activation energies differ markedly for various types of chemical reactions and transport
processes and the temperature dependency of a set of experimental rate data may indicate the prob-
able reaction mechanism (Table 4.8). For example, the relatively small temperature effect on diffu-
sion in water suggests that the temperature dependency of a transport controlled dissolution reaction
is much less than of a surface controlled reaction. The activation energy is very high for diffusion in
solids which implies a huge effect for a small temperature increase from 10 to 20°C. However, the
diffusion coefficient at this temperature is so small, that the actual movement of the atoms in the
solid at 20°C is still negligible.

4.6.4 Mechanisms of dissolution and crystallization

The dissolution or growth of minerals proceeds through a chain of processes, including transport
of solutes between the bulk solution and the mineral surface, the ad- and desorption of solutes at
the surface, hydration and dehydration of ions and surface migration. These processes must take place
simultaneously but some may proceed at a faster rate than others. Since the overall rate will depend on
the rate of the slowest process, it is of importance to identify the rate limiting process or mechanism.
There is an essential distinction between an overall rate control by transport of solutes from the
bulk solution to the mineral surface, and by the reaction occurring at the mineral surface (Nielsen,
1986). The difference is illustrated in Figure 4.16 for a dissolving mineral. If transport of ions to and
from the mineral surface is rate limiting, equilibrium will be approached at the mineral surface and
a concentration gradient develops from the mineral surface to the bulk solution. On the other hand
when a surface reaction is rate limiting, any ion released from the crystal lattice will be carried away
quickly and no concentration gradient develops. Also intermediate cases of surface and transport
controlled reactions can be found.

A B C
Ceq Ceq Ceq

C∞ C∞ C∞

0 r 0 r 0 r

Figure 4.16. Schematic concentration profiles from the mineral surface to the bulk of solution, for different
rate controlling processes during dissolution. C% is the concentration in the bulk solution and Ceq the concen-
tration in equilibrium with the mineral. A. Transport control, B. Surface reaction control, C. Mixed control
(Berner, 1978).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 161

How to decide whether a mineral dissolution reaction is transport or surface reaction controlled? In a
dissolution experiment the stirring rate can be changed. If the reaction is transport controlled, it affects
the concentration gradients at the mineral surface and the rate changes. The reverse is not always true,
since very small crystals (about 5 m) are carried along with the whirling water, so that the convec-
tion at the mineral surface remains the same (Nielsen, 1986). A more elegant way is to calculate whether
the reaction is slower or faster than molecular diffusion (Example 4.13). Since molecular diffusion is the
slowest form of transport possible, slower rates are necessarily controlled by surface reaction.

EXAMPLE 4.13. Dissolution of hydroxyapatite; transport or surface reaction controlled?

Christoffersen and Christoffersen (1979) studied the dissolution kinetics of hydroxyapatite (HAP) using
suspensions of fine crystals. The overall dissolution rate is expressed as R  dn/dt, where n is the amount
of undissolved HAP. For steady state diffusion through a sphere circumscribing the crystal we may use the
equation:
R  4$Dapp N r 2 (ceq  c )

where Dapp is the apparent diffusion coefficient, N the number of crystals, r the radius of the crystals, ceq the
concentration at equilibrium and c the concentration in the bulk solution. Substituting measured values of
R, N, r, ceq and c, yields a value of Dapp
109 cm2/s. Diffusion coefficients in aqueous solution are in the
order 105 cm2/s (Section 3.5.1). Since the apparent diffusion coefficient Dapp, is four orders of magnitude
lower, one must conclude that dissolution of HAP is much slower than predicted by molecular diffusion.
Therefore, the reaction is surface reaction controlled.

The solubility of minerals appears to be related to the dissolution mechanism (Berner, 1980). The dis-
solution of sparingly soluble minerals is generally controlled by surface processes whereas the disso-
lution of soluble minerals predominantly is controlled by transport processes (Table 4.9). Silicate
minerals, like feldspars, pyroxenes and amphiboles are also insoluble and their dissolution is controlled
by surface reactions while most salt minerals are soluble and their dissolution is transport controlled.

Table 4.9. Rate controlling dissolution mechanism and solubility for various substances (Berner, 1980).

Substance Solubility (mol/L) Dissolution rate control

Ca5(PO4)3OH 2 108 Surface-reaction

KAlSi3O8 3 107 Surface-reaction
NaAlSi3O8 6 107 Surface-reaction
BaSO4 1 105 Surface-reaction
AgCl 1 105 Transport
SrCO3 3 105 Surface-reaction
CaCO3 6 105 Surface-reaction
Ag2CrO4 1 104 Surface-reaction
PbSO4 1 104 Mixed
Ba(IO3)2 8 104 Transport
SrSO4 9 104 Surface-reaction
Opaline SiO4 2 103 Surface reaction
CaSO4 2H2O 5 103 Transport
Na2SO4 10H2O 2 101 Transport
MgSO4 7H2O 3 100 Transport
Na2CO3 10H2O 3 100 Transport
KCl 4 100 Transport
NaCl 5 100 Transport
MgCl2 6H2O 5 100 Transport

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

162 Minerals and water

T N

N
E
S

K S X0

G T

Figure 4.17. The schematic surface of a simple cubic crystal. Energetically favorable sites include kinks (K),
steps (S), clusters (N) and etch pits (E) (Zhang and Nancollas, 1990).

During a surface controlled reaction, dissolution or growth only takes place at energetically favor-
able sites on the mineral surface. A schematic picture of a mineral surface (Figure 4.17) identifies
the favorable sites as irregularities on the crystal face. Clearly, ions or atoms which are not sur-
rounded by other crystal units on three sides have a lower binding energy and are the first to be
released upon dissolution. Also crystal growth will take place preferentially at those sites where
bonds in several directions can be established. Selective dissolution will yield beautiful etch patterns
on mineral surfaces that can be observed by SEM on a microscopic level. Such etch patterns can be
used as an indicator for surface reaction dissolution control (Berner, 1978), and may be used to
establish rates of dissolution (Gratz et al., 1990).
Various models have been derived to describe crystal growth or dissolution mechanisms at reac-
tive surface sites. These include spiral growth or dissolution, polynuclear growth/dissolution, etc.,
which are discussed in more detail in texts like Nielsen (1964, 1984), Zhang and Nancollas (1990)
and Lasaga (1998).

4.6.5 Rate laws for mineral dissolution and precipitation

A general rate law for the change in solute concentration due to mineral dissolution/precipitation
reactions can be written as:
n
A  m
Rk 0  g (C ) (4.66)
V  m0 

where R is the overall reaction rate (mol/L/s), k is the specific rate (mol/m2/s), A0 is the initial surface
area of the solid (m2), V is the volume of solution (m3), m0 is the initial moles of solid, and m is the
moles of solid at a given time. The overall rate depends first of all on the specific rate and the ratio of
surface area A0 over the solution volume V. Next, the factor (m/m0)n accounts for changes in surface
area, or more precise, of reactive surface sites, during dissolution. These changes may be due to
changes in crystal size during dissolution/precipitation, changes in the size distribution of the crystal
population, but also factors like selective dissolution and aging of the solid may be important. For a
monodisperse population of uniformly dissolving or growing spheres and cubes, n  2/3, because m
is proportional to the volume, or r3 (here r is the radius of the sphere or the side of the cube) while the

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 163

surface area is proportional to r2. For polydisperse crystal populations, the finest crystals will dissolve
selectively and the size distribution is modified. Therefore, n becomes a function of the initial grain
size distribution and for log-normal size distributions values of up to n  3.4 are possible (Dixon
and Hendrix, 1993). Also the disintegration of the crystals during dissolution may modify n (Larsen
and Postma, 2001). Finally, g(C) is a function that comprises the effects of the solution composition
on the rate, like the pH, the distance from equilibrium, and the effects of catalysis and inhibition
(Lasaga, 1998).

Dissolution of minerals
Let us use the dissolution of quartz (SiO2) as example. The reaction is
SiO 2  2H 2O ↔ H 4SiO04

In an early study, Van Lier et al. (1960) measured a zeroth order rate for dissolution of quartz,

rquartz  1013.5 mol/m 2 /s  106 mol/m 2 /yr. (4.67)

Let’s calculate the concentration of H4SiO04 in soil water (Example 4.14).

EXAMPLE 4.14. Dissolution rate of quartz

Estimate the concentration of H4SiO40 in water in a quartz soil after one year, quartz  2.65 g/cm3, w  0.3,
grain-size  0.1 mm.

ANSWER:
First calculate the surface to volume factor A/V: determine the number of spherical grains in 1 kg soil and
the surface area of a single grain.
volume of 1 grain  4⁄3$r 3  4⁄3  3.14  (0.05103)3  5.241013 m3
volume of 1 kg quartz grains  1 kg / (2650 kg/m3)  3.77104 m3
number of grains in 1 kg  3.77104/5.241013  7.21108
surface of 1 grain  4$r 2  4  3.14  (0.05103)2  3.14108 m2
→ surface area of 1 kg soil, A  7.21108  3.14108  22.6 m2/kg
The volume of water in contact with 1 kg soil, V  3.77104 m3 w /(1  w)
→ V  3.77104  0.3/(0.7)  1.62104 m3
The increase in concentration is Rqu  dmH4SiO04 /dt  rqu A / V  106  22.6 / (0.162)  0.14 mmol/L/yr.

The solubility of quartz is about 0.1 mM H4SiO04, and for the quartz soil in Example 4.14, Equation
(4.67) predicts that this concentration should be reached in less than one year. However, a zero order
rate equation is perhaps not appropriate since the rate must slow down as equilibrium is approached.
Rimstidt and Barnes (1980) therefore proposed a rate equation that includes the saturation state
  IAP/K:
rqu  kqu (1  IAP /K ) (4.68)

where kqu  1013.7 mol/m2/s at 25°C. The rate as defined by Equation (4.68) is positive at
subsaturation, zero at equilibrium, and negative at supersaturation which intuitively appears
correct. The rate equation can be integrated numerically with PHREEQC as shown in
Example 4.15.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

164 Minerals and water

EXAMPLE 4.15. Kinetic dissolution of quartz with PHREEQC

The rate equation is defined under keyword RATES using Equation (4.68) as part of Equation (4.66). The
keyword KINETICS invokes the rate for the case of the quartz soil of Example 4.14.

RATES
Quartz # Rate name
-start
10 A0 = parm(1)
20 V = parm(2)
30 rate = 10^-13.7 * (1 - SR("Quartz")) * A0/V * (m/m0)^0.67
40 moles = rate * time
50 save moles # time and save must be in the rate definition
-end

KINETICS 1
Quartz
-formula SiO2
-m0 103 # initial quartz, mol/L
-parms 22.6 0.162 # A0 in m2, V in dm3
-step 1.5768e8 in 15 # 5 years in seconds, in 15 steps
INCREMENTAL_REACTIONS true
SOLUTION 1 # distilled water
USER_GRAPH
-axis_titles Years "mmol Si / l"
-axis_scale x_axis 0 5
-start
10 graph_x total_time/3.1536e7 # time in years on x-axis
20 graph_y 1000 * tot("Si") # Si conc. on y-axis
-end
END

RATES starts by defining the name ‘Quartz’. Then, the rate equation is written in a BASIC program
between ‘-start’ and ‘-end’. The BASIC commands are given in the PHREEQC manual under keyword
RATES. The initial surface area ‘A0’ in line 10 and the solution volume ‘V’ in line 20 are defined as the vari-
ables parm(1) and parm(2). KINETICS starts by calling Quartz and then provides values for these variables
in the rate equation.
The keyword ‘INCREMENTAL_REACTIONS true’ signals PHREEQC to do the calculations incre-
mentally for each timestep, adding each time to the results of the previous calculation; ‘false’ would mean
that the calculation starts from zero for each timestep, defined with ‘-step’ in KINETICS. Running the file
produces the plot shown in Figure 4.18.

0.12
0.1
Si (mmol/L)

0.08
0.06
0.04
0.02
0
0 1 2 3 4 5
Years

Figure 4.18. Kinetic dissolution of quartz based on Equation (4.68) for a quartz soil.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 165

Figure 4.19. Scanning electron microscopy pictures of mineral grains showing etch pits and surface irregu-
larities. A is from a 10 ka old soil, C stems from a 250 ka old profile (White, 1995).

The calculated Si release in the quartz soil defined in Example 4.14 is shown in Figure 4.18. The initial
rise of the H4SiO04 concentration slows down when the equilibrium concentration is approached, which
takes about 4 years. The term (1  IAP / K) has a small influence when IAP / K  0.1 and in that domain
the kinetics are linear in time, but the term gains importance as the quartz solubility is approached. This
seems to be a general feature for the dissolution of silicate minerals (Nagy and Lasaga, 1992).
The calculated concentration trend not necessarily complies with field observations since the
conditions in a real soil are much more complex. First, other minerals such as clay minerals and
feldspars are present in the soil and may dissolve more quickly than quartz, impeding quartz dis-
solution as soon as its solubility has been reached. Second, the surface area of quartz was calculated
based on a smooth, geometric surface of an average grain; in reality, the grains have pits and irregu-
larities that increase the surface area, as illustrated in Figure 4.19.
In laboratory studies, the surface area is usually measured with the BET method which is based
on gas adsorption (Atkins and de Paula, 2002). Figure 4.20 shows a comparison of BET surface
areas with surface areas estimated from geometric calculations. The BET surface area of freshly
crushed grains is approximately 7 times higher than the geometric estimate, due to sub-microscopic
surface roughness and crevices (Sverdrup and Warfvinge, 1988; Anbeek, 1992; White, 1995).

100
Merced soils
3 Ka
40
250
log surface area A (m2/g)

10 600
3000

1
Qu
ar t
z
K-f
Ge eld
0.1 om spa
etr r
ic s
urfa Alb
ce ite
are
a
0.01
10 100 1000
log particle size (µm)

Figure 4.20. Surface areas as a function of particle diameter for a) Geometric areas calculated for smooth
spheres (dashed line), b) BET surface areas for freshly crushed grains of quartz, K-feldspar and albite (solid
symbols) and c) BET surface areas of soil grains, cleaned for clay and Fe-oxides, of different age (White, 1995).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

166 Minerals and water

The BET surfaces of mineral grains isolated from soils are significantly higher and tend to increase
with age due to further development of dissolution etch pits.
Also the solution composition may influence the rate of quartz dissolution. Both the dissolution
and precipitation kinetics of quartz may increase by up to 35-fold when the concentration of
alkalis in solution increases (Dove and Rimstidt, 1993), and probably similar effects can be con-
nected with other electrolytes which modify the surface properties of quartz. The sub-processes
involved are:
background dissolution: kqu, bck
proton promoted dissolution: kH [ OH 2]
hydroxyl assisted dissolution: kOH [ O ]
ligand promoted dissolution: k L [ OL]
dissolution inhibition: k IB [ OIB]

where [Oi] indicates the activity of a species adsorbed on the surface (dimensionless),
which may be found by a sorption model, or in first approximation be related to the activity
of the driving solute species [i], and ki is the rate constant for the sub-process. Since the
sorbed species are activities (dimensionless), the ki’s all have the dimension mol/m2/s.
The rate constant in Equation (4.68) is summarizing all the sub-processes of (4.69):

kqu  ff (kqu,bck  ∑ ki [ Oi]) (4.70)

Here ff is a factor that relates the field and laboratory dissolution rates. Judged from the greater sur-
face roughness of field grains compared with fresh, crushed grains (Figure 4.20), this factor is
expected to be higher than 1, but in reality it is found to be less than 1, in the order of 0.1 to 0.01.
Apparently, the grains used in laboratory studies have unstable surfaces and sites which dissolve rel-
atively quickly on the timescale of the laboratory experiment, while these highly reactive sites have
disappeared over time in the field.

80

70 Phosphate: 0

60

50
aq ] (µM)

40
[Fe2

10 M
30

20

10 100 M

0
0 2 4 6 8
Time (h)

Figure 4.21. The effect of phosphate on the reductive dissolution of hematite (Fe2O3, 0.017 g/L) H2S (partial
pressure of H2S  102 atm.) (Biber et al., 1994).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of geochemical processes 167

Possibly the field rates are also impeded by the presence of inhibitors which are substances that reduce
the rate of reaction (Ganor and Lasaga, 1998). For quartz dissolution, and also other silicate minerals,
aluminum is a well known inhibitor, just as phosphate inhibits fluorite and calcite dissolution even at
micromolar concentrations (Morse and Berner, 1979; Christoffersen et al., 1988). Inhibitors are able to
block active sites on the mineral surface; an example is shown in Figure 4.21 for the release of Fe2
during reductive dissolution of hematite (Fe2O3) by H2S. In the absence of phosphate there is a steep
immediate release of Fe2 over time. However in the presence of a phosphate concentration as low as
10 M, hematite dissolution is strongly inhibited. It is the recognition that dissolution and growth only
takes place at specific sites on the crystal surface that explains why substances even at very low con-
centrations may have an inhibiting effect on the dissolution or precipitation rate.

Crystal formation
Crystal formation is a two step process: the first step is the formation of a crystal embryo by nucleation,
and the second step is crystal growth. The distinction is based on the fact that very small crystals are
dominated by edges and corners, and since atoms located at such positions are not charge compensated,
very small crystals are less stable than larger crystals. In terms of energy this can be described as:
Gn  Gbulk  Ginterf

where Gn is the free energy of formation of a crystal as a function of the number of atoms n in the
crystal, Gbulk the bulk free energy identical to the one used in Section 4.3.1, and Ginterf the inter-
facial free energy which reflects the excess free energy of very small crystals and is determined by
the crystal-water surface tension. A direct result of this equation is that very small crystals are more
soluble than larger crystals. However, already at crystal sizes larger than 2 m, the effect of the inter-
facial free energy becomes negligible.
The change in Gn with increasing crystal size is shown in Figure 4.22. During the growth of the
crystal embryo an energy barrier has to be passed in order to allow further growth. The crystal size
corresponding to this energy barrier is known as the critical nucleus and its size is dependent on the
saturation state . For increasing , both the energy barrier and the size of the critical nucleus
decreases. Once the size of the critical nucleus has been passed, the crystal enters the domain of
crystal growth.

2 !  1

G1 1

G2 2
Gn O
O n2 n1 n

Figure 4.22. Free energy of a single crystal as a function of the number of atoms (n) in the crystal (Nielsen,
1964; Berner, 1980).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

168 Minerals and water

Qualitatively these relations may explain why a high degree of supersaturation yield poor crystals.
In this case the nucleation threshold is insignificant which allows many crystal nuclei to form, but
few to grow larger. Typical natural examples are Fe-oxyhydroxides and Al-hydroxides. On the other
hand, at low supersaturation, crystal growth will dominate over nucleation and result in few but
larger crystals. The energy barrier also explains why minerals in aquifers rarely form by spontaneous
nucleus formation from solution (homogeneous nucleation), but preferentially do so on pre-existing
surfaces (heterogeneous nucleation) which decrease the energy barrier considerably.
The energy barrier decreases with the crystal-water surface energy, which in turn tends to decrease
with increasing solubility of the mineral (Nielsen, 1986). This effect may explain the precipitation in
sequential order of lesser and lesser soluble minerals constituted of the same elements, known as the
Ostwald step rule (Steefel and Van Cappellen, 1990). First, the mineral with the lowest surface energy
precipitates, as it needs less supersaturation. This mineral may reduce the surface energy term of the
less soluble mineral by acting as template for nucleation and decreasing the surface tension with water.
Steefel and Van Cappellen (1990) explained the presence of halloysite in soils, by defining the inter-
facial energy of the less soluble kaolinite to be so high that the homogeneous nucleation rate would
be very small, and Lasaga (1998) demonstrates why quartz is sluggish to precipitate compared with
amorphous silica.

1
Sa
tu

2 Under-
ra
t
ed

Saturated
–
So
lu
bi
lit
y

3
eq

ed
ui

pB
lib

at Metastable
riu

ur
m

s at
p er
4
Su Hetero-
geneous
Homo- nucleation
geneous
nucleation

pA
Figure 4.23. Regimes of crystal growth for ionic substances A and B that may form mineral AB.
pA  log[A] and pB  log[B]. (Nielsen and Toft, 1984).

The different regimes of crystal formation are illustrated schematically in Figure 4.23 for the activities
of ionic substances A and B that may form a mineral AB. In subsaturated solutions, region (1), only dis-
solution may take place. The line separating region (1) and (2) corresponds to the solubility product. In
region (2), existing crystals may grow but no nucleation can take place. With increasing supersatura-
tion, region (3), crystal growth can be accompanied by nucleation on existing surfaces of other miner-
als etc. And finally in region (4) the degree of supersaturation has become high enough to allow
homogeneous nucleation.
An illustration of crystal growth is shown for fluorite in Figure 4.24. The rate of growth is plot-
ted versus the distance from equilibrium and since the straight line describing the data points has a
slope of 3, the rate is proportional to (  1)3.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 169

m/m0  2
Cs 0.22 mM
5 25°C

log ————
mol/s g
J/m0

6

7
0 0.2 0.4 0.6
log(1)

Figure 4.24. The rate (J) of fluorite crystal growth as a function of the distance from equilibrium (  IAP / K)
(Christoffersen et al., 1988).

PROBLEMS
4.1. The following table lists the composition of some fluoride rich waters from A. Maarum, Denmark, B.
Rajasthan, India, C. Lake Abiata, Ethiopia.

Sample: A B C

pH 7.8 7.3 9.62

Na (mM) 19.1 47.9 194
K (mM) 0.358 0.150 4.91
Mg2 (mM) 1.19 0.785 0.023
Ca2 (mM) 1.05 0.675 0.042
Cl (mM) 5.67 17.4 53.9
HCO 3 (mM) 17.8 14.8 138
SO 2
4 (mM) 0.00 5.20 0.15
NO3 (mM) 0.032 7.80 –
F (mM) 0.089 0.356 6.28

a. Calculate the saturation state of these waters for fluorite without making any corrections for ionic
strength and complexes.
b. Do the same but correct for the ionic strength.
c. Do the same with PHREEQC.

4.2. The mineral villiaumite (NaF) has been proposed as a possible solubility control on dissolved fluoride.
While the solubility of fluorite can be found in any geochemistry book, this is not the case for villiaumite.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

170 Minerals and water

Fortunately, Wagman et al. (1982) list the following thermodynamic data for villiaumite and associ-
ated species.

G 0f H 0f

Villiaumite 543.5 kJ/mol 573.6 kJ/mol

Na 261.9 kJ/mol 240.1 kJ/mol
F 278.8 kJ/mol 332.6 kJ/mol

a. Calculate Kvilliaumite at 25°C

b. Calculate Kvilliaumite at 10°C
c. Check the saturation state of the water samples in Problem 4.1 for villiaumite and consider whether this
mineral could limit dissolved fluoride concentrations.

4.3. One could consider whether it is possible to decrease the fluoride content of the Rajasthan water in 4.1 by
addition of gypsum.
a. Calculate the saturation state of the Rajasthan water for gypsum. Would it increase or decrease when
complex CaSO40 is included. (The complete calculation considering both ionic strength and complexes
is to cumbersome to carry out by hand calculations.)
b. To what level could the fluoride concentration be reduced by gypsum addition if you neglect the effects
of ionic strength and complexes. Would these additional corrections increase or decrease the molar flu-
oride concentration?
c. Check your last conclusion by doing the calculation with PHREEQC.

4.4. It has been argued in the literature that ion exchange of Ca2 replacing adsorbed Na may increase fluo-
ride concentrations in groundwater. Inspect the water analyses of 4.1 to find evidence that supports such a
mechanism. Which additional condition is required?

4.5. Calculate the solubility (mg/L) of atmospheric O2 in water at 5°C, 15°C and 25°C, check with
PHREEQC.

O2(g) O2(aq)

G 0f 0 16.5 kJ/mol
H 0f 0 10.0 kJ/mol

4.6. Redo Example 4.9 and the questions using smithsonite (ZnCO3) and calcite

4.7. Redraw Figure 4.16 for mineral precipitation.

4.8. Pyrite forms in sediments mostly by a two step process: First metastable FeS is formed, which is then trans-
formed to FeS2 by the overall reaction:

FeS  S0 → FeS2

Two reaction mechanisms have been proposed: 1) A solid state reaction between FeS and S0.
2) Dissolution of both FeS and S0 and subsequent reaction between Fe2 and polysulfides. (Polysulfides
are Sn_S2 compounds which form by the reaction between H2S and S0). Experimental data of Rickard
(1975) are listed below. Rates are given as d(FeS2)/dt.

1) PH 2S variable (pH  7, FeS surf. area  1.6105 cm2, S0 surf. area  1.4103 cm2)
PH2S (atm) Rate (mol/L/s)
1.00 8.3107
0.50 4.8107

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 171

0.25 2.3107
0.10 9.5108
0.05 4.9108

2) S0-surf. area variable (pH  7, PH2S  1 atm, FeS-surf. area  1.6105 cm2).
S0-surf.area (cm2) Rate (mol/L/s)
1.4103 8.3107
7.0102 4.8107
3.5102 1.9107
1.4102 5.4108

3) FeS-surf. area variable (pH  7, PH2S  1 atm, S0-surf. area  1.4103 cm2).
FeS-surf.area (cm2) Rate (mol/L/s)
2.5105 2.0106
1.6105 8.3107
8.2104 1.5107
4.1104 4.8108
4) pH variable (PH S  1 atm, S0-surf. area  1.4103 cm2, FeS-surf. area  1.6105 cm2).
2

pH Rate (mol/L/s)
7.0 8.3107
7.5 7.0107
8.0 6.4107
a. Determine the reaction order for PH2S, pH, FeS-surface area and S°-surface area.
b. Construct a rate equation for the reaction.
c. Which of the two mechanisms is supported by the experimental data.

4.9. Compare CO2 emission by activity of man, with natural escape from soil in the Netherlands. Use: 0.1 m
thick humic soil (40% OC, b  1.4 g/cm3), aerobic respiration rate dC/dt k C, k  0.025/yr. Surface
of the Netherlands is 31010 m2, 16106 inhabitants emit 10 ton/yr/person.

4.10. A reaction flask is filled with harbor sludge and 20 mM Na2SO4 solution. Use PHREEQC to model the
reduction of SO42 with the rate for Organic_C given under RATES in PHREEQC.DAT. The flask con-
tains 2.68 M OC which is 8104 times more reactive than the terrestric organic matter of the Organic_C
rate (After data from Westrich and Berner, 1984).
a. Calculate the first order decay rate for OC in the experiment.

REFERENCES
Allison, J.D., Brown, D.S. and Novo-Gradac, K.J., 1991. MINTEQA2, a geochemical assessment data base and
test cases for environmental systems. Report EPA/600/3-91/-21. US EPA, Athens, GA.
Anbeek, C., 1992. Surface roughness of minerals and implications for dissolution studies. Geochim.
Cosmochim. Acta 56, 1461–1469.
Anderson, G.M. and Crerar, D.A., 1993. Thermodynamics in geochemistry, the equilibrium model. Oxford
Univ. Press, 588 pp.
Atkins, P.W. and de Paula, J. 2002. Atkins’ Physical chemistry, 7th ed. Oxford Univ. Press, 1149 pp.
Berner, R.A., 1978. Rate control of mineral dissolution under earth surface conditions. Am. J. Sci. 278,
1235–1252.
Berner, R.A., 1980. Early diagenesis – a theoretical approach. Princeton Univ. Press, 241 pp.
Berner, R.A., 1981. Authigenic mineral formation resulting from organic matter decomposition in modern sed-
iments. Fortschr. Miner. 59, 117–135.
Biber, M.V., Dos Santos Afonso, M. and Stumm, W., 1994. The coordination chemistry of weathering: IV.
Inhibition of the dissolution of oxide minerals. Geochim. Cosmochim. Acta 58, 1999–2010.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

172 Minerals and water

Böttcher, M.E., 1997. Comment on “Solid solution partitioning of Sr2+, Ba2+, and Cd2+ to calcite”. Geochim.
Cosmochim. Acta 61, 661–662.
Boudreau, B.P. and Westrich, J.T., 1984. The dependence of bacterial sulfate reduction on sulfate concentration
in marine sediments. Geochim. Cosmochim. Acta 48, 2503–2516.
Buffle, J.D., 1990. Complexation reactions in aquatic systems. Ellis Horwood, Chichester, 692 pp.
Burgess, J., 1988. Ions in solution. Ellis Horwood, Chichester, 191 pp.
Busenberg, E. and Plummer, L.N., 1989. Thermodynamics of magnesian calcite solid-solutions at 25°C and 1
atm. total pressure. Geochim. Cosmochim. Acta 53, 1189–1208.
Christoffersen, J., Christoffersen, M.R., Kibalczyc, W. and Perdok, W.G., 1988, Kinetics of dissolution and
growth of calcium fluoride and effects of phosphate. Acta Odontol Scand. 46, 325–336.
Christoffersen, J. and Christoffersen, M.R., 1979. Kinetics of dissolution of calcium hydroxy apatite. J. Crystal
Growth 47, 671–679.
Cox, J.D., Wagman, D.D. and Medvedev, V.A., 1989. CODATA key values for thermodynamics. Hemisphere
Publ. Corp., Washington, D.C.
Curti, E., 1999. Coprecipitation of radionuclides with calcite: estimation of partition coefficients based on a
review of laboratory investigations and geochemical data. Appl. Geochem. 14, 433–455.
Denbigh, K., 1971. The principles of chemical equilibrium. 3rd ed. Cambridge Univ. Press, 496 pp.
Dixon, D.G. and Hendrix, J.L., 1993. Theoretical basis for variable order assumption in the kinetics of leaching
of discrete grains. AIChE J. 39, 904–907.
Dove, P. and Rimstidt, J.D., 1993. Silica-water interactions. Rev. Mineral. 29, 259–308.
Fitzpatrick, R.W. and Schwertmann, U., 1982. Al-substituted goethite – an indicator of pedogenic and other
weathering environments in South Africa. Geoderma 27, 335–347.
Ganor, J. and Lasaga, A.C., 1998. Simple mechanistic models for inhibition of a dissolution reaction. Geochim.
Cosmochim. Acta 62, 1295–1306.
Garrels, R.M. and Christ, C.L., 1965. Solutions, minerals, and equilibria. Harper and Row, New York, 450 pp.
Glynn, P.D., 1991. MBSSAS: A code for the computation of Margules parameters and equilibrium relations in
binary solid-solution aqueous-solution systems. Comp. Geosc. 17, 907–920.
Glynn, P.D., 2000. Solid-solution solubilitites and thermodynamics: sulfates, carbonates and halides. Rev.
Mineral. 40, 481–511.
Glynn, P.D. and Reardon, E.J., 1990. Solid-solution aqueous-solution equilibria: thermodynamic theory and
representation. Am. J. Sci. 278, 164–201.
Glynn, P.D. and Reardon, E.J., 1992. Reply to a comment by Königsberger, E. and Gämsjager, H. Am. J. Sci.
292, 215–225.
Grassi, S. and Netti, R., 2000. Sea water intrusion and mercury pollution of some coastal aquifers in the
province of Grosseto (Southern Tuscany-Italy). J. Hydrol. 237, 198–211.
Gratz, A.I., Bird, P. and Quiro, G.B., 1990. Dissolution of quartz in aqueous basic solution. Geochim.
Cosmochim. Acta 54, 2911–2922.
Guggenheim, E.A., 1967. Thermodynamics. Elsevier, Amsterdam, 390 pp.
Handa, B.K., 1975. Geochemistry and genesis of fluoride-containing ground waters in India. Ground Water 13,
275–281.
Harvie, C.E., Eugster, H.P. and Weare, J.H., 1982. Mineral equilibria in the six-component seawater system, Na-
K-Mg-Ca-SO4-Cl-H2O at 25°C. II: Compositions of the saturated solutions Geochim. Cosmochim. Acta 46,
1603–1618.
Helgeson, H.C., Delany, J.M., Nesbitt, W.H. and Bird, D.K., 1978. Summary and critique of the thermodynamic
properties of rock forming minerals. Am. J. Sci., 278A, 1–229.
Hochella, M.F. and White, A.F. (eds), 1990. Mineral-water interface geochemistry. Rev. Mineral. 23, 603 pp.
Jacks, G., Rajagopalan, K., Alveteg, T. and Jönsson, M., 1993. Genesis of high-F groundwaters. Southern India.
Appl. Geochem. Suppl. Issue 2, 241–244.
Jacks, G., Bhattacharya, P. and Singh, K.P., 2000. High-fluoride groundwaters in India. In P.L. Bjerg, P.
Engesgaard and Th.D. Krom (eds), Groundwater 2000, 193–194, Balkema, Rotterdam.
Langmuir, D., 1997. Aqueous environmental geochemistry. Prentice-Hall, Englewood Cliffs, 600 pp.
Larsen, O. and Postma, D., 2001. Kinetics of bulk dissolution of lepidocrocite, ferrihydrite, and goethite.
Geochim. Cosmochim. Acta 65, 1367–1379.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 173

Lasaga, A.C., 1998. Kinetic theory in the earth schiences. Princeton Univ. Press, 811 pp.
Lewis, G.N. and Randall, M., 1961. Thermodynamics. 2nd ed. McGraw-Hill, New York, 723 pp.
Lippmann, F., 1980. Phase diagrams depicting aqueous solubility of binary mineral systems. N. Jb. Min. Abh.
139, 1–25.
McIntire, W.L., 1963. Trace element partition coefficients – a review of theory and applications to geology.
Geochim. Cosmochim. Acta 27, 1209–1264.
Millero, F.J. and Pierrot, D., 1998. A chemical equilibrium model for natural waters. Aq. Geochem. 4, 153–199.
Monnin, C. and Schott, J., 1984. Determination of the solubility products of sodium carbonate minerals and an
application to trona deposition in Lake Magadi (Kenya). Geochim. Cosmochim. Acta 48, 571–581.
Morse, J.W. and Berner, R.A., 1979. Chemistry of calcium carbonate in the deep oceans. In E.A. Jenne (ed.),
Chemical Modelling in Aqueous Systems. ACS Symp. Ser. 43, 499–535.
Nagy, K.L. and Lasaga, A.C., 1992. Dissolution and precipitation kinetics of gibbsite at 80°C and pH 3: The
dependence on solution saturation state. Geochim. Cosmochim Acta 56, 3093–3111
Nielsen, A.E., 1964. Kinetics of precipitation. Pergamon Press, Oxford, 151 pp.
Nielsen, A.E., 1984. Electrolyte crystal growth mechanisms. J. Crystal Growth 67, 289–310.
Nielsen, A.E., 1986. Mechanisms and rate laws in electrolyte crystal growth. In J.A. Davis and K.F. Hayes (eds),
Geochemical Processes at Mineral Surfaces. ACS Symp. Ser. 323, 600–614.
Nielsen, A.E. and Toft, J.M., 1984. Electrolyte crystal growth kinetics. J. Crystal Growth 67, 278–288.
Nordstrom, D.K. and Jenne, E.A., 1977. Fluorite solubility equilibria in selected geothermal waters. Geochim.
Cosmochim. Acta 41, 175–188.
Nordstrom, D.K. and Munoz, J.L., 1994. Geochemical thermodynamics, 2nd ed. Blackwell, Oxford, 493 pp.
Parkhurst, D.L., 1990. Ion association models and mean activity coefficients of various salts. In D.C. Melchior
and R.L. Basset (eds), Chemical modeling of aqueous systems II. ACS Symp. Ser. 416, 30–43.
Parkhurst, D.L. and Appelo, C.A.J., 1999. User’s guide to PHREEQC (version 2)-a computer program for spe-
ciation, batch-reaction, one-dimesnional transport, and inverse geochemical calculations. U.S. Geol. Surv.
Water Resour. Inv. Rep. 99–4259, 312 pp.
Pitzer, K.S., 1981. Characteristics of very concentrated aqueous solutions. In F.E. Wickman and D.T. Rickard
(eds), Chemistry and geochemistry of solutions at high temperatures and pressures. Phys. Chem. Earth 3
and 4, 249–272.
Plummer, L.N. and Busenberg, E., 1982. The solubilities of calcite, aragonite and vaterite in CO2-H2O solutions
between 0 and 90°C, and an evaluation of the aqueous model for the system CaCO3-CO2-H2O. Geochim.
Cosmochim. Acta 46, 1011–1040.
Plummer, L.N., Parkhurst, D.L., Fleming, G.W. and Dunkle, S.A., 1988. A computer program incorporating
Pitzer’s equations for calculation of geochemical reactions in brines. U.S. Geol. Surv. Water Resour. Inv.
Rep. 88–4153.
Postma, D., 1983. Pyrite and siderite oxidation in swamp sediment. J. Soil Sci. 34, 163–182.
Rickard, D.T., 1975. Kinetics and mechanism of pyrite formation at low temperatures. Am. J. Sci. 275, 636–652.
Riehl, N., Sizmann, R. and Hidalgo, P., 1960. Zur Deutung der Verteilung kleinster Fremdsubstanzmengen
zwischen einem wachsenden Kristall und der Lösung. Z. Phys. Chem. 25, 351–359.
Rimstidt, J.D. and Barnes, H.L., 1980. The kinetics of silica water reactions. Geochim. Cosmochim. Acta 44,
1683–1699.
Rimstidt, J.D., Balog, A. and Webb, J., 1998. Distribution of trace elements between carbonate minerals and
aqueous solutions. Geochim. Cosmochim. Acta 62, 1851–1863.
Robie, R.A., Hemingway, B.S. and Fisher, J.R., 1978. Thermodynamic properties of minerals and related sub-
stances at 298.15 K and 1 bar (105 Pascals) pressure and at higher temperatures. U.S. Geol. Surv. Bull.
1452, 456 pp.
Saxena, V.K. and Ahmed, S., 2003. Inferring the chemical parameters for the dissolution of fluoride in ground-
water. Env. Geol. 43, 731–736.
Schwartz, M. O. and Ploethner, D., 1999. Removal of heavy metals from mine water by carbonate precipitation
in the Grootfontein-Omatako canal, Namibia. Env. Geol. 39, 1117–1126.
Schwarzenbach, R.P., Gschwend, P.M. and Imboden, D.M., 1993. Environmental organic chemistry. Wiley and
Sons, New York, 681 pp.
Sparks, D.L., 1989. Kinetics of soil chemical processes. Academic Press, San Diego, 210 pp.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

174 Minerals and water

Sparks, D.L. and Jardine, P.M., 1984. Comparison of kinetic equations to describe K-Ca exchange in pure and
in mixed systems. Soil Sci. 138, 115–122.
Steefel, C.I. and Van Cappellen, P., 1990. A new approach to modeling water-rock interactions: the role of pre-
cursors, nucleation, and Ostwald ripening. Geochim. Cosmochim. Acta 54, 2657–2677.
Stipp, S.L., 1994. Understanding interface processes and their role in the mobility of contaminants in the
geosphere: The use of surface sensitive techniques. Eclogae geol. Helv. 87, 335–355.
Stumm, W. and Morgan, J.J., 1996. Aquatic Chemistry. 3rd ed. Wiley and Sons, New York, 1022 pp.
Sverdrup, H. and Warfvinge, P., 1988. Weathering of primary silicate minerals in the natural soil environment
in relation to a chemical weathering model. Water Air Soil Poll. 38, 387–408.
Tesoriero, A.J. and Pankow, J.F., 1996. Solid solution partitioning of Sr2+, Ba2+, and Cd2+ to calcite. Geochim.
Cosmochim. Acta 60, 1053–1063.
Truesdell, A.H. and Jones, B.F., 1973. Wateq, a computer program for calculating chemical equilibria of natu-
ral waters. U.S. Geol. Surv. NTIS PB-220 464, 73 pp.
Van Lier, J.A., De Bruyn, P.L. and Overbeek, J.T.G., 1960. The solubility of quartz. J. Phys. Chem. 64,
1675–1682.
Wagman, D.D., Evans, W.H., Parker, V.B., Schumm, R.H., Halow, I., Bailey, S.M., Churney, K.L. and Nuttall,
R.L., 1982. The NBS tables of chemical thermodynamic properties: Selected values for inorganic and C1
and C2 organic substances in SI units. J. Phys. Chem. Ref. Data Suppl. 2, 392 pp.
Westrich, J.T. and Berner, R.A., 1984. The role of sedimentary organic matter in bacterial sulfate reduction: the
G model tested. Limnol. Ocanogr. 29, 236–249.
White, A.F., 1995. Chemical weathering rates of silicate minerals in soils. Rev. Mineral. 31, 407–461.
White, A.F. and Brantley, S.L. (eds), 1995. Chemical weathering rates of silicate minerals. Rev. Mineral. 31, 583
pp.
Wolery, T.J., 1992. EQ3/6, A solftware package for geochemical modeling of aqueous systems. UCRL-MA-
110662 Pts I-III, Lawrence Livermore Nat. Lab., Cal.
Woods, T.L. and Garrels, R.M., 1987. Thermodynamic values at low temperature for natural inorganic materi-
als. Oxford Univ. Press, New York, 242 pp.
Zack, A.L., 1980. Geochemistry of fluoride in the Black Creek Aquifer system of Horry and Georgetown coun-
ties, South Carolina-and its physiological implications. U.S. Geol. Surv. Water-Supply Paper 2067, 40 pp.
Zhang, J.W. and Nancollas, G.H., 1990. Mechanisms of growth and dissolution of sparingly soluble salts. Rev.
Mineral. 23, 365–396.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

5

Carbonates and Carbon Dioxide

Limestones and dolomites often form productive aquifers with favorable conditions for groundwa-
ter abstraction. The main minerals in these rocks are Ca- and Mg-carbonates which dissolve easily
in groundwater and give the water its ‘hard’ character. Usually, the carbonate rock consists of recrys-
tallized biological material with a high porosity and a low permeability. Groundwater flow is then
restricted to more permeable fracture zones and karst channels produced by carbonate dissolution.
Prolonged dissolution over periods of thousands of years may result in the development of karst
landscapes and fantastic features in caves. Apart from the carbonate rocks that consist exclusively of
carbonate minerals, sands or sandstones, and marls and clays may contain carbonate minerals as
accessory minerals or as cement around the more inert grains.
On a world-wide basis, the effect of carbonate dissolution on water compositions is quite
conspicuous. Figure 5.1 compares the average river water composition from different conti-
nents by the relation between total dissolved solids (TDS) and the concentrations of individual ions.

1.6
North America

Europe
Asia

1.4

1.2
Africa

1.0
meq or mmol/L

South America

0.8 Ca2

HCO
3
0.6
SO42 

Mg2


0.4
Rain water

Na

Cl
0.2
SiO2




   K 

0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Total dissolved solids (meq or mmol/L)

Figure 5.1. The relation between total dissolved solids and concentrations of individual ions in average river
water compositions from different continents (Modified from Garrels and Mackenzie, 1971; mmol/L for SiO2).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

176 Carbonates and carbon dioxide

The high TDS values are mainly related to increases of Ca2 and HCO 3 , which predominantly are
due to carbonate dissolution. The abundance of carbonate rocks in Europe is clearly reflected in the
river water composition.
This chapter presents aspects of the mineralogy, the aqueous chemistry and the isotope
chemistry required to appreciate the role of carbonate dissolution as regulator of natural water
compositions.

5.1 CARBONATE MINERALS

Some carbonate minerals and their solubility products are listed in Table 5.1. Carbonate minerals
crystallize with either a trigonal or an orthorhombic crystal structure, depending on the ionic radius
of the cation. The smaller ions form trigonal minerals, while the larger ones form orthorhombic min-
erals. The Ca2 ion is intermediate with the trigonal mineral calcite, being the main component of
older limestones, and the orthorhombic aragonite, which is common in recent carbonate sediments
and the mineral forming pearls and mother of pearl. Rock-forming carbonate minerals are calcite,
dolomite and aragonite. The other carbonate minerals, like siderite or rhodochrosite, occur only in small
amounts if present at all. Still, they may exercise an important control on dissolved concentrations of
Fe2 and Mn2 (Al et al., 2000; Jurjovec et al., 2002).
Solid solutions of different carbonate minerals are common (Reeder, 1983; Chapter 4). Solid
solution may occur among minerals that belong to the same crystal group, within the limits determined
by the degree of deformation allowed in the crystal lattice. For example, Mg atoms may substitute for
Ca in calcite, and Sr for Ca in aragonite, but Sr will not replace much of Ca in calcite because the ionic
radii are too different. Solid solutions between calcite and magnesite are called Mg-calcites. Usually,
they are subdivided into low Mg-calcites with 5 mol% and high Mg-calcites with 5–30 mol% Mg.
The composition of Mg-calcites in recent marine sediments, mostly from tropical to subtropical areas,
is displayed in Figure 5.2, showing a composition with 10–15 mol% MgCO3. The biological uptake
of Mg in the shells of marine organisms is temperature dependent (Morse and Mackenzie, 1990).
Carbonate shells sampled along the coast of Holland where the summer seawater temperature is about
18°C, contain less than 1% Mg.
Another important rock-forming carbonate mineral is dolomite with a structure displayed
in Figure 5.3. Dolomite shows a high degree of ordering with Ca and Mg atoms located in
separate layers in the crystal structure. This is different from the structure of high Mg-calcite in
which Ca and Mg atoms are distributed more or less at random in the Ca layer (Figure 5.3).
A 50 mol% Mg-calcite is therefore not the same as dolomite. Thermodynamically, dolomite
should be treated as a pure phase while for Mg-calcites the rules of solid solutions apply.

Table 5.1. Mineralogy and solubility of some carbonates. Log K’s from Nordstrom et al., 1990; cation radii
from Dowty, 1999 except for Sr.

Trigonal Formula log K Cation Orthorhombic Formula log K Cation

radius (Å) radius (Å)

Calcite CaCO3 8.48 1.12 Aragonite CaCO3 8.34 1.12

Magnesite MgCO3 8.24 0.72 Strontianite SrCO3 9.27 1.18
Rhodochrosite MnCO3 11.13 0.84 Witherite BaCO3 8.56 1.42
Siderite FeCO3 10.89 0.74 Cerussite PbCO3 13.1 1.18
Smithsonite ZnCO3 10.01 0.83
Otavite CdCO3 13.74 0.99
Dolomite CaMg(CO3)2 17.09

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate minerals 177

40
107 Samples
Shallow-water limestones
Av mol % MgCO3
12.7  3.9 Cemented carbonate sands

Ooids
30
Deep-sea limestones from
semi-enclosed basins
Deep-sea limestones
No. of samples

Beachrock cements
20 Intergranular cementation
(cryptocrystallization)
Seamounts and guyots

10

1 5 10 15 20
Mol % MgCO3

Figure 5.2. The composition of Mg-calcite cements precipitated from seawater (Mucci 1987).

On the other hand, dolomite does show solid solution like most other minerals. The Ca/Mg ratio may
vary, and substitution of Fe and Mn is common. The highly ordered structure of dolomite compli-
cates its precipitation from aqueous solutions at low temperatures and the formation of dolomite is
still a matter of lively debate (Hardie, 1987; Budd, 1997; Van Lith et al., 2002; Deelman, 2003).

Ca
Dolomite
CaMg(CO3)2
C axis

Calcite
Substitution Substitution
CaCO3 of Fe, Sr, Na of Ca
C
Ca
Ca Ca 1 CO3 plane
A 2 Ca plane Layered
5
4 3 CO3 plane superstructure
Ca 3 4 Mg plane
2
1
0 Ca
Ca
Ca CO3
Ca
C Mg

Ca

Figure 5.3. The structures of calcite and dolomite. Note the separation of Ca and Mg atoms in different layers
in the dolomite (Morrow, 1982; Deer, Howie and Zussman, 1966). The structure of minerals can be displayed
easily with the computer program Atoms (Dowty, 1999; www.shapesoftware.com).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

178 Carbonates and carbon dioxide

Figure 5.4. Scanning electron microscope photographs of recent chalk ooze (left) consisting skeletal remains
of organisms and (right) partly recrystallized chalk ooze where inorganic calcite crystals become dominant.
Picture size is 3  3 m (Fabricius, 2003).

Carbonate rocks are with few exceptions, like travertine, biogenic in origin. They consist of
the skeletal remains of different organisms as is illustrated in Figure 5.4 showing a subrecent
chalk ooze. Older carbonate rocks still bear traces of their organic origin but generally are exten-
sively recrystallized (cf. Reeder, 1983). Recrystallization is also reflected by the observation that
recent carbonate sediments generally consist of aragonite and high Mg-calcite while older carbonate
rocks (Pre-Pleistocene (!106 yr)) contain the more stable calcite and dolomite (Hanshaw and
Back, 1979).

5.2 DISSOLVED CARBONATE EQUILIBRIA

If we put small calcite crystals in a beaker with distilled water, cover the beaker to exclude CO2
from air, and return the next day to analyze the solution, we would expect from the dissolution
reaction:

CaCO3 → Ca 2  CO32 (5.1)

that:

[Ca 2 ]  [CO32 ]

and in combination with the solubility product (Table 5.1)

Kcalcite  [Ca 2 ][CO32 ]  [Ca 2 ]2  108.48

Thus, the predicted calcium concentration is

[Ca 2 ]  104.25  0.06 mmol/L.

For comparison, Table 5.2 shows the composition of two groundwaters in a carbonate aquifer. The
Ca2 concentration in the groundwater of the Alpujarras aquifer (Table 5.2) is 63 and 24 mg/L, cor-
responding to 1.6 and 0.60 mmol/L respectively. Both values are many times higher than the value
of 0.06 mmol/L estimated above and the chemistry involved is clearly more complex.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dissolved carbonate equilibria 179

Table 5.2. Groundwater composition in the Triassic Alpujarras aquifer, southern Spain (Cardenal
et al., 1994). Units are mg/L. Alkalinity is listed as HCO
3 in mg/L.

Sample Temp°C pH Na Mg2 Ca2 Cl Alkalinity SO42

2 18.3 7.23 9 20 63 16 269 24

44 18.5 7.91 10 20 24 13 133 20

The higher Ca2 concentration in the groundwater is due to the reaction of calcite with carbon diox-
ide derived from respiration or oxidation of organic matter. Carbon dioxide reacts with H2O to form
carbonic acid (H2CO3). The acid provides protons (H) which associate with the carbonate ion
(CO32) from calcite to form bicarbonate (HCO 3 ). This is similar to complexation of Ca
2
and
2
SO4 ions leading to an increase of the solubility of gypsum (Example 4.3). The overall reaction
between carbon dioxide and CaCO3 is:

CO 2(g)  H 2 O  CaCO3 → Ca 2  2HCO

3 (5.2)

This reaction is fundamental for understanding the behavior of CaCO3 dissolution and precipitation
in nature. An increase of CO2 results in dissolution of CaCO3. Removal of CO2 causes CaCO3 to pre-
cipitate. An example is CO2 degassing from carbonate water in springs and stream rapids where
CaCO3 precipites in the form of travertine. Reaction (5.2) also demonstrates the direct coupling
between the biological cycle and carbonate mineral-reactions. Degradation of organic matter, or res-
piration, produces CO2 and enhances the dissolution of carbonates in soils, while photosynthesis
consumes CO2 and results, for example, in CaCO3 precipitation in lakes.

5.2.1 The carbonic acid system

When CO2 dissolves in water, gaseous CO2(g) becomes aqueous CO2(aq), and some of this associates
with water molecules to form carbonic acid, H2CO3:

CO 2(g) → CO 2(aq) (5.3)

and subsequently:

CO 2(aq)  H 2O → H 2CO3 (5.4)

Actually CO2(aq) is at 25°C about 600 times more abundant than H2CO3, but to facilitate calculations
a convention is adopted in which the two species are summed up as H2CO*3. The overall reaction
becomes then:

CO 2(g)  H 2O → H 2CO*3 (5.5)

where H2CO*3  CO2(aq)  H2CO3.

During dissociation, carbonic acid stepwise releases two protons. The concentrations of the dis-
solved carbonate species therefore depend on the pH of the solution. The reactions and mass action
constants that are needed to calculate the composition of the solution are given in Table 5.3. The con-
stants tabulated are approximate values, adequate for manual calculations.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

180 Carbonates and carbon dioxide

Table 5.3. Equilibria in the carbonic acid system with approximate equilibrium constants at 25°C. More
precise values can be found in PHREEQC.DAT.

H 2 O ↔ H  OH K W  [H ][OH]  1014.0

CO 2(g)  H 2 O ↔ H 2 CO*3 K  [H CO* ] / [P ]  101.5
H 2 3 CO2 (5.5)
H 2 CO*3 ↔ H  HCO
3 K1  [H ][HCO *
3 ] / [H 2 CO3 ]  10
6.3 (5.6)
HCO
3 ↔ H  CO32 K 2  [H ][CO32 ] / [HCO 3 ]  10
10.3 (5.7)

Equation (5.6) can be rewritten as

[HCO3] 106.3
 (5.6a)
[H 2CO*3 ] 10pH

At a pH of 6.3, the activities of HCO 3 and H2CO3 are equal. With pH ! 6.3, HCO3 is the


predominant species and at pH  6.3 there is more H2CO 3 . The same relation for the CO3 / HCO
* 2
3
couple shows that the two species have equal activity at pH  10.3. At a pH ! 10.3, CO32 becomes
the predominant species while HCO 3 is more abundant at pH  10.3. Figure 5.5 summarizes how
the different aqueous carbonate species vary with pH.

100

80

Percentage HCO3

 2
60 H2CO*3 HCO3 CO3

40
6.3 10.3

20

0
3 5 7 9 11 13
pH

Figure 5.5. Percentage of HCO

3 of total dissolved carbonate as function of pH.

The distribution of dissolved carbonate species can be calculated for two idealized cases. In the first
case the CO2 pressure is known and constant, a system that is termed open with respect to
carbon dioxide gas. In the second case, total inorganic carbon (TIC) is known and constant and the
system is called closed with respect to CO2.
The system open with respect to CO2 gas considers the [PCO ] as constant. Using a log- 2
transformation of the mass action Equation (5.5) we obtain:

log[H 2CO*3 ]  log[PCO ]  1.5 (5.8)

2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dissolved carbonate equilibria 181

indicating that also [H2CO*3] is constant and independent of the pH. The activity of HCO
3 as a func-
tion of carbonic acid activity and pH is derived from Equation (5.6):

log[HCO *
3 ]  6.3  log[H 2 CO3 ]  pH (5.9)

To calculate the activity of CO2 *

3 as a function of pH at constant [H2CO 3 ] we combine Reactions
(5.6) and (5.7). Adding two reactions, means multiplication of the two mass action constants, and
subtraction requires similarly that the two be divided.

H 2CO*3 ↔ H  HCO 3; K1  106.3

 
HCO3 ↔ H  CO3 ; 2 K 2  1010.3
 (5.10)
+
H 2CO*3 ↔ 2H  CO32; K  1016.6

And the law of mass action is, after log-transformation:

log[CO32 ]  16.6  log[H 2CO*3 ]  2pH (5.11)

When the gas pressure of CO2, or the activity [PCO ], is known, the activities of all dissolved car-
2

bonate species can be calculated accordingly. Figure 5.6 shows a pH-log[activity] diagram in which
the activities of the CO2-species are given for a constant gas pressure of CO2 of 0.01 atm. The bold
line indicates the sum of the dissolved carbonate species (ΣCO2 or TIC). The log(ΣCO2) increases
1:1 with pH for log K1  pH  log K2, and 2:1 for pH ! log K2.
The sum of the dissolved carbonate species is given by the mass balance equation:
TIC  mH *  mHCO  mCO2 (5.12)
2 CO3 3 3

The mass of a species is expressed in molal concentration while the mass action equations in Table
5.3 contain activities. For simplicity we equate activity and molal concentration, and then the con-
centrations can be calculated and added up as shown in Example 5.1.

pH
0 2 4 6 8 10 12 14

0 P CO2 (ΣCO2)

2
H2CO3
log i

4

HCO3
6
CO32

8

10

Figure 5.6. The activity of CO2 species in water as function of pH, at a constant CO2 pressure of 0.01 atm. The
bold line indicates the sum of the carbonate species.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

182 Carbonates and carbon dioxide

EXAMPLE 5.1. Calculate TIC in water at pH  7 and pH  10, when CO2 pressure is 0.01 atm

ANSWER:
We calculate the concentrations of the individual species, and add them up:
pH  7 pH  10
1.5 3.5
[H2CO*3]  10 [PCO2] 10 103.5
[HCO 3 ]  10
6.3
[H2CO*3] / [H] 102.8 100.2
[CO32]  1010.3 [HCO 
3 ] / [H ] 106.1 100.1
ΣCO2  102.72 100.38

In a system closed with respect to CO2, the sum of the dissolved carbonate species is considered
constant. The distribution of the dissolved carbonate species, as a function of pH, can be calculated by
substituting the mass action expressions (5.6) and (5.7) in (5.12). The results are shown in Figure 5.7
and the line indicating TIC (or ΣCO2) is now parallel to the pH axis. Again we note that carbonic acid
is dominant at pH  6.3 while CO32 becomes dominant at pH ! 10.3, and in between these pH
values HCO 3 is the major dissolved carbonate species in water. At pH ! 6.3, the activity of carbonic
acid, and the PCO decreases one log-unit with each unit of pH increase; above pH  10.3, it decreases
2

two log-units with each unit pH increase. At pH’s in the near vicinity of the equivalence pH (where
two species have equal activity) a light rounding of the curves indicates that two species have a signi-
ficant contribution to the TIC.

pH
0 2 4 6 8 10 12 14

P CO2
2 
H2CO3 HCO3 CO32
Σ CO2
4

 HCO3
HCO3
6
log i

8
P CO2
10 CO32 H2CO3

12

14

Figure 5.7. The activity of CO2 species in water at constant total inorganic carbon (TIC  1 mmol/L).

The pH-log[activity] diagram is illustrative for depicting effects of pH on the relative concentra-
tions of solute species. The actual concentration of an individual species as a function of pH can be
found quickly by using the speciation factor . For example for HCO 3:

mHCO   CO 2 (5.13)

3

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dissolved carbonate equilibria 183

where  gives the fraction of TIC made up by HCO

3 . The speciation factor   mHCO3 / ΣCO2, or:

1 
∑ CO2 
mH
2 CO3

mHCO
3

mCO2
3

mHCO mHCO mHCO mHCO

3 3 3 3

[H 2CO*3 ]1 [HCO

3] [CO32 ]1
  
[HCO
3 ]0

[HCO3 ] [HCO 3 ]2

[H ]1 K 
  1  2 1
K10 [H ]2

where 0,1,2 are the activity coefficients for the neutral molecule and the singly and doubly charged
ion, respectively. This equation was used to calculate the distribution of HCO 3 as function of pH
shown in Figure 5.5.

QUESTIONS:
Using Equation (5.7), give the pH where HCO 2
3 and CO3 have equal activity
ANSWER: 10.3.
Give the ratio [CO32] / [HCO3 ] at pH  9.3 and 11.3?
ANSWER: 1:10 and 10:1.
Estimate the acidity constant of H2CO3 (which is about 1/600 part of H2CO*3 , cf. Kern, 1960)?
ANSWER: 600106.3  103.5

5.2.2 Determining the carbonate speciation in groundwater

This section describes how the speciation of dissolved carbonate is determined for specific ground-
water samples. The four equations (5.5), (5.6), (5.7) and (5.12) together define the carbonate species.
The number of variables in the equations amounts to six. Therefore the determination of any two
variables fixes the remaining variables in the aqueous carbonate system. In usual practice one of the
following pairs of variables are analyzed: pH and alkalinity, TIC and alkalinity, PCO and alkalinity, 2
or PCO and pH. For each of these couples PHREEQC calculates the same carbonate species
2
concentrations, as shown in Example 5.2.

EXAMPLE 5.2. Calculation of the aqueous carbonate system with PHREEQC

Given the aqueous carbonate composition (temperature is 25°C):
pH  7.23; Alk  4.41 meq/L; TIC  4.959 mmol/L; log PCO2  1.789
The following input solutions all produce identical results (note that meq is not a valid input format, mmol
is used for Alk):
SOLUTION 1 # pH and Alkalinity...
pH 7.23
Alkalinity 4.41
END

SOLUTION 2 # TIC and Alkalinity...

C(4) 4.959
Alkalinity 4.41
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

184 Carbonates and carbon dioxide

SOLUTION 3 # PCO2 and Alkalinity...

pH 7 CO2(g) -1.789 # pH is adapted to equilibrium with [PCO2] = 10-1.789
Alkalinity 4.41
END

SOLUTION 4 # pH and PCO2...

pH 7.23
C(4) 1 CO2(g) -1.789 # Total C(4) is adapted to equilibrium with
# [PCO2] = 10-1.789
END

Enter these input files in PHREEQC and inspect the results.

In most cases the carbonate speciation is determined by measuring pH and alkalinity. This is best
done in the field immediately after sample retrieval, to avoid degassing of CO2. The alkalinity of a
water sample is equal to the number of equivalents of all dissociated weak acids. Phosphoric acid
and other weak acids may contribute to some extent, and in leachates from waste site various organic
acids may be important. However, normally only the carbonate ions are of quantitative importance
for the measured alkalinity. It is then equal to:
Alk  mHCO  2 mCO2 (5.14)
3 3

At pH values below 8.3, less than 1% of the carbonic acid is present as CO32, and then only the first
term is of importance.
The alkalinity is often determined by titration with a HCl or H2SO4 solution of known normality
towards an endpoint pH of about 4.5, as indicated by a color reaction. The most accurate way to
determine alkalinity is the so-called Gran titration (Stumm and Morgan, 1996). The principle of the
method is that past the point where all HCO 3 has been converted to H2CO3 (at a pH of about 4.3),
the concentration of H increases linearly with the amount of H added. In practice, the volume of
HCl added is plotted against the Gran function

F  (V  V0 ) 10pH (5.15)

where V is volume of acid added, and V0 is start volume of the sample. The equivalence point is
obtained by backwards extrapolation from the linear part of the curve as shown in Figure 5.8.

2.4
(V  V0) . 10pH . 102

2.0

1.6

1.2
Equivalence
0.8 point
0.4

0
0.2 0.3 0.4 0.5 0.6 0.7 0.8
V (mL)

Figure 5.8. Illustration of the Gran plot method for determining alkalinity by titration with acid.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Dissolved carbonate equilibria 185

The advantages of this method are several. First, the equivalent point need not be determined accurately
during the titration as such, second, the equivalent point is determined by linear regression on a num-
ber of pH readings, and third, calibration errors of the pH meter are not important since only changes
in [H] are used. Once the pH and alkalinity of the water sample are known, the activity of CO32, the
PCO and the saturation index for carbonate minerals can be calculated as is illustrated in Example 5.3.
2

EXAMPLE 5.3. Manual calculation of carbonate speciation and SIcalcite

For water sample 2 of the Alpujarras aquifer (Table 5.2) calculate the SIcalcite and PCO2 at 25°C:
pH  7.23; Alk  269 mg HCO
3 /L  10
2.36
mol/L; Ca2  63 mg/L  102.80 mol/L

ANSWER:
Using the equations from Table 5.3:
log[CO32 ]  log K 2  pH  log[HCO
3 ]  10.3  7.23  (2.36)  5.43
→ SI calcite  log([Ca 2 ][CO32 ])  log K cc  2.80  (5.43)  (8.48)  0.25
log[H 2 CO* 
3 ]  log K1  pH  log[HCO3 ]  6.3  7.23  (2.36)  3 .29
→ log [P ]  log K  log[H CO* ]  1.5  3.29  1.79
CO2 H 2 3

The more accurate calculation which includes activity coefficients, complexing and correction of
mass action constants to in situ temperature is carried out by PHREEQC for both samples of the
Alpujarras aquifer (Table 5.2) in Example 5.4.

EXAMPLE 5.4. PHREEQC calculation of carbonate speciation and SIcalcite

Speciate the groundwater analyses of the Alpujarras aquifer listed in Table 5.2.

ANSWER:
For sample 2, the PHREEQC input file is
SOLUTION 1 # Cardenal et al 94 sample 2
pH 7.23
temp 18.3
units mg/L
Na 9
Mg 20
Ca 63
Cl 16
Alkalinity 269 as HCO3
S(6) 24
END

and similar for sample 44. In the output file we find:

Sample TIC (mmol/L) log[PCO2] log[CO32] log[Ca2] SIcalcite

2 4.98 1.85 5.566 2.979 0.1

44 2.22 2.83 5.185 3.364 0.1

The PHREEQC calculations indicate that both samples are very slightly subsaturated for calcite
even though their water compositions actually are quite different. The manual calculation for sam-
ple 2 shows slight supersaturation, the difference resulting from the effects of temperature, activity
coefficients and complexing.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

186 Carbonates and carbon dioxide

pH measured

5
5 6 7 8
pH calculated

Figure 5.9. A comparison of pH calculated with PHREEQC from TIC/alkalinity measurements, with pH
values measured in the field.

A reliable estimate of [CO32] requires a good measurement of pH and alkalinity. Sometimes the pH
measurement of field samples is not reliable due to temperature differences of electrode, buffers and
sample, degassing of CO2, sluggish or faulty electrodes, etc. As seen above, errors in measured pH
directly influence the calculation of the CO32 activity from alkalinity. As an alternative the TIC
(cf. Table 1.6) and alkalinity can be measured, and the pH and the aqueous carbonate speciation is
then calculated using PHREEQC (cf. Example 5.2). A comparison of measured field pH values and
pH values calculated with PHREEQC according to this method is shown in Figure 5.9. Generally
there is good agreement between measured and calculated pH which strengthens the confidence in
both measurements and theory. However, at low pH, where pH buffering by alkalinity is poorest, the
measured pH values tend to be slightly higher, perhaps due to some degassing.

QUESTION:
Make a PHREEQC input file to calculate the speciation for sample 44, Table 5.2.
Find SIcalcite when the samples’ temperature is 25°C?
ANSWER: 0.01

5.3 CARBON DIOXIDE IN SOILS

The CO2 concentration in the atmosphere is low, 0.03 vol%, which is equivalent to PCO  3104 2

of 1 atm  103.5 atm. The CO2 pressure of groundwater is easily one or two orders of magnitude
higher, primarily as result of uptake of carbon dioxide during the infiltration of rainwater through
the soil. In the soil, CO2 is generated by root respiration and decay of labile organic material (Hanson
et al., 2000):
CH 2O  O 2 → CO 2  H 2O (5.16)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon dioxide in soils 187

CO2 (vol%)
0 1 2 3 4 5 6
0.0
Gravel pit
Woodland
0.5 Grass land
Golf course
1.0

Depth (m) 1.5

2.0

2.5

3.0

3.5

4.0

Figure 5.10. Concentration of CO2 gas in sandy unsaturated zone with different land use. The data were
collected in the growth season, July 1992. The concentration at 0 m depth corresponds to the atmospheric
CO2 concentration. Modified from Lee 1997.

The mean residence time of organic carbon in soils is short, 10–100 years for a temperate climate
soil (Russell, 1973; Keller and Bacon, 1998). Most of the CO2 that is produced in the soil escapes to
the atmosphere, where it has a residence time of about 10 years. The short residence time means that
permutations in agricultural practice may have important effects on the CO2 content of the atmos-
phere, much like the burning of fossil carbon, which adds to the complexity of assessing the global
CO2 budget and the effects of CO2 on climatic change.
Figure 5.10 shows the concentration of CO2 gas in unsaturated sandy soils below different vege-
tations. The gravel pit has a sparse vegetation cover and the CO2 concentration is low but still nearly
an order of magnitude higher than in the atmosphere. The woodland consists of mixed hardwood and
conifer trees and has a higher CO2 concentration. Below uncultivated grass lands the CO2 concen-
tration may be 1 vol% and higher. Finally, below the cultivated grass fields of the golf course CO2
reaches as much as 5 vol% CO2, probably an effect of high density growth of hybrid grasses in com-
bination with intense irrigation and fertilization. In all the profiles, a steep gradient of the CO2
distribution towards the surface (0 m depth) indicates a loss of CO2 to the atmosphere. Globally this
loss of CO2 was estimated to be 77 Gt C/yr (Raich and Potter, 1995).
Seasonal variations of CO2 in soil air are apparent in Figure 5.11. During winter the biological CO2
production stops and the CO2 gradient indicates that groundwater is degassing to the atmosphere. In
summer the CO2 production is high and the CO2 distribution suggests both upward and downward
diffusion. The CO2 concentrations in the atmosphere also fluctuate seasonally, concentrations are
lowest during the late summer (September on the northern hemisphere, and March on the southern
hemisphere) mainly as result of the photosynthesis of land plants (Bolin et al., 1979).
Due to higher biological activity, the CO2 concentration in soil air increases with temperature
(Hamada and Tanaka, 2001; Yoshimura et al., 2001; Drake, 1983). However, temperature cannot be
the only controlling variable for biological productivity, also the availability of water and the soil
conditions must play a role. Accordingly, it has been suggested to use evapotranspiration as a meas-
ure of biological production and hence of the average CO2 pressure in the soil. Apart from biologi-
cal activity, there are also physical feedbacks between the soil CO2 content and evapotranspiration.
When soils dry up, CO2 diffusion out of the soil increases and the CO2 concentration goes down.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

188 Carbonates and carbon dioxide

CO2 (vol%)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

1978
1.0 May 24
June 28
2.0 July 27
August 10
3.0 August 24
4.0

5.0
Depth (m)

6.0

7.0

8.0

9.0

10.0

11.0
Water table

Figure 5.11. Seasonal fluctuations of soil CO2 pressures in an unsaturated sand (Reardon et al., 1979).

Figure 5.12 shows the relation between evapotranspiration and soil PCO (Brook et al., 1983). Cool 2

continental climates like Nahanni and Saskatchewan, Canada, have a low evapotranspiration and
PCO , while warm and humid places like Puerto Rico and Thailand are found at the other end of
2

the scale. The relationship has been used to produce a world map of CO2 pressure based on actual
evapotranspiration (Figure 5.13), showing the highest PCO values in the tropics. 2

1.0

TR TH
Mean growing season soil log(PCO2)

1.5 CH SU
MG PR
FG
JA
2.0 BP KY
TN
NF FL
VA
TC
2.5 NA
SA log(PCO2)  3.47  2.09 (1e(0.00172 AET))
AL RM Asymptote log(PCO2)  1.38
r  0.82
3.0

3.5
0 200 400 600 800 1000 1200 1400 1600 1800
Mean annual actual evapotranspiration (mm)
The relationship between mean growing-season soil log(P CO2) and mean annual actual evapotranspiration predicted by
equation (7). NA  Nahanni, Canada; SA  Saskatchewan, Canada; RM  Rocky Mountains, Canada; NF Newfoundland, Canada;
BP  Bruce Peninsula, Canada; TC  Trout Creek, Ontario, Canada; AL  Alaska, USA; VA  Reston, Virginia, USA; TN  Sinking
Cove, Tennessee, USA; KY  Mammoth Cave, Kentucky, USA; FL  South Florida, USA; FG  Frankfurt-Main, W. Germany;
MG  Müllenbach, W. Germany; JA  Jamaica; TR  Trinidad; PR  Puerto Rico; CH  Yunan, China; SU  Sulawesi;
TH  Phangnga, Thailand

Figure 5.12. Relationship between soil CO2 pressure and actual evapotranspiration (Brook et al., 1983).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PACIFIC OCEAN

PACIFIC OCEAN

EQUATOR

INDIAN OCEAN

PACIFIC OCEAN ATLANTIC OCEAN

Carbon dioxide in soils

3.00 2.60 2.20 1.80
Below
Soil log [PCO2] to to to to
3.00
2.60 2.20 1.80 1.40

Figure 5.13. Map of soil CO2 pressure produced with the relationship shown in Figure 5.12 (Brook et al., 1983).

189
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
190 Carbonates and carbon dioxide

The data have also been used to estimate the global flux of dissolved carbonate moving down-
ward into groundwater, by combining the world PCO map with maps of recharge and soil pH
2

(Kessler and Harvey, 2001). Using PCO and pH, TIC is calculated and multiplied with recharge.
2

The resulting estimate amounts to 0.2 Gt C/yr, which is much smaller than the upward loss of
CO2 to the atmosphere of 77 Gt C/yr, but lies in the range of the 0.3 Gt C/yr carried as TIC by
rivers to the oceans (Suchet and Probst, 1995). For comparison, the total human emission of
CO2 is about 10 Gt C/yr, which augments the natural CO2 flow into the atmosphere by about
one seventh.
Soil respiration is not the only source of CO2 entering groundwater. Also the oxidation of
organic carbon present in sediments (Keller and Bacon, 1998; see also Chapter 9) or of DOC
carried with the groundwater (Alberic and Lepiller, 1998) is an important source of TIC.
Furthermore, CO2 derived from deep sources, which include degassing of magma or thermal
metamorphosis of oceanic carbonate rock, may migrate upward along fracture zones and
contribute to the groundwater CO2 input (Etiope, 1999; Yoshimura et al., 2001; Chiodini et al.,
1998).
Figures 5.10 and 5.11 indicate the range of CO2 content in soil gas to lie in between the atmos-
pheric value of 0.03 vol% (PCO  103.5) and 3 vol% (PCO  101.5). Let us approximate the soil
2 2

water as pure water in which CO2 dissolves, and calculate the composition of this hypothetical “soil
moisture” in equilibrium with a constant CO2 pressure. When CO2 dissolves in water the aqueous
species are H2CO3, HCO 2
3 and CO3 and the activity ratios among the species are given by the mass
action equations in Table 5.3. The electroneutrality condition is:
mH  mHCO  2mCO2  mOH (5.17)
3 3

Dissolution of CO2 in water forms carbonic acid and we may surmise that the pH will be slightly
acidic. In first approximation we therefore neglect mCO32 and mOH. This reduces the electroneu-
trality equation to:
mH  mHCO (5.18)
3

We can express [HCO 

3 ] as function of [H ] and [PCO ] with Equations (5.5) and (5.6) to find:
2

[HCO
3 ]  10
7.8 [P 
CO ] / [H ]
(5.19)
2

By combining this relation with Equation (5.18), and neglecting the differences between activity and
molality, we obtain the concentration of H:

mH  107.8 [PCO ] (5.20)

2

Table 5.4. Calculated CO2 species in soil water and rainwater.

[PCO2] 101.5 103.5

H 104.6 105.6 mol/L

pH 4.6 5.6
H2CO3 103.0 105.0 mol/L
HCO 3 104.6 105.6 mol/L
CO32 1010.3 1010.3 mol/L
OH 109.4 10 8.4 mol/L

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Calcite solubility and PCO2 191

Back substitution in the equations of Table 5.3 gives the concentrations entered in Table 5.4.
The PCO of the atmosphere is 103.5 atm and the pH of unpolluted rainwater should therefore
2

be 5.6 (compare with the values in Table 2.2). A PCO of 101.5 atm is near the upper limit for respi-
2

ration and organic matter degradation, which means that the pH of soil water should not become lower
than 4.6 (actual soil waters can be more acidic due to acid rain and dissolved organic acids,
cf. Chapter 8).
Furthermore, note that our approximation to neglect mCO 23 and mOH  in the electroneutrality
equation is acceptable. PHREEQC could easily calculate the pH of a solution in contact with a
constant [PCO2], without approximations, following Example 5.2.

QUESTIONS:
Estimate the CO2 pressure in soil below grassland, Figure 5.10?
ANSWER: CO2  1.1 vol%  101.96 atm
Check the results of the manual calculation in Table 5.4 with PHREEQC (temp  25°C).
ANSWER: SOLUTION; EQUILIBRIUM_PHASES; CO2(g) -1.5; END gives
pH  4.66

5.4 CALCITE SOLUBILITY AND PCO2

Intuitively we expect that a higher PCO and the associated increase of dissolved carbonic acid allows
2

more calcite to dissolve (Equation 5.2). More formally we may combine the dissociation reaction for
calcite with the reactions for aqueous carbonate:

CaCO3 ↔ Ca 2  CO32 K cc  108.5

CO32  H ↔ HCO 3 K 21  1010.3
H 2CO*3 ↔ H  HCO 3 K1  106.3
CO 2(g)  H 2O ↔ H 2CO*3 KH  101.5
 +
CO 2(g)  H 2O  CaCO3 ↔ Ca 2  2HCO
3 K  106.0

The corresponding mass action equation is:

[Ca 2 ][HCO
3]
2
 106.0 (5.21)
[PCO ]
2

For a system containing only calcite, CO2 and H2O, the electroneutrality equation in the solution is:

2mCa 2  mH  mHCO  mOH  2mCO2 (5.22)

3 3

Assume the pH of the soil water to be less than 8.3, then mCO 23 and mOH  can be neglected in (5.22).
Furthermore, assume that mH is negligible in comparison to mCa 2 and the electroneutrality is reduced to:

2mCa 2  mHCO (5.23)

3

Combining with Equation (5.21), and assuming that activity equals molality yields:

mCa 2  3 106.0 [PCO ] / 4 (5.24)

2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

192 Carbonates and carbon dioxide

Table 5.5. Solution speciation in equilibrium with a given PCO2 and calcite.

[PCO2] 101.5 103.5

Ca2 2.0 0.44 mmol/L

CO32 1.64103 7.6103 mmol/L
HCO 3 4.0 0.87 mmol/L
H 1.24104 5.7106 mmol/L
pH 6.9 8.2

Now we have obtained a simple relationship between the Ca2 concentration and the CO2 pressure
for equilibrium with calcite in pure H2O. Once mCa2 ( [Ca2]) is known, [CO2
3 ] can be calculated
from the solubility of calcite; the amount of HCO3 can be calculated with Equation (5.23), and
the other species follow from the equations provided in Table 5.3. The results of such calculations
have been entered in Table 5.5 for CO2 pressures of 101.5 and 103.5 atm. Note that concentrations
of H, CO32 and OH are small compared to concentrations of HCO 2
3 and Ca , and were rightly
neglected.
Equation (5.24) predicts that the Ca2 concentration increases with the cube root of the CO2
pressure as shown in Figure 5.14. The relation has an important consequence. Mixing of two waters
which are both at equilibrium with calcite but which have a different CO2 pressure (A and B in
Figure 5.14), produces a water composition on an almost straight line that connects A and B.

B
2
Ca (mM)

0
0 0.5 1 1.5 2 2.5 3
[PCO ] / 102
2

Figure 5.14. The solubility of calcite in H2O as a function of PCO2. The curved line indicates equilibrium with
calcite according to Equation (5.24). Mixtures of waters A and B are positioned on the (nearly) straight line
connecting A and B.

The straight line lies below the equilibrium curve, indicating that the Ca2 concentration is smaller
than is required for equilibrium. In other words, a mixture of two waters which are both saturated for
calcite, becomes subsaturated for calcite. This effect may cause calcite dissolution in carbonate
aquifers and has been called Mischungskorrosion in the German literature (Bögli, 1978). It is probably

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Calcite solubility and PCO2 193

the dominant process in cave formation in karst, and even plays a role when seawater (that is supersat-
urated for calcite) is mixed with fresh groundwater in equilibrium with calcite, but with a higher CO2
pressure (Section 5.5.2). Example 5.5 shows how to calculate the effects of mixing with PHREEQC.

EXAMPLE 5.5. Calculate mixing effects on calcite saturation state with PHREEQC
The input file to calculate and plot Figure 5.14 is listed below. In the first simulation, the equilibrium line is
calculated. CO2 is added stepwise to distilled water with keyword REACTION, while equilibrium with cal-
cite is maintained. In the second and third simulation the two solutions B and A are defined. In the fourth
simulation, keyword MIX is used to mix the two solutions to illustrate that the mixing line is slightly curved.
Note how in USER_GRAPH the special function sr (“CO2 (g)”) is used. It gives the Saturation Ratio of
CO2(g), or the partial pressure of this gas with respect to the standard state of 1 atm.

SOLUTION 0 # Calculate the equilibrium line...

EQUILIBRIUM_PHASES 1
Calcite
REACTION 1
CO2 1; 3.5e-3 in 30
USER_GRAPH; -head CO2/vol% Ca/mM
-start; 10 graph_x 100 * sr("CO2(g)"); 20 graph_y 1e3 * tot("Ca") ; -end
END

SOLUTION 2 # point B in Figure 5.14...

pH 7 charge; Ca 1 Calcite; C(4) 1 CO2(g) -1.7
USER_GRAPH; -connect_simulations
END

SOLUTION 1 # point A in fig 5.14...

pH 7 charge; Ca 1 Calcite; C(4) 1 CO2(g) -3.5
END

MIX
1 0.97; 2 0.03
END

QUESTIONS:
Why is the mixing line in Figure 5.14 slightly curved near point A?
ANSWER: H2CO*3 does not behave conservatively upon mixing (when pH changes)
Calculate SIcalcite when solution A and B (Figure 5.14) are mixed 1:1, and how much calcite
dissolves in the mixture?
ANSWER: SIcc  0.27, 0.12 mM calcite dissolves

5.4.1 Calcite dissolution in systems open and closed for CO2 gas
Figure 5.15 sketches two situations for calcite dissolution in the field. In both cases root respiration
provides a continuous source of CO2. In the case at left, CaCO3 is present above the water table, and
it will dissolve in contact with the CO2 charged water. The CO2 that is consumed is resupplied by
root respiration, and the system is open with respect to CO2. In the calculations this is simulated by
keeping [PCO ] constant. The open system extends through most of the unsaturated zone due to rapid
2

gas exchange through the gas phase (Laursen, 1991). The water composition can be calculated for a
given [PCO ] with Equation (5.24) (see also Table 5.5).
2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

194 Carbonates and carbon dioxide

CO2 CO2

Open system:
CO2  H2O H2CO3
Open system:
CO2  H2O  CaCO3

Ca2  2HCO3

Closed system:
H2CO3  CaCO3 Ca2  2HCO
3

Figure 5.15. Calcite dissolution in systems open (left) and closed (right) with respect to CO2. In the open
system calcite dissolution proceeds in contact with CO2 production. During closed system dissolution CO2
production and calcite dissolution are spatially separated.

On the situation at right (Figure 5.15), calcite is absent in the unsaturated zone. The CO2 charged
water first encounters a calcite layer below the groundwater table. When calcite dissolves there, the
CO2 consumed is not replenished, because diffusion of CO2 through water is slow. The system is
closed with respect to CO2 gas and Equation (5.24) is not applicable. However, the water composi-
tion can be calculated assuming that all the initial CO2 is used up by dissolution of calcite:

CO 2  H 2 O  CaCO3 → Ca 2  2HCO
3 (5.25)

The amount of Ca2 released equals the initial amount of CO2 ((CO2)root):

mCa 2  (CO 2 ) root (5.26a)

while the amount of HCO

3 is twice as much:

mHCO  2(CO 2 ) root (5.26b)

3

Here it is assumed that pH 8.3, so that CO32 is unimportant. For a given initial [PCO2] the concen-
tration of H2CO*3 can be calculated from Henry’s law (Equation 5.5) which is equal to ( ΣCO2)root.
The calculation of the other species in a hand-calculation follows the same route as in the open case.
The analogue calculations with PHREEQC are shown in Example 5.6.

EXAMPLE 5.6. PHREEQC calculation of open and closed system calcite dissolution
In open system dissolution, water is equilibrated with both calcite and CO2 gas at the desired pressure. With
closed system dissolution, water is first equilibrated with CO2 gas. This solution is SAVEd, and USEd in the
next simulation where it is equilibrated with calcite. This stepwise calculation closely follows the sequence
in the field where the water first is charged with CO2 at a given pressure in the soil, and then contacts cal-
cite below the groundwater table where the CO2 pressure will decrease as calcite dissolves.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Calcite solubility and PCO2 195

SOLUTION 1 # open system dissolution of CO2 with calcite...

EQUILIBRIUM_PHASES 1
CO2(g) -1.5; Calcite 0.0
END

SOLUTION 1 # open system dissolution of CO2 in the

# unsaturated soil...
EQUILIBRIUM_PHASES 1
CO2(g) -1.5
SAVE solution 2
END

USE solution 2 # equilibrate with calcite below the groundwater

# table...
EQUILIBRIUM_PHASES 2
Calcite
END

Table 5.6 summarizes the effect of CO2 pressure and calcite dissolution on the water com-
position for [PCO ] ranging from 10
1.5 in a productive soil to 10
3.5 in a desert sand. The
2

resulting pH varies from 7.0–8.3 for open system calcite dissolution, and from 7.6–10.1 for closed
system calcite dissolution. With the same initial [PCO ], calcite dissolution in an open system
2

yields higher Ca2 and HCO3
concentrations than if the system is closed with respect to
CO2 gas.

Table 5.6. Summary of possible concentrations in water in which calcite

dissolves at 15°C (PHREEQC calculations).

Open system, Closed system, with

constant [PCO 2] known initial [PCO2]

[PCO2] Initial 10
1.5 10
3.5 10
1.5 10
3.5

Final 10
1.5 10
3.5 10
2.5 10
6.4
pH 6.98 8.29 7.62 10.06
Ca2 2.98 0.58 1.32 0.12 mmol/L
Alk 5.96 1.16 2.65 0.24 mmol/L
EC 600 120 265 25 S/cm

QUESTION:
Check the concentrations in Table 5.6

5.4.2 Two field examples

Figure 5.16 shows the dissolution of carbonate in unsaturated sand that contains calcite and
dolomite at depths greater than 1.7 m. The water table at the site is located at 12 m depth.
As soon as infiltrating water reaches the level where calcite is present, the concentrations of Ca2
and HCO
3 , and the pH increase, and the saturation index for calcite approaches equilibrium. The
PCO in the profile remains constant which is typical for open system dissolution.
2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

196 Carbonates and carbon dioxide

0 Calcite 79/03/28 79/03/28 79/03/28

Dolomite 79/06/14 79/06/14 79/06/14

1.5
Depth (m)

3.0

4.5

6.0
0 10 20 0 0.75 1.50 2.25 2.50 2.75 4.0 2.0 0 2.0
WT % Ca (mmol/L) log(P CO2) SIcalcite

Figure 5.16. Calcite and dolomite distribution and calcite dissolution at constant CO2 pressure in unsaturated
sand (open system), at Trout Creek, Canada, (Reardon et al., 1980). PCO2 values given by crosses were calcu-
lated from pH and alkalinity while the open symbols are direct measurements of soil gas.

The depth distributions of the two carbonates calcite and dolomite are different in the profile. For
calcite, the transition in the sediment concentration is very sharp, from zero in the upper part to
20 wt% at 1.7 m. This suggests that calcite dissolves rapidly until saturation is reached, whereafter
dissolution stops as indicated by the sudden jump to SI
0. The dissolution front moves slowly
downward as calcite is depleted. On the other hand, the concentration of dolomite increases gradu-
ally with depth beyond 1.7 m, apparently because dolomite dissolves slower and needs more time
before saturation is attained. The kinetic dissolution of dolomite causes supersaturation for calcite
which will, in principle, precipitate (cf. Problem 5.15). The result is that dolomite is transformed into
calcite, in other words, the sediment is dedolomitized.
Figure 5.17 illustrates calcite dissolution in groundwater in the Rømø aquifer (Jakobsen and
Postma, 1999). The upper 4 m of the saturated zone is free of CaCO3, the pH of groundwater is around
5, the H2CO*3 concentration is about 1 mM, and the groundwater is strongly subsaturated for calcite.

Total inorganic C mmol/kg dry wt. Ca2(meq/L) TIC, H2CO3* (mmol/L) log(IAPcalcite/Kcalcite)
0 100 200 300 400 0.0 2.0 4.0 6.0 8.0 6.0 4.0 2.0 0.0
2 2

3 3
Meters below surface

pH
4 4

5 5
TIC

6 H2CO3* 6
Ca2

7 7

Figure 5.17. Calcite dissolution under conditions closed with respect to CO2 in the Rømø aquifer, Denmark
(Jakobsen and Postma, 1999).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate rock aquifers 197

Below 4.5 m depth, CaCO3 is present and dissolves. In a distinct dissolution front, pH, Ca2 and TIC
increase, while H2CO*3 decreases until the groundwater reaches saturation with calcite at 5 m depth. In
this case, calcite dissolves spatially separated from the point where CO2 is generated in the soil, a situ-
ation that conforms to calcite dissolution in a system closed with respect to CO2 gas. The important dif-
ference with the situation at Trout creek (Figure 5.16) is that H2CO*3, and hence [PCO ], is decreasing 2

when calcite dissolves.

EXAMPLE 5.7. Propagation of the calcite dissolution front in the Rømø aquifer
Calculate the time needed to decalcify the upper 4 m of the Rømø aquifer. The pristine sediment contains
50 mmol CaCO3/kg and has b  1.86 g/cm3,   0.3. The downward flow velocity vH2O  0.8 m/yr.

ANSWER:
The downward rate of the CaCO3 front is calculated with the retardation equation:

vCaCO  vH O / R  vH / (1  q /c )
3 2 2O

where q and c are the change in CaCO3 and in solute TIC across the dissolution front.
q  (50 mmol/kg)  (1.86 kg/L) / 0.3  310 mmol/L pore water. c  1.5 mmol/L (for TIC, Figure 5.17;
note that Ca2 increases more). Hence,

vCaCO  0.8 / (1  207)  0.004 m/yr

3

Accordingly, the present day profile has developed during the last 4/0.004  1000 years. While this value is
only approximate since considerable uncertainty is involved, particularly in q, given the young age of the
sediment (1500 yrs old), it is not unreasonable.

The concentration of Ca2 is higher in the Rømø aquifer (Figure 5.17, closed system) than at the
Canadian site (Figure 5.16, open system), which is the opposite of the expected trend (Table 5.6).
Primarily, the reason lies in a different PCO in the unsaturated zone of the two studied plots. Within a
2

single area, the concentration contrast is expected to follow the theory more closely. Open vs closed
system dissolution was invoked to explain the different compositions in the Triassic Alpujarras
aquifer (Table 5.2 and Example 5.4) and was used to identify the environment where Ca2 entered the
water and to discern infiltration areas (Langmuir, 1971; Pitman, 1978; Hoogendoorn, 1983; Problem
5.23). In addition, 13C can be used to distinguish chemical processes in the source area (Section 5.8).

QUESTIONS:
Estimate the CO2 pressure in the unsaturated zone above the Rømø aquifer? (Hint: look at the H2CO*3
concentration)
ANSWER: PCO 2
101.5 atm (compare with the Canadian site: 102.5 atm).
Compare concentrations in Figure 5.17 with values in Table 5.6?
ANSWER: concentrations in Figure 5.17 are somewhat higher, but temperature in the
field is lower (10°C).

5.5 CARBONATE ROCK AQUIFERS

There are two modes of groundwater transport through a limestone aquifer, matrix or diffuse flow
and conduit flow. In the case of matrix flow, groundwater is transported through the rock pores or

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

198 Carbonates and carbon dioxide

through a possibly dense network of micro-fissures in a manner not unlike Darcy flow. Conduit flow
occurs along larger fissures and openings that result from dissolution of carbonate rock and is much
quicker, and may be even turbulent. The two flow types interchange water continuously as illustrated
in Figure 5.18. Due to fast conduit flow, pollutants may arrive earlier than with plug-flow which
considers displacement of the whole resident mass of matrix and conduit water together. The peak
concentrations in conduits are slightly diminished by exchange with the matrix, while slow bleeding
from a polluted matrix yields a continuous source of contaminants, albeit at a possibly low concen-
tration. Atkinson (1975) derived from hydrograph-analysis and other parameters in the Mendip
Hills, England, that conduit flow probably accounts for between 60% and 80% of the transmission
of water in the aquifer, although conduits contain only 1/29 of all the phreatic groundwater when the
aquifer is fully recharged.

(A) Contaminant source (B)

Single porosity
plug flow

Concentration
Diffusion into
porous rock
Double porosity
flow

Sampling
point 0
Karst river

Time

Figure 5.18. Dual porosity character of limestones (A) results in two flow modes with typical effects on solute
(pollutant) transport and arrival times of a pulse injection (B).

The flowpath has an important effect on the solute concentrations in the upper, phreatic zone where dis-
solution is going on. Figure 5.19 shows different flowpaths in carbonate karst in New Zealand, together
with mean Ca2 and Mg2 concentrations (Gunn, 1981). The concentrations are lowest in overland
flow and throughflow, where contact-time and -surface are small. Water in the subcutaneous zone has
the highest concentrations, and this water seems to feed the larger karst pipes (termed shafts in Figure
5.19). The presence of lower concentrations in smaller fissures seems illogical, but perhaps there is a
rapid flow component involved which brings water rapidly downwards in these particular fissures.
Seasonal variations in water qualities of carbonate rock springs have been related to the type
of flow mechanism in limestone aquifers. Shuster and White (1972) indicated that conduit
flow springs have large variations in concentrations, temperature and discharge over the year,
whereas diffuse flow springs have in all aspects a more stable regime (Figure 5.20). Martin and Dean
(2001) found that during low flow the concentrations generally increased due to a greater matrix
contribution.
The dissolution of carbonate along conduits may result in spectacular features such as sinkholes
(cenotes) caves, rivers disappearing underground and a karstic landscape may develop over time. Conduit
flow is often rapid, giving sometimes even turbulent flow with flow rates of up to 150 m/h (Alberic and
Lepiller, 1998). Because of their highly dynamic character, karst systems are vulnerable towards pollu-
tion (Katz et al., 1998). Solution features are also visible on the surface of all barren limestones.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate rock aquifers 199

Figure 5.21 shows the typical solution flutes (sometimes denoted with the German term
“Rillenkarst”) that are often observed. Surface solution phenomena as well as cave formation are
likewise, though more seldom, found in dolomite rock (Zötl, 1974; Ford and Williams, 1989;
Figure 5.21).

Bare
rock

e
il zon
1 So us
Vertical n eo
2 uta
shaft bc
Su
3

6 5 4

Cave stream

Flow component Number of Standard Coefficient of

(key below) observations Mean deviation variation (%) Range

(a) Calcium ion concentrations (mg/L)

1 6 7.5 2.49 33.2 4.2–10.4
2 162 18.7 5.04 26.9 8.0–31.7
3 154 46.5 6.75 14.5 34.8–61.4
4 61 46.8 6.66 14.2 30.0–56.4
5 186 30.6 17.86 58.3 10.4–61.4
6 1714 36.1 6.72 18.6 20.6–59.0

(b) Magnesium ion concentrations (mg/L)

1 6 0.6 0.308 47.4 0.2–1.0
2 154 1.1 0.270 23.9 0.5–1.8
3 139 1.3 0.212 16.2 0.8–1.5
4 58 1.3 0.224 16.9 0.7–1.8
5 183 1.0 0.294 30.4 0.5–1.5
6 1754 1.5 0.427 28.0 0.7–3.0

(c) Temperature (C)

2 78 10.3 2.43 4.9–14.7
3 95 10.7 0.58 9.4–12.2
4 34 10.0 1.33 8.3–12.8
5 105 11.1 1.53 8.1–13.4

Key to flow components: 1. Overland flow; 2. Throughflow; 3. Subcutaneous flow; 4. Shaft flow; 5. Vadose
flow. 6. Vadose seepage.

Figure 5.19. Types of flow in New Zealand karst, and properties of flow components (Gunn, 1981).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

200 Carbonates and carbon dioxide

250 12 0.2

Total hardness Saturated Spruce

Spruce 11 0.0 creek
200 aggressive
creek spring
Temperature spring

Temperature (°C)
Hardness (ppm)

Saturation index
10 0.2
150 Temperature
9 0.4
Hardness
100
8 0.6
Rock Rock spring
spring
50 7 0.8

6
0

.
ar
r.
ay

ne

ly

g.

.
n.

b.

r.

ay

ne

ly

g.

n.
pt

ct

ov

ec

pt

ct

ov

ec
M
Ap

Ap
Ju

Ju
Au

Ja

Fe

Au

Ja
M

O
Ju

Se

Ju

Se
N

D
Figure 5.20. (Left) Comparison of seasonal variation of hardness and temperature for a typical diffuse-flow
spring (Spruce Creek spring) and a typical conduit-flow spring (Rock spring). (Right) Seasonal variation in sat-
uration index for the two types of springs (Shuster and White, 1972).

Further contributing to dissolution along conduits in a karst system is the variability of water com-
positions, illustrated in Figure 5.19. Mixing of waters with different compositions will produce
waters subsaturated for calcite (Figure 5.14), allowing renewed dissolution in the subsurface. Finally,
surface water that is transported rapidly into a karst system may contain reactive organic carbon.
Decomposition of the organics down in the karst channels causes CO2 production and allows calcite
dissolution (Batoit et al., 2000; Alberic and Lepiller, 1998).

Figure 5.21. (Left) Rillenkarst on Limestone in the N.-Italian Alps. (Right) Karstic developments of shafts in
the Dolomites (N.-Italy) (Appelo, priv. coll.).
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Carbonate rock aquifers 201

5.5.1 Dolomite and dedolomitization

Aquifers of dolomite rock behave in many ways similar to limestones. For the dissolution of
dolomite we can write:
2CO 2(g)  2H 2 O  CaMg(CO3 )2 ↔ Ca 2  Mg 2  4HCO
3
(5.27)

Groundwater in a monomineralic dolomite aquifer should ideally contain equal amounts of Ca2 and
Mg2 while the bicarbonate concentration should be four times the Ca2 and Mg2 concentrations.
Models for dissolution in systems open or closed with respect to CO2 gas apply also to dolomite. The
law of mass action for Reaction (5.27) is (cf. Equation 5.21):

[Ca 2 ][Mg 2 ][HCO

3]
4
 Kdol (KH K1 / K 2 )2 (5.28)
[PCO ]2
2

The electroneutrality equation 2mCa2  2mMg2  mHCO3 and the mass balance mCa2  mMg2
combine to:
4mCa 2  mHCO (5.29)
3

which yields together with Equation (5.28):

2
6
1  K 
mCa 2  Kdol  [ PCO ]KH 1  (5.30)
256  2 K2 

where Kdol  1016.7 at 10°C.

EXAMPLE 5.8. Dissolution of dolomite in the Italian Dolomites

The Dolomites in N.-Italy form isolated, barren plateaus at levels up to 3000 m. Scree slopes which are veg-
etated at lower altitudes, extend from the plateaus. The timberline is at around 2000 m. The situation,
depicted in Figure 5.22, gives low concentrations in water infiltrated on the plateaus (flowline A), and
higher concentrations in water which infiltrates in the forests on the slopes (flowline B).

3000 m

PCO2  103.5 atm

Dolomite reef
2000 m CaMg(CO3)2
PCO higher
2

A
B
B

A
1000 m

Figure 5.22. Schematic cross section through a dolomite plateau.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

202 Carbonates and carbon dioxide

The difference suggests rapid dissolution of dolomite to near saturation levels, dictated by the CO2 pressure
of the infiltration area.
When discharge of the springs increases in response to rain, it is often observed that the concentrations in
the discharge increase as well. Apparently it is primarily the contribution of runoff from the scree slopes that
increases in spring discharge. The corresponding flowline goes through highly permeable scree slopes
which can give very rapid response to rainfall, while at the same time the soil slopes have a higher CO2 pres-
sure than the plateaus at 3000 m.
For a CO2 pressure at the plateau of 103.5 atm, similar to the atmosphere, and for the forested scree
PCO2  102 atm, the calculated concentrations in flowline A and B are as given in the table, in close
correspondance to observed concentrations.

Flowline A B

Ca2  Mg2 0.33 1.04 mmol/L

HCO3 1.31 4.16 mmol/L
pH 8.42 7.42
EC 131 416 S/cm

For closed system dissolution of dolomite identical arguments provide us with:

HCO
3  2( CO 2 ) root ; mCa 2  mMg2  1⁄4 mHCO 3; etc.

Aquifer rocks often contain both calcite and dolomite either in separate layers or as disseminated crys-
tals. Figure 5.23 shows the saturation state of groundwater in Paleozoic rock with respect to both cal-
cite and dolomite. The spring waters generally have a short residence time or consist of mixed waters
and are subsaturated, while the well waters, representing a longer residence time, approach saturation.

0.1
Calcite saturation
0
0.1
Dolomite saturation
SIcalcite

0.5

1.0

1.5 1.0 0.5 0.1 0 0.1

SIdolomite

Figure 5.23. The saturation state of groundwater in paleozoic limestones with respect to calcite and dolomite.
Open symbols are spring waters and filled symbols are groundwaters (Langmuir, 1971).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate rock aquifers 203

The point where the groundwater is in equilibrium with both calcite and dolomite has special chemical
properties. For his situation we can write the reaction:

Ca 2  CaMg(CO3 )2 ↔ 2CaCO3  Mg 2 (5.31)

The corresponding mass action equation is:

K  [Mg 2 ] / [Ca 2 ]  Kdol / (K cc )2  1017.09 / (108.48 )2  0.8 (5.32)

at 25°C. In other words, in case of equilibrium with both calcite and dolomite, the [Mg2] / [Ca2]
ratio becomes invariant (at constant temperature and pressure).
When simultaneous equilibrium with calcite and dolomite is combined with the dissolution of
gypsum or anhydrite, the process of dedolomitization may take place. For dissolution of anhydrite or
gypsum we write:

CaSO 4 → Ca 2  SO 24 (5.33)

The increasing Ca2 concentration due to gypsum dissolution causes calcite to precipitate, which
is similar to the effect of gypsum dissolution on fluorite precipitation discussed in Section 4.1.
The CO32 concentration decreases as calcite precipitates, and this provokes the dissolution of
dolomite and an increase of the Mg2-concentration. When Mg2 increases, Ca2 must increase as
well according to relation (5.32). The net result is that the dissolution of gypsum induces the trans-
formation of dolomite to calcite in the rock and produces waters with increased Mg2, Ca2 and
SO24 concentrations. Dedolomitization may occur in aquifers containing dolostones (dolomite-type
carbonates) associated with gypsiferous layers (Back et al., 1983; Plummer et al., 1990; Saunders
and Toran, 1994; Cardenal et al., 1994; Sacks et al., 1995; Capaccioni et al., 2001; López-Chicano
et al., 2001). Also groundwater pumping may draw sulfate rich waters from gypsum layers into
dolomite rock and thereby cause dedolomitization in a well field (López-Chicano et al., 2001).
The process of dedolomitization is illustrated for the Madison aquifer, USA, in Figures 5.24 and 5.25
(Plummer et al., 1990). The sulfate concentration was used to monitor the extent of gypsum dissolution.
The groundwater is at saturation or slightly supersaturated for calcite throughout the aquifer (Figure 5.24),
regardless of the sulfate concentrations. The data scatter for saturation with respect to dolomite is some-
what larger, but still within a SI variation of 0.5 from equilibrium. (The calculation of SIdol involves 4
solutes, increasing the effect of analytical errors by a factor of 2 compared to SIcc). For gypsum, the satu-
ration state ranges from strongly subsaturated to equilibrium. When equilibrium with both calcite and
dolomite is maintained, the ratio of [Mg2] / [Ca2] must remain about 0.8. Since the two ions behave
similarly with regard to complexation and activity coefficient corrections, the ratio of total concentrations
of the two ions should also remain about 0.8. The total mass transfer is thus given by the reaction:

1.8CaSO 4  0.8CaMg(CO3 )2 → 1.6CaCO3  Ca 2  0.8Mg 2  1.8SO 24 (5.34)

For each mole of dolomite that dissolves, two moles of calcite are precipitated. The reaction scheme
predicts that the SO 42 concentration must increase with both the concentration of Mg2 and Ca2,
as occurs in the Madison aquifer (Figure 5.25). The ideal stoichiometry according to Reaction (5.34)
is indicated by lines, labelled with 1.8 as the coefficient for SO42. The suggested trend is closely fol-
lowed by Ca2, but not by Mg2. However, Figure 5.24 shows the groundwater to be supersaturated
with respect to calcite, which influences the [Mg2] / [Ca2] ratio. If we take into account that the
waters on the average have SIcc  0.2 and SIdol  0, the ratio becomes:

[Mg 2 ] / [Ca 2 ]  1017.09 / (108.480.2 )2  0.3

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

204 Carbonates and carbon dioxide

1.0

Calcite saturation index 0.5

0.0 Equilibrium

0.5

1.0
0 5 10 15 20 25
1.0

0.5
Dolomite saturation index

0.0 Equilibrium

0.5

1.0

1.5

2.0
0 5 10 15 20 25

1.0

0.5

0.0 Equilibrium

0.5
Gypsum saturation index

1.0

1.5

2.0

2.5

3.0

3.5

4.0
0 5 10 15 20 25
Sulfate concentration (mmol/kg H2O)

Figure 5.24. Calcite, dolomite and gypsum saturation indices as a function of total dissolved sulfate content
for wells and springs in the Madison aquifer (Plummer et al., 1990).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate rock aquifers 205

15 6

Magnesium concentration (mmol/kg H2O)

8
Calcium concentration (mmol/kg H2O)

1.
.8

8:
1:1 5

0.
1.
3 


1: O 4
a:S

4


O
3
C :1.

S
10 O4 4 0.3

g:
:S

M
Ca 
4
:SO
Mg
3

5 2

0 0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
Sulfate concentration (mmol/kg H2O) Sulfate concentration (mmol/kg H2O)

Figure 5.25. Dissolved calcium and magnesium concentrations as a function of dissolved sulfate content for
waters from the Madison aquifer. Lines refer to the mass transfer predicted by Equations (5.34) and (5.35).
(Modified from Plummer et al., 1990).

The overall mass transfer now changes to

1.3CaSO 4  0.3CaMg(CO3 )2 → 0.6CaCO3  Ca 2  0.3Mg 2  1.3SO 24 (5.35)

Lines according to this stoichiometry have also been drawn in Figure 5.25, and are labelled with 1.3
as the coefficient for SO42. The relation presents an upper boundary for the increase of Ca2 with
SO42, while the increase of Mg2 with SO42 is fitted quite well.
Apparently the mass transfer coefficients are determined by the failure of calcite to attain com-
plete equilibrium during precipitation. This may be caused by precipitation inhibitors present in the
groundwater. Plummer et al. (1990) found that the average apparent rates remain very low in this
aquifer: 0.59 mol/L/yr for calcite precipitation, 0.25 mol/L/yr for dolomite dissolution and
0.95 mol/L/yr for gypsum dissolution. These rates have no mechanistic kinetic significance, but
depend on the haphazard encounter of water with gypsum, which dissolves immediately.

QUESTION:
Calculate the dedolomitization reaction with PHREEQC, plot Ca2 and Mg2 vs SO42 concentrations for
SIcc  0 and 0.2
ANSWER: SOLUTION 1; pH 7 charge; C 2
EQUILIBRIUM_PHASES 1; Calcite 0.; Dolomite
REACTION; CaSO4; 20e-3 in 20; END

5.5.2 Pleistocene carbonate aquifers

Recent marine carbonate sediments found in tropical areas like the Caribbean and the South Pacific
consist of aragonite and high Mg-calcite of biogenic origin, such as shells, skeletal debris, carbonate
mud, etc. A minor part is played by abiotic precipitation at high supersaturation which also gives
aragonite and high Mg-calcite (Given and Wilkinson, 1985). Since these sediments are formed in the
marine environment we first inspect their behavior in contact with seawater. Using PHREEQC we
can calculate the saturation states for surface ocean seawater at 25°C:
SI calcite  0.74 SI aragonite  0.60 SI dolomite  2.37.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

206 Carbonates and carbon dioxide

30 30 Aragonite
Calcite
Mg-free
Mg-free log PCO2 SW SW
log PCO SW SW
2
1.51
20 1.51 20 2.01
2.01 2.54
2.54 ~3.5
Rate

~3.5 (closed)
(closed)
 of SW
10 10
 of SW in ocean
in ocean

0 0
0 0.20 0.40 0.60 0.80 1.00 0 0.20 0.40 0.60 0.80
log  Calcite log  Aragonite

Figure 5.26. The rate of precipitation of calcite and aragonite vs the degree of saturation in artificial seawater
with or without Mg. log   SI  log IAP/K, or the saturation index (Berner, 1975).

Surface seawater is apparently strongly supersaturated with respect to calcite, aragonite, and
dolomite. These minerals should therefore precipitate extensively from seawater, but this does not
happen because of kinetic inhibition. Calcite and aragonite precipitate only slowly and locally while
dolomite does not seem to precipitate at all in the surface layers of the ocean.
The lack of precipitation is due to inhibitors present in seawater. Seawater contains about five times
as much Mg2 as Ca2 on a molar base. Berner (1975) showed in a series of precipitation experiments
with calcite and aragonite, that the rate of calcite precipitation in the presence of Mg2 at seawater con-
centration becomes negligible (Figure 5.26). Apparently, Mg2 disturbs the crystal growth of calcite. An
alternative explanation is that incorporation of Mg2 enhances the solubility and thereby inhibits calcite
growth (Davis et al., 2000). The Mg2-ion can become incorporated in the structure when supersatura-
tion becomes very high and the resulting calcites may contain up to 30% MgCO3 (Figure 5.2). In
contrast, aragonite, having another crystal structure is not affected by the presence of Mg2.
Organic acids (Berner et al., 1978) and phosphates (Walter and Hanor, 1979; Mucci, 1986) are
also well known inhibitors and may also inhibit the precipitation of aragonite. The overall conclusion
is, that in contact with seawater the most stable minerals, low Mg-calcite and dolomite, cannot form
because of kinetic inhibition. Instead, metastable minerals like aragonite and Mg-calcite are formed.
What happens to these minerals when they enter the freshwater environment? Sea level varia-
tions, related to glaciations and deglaciations, may amount to several hundreds of meters and can lift
marine sediments into the freshwater zone, for example forming the coastal aquifers on the Yucatan
peninsula in Mexico. In contact with fresh water the precipitation of calcite is no longer inhibited
and a giant recrystallization process starts whereby aragonite and high-magnesium calcite dissolve
and calcite precipitates.
The transition of aragonite into calcite is predictable from the solubility products (Table 5.1). The
reaction:

CaCO3 ↔ Ca 2  CO32

has

Kcalcite  [Ca 2 ][CO32 ]  108.48  3.3109 (25C)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate rock aquifers 207

or

Karagonite  [Ca 2 ][CO32 ]  108.34  4.6109 (255C)

Water in equilibrium with aragonite is therefore supersaturated for calcite so that calcite precipitates and
aragonite dissolves. The reaction is a two step process with sequential dissolution and precipitation:

CaCO3 aragonite → Ca 2  CO32 → CaCO3 calcite (5.36)

For high Mg-calcite the situation is more complex because it is a solid solution. Figure 5.27 shows
what happens when a high Mg-calcite is placed in distilled water. Initially both the Mg2 and Ca2
concentrations increase, corresponding to congruent dissolution. However, after some time, the Ca
concentration starts to fall. Apparently a new Mg-calcite with a lower MgCO3 content precipitates,
and the process can be formalized as:

Ca1 x Mg x CO3 → a Ca1 y Mg y CO3  (x  ay ) Mg 2 

(5.37)
 (1  x  a  ay ) Ca 2  (1  a) CO32

where y  x.
In this fashion, a row of Mg-calcites with slowly decreasing MgCO3 contents will precipitate,
and the water composition is enriched in Mg2. The process of dissolution whereby another, less sol-
uble solid precipitates, is termed incongruent dissolution.
Figure 5.27 shows that it is quite impossible to measure the solubility of a Mg-calcite with
a specific composition from such dissolution experiments since the solid phase composition
changes during the experiment. This inability to reach equilibrium in experiments with
Mg-calcite, and which probably also is valid in many natural settings, led Thorstenson and
Plummer (1977) to reevaluate the equilibria criteria for two-component solid solutions.

16

Stage 1 Stage 2 Stage 3

14

12
Calcium
10
sium
mmol/L

Magne
8

0
0 2 4 6 8 10 12 14 16 18 20 22 24 26
(Hours)1/2

Figure 5.27. The dissolution of a high Mg-calcite in distilled water (Plummer and Mackenzie, 1974).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

208 Carbonates and carbon dioxide

They proposed the concept of stoichiometric saturation, which defines equilibrium between an
aqueous solution and a homogeneous solid solution of fixed composition. For Mg-calcite the
stoichiometric solubility expression becomes:

Ca1 x Mg x CO3 → (1  x )Ca 2  x Mg 2  CO32 (5.38)

for which the stoichiometric solubility product is:

K(x )  [Mg 2 ]x [Ca 2 ]1 x [CO32 ] (5.39)

For stoichiometric saturation as expressed by Equation (5.39), the [Mg2] / [Ca2] ratio in solution
is decoupled from the MgCO3 content of the solid, which means that saturation with any Mg-calcite
can be attained (in principle) in a solution with a very low Mg2 concentration. This differs from
solid solution behavior where the [Mg2] / [Ca2] ratio of the solution determines the composition
of the solid and vice versa (Equation 4.35). The stoichiometric saturation concept has been applied
with some success to the Mg-calcite system (Busenberg and Plummer, 1989) and the strontianite-
aragonite system (Plummer and Busenberg, 1987).
Criteria for attaining stoichiometric saturation are, that dissolution proceeds congruent, as in
stage 1 of Figure 5.27, that a constant IAP (corresponding to Equation 5.39) is maintained over a con-
siderable range of time, and that the solution composition does not change when the crystals are
placed in a solution with the same IAP  K(x). Busenberg and Plummer (1989) prevented reprecipi-
tation of lower Mg-calcite, as happens during stage 3 in Figure 5.27, by adding phosphate to their
experiments which strongly inhibits calcite precipitation. The stability of Mg-calcite, obtained with
the stoichiometric saturation concept, is shown in Figure 5.28. It may be noted that high Mg-calcite,
as present in recent marine carbonates, is indeed unstable relative to low Mg-calcite. Furthermore,
calcite with a few mol% MgCO3 is more stable than pure calcite.

8.35
 Inorganic
 Synthetic
 Bischoff (1985)
8.40  Mucci & Morse (1984)
log IAPCa1xMgxCO3

8.45

8.50

8.55
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
Mole fraction MgCO3

Figure 5.28. The stoichiometric solubility of well crystallized Mg-calcite From Busenberg and Plummer (1989).

The recrystallization process can be traced in the present-day groundwater chemistry of the Bermuda
aquifer. The local sediments consist of calcarenites (beach deposits, dunes, etc.). The sediments are
oldest in the central part of the island and become younger towards the marginal parts. Recent
sediments contain about 35% aragonite and 50% high Mg-calcite with 14 mol% magnesite on average.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbonate rock aquifers 209

90

Upper Paget
80

Lower Paget member

Recent sediments

member
70 Calcite

60

Weight percent 50

Belmont formation
40
Magnesium–Calcite

Walsingham formation
30

20
Aragonite
10

0
0 20 40 60 80 100 120 500 1000
Geologic age (103 years before present)

Figure 5.29. Relative composition by weight percent of Bermuda limestones as a function of geological age
(Plummer et al., 1976).

At the other end of the scale, 106 yr old limestones consist exclusively of low Mg-calcite (Figure 5.29).
In the central parts of the island, groundwater is near equilibrium with calcite but subsaturated for arag-
onite (Figure 5.30). In the marginal parts, the groundwaters are close to equilibrium with aragonite
(which dissolves), but supersaturated for calcite (which will precipitate). An additional indication for
aragonite dissolution is the high strontium concentration in the marginal groundwaters (Plummer et al.,
1976). Small amounts of strontium are easily incorporated in aragonite but not in calcite (cf. Table 5.1),
and Sr 2 will be enriched in groundwater when aragonite recrystallizes to calcite.

Harrington Harrington
SI Aragonite Sound SI Calcite Sound
Flatts Flatts

0.02 0.09
0.17 0.07
0.02 0.10
0.20 0.10
n

0.20 0.78 io 0.05 0.09 0.67

at ti o 0.05
S u b s atur ra
m

0.40 0.05 S a t u 0.30 0.06

u

ri
0.02 lib 0.09
io
n
q ui at
r E Devonshire Bay ur Devonshire Bay
0.09 N e a 0.02 at
0.07 ers 0.04
Sup

Figure 5.30. Saturation indexes of Bermuda groundwater with respect to calcite and aragonite (Plummer
et al., 1976).

Figure 5.31 presents an overall summary of the processes in coastal carbonate aquifers, indicating
both dissolution (“corrosion”) and precipitation. Three corrosion processes are discerned, of which
mixing corrosion is probably the most important for cave formation. Mixing corrosion takes place
when waters with different PCO2 mix (Figure 5.14) and applies for mixing seawater and groundwater
(Wigley and Plummer, 1976; Back et al., 1986; Sanford and Konikow, 1989; Problem 5.16).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

210 Carbonates and carbon dioxide

Biogenic
2
corrosion
Precipitation
Simple
1 corrosion (CO2 loss)

Degassing
Water table
S ea
w a ter
PCO2
Fresh water
3 Mixing
corrosion Marine
Salinity water

Mixed water

Figure 5.31. Dissolution features in a coastal carbonate aquifer (James and Choquette, 1984).

However, the mixing effects become often more complicated when other geochemical processes par-
ticipate. Particularly redox processes may affect the dissolved carbonate composition and change
both the mixing endmember and the saturation state of the mixed waters (Stoessel et al., 1993;
Whitaker and Smart, 1997; Andersen, 2001).

QUESTION:
Estimate the difference in log K for calcite and aragonite from Figure 5.30?
ANSWER: log Kcc  log Karg  0.11 (varies a bit from 0.10 to 0.12)

5.6 KINETICS OF CARBONATE REACTIONS

5.6.1 Dissolution
The general pattern of calcite dissolution kinetics as a function of pH and CO2 pressure is shown in
Figure 5.32 (Plummer et al., 1978). Three regions are discerned as a function of pH. The first region is
confined to pH values below 3.5, where the rate is proportional to [H]. At these low pH values a strong
dependence on the stirring rate is found which indicates that transport of H to the calcite surface is rate
controlling. At higher pH, the rate becomes independent of pH but instead dependent on PCO . Here, the 2

rate appears to be controlled by both transport and surface reaction (Rickard and Sjöberg, 1983).
The third region indicates a sharp drop of the dissolution rate when the pH of saturation is approa-
ched, depending on the specific CO2 pressure.
Different approaches have been used to model these general results (Morse and Arvidson, 2002).
Sjöberg (1978), Rickard and Sjöberg (1983) and Morse (1983) described calcite dissolution kinetics
in seawater by empirical rate expressions of the type
R  k (A / V )(1   )n (5.40)

where A is the calcite surface area, V is the volume of the solution,  is the saturation state IAP/K,
and k and n are coefficients which depend on the composition of the solution, and are obtained by
fitting with observed rates.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of carbonate reactions 211

3.0
25°C Method PCO Symbol
2
“pH·Stat” 0.00
“pH·Stat” 0.30
“pH·Stat” 0.97
4.0 “Free drift” 0.03
“Free drift” 0.30
“Free drift” 0.97
Region 1 Region 2 Region 3
log rate (mmoles/cm2/sec)

5.0

0.9
7
0.
0.030
6.0 0

7.0

8.0

2.0 3.0 4.0 5.0 6.0 7.0

pH

Figure 5.32. Dissolution rates of calcite as a function of pH and PCO2 (Plummer et al., 1978).

Plummer, Wigley and Parkhurst (1978, denoted further as “PWP”) developed a rate model for cal-
cite dissolution based on three dissolution reactions:

CaCO3  H → Ca 2  HCO
3 (5.41a)

CaCO3  H 2CO*3 → Ca 2  2HCO

3 (5.41b)

CaCO3  H 2O → Ca 2  HCO
3  OH
 (5.41c)

The first reaction reflects the dominating process of proton attack at pH  3.5, and the second reac-
tion incorporates the effect of H2CO*3 at higher pH. At still higher pH, above 7, the third reaction
becomes important, which reflects simple hydrolysis of calcite. Finally the backward precipitation
reaction was added:

Ca 2  HCO
3 → CaCO3  H

(5.42)

These reactions form the basis for the rate equation which covers both dissolution and precipitation:

r  k1[H ]  k 2 [H 2CO3 ]  k3 [H 2O]  k 4 [Ca 2 ][HCO

3]
r fw rb (5.43)

where r is the specific rate of calcite dissolution (mmol/cm2/s), separated in a forward rate rfw and a
backward rate rb, and k1, …, k4 are the rate constants. The forward rate constants have been fitted to

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

212 Carbonates and carbon dioxide

the experimental data as function of temperature (T, in K):

log k1  0.198444 / T
log k2  2.842177 / T
log k3  5.86317 / T for T ' 298
log k4  1.11737 / T for T ! 298 (5.44)

The backward rate was derived from the principle of microscopic reversibility of the individual forward
reactions as a function of [H]sf and [H2CO*3]sf, which are activities on the crystal surface. It was
assumed that saturation exists at the crystal surface during the entire dissolution process, so that [H]sf
and [H2CO*3]sf are given by their saturation values at the bulk solution [PCO ]. This rather complicated
2

reasoning was necessary, since the backward reaction was observed to depend on [Ca2] and [HCO 3]
only, i.e. giving only the backward reaction of the first dissolution reaction. Strict microscopic
reversibility would require a likewise dependence of the backward reaction on [OH] from the third
term, and a quadratic dependence on [HCO 3 ] from the second term (Chou et al., 1989). Anyhow,
Plummer et al. (1978) argued that as long as [H]sf and [H2CO*3]sf remain constant during the dissolu-
tion process, the value of k4 remains constant as well. This allows a straightforward calculation of k4
since at saturation the net rate r  rfw  rb  0. Constant [H2CO*3]sf is true when CO2 pressure is con-
stant (open system dissolution). For example, at constant [PCO2]  101.5 atm, the water composition
at equilibrium with calcite was calculated to be (Table 5.5): [H]  106.9; [H2CO*3]  103.0;
[Ca2]  102.7, and [HCO  3 ]  2[Ca ]. Hence we find, at 10°C ( 283 K):
2

k1 [H ]  101.4 106.9  108.3

k 2 [H 2CO*3 ]  104.9 103.0  107.9
k3 [H 2O]  107.0 1  107.0
r fw  106.9

And with rb  k4 2 [Ca2]2  rfw, we obtain k4  106.9 / (2 (102.7))2  101.8. We furthermore

define k4  2 k4  101.5.
The contribution from the first term, k1[H], rapidly diminishes as pH increases. The term can
be neglected when pH is more than 6: the forward rate then becomes a constant which increases
slightly with PCO2. The overall rate which describes calcite dissolution in a system with a constant
CO2-pressure of 101.5 atm is therefore simplified to:

dmCa
r  r fw  k 4 mCa
2  106.9  101.5m 2
Ca
(5.45)
dt
Equation (5.45) can be integrated to provide the Ca2-concentration as a function of time:
mCa
t
t
dmCa
∫  ∫ dt (5.46)
0 (r fw  k 4 mCa
2 )
0

where mCa t is the Ca2 concentration reached at time t. The solution for Equation (5.46) is:

  
1  r fw  k 4 mCat  (5.47)
ln  t
2 r fw k 4  r fw  k 4 mCa 
t

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of carbonate reactions 213

Inserting values for rfw and k, 4 and for A/V  1 cm /cm , the timespace is obtained to reach a
2 3

concentration in solution of, say, 95% of the value at saturation. (The time to reach 100% saturation is
infinite). In the case of [PCO2]  101.5, for which we calculated mCa  0.002 mol/L at saturation, it is:
2

t.95  30 000 s  8.3 hr (5.48)

Let’s compare this result with a PHREEQC calculation, first integrating Equation (5.45) by defining
it as rate under keyword RATES, and then also apply the parent Equation (5.43) which is available
in PHREEQC.DAT (Example 5.9).

EXAMPLE 5.9. Kinetic dissolution of calcite calculated with PHREEQC, comparing simplified and par-
ent “PWP” rates
The simplified rate of calcite dissolution is defined as “Calcit2” in 2 BASIC lines under keyword RATES. Line
10 contains the formula (Equation 5.45) in which mCa2 is obtained with the function tot (“Ca”). Line 20 mul-
tiplies with 1/cm to convert mmol/cm2 to mol/dm3 (PHREEQC adds moles and the solution contains 1 dm3 by
default). The functions “save” and “time” in line 20 signal the start of the integration (cf. Chapter 4). The rate
is invoked with keyword KINETICS where the chemical formula is entered for Calcit2, the initial amount of
1 mol reactant and the timespace of 30 000 s. SOLUTION 1 is pure water with a [PCO2]  101.5, and the CO2
pressure is kept constant with EQUILIBRIUM_PHASES 1. Finally, keyword USER_GRAPH defines the plot.
The next simulation invokes the rate for Calcite which is defined in PHREEQC.DAT. The rate (look in
the database) requires the surface to volume ratio (A/V in 1/dm) and the exponent for surface change (n in
(m / m0)n) which are defined as ‘-parms 10 0.67’ under KINETICS. Thus, A/V  10/dm  1/cm, equal to the
ratio in the previous rate, and the exponent n  0.67.
Running the file, and inspecting the results of the simple rate in the grid, shows that the Ca2 concentra-
tion after 30 000 s is 1.91 mM according to the simple rate, which is indeed 95% of the estimated saturation
concentration of 2.0 mM.

DATABASE phreeqc.dat # contains the PWP calcite rate

RATES
Calcit2 # simplified rate cf. Equation (5.45)...
-start
10 rate  10^-6.91 - 10^-1.52 * (tot("Ca"))^2 # mmol/cm2/s
20 save 1 * rate * time # integrate in moles/L
-end
SOLUTION 1
temp 10; pH 6 charge; C 1 CO2(g) -1.5
EQUILIBRIUM_PHASES 1
CO2(g) -1.5

KINETICS 1
Calcit2; formula CaCO3; -m0 1; -step 30000 in 20
INCREMENTAL_REACTIONS

USER_GRAPH
-head time Ca pH; -axis_titles "Time / 1000s" "Ca / mM" "pH"
-start
10 graph_x total_time / 1e3; 20 graph_y tot("Ca") * 1e3; 30 graph_sy -la("H+")
-end
END
# Simulation 2, with the parent PWP rate...
USE solution 1; USE equilibrium_phases 1
KINETICS 1
Calcite; -m0 1; -parms 10 0.67; -step 30000 in 20
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

214 Carbonates and carbon dioxide

7
3
pH

Ca (mM) 6

pH
2

1 5
Ca

0 4
0 5 10 15 20 25 30
Time (1000 s)

Figure 5.33. pH and Ca increase with kinetic dissolution of calcite. Full line indicates the PWP rate,
dotted line is for the simplified rate.

The plot in Figure 5.33 shows for the simple formula a smaller final Ca concentration, since complexes
and activity coefficients and the increase of the calcite solubility at 10°C were all neglected. Also, pH and
Ca2 increase slower than with the PWP calcite rate because the proton effect on the rate is neglected.
Similar differences will be found with all other rate equations such as Equation (5.40) that do not account
for specific reactants.

QUESTIONS:
Calculate the concentrations for 25°C instead of 10°C, comment on the change in shape of the Ca/time
curves.
ANSWER: the simple formula yields the same curve. In the calcite rate the proton
contribution is enhanced (cf. temperature dependence of k1), the solubility is
reduced.
Calculate the concentrations with time for [PCO2]  103.5. Is the reaction quicker?
ANSWER: saturation is attained about 4.6 times faster.
Compare the solubility of calcite at 103.5 and 101.5 [PCO2]?
ANSWER: About (102) ⁄  4.6 times smaller at [PCO2]  103.5 (hence, saturation is
1
3

reached 4.6 times faster).

Plot the log(rate) vs pH, and compare the two rate formulas. (Hint: in calcit2 add line 12 if rate > 0 then
put (log10(rate), 1). In USER_GRAPH, graph_x -la(“H1”) and graph_y get (1). For the calcite rate, copy
the rate definition from PHREEQC.DAT into the input file, add the line: 122 if moles > 0 then put
(log10(moles), 1).
ANSWER: The simple rate is independent of pH  6, in contrast to the calcite rate.
Plot log (rate) vs pH, starting at pH  3 (Hint: define pH 3 in solution 1, add Cl 1 charge).
Consider the effect of changing the exponent on the saturation ratio in the PWP calcite rate from 2⁄3 to 1 (the
exponent of 1 is given by strict microscopic reversibility).
ANSWER: with an exponent of 1 the log(rate)/pH curve turns more sharply to
saturation.

We may apply the calculated saturation time to a real world situation, for example a shaft in lime-
stone along which water flows in small films (Weyl, 1958; Dreybrodt, 1981; Gabrovšek and

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of carbonate reactions 215

Dreybrodt, 2000), or a small film of water which covers calcite fragments in a soil. The reaction rate
is in units mmol/cm2/s and must be multiplied by the available surface area A (cm2) of calcite, and
divided by the solution volume V (cm3) to obtain the actual rate in mol/L/s. Take a waterfilm of
0.05 cm thickness, which gives A/V  20/cm, and t.95 becomes 25 minutes only. The distance cov-
ered during that timespan can be calculated for a specific hydrological situation. For a soil with a net
precipitation of 0.3 m/yr and 10% water filled porosity, the percolation velocity is 3 m/yr. During 25
minutes a flowpath of only 0.2 mm is covered, sufficient to reach concentrations corresponding to
95% of calcite saturation values.
The average flow velocity in a shaft can be calculated from:

2 g d h
H  (5.49)
2O 3 dz

where
is waterfilm thickness (m),  is density of water (1000 kg/m3), g is the acceleration in the
earth’s gravitation field (10 m/s2), is viscosity of water (1.4103 Pa s at 10°C), and dh/dz is the
hydraulic gradient (m H2O/m). The average flow velocity in a waterfilm of 0.5 mm, with a hydraulic
gradient of 1, is v  0.6 m/s (or about 2 km/hr). The distance, covered in 25 minutes is accordingly
about 890 m.
Clearly, the dissolution kinetics of calcite are fast enough to reach equilibrium in a soil containing
calcite fragments, whereas the flow of water over barren rock may be so fast that the water reaches
the inner realms of the rock while still undersaturated, and thereby contributing to karst formation.
However, the dominant process that explains cave formation in pure carbonate rocks is undoubtedly
the mixing of waters with different CO2 pressures discussed before.
The dissolution kinetics of dolomite are much slower than for calcite. Busenberg and Plummer
(1982) interpreted dolomite dissolution experiments with a rate expression similar to the one for cal-
cite. However, the resulting equation predicts negative rates while the solution is still strongly sub-
saturated (SI  0.5), and allows no extrapolation from the highly subsaturated pH-stat
experiments to the “free-drift” conditions which apply in nature. Other experiments at low CO2 pres-
sure indicate that the dissolution rate can be appreciable, but still 10–20 times slower than for calcite
(James, 1981; Appelo et al., 1984; Schulz, 1988; Chou et al., 1989; Pokrovsky and Schott, 2001).
Even equilibrium may be reached, permitting the derivation of the dolomite solubility product from
the solution composition (Sherman and Barak, 2000). Appelo et al. (1984) related the dissolution
rate to the natural logarithm of the saturation ratio:

r  dmCa / dt  k d ln(IAP / K dol ) (5.50)

It is possible to integrate this equation analytically, but it is easier to let PHREEQC do the work. For
the conditions used previously for calcite (10°C, [PCO2]  101.5, A / V  1/cm) and kd 
1.21010 mmol/cm2/s, the time to reach 95% saturation is t.95  1.4107 s (160 days; Problem
5.18). This is about 450 times longer than required for calcite.
In contrast, field observations indicate that karst phenomena are identical for both calcite and
dolomite rocks, with similar dissolved concentrations in springs and streams. How can these field
observations be reconciled with the difference in laboratory rates? The main reason is that dolomite
rock weathers (or rather, falls apart) into a grainy, friable material which almost consists of individ-
ual crystals. This material gives the magnificent scree slopes which are renowned in the Dolomites
of N.-Italy and alps all over the world. The grainy texture gives a high ratio of surface area over water
volume and increases the overall dissolution rate. Karst rivers draining dolomite rocks may in
fact show higher and more steady concentrations during runoff peaks than rivers draining calcitic
terrains (Zötl, 1974).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

216 Carbonates and carbon dioxide

7 8.5

July 28, 1984

3 (mmol/L)
6 8.0
pH
Ca2 and HCO

HCO

pH
3 7.5
5

Ca2
4 7.0

Falls
Cave Surface stream
3 6.5

1.8 1.0
July 28, 1984
1.4 SIcc 1.4

1.0 log PCO 1.8

log PCO2
2
SIcc

0.6 2.2
SIcc
0.2 2.6
0.0
0.2 Falls 3.0
Cave Surface stream
0.4 3.4
0 0.5 1.0 1.5 2.0 2.5 5.7
Distance downstream (km)

564
120

100 October 14, 1984

Calcite precip. rate (mmol/kg/s)

80

60

40 Plummer et al., 1978

20

0
F-1
F-3
F-4

F-5

F-2

F-6
D-1

D-3

20
0 2 4 6 8 10 12 14 42
Distance/velocity (ks)

Figure 5.34. Calcite precipitation along Falling Spring Creek. (Upper 2 graphs) Change of concentrations
along the reach. (Lower graph) Observed mass which is deposited, compared with predicted mass according to
the Plummer et al. (1978) rate equation. From: Herman and Lorah, 1988.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of carbonate reactions 217

QUESTIONS:
The solubility of dolomite for [PCO2]  103.5 is 5.2 times smaller than for [PCO2]  101.5 (calculated
with PHREEQC). Estimate the time to reach 95% saturation for [PCO2]  103.5.
ANSWER: (160 / 5.2)  31 days. (check with PHREEQC, Problem 5.18!)
Estimate the solubility decrease of dolomite, when [PCO2] decreases 100-fold.
ANSWER: 100 ⁄  4.6 times smaller solubility.
1
3

5.6.2 Precipitation
The calcite dissolution rate model of Plummer et al. (1978) contains a backward term, which has
been used successfully to describe precipitation of calcite (Plummer et al., 1979; Reddy et al., 1981;
Busenberg and Plummer, 1986). An interesting study by Herman and Lorah (1988) compared the
mass of calcite deposited by a karst stream with the amounts predicted by the Plummer et al. (1978)
rate equation. The actual deposited mass was calculated from decline of concentrations at sampling
points along the stream, as shown in Figure 5.34. The theoretical mass was obtained by multiplying
the rate, calculated for the streamwater composition, with the wetted perimeter of the streambed. The
comparison is shown in Figure 5.34, and we may note some discrepancies; the predicted amounts are
lower where a waterfall increases outgassing of CO2 and precipitation, and are higher in the other
stretches of the stream. However, agreement is generally within a factor of 2–3.
One problem is to estimate the correct reactive surface area. It is clear that irregularities in the
deposited travertine present a much larger microscopic surface than is given by the wetted perimeter
of the stream. Experiments where small blocks of calcite were placed in a calcite depositing stream
and the weight increase of the block was measured (Dreybrodt et al., 1992; Zaihua et al., 1995)
yielded much less precipitation than predicted from the laboratory rate equation, even though the ini-
tial surface area of these blocks rapidly increased as they became covered by a myriad of small calcite
crystals. The slower rates in streams may also be due to inhibition by foreign ions such as Mg2
(Zaihua et al., 1995) or organic matter (Plummer et al., 2000). In relatively simple systems where the
limestone is pure, the field data agree better with predictions from the laboratory rate equation (Baker
and Smart, 1995).
Organic acids and phosphates inhibit the precipitation of carbonate in fresh water environments
(Inskeep and Bloom, 1986). The inhibitory action already starts at low concentrations of phosphates
(1.0 mol/L) and organic acids (10 mol/L, Reddy, 1977; Lebron and Suarez, 1996). The inhi-
bition of calcite precipitation by the high Mg2 content of seawater has already been mentioned.
Generally, divalent cations that form solid carbonates tend to inhibit the growth rate of calcite
increasingly more as the solubility of their solid carbonate form decreases (Terjesen et al., 1961;
Gutjahr et al., 1996). Figure 5.35 shows that the growth rate is halved by a concentration of Mg2 of
103.8 M (0.16 mM) and of Fe2 of 107.3 M (0.05103 mM) while the pK’s of their carbonates are
8.24 and 10.89, respectively (Table 5.1). Brown and Glynn (2003) suspended sacks with calcite and
dolomite in wellscreens in acid groundwater and noted dissolution rates that were 1000 times
smaller than theoretically expected, which may be related to the high concentrations of heavy metals
in the acid mine drainage that was the source of the groundwater.

QUESTION:
Estimate the inhibitory action of Cd2 on calcite precipitation? (Hint: find Kotavite in Table 5.1)
ANSWER: Figure 5.35 suggests very strong inhibition for pKotavite  13.7.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

218 Carbonates and carbon dioxide

2
Calcite
3 Sr Ba Mg Aragonite

log c0.5 (mol/L)

4
Mn Co Ni
5
Pb
6 Cu
Fe
7
Zn
8
14 12 10 8 6
log K

Figure 5.35. Concentrations of divalent cations by which the calcite and aragonite precipitation rates are
halved as a function of the solubility product of the metal-carbonate (Modified from Gutjahr et al., 1996).

5.7 CARBON ISOTOPES

Carbon occurs in nature with two stable isotopes, 12C and 13C (Table 2.4). The
13C of dissolved carbon
species in groundwater depends on the
13C signature of the dissolving C-source and the fractionation
among the carbonate species in the solution (cf. Table 5.7). Atmospheric CO2(g) has a
13C
7‰.
When this CO2(g) dissolves in water, 13C is slightly depleted, resulting in a
13C
8‰ for CO2(aq).
However, subsequent hydration of CO2(aq) favors the heavier isotope and produces a
13C
2‰
in HCO 3 . The overall
C of the dissolved carbonate depends therefore on the relative proportions
13
 2
of CO2(aq), HCO3 and CO3 .

Table 5.7. Enrichment of 13C in carbonate species with respect to CO2(gas), and

13C in species in equilibrium with a given
13C of CO2(gas).

Species ‰ enrichment
13C of species

i 13Ci / CO2(g)

13CCO2(g)  7‰
13CCO2(g)  23‰

CO2(aq) 1 8 24
HCO3 9 2 14
CO2
3 8 1 15
CaCO3 11 4 12

QUESTIONS:
2
What is the enrichment of 13C in the reaction HCO3 → CO3  H?
ANSWER:  CCO32 / HCO
13
3
  CCO32 / CO2(g)  13CHCO
13
3 / CO2(g)

1‰
Water in equilibrium with calcite at [PCO2]  101.5 and
13CCO2(g)  23‰ has pH  7, H2CO*3  1.1 mM,
HCO3  5.0 mM and CO2 3  8 M. Estimate
C of the solution?
13

ANSWER:
13C  ((24  1.1)  (14  5)) / 6.1  16‰

Estimate (roughly)
13C of water in equilibrium with calcite and [PCO2]  101.5?
ANSWER: the proportion of H2CO*3 decreases, hence
13C ! 16‰

The majority of terrestrial plants (C3 plants) uses CO2 for photosynthesis. During diffusion
of atmospheric CO2 (
13C
7‰) through the stomata and with enzymatic conversion of CO2
to CH2O in the chloroplast, 13C is depleted by about 20‰ to give
13C
27‰ for carbon in the C3
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Carbon isotopes 219

plant tissue. Some arid zone vegetations (C4 plants) are able to close off the stomata to reduce
transpiration and incorporate HCO 3 from the cell fluid, which is isotopically heavier and results
in a heavier
13C
13‰ of C4 plant material (Farquhar et al., 1989; Vogel, 1993). The name
C4 plant refers to the 4-carbon molecule malate that forms in the cells by reaction of HCO 3 with pyru-
vate and transports the CO2 to the sites where conversion to CH2O takes place (cf. Berg et al., 2002).
When the plants and roots decay in the soil, CO2-gas with
13C
27‰ from C3 plants diffuses
upward to the atmosphere. The diffusion coefficient of 12CO2(g) is 4.4‰ higher than of the heavier
13
CO2(g) (Chapter 3). To obtain fluxes into the atmosphere which balance the production of the two iso-
topes, the concentration gradient of 13CO2(g) must be larger by about 4.4‰, which increases the
13C of
soil CO2 gas to
23‰ in the temperate climate (Dörr and Münnich, 1980; Cerling, 1984). In very poor
soils with a CO2 pressure lower than 102.7 atm,
13C of soil gas increases with the reciprocal of the CO2
gas pressure to 7‰, the value of the atmosphere (Van der Kemp et al., 2000; Amundson et al., 1998).
During open system dissolution of calcite, the isotopes in CO2 gas determine the
13C of water
because the CO2 gas/water exchange is quick (Chapter 10). Thus, water in equilibrium with calcite
and [PCO2]  101.5 will have
13C
16‰ (Question below Table 5.7). In the case of closed system
calcite dissolution, the TIC is derived in about equal proportions from CO2 gas (
13C
24‰
in H2CO*3 which is mostly CO2(aq)) and from CaCO3 (
13C
0‰ in marine carbonate, similar to the
PDB standard), which results in
13C
12‰. We can calculate the change of
13C of TIC during
dissolution precisely with PHREEQC (Example 5.10).

EXAMPLE 5.10. PHREEQC calculation of
13C during calcite dissolution

For the case of open system dissolution, the principle is to calculate the isotopic composition of the carbonate
species with respect to CO2 gas. Summing them up in proportion to their contribution to TIC yields
13C of the
solution. The
13C of the carbonate species are calculated with respect to the constant
13C of CO2 gas using
enrichment formulas selected by Clark and Fritz (1997). The concentrations of the carbonate species and of TIC
(which all change during calcite dissolution) are obtained from PHREEQC (lines 40–70 in USER_GRAPH in the
input file). Dissolution is modeled using keyword KINETICS.
In the case of closed system dissolution, the isotopic composition of the initial (calcite free) solution
is calculated as in the open system. The mass of
13C  TIC0 (d13 ms in line 90 in USER_GRAPH) is stored
in memory location 1 (line “92 put(d13 ms, 1)”). Adding
13C from calcite to the inital mass, and dividing by
TIC, gives
13C of the solution. The calculation of initial
13C is bypassed during stepwise addition of the
reaction (line “2 if step_no ! 0 then goto 100”).
The figure produced by USER_GRAPH (Figure 5.36) shows that closed systems have higher
13C values
than open systems (this is also true when the final Ca2 concentrations are the same, cf. Question). The
13C
value will increase when initial [PCO2] is smaller (cf. Question).

12
Equilibrium
14
Closed
16

13C (‰)

Open
18

20

22

24
0 1 2 3 4
Calcite (mmol/L)

Figure 5.36.
13C of water when calcite dissolves in open or closed conditions with respect to CO2 gas.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

220 Carbonates and carbon dioxide

DATABASE phreeqc.dat
# Find d13C. Open system dissolution of calcite...
PRINT; -reset false
SOLUTION 1
-temp 10
EQUILIBRIUM_PHASES 1; CO2(g) -1.5
KINETICS 1
Calcite; -m0 1; -parms 200 0.67; -steps 3600 in 100
INCREMENTAL_REACTIONS true

USER_GRAPH
-head cc open_d13C
# Find d13C in solute species 1 = CO2aq, 2 = HCO3-,3 = CO3-2
-start
2 if step_no = 0 then goto 120
# Define enrichments (eps) relative to g  CO2_gas from Clark and Fritz, 1997, p. 121 ...
4 T = TK
10 eps_1_g = 0.19 - 373 / T
20 eps_2_g = -24.1 + 9.552e3 / T
30 eps_3_g = -3.4 + 0.87e6 / T^2
# Find carbonate species...

40
co2 = mol("CO2")
50
hco3 = mol("HCO3-") + mol("CaHCO3+")
60
co3 = mol("CO3-2") + mol("CaCO3")
70
tic = co2 + hco3 + co3
# and d13 of the species...
80 d13g = -23
90 d13tot = (co2 * (eps_1_g + d13g) + hco3 * (eps_2_g + d13g) +\
co3 * (eps_3_g + d13g)) / tic
# chart...
100 graph_x tot("Ca") * 1e3
110 graph_y d13tot
120 end
-end
END

# simulation 2. Find d13C. Closed system dissolution...

SOLUTION 2
-temp 10; pH 6 charge; C 1 CO2(g) -1.5
KINETICS 2
Calcite; -m0 1; -parms 200 0.67; -steps 1200 in 30
USER_GRAPH
-head cc closed_d13C
-start
2 if step_no > 0 then goto 100
# Copy lines 4 .. 80 from the open system above ...
# 4 ...
# ...
# 80 ...
# Calculate
13C x TIC of the initial solution...
90 d13ms  co2 * (eps_1_g + d13g) + hco3 * (eps_2_g + d13g) +
\co3 * (eps_3_g + d13g)
# and store in memory 1...
92 put(d13ms, 1)
94 goto 120
# chart...

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon isotopes 221

100 graph_x tot("Ca") * 1e3

# d13C from calcite (d13C = 0) and from initial
solution (memory 1)...
110 graph_y (tot("Ca") * 0 + get(1)) / tot("C(4)")
120 end
-end
END

QUESTIONS:
Find
13C for closed system dissolution of calcite that gives gives the same Ca2 concentration as for con-
stant [PCO2]  101.5 (Hint: use initial [PCO2]  101.05 in solution 2).
ANSWER: mCa2  3.2 mM;
13C  15.6‰ (open for CO2 gas);
13C  14.5‰
(closed for CO2 gas)
Find
13C for open and closed system dissolution of calcite when initial [PCO2]  102.5, 103.5?
ANSWER: 13.99 and 10.80‰ for [PCO2]  102.5 and 13.54 and 3.36‰ for
[PCO2]  103.5.

5.7.1 Carbon-13 trends in aquifers

Figure 5.37 shows a plot of
13C of various karst water types (pools, fast drip, and stalactite and film
waters) sampled in the Soreq cave in Israel (Bar-Matthews et al., 1996). The
13C values range
from 15 to 5‰ and are hyperbolically related to TIC (grey outline in Figure 5.37). When calcite
precipitates, TIC decreases and
13C of the water increases. This is illustrated by the lines in Figure
5.37 which connect the compositions of three fast-dripping waters with the pool waters
in which they fall and from which calcite precipitates. The reaction is:

Ca 2  2HCO
3 → CaCO3  CO 2(g)  H 2 O (5.51)

The 13C enrichment is largely due to the escape of isotopically light CO2 that enriches TIC by 9‰ (Table
5.7), whereas the precipitation of calcite will lighten
13C of TIC by only about 1‰. Because calcite and
CO2(g) are produced in equal proportion, the water is enriched by (9  (1)) / 2  4‰ in a Rayleigh
process that generates the hyperbolic relation between TIC and
13C (Bar-Matthews et al., 1996).

3
Pool water
5 Fast-drip water
Stalactite & film water
δ13C (‰ PDB)

7

9

11

13

15
0 2 4 6 8 10 12
TIC (mM)

Figure 5.37. Plot of
13C vs TIC of waters from the Soreq cave, Israel. The arrows connect pool water com-
positions with their fast-drip source (Bar-Matthews et al., 1996).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

222 Carbonates and carbon dioxide

The process is modeled with PHREEQC in Chapter 11. The enrichment of 13C in calcite depends on the
precipitation rate (Turner, 1982), in the same way as the composition of solid solutions is a function of
kinetics (Chapter 4). Bar-Matthews et al. (1996) found an enrichment near 1‰, but for slowly precipi-
tating calcite the enrichment may become as low as 3‰.
Groundwaters, which have equilibrated with calcite and are closed with respect to gas exchange,
are expected to maintain a stable chemical and isotopic composition during flow through the aquifer.
Nevertheless, groundwaters in many aquifers show a trend of increasing
13C with increasing age
along the flowpath (Bath et al., 1979; Chapelle and Knobel, 1985; Plummer et al., 1990; Aravena
et al., 1995; Kloppmann et al. 1998). The increase may be the result of renewed dissolution of
carbonates, due to dedolomitization (Edmunds et al., 1982; Plummer et al., 1990, cf. Section 5.5.1)
or loss of Ca2 from solution into the cation exchanger (Chapelle and Knobel, 1985; Chapter 6).
Methanogenesis, whereby CO2 is converted into CH4 by bacteria, will also increase
13C of the
remaining CO2 (Aravena et al., 1995) and is discussed in Chapter 9.
As an example of dedolomitization, let’s consider the Triassic Sherwood sandstone aquifer in the
English East Midlands (Figure 5.38). The aquifer is about 200 m thick and consists of quartz with
some detrital and authigenic feldspars, and 1–4% detrital dolomite and calcite. Figure 5.38 presents
concentrations plotted as a function of the water temperature, which increases downdip in the aquifer
and is used as a proxy for the residence time of the groundwater (Edmunds et al., 1982; Smedley and
2
Edmunds, 2002). Modern water has high Cl and SO4 concentrations compared with older Holocene
 2
water in which Cl is generally less than 30 mg/L. In older water, SO4 increases from 10 to
100 mg/L as result of gypsum dissolution. The Mg / Ca ratio and the HCO
2 2
3 concentration
remain more or less constant which suggests that dedolomitization takes place (Section 5.5.1). The
evolution in groundwater composition has been attributed to 1) congruent dissolution of dolomite,
2) dedolomitization and 3) reduction which introduces ferrous iron and depletes oxygen. The molar
Mg2 / Ca2 ratio tends to decrease somewhat with increasing temperature, which is expected when
the groundwater remains in equilibrium with calcite and dolomite. The Mg2 / Ca2 ratio is smaller
than predicted for pure dolomite, perhaps because ankerite (CaFe(CO3)2) is also present. Finally,

13C increases from 13‰ to 10‰ in conjunction with the increase of SO2 4 .
The evolution of
13C during dedolomitization can be calculated from the mass balance in which
calcite precipitates and dolomite dissolves:

13Ci  TICi 
13Cdol  2  Ddol  (
13C  ecc/sol )  Pcc

13C  (5.52)
TICi  Ddol  Pcc

where D indicates moles of dissolution, P is moles of precipitation, and subscripts i, dol and cc stand
for the initial solution composition, dolomite and calcite, respectively. The
13C of the solution is
also present on the right of the equal sign where it determines the isotopic composition of the calcite
precipitate. Therefore, the equation must be integrated in small steps from 0 to the amount of pre-
cipitate for a given amount of gypsum. However, if we neglect the fractionation in the precipitate
(assume cc/sol  0), we can directly calculate the isotopic composition of the solution from the mass
balance of the original solution and the dissolving dolomite. If as much carbon from dolomite is
added as was present as TIC in the initial solution,
13C of the solution will become 7.5‰, if

13Ci  13‰ and
13Cdol  2‰. Thus, the solution becomes heavier by dissolution of dolomite,
and the calcite precipitate (which has the
13C of the solution) is lighter than the dolomite.
This behavior was confirmed in simple dissolution experiments with Sherwood aquifer rock samples
in acid (Figure 5.39, Edmunds et al., 1982). Initially, the acid dissolves calcite with a low
13C, followed
by the slower reaction of dolomite with a higher
13C. Thus,
13C of the solution increases over time, as
illustrated in Figure 5.39. In the experiment,
18O of CO2 increases together with
13C, indicating that
fresh water carbonate (with the low 18O of fresh water) dissolves at the onset of the experiment.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon isotopes 223

East
NNW Worksop Retford R.Trent Lincoln SSE
Whisker
Hill Egmanton
meters Markham meters
Sunnyside Elkesley Clinton Rampton Newton South Boultham Welton
200 Grove Cottam Scarle 200
Ordsall
Modern
O.D Cl ! 20 mg/L L&M. O.D
Juras
sic
Holocene Colw Pen
ick a r th L.L
Cl  10 mg/L FM Me grou ias
200 rcia p 200
Redox mu
Pleistocene ds ton
boundary 18 e
δ O8.5 to 9.6
400 Cl 10–20 mg/L 400
She Saline
rwoo
d sand interface
Up ston
per M.P e
600 c a rb erm 600
oni ian U.Ma U
fero ma gne .Pe
us L.M rl sia
n rmia
a gne lim nm
sia esto arl Cl 500 mg/L
n li ne
me
800 0 kms 10 sto
ne 800

3 9
Mg/Camol

pH
8
1

0 7
10 15 20 10 15 20
300
1000 RB RB
HCO3 (mg/L)
SO4 (mg/L)

200
100

100
10
0
10 15 20 10 15 20
10000 8
RB
10
δ13C (‰)
Cl (mg/L)

1000
12

100 14

16
10
18
10 15 20 10 15 20
Temperature (°C) Temperature (°C)

Figure 5.38. Cross section over the Sherwood aquifer and variation of waterquality parameters as a function
of the water temperature, which is used as a proxy for residence time. RB indicates the redox boundary where
water becomes depleted of oxygen (Smedley and Edmunds, 2002).

The
18O in Figure 5.39 is given relative to the PDB standard (which is a marine carbonate with about
the same
18OPDB as the dolomite in the aquifer).
The calculation of isotopic compositions which includes fractionation into precipitates can be
done accurately with the KINETICS module of PHREEQC as shown in Example 5.11.

EXAMPLE 5.11. PHREEQC calculation of 13C evolution during dedolomitization

Calculate
13C when 1 mmol gypsum/L is added to a solution with 3.3 mM TIC with
13C  13‰, in
equilibrium with calcite and dolomite.

ANSWER:
We use the kinetic integrator of PHREEQC, which automatically adjusts the steps in the integration process to
attain the required accuracy of the answer. The reactions are programmed for 1 second overall time since we only
want the net result. Thus, line 10 in the rate for gypsum in the input file below will add 1 mM gypsum in total.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

224 Carbonates and carbon dioxide

3
(30) (46)
2
1
0

δ18O (‰ PDB)
1
2
3
4
5
6
7
8
9
0
(30)
1
δ13C (‰ PDB)

2
3
4 (46)
5
6
7
8
0 10 20 40
Reaction time (hours)

Figure 5.39. Trends in
13C and
18O in CO2 released during dissolution experiments with Sherwood aquifer
samples. The increase with time indicates the transient dissolution of fresh water carbonate followed by marine
dolomite (Edmunds et al., 1982).

The rates for calcite and dolomite provide reaction amounts in small steps. The reaction amounts are stored in
memory and used by the rate “Carbon13” which keeps track of the concentration of 13C (in mol/L). Note that
the rates for calcite and dolomite are so high (0.1 M/s) that equilibrium is always attained when gypsum is added.
In the rate “Carbon13” we use isotope fractionation factors rather than enrichments since the actual concen-
tration of the individual isotopes is calculated. With the speciation formula (Equation 5.13) the fraction
of HCO 12 13 
3 of TIC is obtained (this is C), and similarly, the fraction of H CO3 of the total moles of C in
13

solution (the total is given by “m” in line 90). The ratio of the two, and the fractionation factor for calcite and
solution, determine the loss of 13C. The gain in 13C from dolomite dissolution is found in line 120, where the
reaction amount of dolomite is multiplied with the moles of 13C in the dolomite and with 2 (the formula is
CaMg(CO3)2).
2
As initial solution we define a water with 3.3 mM TIC (195 mg HCO 3 /L) and 0.1 mM SO4 in equilib-
rium with calcite and dolomite.
# Find 13C when gypsum is added to calcite/dolomite equilibrated solution
RATES
Gypsum # Add 1 mmol gypsum in 1 s
-start
10 save 1e-3 * time
-end

Calcite # rate is for 13C integration only. . .

-start
10 moles  1e-1 * (1 - sr("Calcite")) * time
20 put(moles, 1) # . . .for C13 kinetics
30 save moles
-end

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon isotopes 225

Dolomite # rate is only for 13C integration. . .

-start
10 moles = 1e-1 * (1 - sr("Dolomite")) * time
20 put(moles, 2) # . . .for C13 kinetics
30 save moles
-end

Carbon13 # Distribute 13C among solute species

# 1 = CO2aq, 2 = HCO3-,3 = CO3-2, and in calcite.
-start
# Define alpha’s from Clark and Fritz, 1997, p. 121 ...
4 T = TK
10 aa_1_2 = exp(24.29e-3 - 9.925 / T)
20 aa_3_2 = exp(20.7e-3 - 9.552 / T + 870 / T^2)
30 aa_cc_2 = exp(26.561e-3 - 17.2183 / T + 2988.0 / T^2)
# Find H12CO3- and H13CO3-...
50 aH = act("H +")
60 K1 = aH * act("HCO3-") / act("CO2") / act("H2O")
70 K2 = aH * act("CO3-2") / act("HCO3-")
80 m12C = (tot("C(4)")) / (aH/K1 + 1 + K2/aH)
90 m13C = m / (aa_1_2 * aH/K1 + 1 + aa_3_2 * K2/aH)
# and the ratio 13C/12C in HCO3-. . .
100 R2 = m13C / m12C
# Fractionate 13C into calcite . . .
110 d13C_cc = aa_cc_2 * -get(1) * R2
# Add 13C from dolomite with d13C = -2 permil . . .
120 d13C_dol = get(2) * 2 * 0.0112145
# Integrate. . .
130 save d13C_cc - d13C_dol
-end
KINETICS 1
Gypsum; Calcite; Dolomite
Carbon13; -formula C 0; -m0 0.0366E-3 # d13C  -13 for TIC  3.3e-3
-step 1 # only 1 step in 1 second
SOLUTION 1
-temp 15; pH 7 charge; S(6) 0.1; Ca 1 Calcite; Mg 1 Dolomite; C(4) 3.3
USER_PRINT
-start
10 Rst = 0.011237
20 print 'd13C of solution :', (kin("Carbon13") / tot("C(4)") / Rst - 1) * 1e3
-end
END

The results are presented in Table 5.8. Note that
13C increases from 13‰ to 11‰, that the alkalinity
remains equal, and that pH decreases in agreement with the observed water qualities (Figure 5.38, consider
the unpolluted samples to the right of the RB). Adding 10 mM gypsum will increase
13C to 6.2‰. In the
aquifer, however, concentrations above 1000 mg SO2 4 /L are accompanied by
C
17‰ (Figure 5.38).
13

This low value is indicative of methanogenesis.

Table 5.8.
13C and water composition during dedolomitization of Sherwood aquifer water. Concentrations
in mmol/L, pH units and ‰ for
13C.

Water SO42
13C TIC Alk pH Mg2 Ca2 cc dol

Initial 0.1 13 3.30 3.19 7.76 0.85 0.84 0 0

 reaction 1.1 11.1 3.26 3.11 7.64 1.32 1.33 0.98 0.47

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

226 Carbonates and carbon dioxide

QUESTIONS:
Give an estimate of
13C using only the reaction amounts listed in Table 5.8 (i.e. neglect fractionation in
the precipitate)
ANSWER: (13 3.3  (2)  (0.47  2)) / (3.3  (0.47  2))  10.6‰
Why is the solution, calculated with PHREEQC, lighter?
ANSWER: the solution loses relatively more 13C than 12C to calcite.
Calculate
13C with PHREEQC when 5 mM gypsum enters the solution?
ANSWER: by hand: 6.6‰, with PHREEQC: 7.3‰
Program the calculation for total time 3  109 s (100 yrs, gives about 1.5 km flow distance in the field)
ANSWER: in all rates, redefine “time” to “time/3e9”, in KINETICS use -step 3e9

14
5.7.2 C and groundwater age
The travel time of groundwater can be calculated from the decay of radioactive carbon, 14C accord-
ing to (cf. Chapter 4):

dc / dt  c (5.53)

which solves for 14C to:

t 14 C  ln(14 A / 14 A0 ) / 1.21104  8267 ln(14 A0 / 14 A) (yr) (5.54)

where   1.21104 is the decay constant (yr1), 14A is the activity of 14C in TIC in the down-
stream sample (expressed in % modern carbon, pmc) and 14A0 is the activity in the upstream sample.
The difficulty is to define the 14A0 and the chemical reactions which may have modified 14A during
flow (Kalin, 2000).
Usually, 14A0 must be found for soil water in the recharge area. When soil water equilibrates with
CO2 from the atmosphere and young organic matter, it obtains 14A  100 pmc (since nuclear
bomb testing in the atmosphere this value has augmented, but also over the Pleistocene and
Holocene the values have fluctuated). When “dead” carbon from old carbonates dissolves in the
unsaturated soil, 14A of soil water still remains 100 pmc because of CO2 exchange with soil gas.
Therefore, for open system dissolution of carbonate, 14A0  100 pmc. If the carbonate dissolves
below the water table, gas exchange is absent and TIC is diluted by carbon from the old carbonate.
We calculated for closed system dissolution that half of the TIC comes from H2CO*3 from the soil,
and half from the solid carbonate. Accordingly, 14A0  50 pmc for closed system carbonate dissolu-
tion (50 pmc is correct for PCO2,soil ! 102 atm, otherwise 14A0 will be lower, Problem 5.20).
Empirical evidence so far suggests that some intermediate value in between 50 and 100 pmc should
be used (Clark and Fritz, 1997).
In the Triassic Sherwood aquifer (Figure 5.38), the groundwater samples from the recharge zone
which contain tritium, have 14A ranging from 40 to 80 pmc. The pH of the groundwater of around 8
indicates closed system carbonate dissolution with an initial [PCO2] ! 102, and therefore, a value of
50 pmc for 14A is expected. The low value of 40 pmc may indicate mixing of old and young water in
the long screen of the production well. On the other hand, the high value of 80 pmc suggests a mix-
ture of waters with open and closed system dissolution. Using the average of 14A0  60 pmc for
water in the recharge zone of the Sherwood aquifer, we can calculate the groundwater age, corrected
for dilution by the dedolomitization reaction (Example 5.12).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon isotopes 227

EXAMPLE 5.12. Groundwater age from 14C

Calculate the age of water in the Sherwood aquifer from well Ompton 2 which has 50 pmc, 2.9 mmol TIC/L,

13C  12.7‰ and 5 mg SO42/L. Also for well Gainsborough 2 with 4.8 pmc, 4.3 mmol TIC/L,

13C  10.4‰ and 201 mg SO42/L. Assume 14A0  60 pmc. (
13C from Smedley, pers. comm., 14A from
Bath et al., 1979).

ANSWER:
Water from Ompton 2 with 5 mg SO42 shows no dedolomitization, hence t  8267 ln(60 / 50)  1507 yr.
In water from Gainsborough 2 with 201 mg SO42, 0.92 mM dolomite has dissolved, diluting TIC to
4.3  2  0.92  6.14 mM (calculated according to Example 5.11). Thus, 14C is diluted to
(14A0)r  60  (4.3 / 6.14)  42.02 pmc (the subscript r indicates reaction corrected). The age of the water
is t  8267 ln(42.02 / 4.8)  17,935 yr. Model
13C  10.4‰, equals observed.

In Example 5.12, we have neglected the fractionation of 14C into precipitating calcite. We have seen
already that 13C enrichment will change the
13C of the solution by about 0.5‰ (Questions with
Example 5.11). 14C fractionates 2.3 times more than 13C and changes by 1.3‰, which gives
(14A0)r  42.02  0.13  41.89 pmc. The age then becomes 17,909 instead of 17,935 yr, which is a
negligible difference.
The uncertainties of translating the 14C age into real time comprise dilution and fractionation
effects associated with other reactions such as ion exchange and oxidation-reduction which
are discussed in later chapters. The computer code NETPATH (Plummer et al., 1994) is
useful for calculating the corrections. Often, the reactions can be traced to concentration
changes of specific ions, and 14A0 can be corrected as was done for the dedolomitization
reaction with SO42 (Plummer et al., 1990, 1994; Parkhurst, 1997; Van der Kemp et al., 2000).
For common reactions such as the conversion of aragonite into calcite and of high-Mg calcite
into low-Mg calcite, the correction can be quantified with
13C. Figure 5.40 shows the relation
of 14C and
13C in Chalk groundwaters from the Paris and N.-German Basins, and Figure 5.41
gives the Mg2 / Ca2 ratio of these waters as a function of
13C (Kloppmann et al., 1998).

100
Groundwaters
Un- & semiconfined
chalk, Paris basin
80 Confined chalk,
Paris basin
Chalk, North
A14C (pmc)

60 German basin

40
Solid chalk

20

0
16 14 12 10 8 6 4 2 0 2

13C (‰ vs PDB)

Figure 5.40. Plot of 14

C vs
13C in Chalk groundwaters from the Paris and N.-German Basin (Kloppmann
et al., 1998).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

228 Carbonates and carbon dioxide

2.0
Groundwaters
Un- & semiconfined
chalk, Paris basin
Confined chalk,
1.5
Paris basin
Molar ratio Mg/Ca

Chalk, North
German basin

1.0

0.5

0.0
16 14 12 10 8 6 4 2 0

13C (‰ vs PDB)

Figure 5.41. The Mg/Ca ratio increases with
13C in the groundwaters from the Paris and N.-German Basin
(Kloppmann et al., 1998).

The exponential correlation of 14C and
13C suggests that the decrease of 14C by radioactive decay
is enhanced by the dissolution of dead carbon from Chalk that has
13C of 2.2‰. The increase of the
Mg2 / Ca2 ratio with
13C indicates that the chemical reaction involves the incongruent dissolu-
tion of high-Mg calcite. Because the latter reaction is a gliding one that produces calcites of inter-
mediate compositions, it is still uncertain what the contribution is of these intermediate products on
the water composition, and whether these are really completely “dead”. The process can also be
lumped into a first-order kinetic recrystallization factor that increases  (Gonfiantini and Zuppi,
2003). In the case of the Paris Basin,  increases from 1.21104 to 2.95104/yr, which reduces
the “conventional” 14C age by a factor of 2.4.

5.7.3 Retardation by sorption and stagnant zone diffusion

Sometimes, unexpected low values of 14A are encountered in groundwater which give a much higher
age than follows from the flow velocity calculated with groundwater models, even after the various
chemical corrections have been applied. The discrepancy may be due to adsorption of carbonate,
which retards the flowtime of carbon and hence increases the decay of 14C, but is not visible in
changes of concentrations of other solutes or of 13C if the latter has attained sorption equilibrium
(Mozeto et al., 1984; Garnier, 1985; Garnier et al., 1985; Maloszewski and Zuber, 1991). With sorp-
tion, the age of the water becomes (Chapter 3):
tH  t 14 C / Rc (5.55)
2O

where Rc is the retardation, Rc  1  dq / dc, q is sorbed concentration expressed in the units of

solute concentrations, e.g. mol/L. The retardation can be included in a “working” decay constant
which becomes:
( 14 C ) R  Rc  14 C (5.56)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon isotopes 229

Actually, it is difficult to discern sorption on carbonate minerals from recrystallization. Mozeto et al.
(1984) found that sorption of 13C was reversible on aged, crushed calcite, but irreversible on freshly
crushed calcite which dissolved partly and incorporated the isotope in the reprecipitated crystals.
Garnier (1985) used a mixture of aragonite and calcite to measure the retardation of 14C in column
experiments and noted clogging due to calcite precipitation which apparently resulted from recrystal-
lization of aragonite. One molecular layer of the calcite surface was sufficient to explain the amount
of reversible sorption in the experiments of Mozeto. Labotka et al. (2000) measured carbon diffusion
in calcite in between 600 and 800°C and noted very small diffusion lengths, in agreement with unal-
tering carbon isotope ratios in calcite in metamorphic rocks. Although uncertain because it grossly
surpasses the experimental temperature range, their data extrapolate to D13C  61042 m2/s at 25°C
in calcite. This gives   (2Dt)  21015 m in 10,000 yrs, which is even less than the single layer
of carbonate ions that exchanged isotopes in the experiments of Mozeto et al. Thus, sorption on cal-
cite is limited to the outer surface layer, but probably with notable effects (cf. Example 5.13). In addi-
tion, sorption of carbonate takes place onto iron- and aluminum-oxyhydroxides, which gives
a retardation of about 2 in the case of the earlier discussed Sherwood (Bunter) sandstone aquifer
(of which the grains are red-stained by iron-oxyhydroxides, cf. Problem 7.2). This sorption reduces
the ages calculated in Example 5.12 by a factor of 2.

EXAMPLE 5.13. Estimate exchangeable carbonate on Chalk and the retardation of 14C
The rhombohedral face of a calcite crystal contains 1 molecule CO3/21 Å2 (Reeder, 1983). Estimate the
outer layer of carbonate ions of calcite (mol/L pore water) for Chalk with 10 m calcite crystals and 30%
porosity, and find the retardation of 14C when TIC  5 mM.

ANSWER:
One liter water contacts 1  (1  0.7) / 0.3  2.33 L sediment (solids only). There are 2.33 / (104)3 
2.331012 crystals with 1399 m2 total surface area. Thus, the Chalk has 1399 m2 / (211020 m2 per
molecule) / (6.0221020 molecules per mmol)  10.9 mmol/L exchangeable carbonate ions (q). Assume
linear sorption and negligible fractionation for 14C, and find R  1  dq / dc  1  q/ c 
1  10.9 / 5  3.

Example 5.13 shows that the retardation of 14C by exchange with carbonate groups on the surface of cal-
cite in Chalk may be 3. The same retardation applies for 13C, and therefore the exchange reaction
becomes “invisible” for the usual 14C correction methods after water with a given 13C has flushed the
porosity of the Chalk more than three times. The observed trend in 13C in the Paris Basin Chalk was
attributed to recrystallization of high-Mg calcite which gives a retardation for 14C of 2.4. However, if the
system has been flushed more than 3 times, the exchange reaction should be added, which augments the
overall retardation to R  2.4  2  4.4. The uncorrected 14C ages would then be a factor 4.4 too high.
Another retardation mechanism that may go unnoticed in 14C-age determinations is diffusion into
the matrix of limestone or in the low permeability layers of a sedimentary sequence (Neretnieks, 1981;
Sudicky and Frind, 1981; Maloszewski and Zuber, 1991; Sanford, 1997). Consider the situation of a reg-
ular alternation of sand and clay layers or of conduits in a marly limestone, etc., depicted in Figure 5.42.
Neglecting dispersion in the mobile zone, the steady state transport equation in the zone with mobile
water (x-direction) is (Chapter 3):
∂cm ∂c
 0  H O m  cm  Qz (5.57)
∂t 2 ∂x
where cm is the concentration in the mobile water (mol/m3) and Qz is the mass transfer into the stag-
nant zone (mol/m3/s). The mass transfer equals the diffusive flux into the stagnant zone ( im Fz),

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

230 Carbonates and carbon dioxide

Clastic Volcanic
sedimentary rock Fractured
deposits layers rock
Penetration by

Input (Co)
Dense diffusion
Clay Rock matrix
lava flow
Sand Rubble zone Fracture Porous matrix
Stagnant Dense b
Clay Rock matrix
zone lava flow Layer
widths
Flow zone Sand Rubble zone Fracture a

Clay Dense Rock matrix

C/Co
lava flow No diffusion
Sand Rubble zone Fracture

Clay Dense x
Rock matrix
lava flow

Figure 5.42. A regular alternation of mobile and mobile zones to model diffusion into the matrix of lime-
stones, volcanic rocks, or sand/clay alterations (Sanford, 1997), and concentration profile in the mobile zone,
with and without diffusion into the immobile zone.

counted twice for upper and lower contact, and normalized to the amount of water in the mobile zone
( m a), or:

2eim  ∂c 
Qz  De im  (5.58)
em a  ∂z z0

where a is the width of the mobile zone (m).

The concentration gradient in Equation (5.58) is obtained by first solving the steady state diffu-
sion equation in the stagnant zone:

∂cim ∂2 cim
 0  De  cim (5.59)
∂t ∂z 2

This is an ordinary differential equation (t  %):

d 2 cim 
 c  l 2 cim (5.60)
dz 2 De im

with the general solution:

cim  k1elz  k 2elz (5.61)

The two constants k1 and k2 are obtained from the boundary conditions. At the contact of the mobile
and immobile zone, cim,z0  cm  k1  k2. Midway in the immobile zone of width b, the concen-
tration gradient is zero, (∂cim / ∂z)zb / 2  0  k1lelb/2  k2le lb/2. Together in (5.61) this yields:

 b 
b
l ( z )
b
l (  z ) cosh  l   z 
e 2  e 2  2 
cim  cm  cm (5.62)
l
b
l
b  b
e2 e 2 cosh  l 
 2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Carbon isotopes 231

We can now find the concentration gradient at z  0:

b b
l l
dcim e 2 e 2  b
 cm l  cm l tanh  l   lwcm (5.63)
dz l
b
l
b
 2
e 2 e 2
When Equations (5.57), (5.58) and (5.63) are combined, a simple, ordinary differential equation
(valid for t  %) is obtained:

dcm  2e w 
vH   1  im  cm   Rim cm (5.64)
2O dx em al 


which gives the concentration profile in the mobile zone:

x
 Rim
cm ( x, ∞)  c0 e v  c0 e Rim t (5.65)

where c0 is the concentration (e.g. 14A0) that enters at x  0.

Equations (5.64) and (5.65) state that the decay constant of 14C is multiplied by a retardation fac-
tor which depends on the interaction of mobile and immobile water. When the immobile zone is fully
penetrated by 14C from the conduits, we can sense already that
Rim  (1  (beim)/(aem)),

i.e. the retardation is given by the ratio of the total and the mobile pore volume. Thus, Rim can be quite
high, depending on the volume of water in the matrix and in the conduits of the limestone (Example 5.14).

EXAMPLE 5.14. Estimate the retardation of 14C by matrix diffusion

Calculate Rim in Equation (5.64) given conduit width a  1 mm, fully filled with water m  1, and matrix
block size b  0.3 m with porosity im  0.3. The diffusion coefficient in the matrix De  0.3109 m2/s.

ANSWER:
The various constants solve to:

l  (/ De)0.5  (3.841012 / 3 1010)0.5  0.113/m (note that  is per second)

b/ 2  0.15 m

exp(l  b/ 2)  1.0171  A, exp(l  b/ 2)  0.9832  B

w  tanh(l  b/ 2)  (A  B) / (A  B)  0.01696

Rim  1  (2  im  w/ (m  a  l))  90.99

(which compares with (Rim)max  (1  (bim) / (am))  91).

QUESTIONS:
Find the retardation by sorption of carbonate when calcite crystals in Chalk are 20 m?
ANSWER: 1  5.4 / 5  2.
Find Rim and (Rim)max for b  10 m instead of 0.3 m in Example 5.14?
ANSWER: Rim  2717 and (Rim)max  3001.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

232 Carbonates and carbon dioxide

Give the decay constant for 14C which includes mobile/immobile exchange and linear sorption, q / c  2?
ANSWER: We assumed steady state (t  %), hence the contribution by sorption dqim /
dcim dcim / dt  0, sorption does not add to the retardation at t  %. Transient
profiles must be calculated numerically with PHREEQC using option stagnant
in keyword TRANSPORT (Chapter 6).

PROBLEMS
Laboratory exercise: dissolution of calcite in water and carbon dioxide
To simulate the dissolution process of limestone, 5 g calcite crystals are added to 200 mL distilled water in an
erlenmyer flask that is bubbled with 99% N2 and 1% CO2 gas. After one day, determine the solution pH (pH-
meter) and, after filtration over 0.45 m, Ca2, HCO 3 and CO2 (titrations).
Find:

pH  . . . . . Ca2  . . . . . mmol/L HCO3  . . . . . mmol/L CO2  . . . . . mmol/L

From analyzed CO2 we calculate PCO2  . . . . . atm

From CO2 / HCO 3 equilibrium follows pH  . . . . . , and [CO3 ]  . . . . . The saturation index for calcite,
2

SIcc  ..… (without activity coefficients), and SIcc  . . . . . (with activity coefficients). (Hint: Note how to use
activity coefficients when calculating [CO23 ]).

3 ]  . . . . . The saturation index for calcite, SIcc  . . . . . (without activity coefficients).

From measured pH, [CO2
The calcite crystals consist of rhomboeders. Assume them to be cubes with a side of 50 m with density
2.72 g/cm3. Calculate the time to reach 95% saturation with PHREEQC.

What would be the solution composition, pH, Ca2, CO2 3 , HCO 3, if we used 100 mM NaCl instead of distilled
water (Hint: Derive Equation (5.24) once again for 0.1 M NaCl while correcting for activity coefficients).

pH  . . . . . Ca2  . . . . . mmol/L 3  . . . . . mmol/L HCO 3  . . . . . mmol/L
CO2

5.1. Calculate TIC in Example 5.1 for pH  7, when the solution contains 10 mmol NaCl/L. Na balances the
alkalinity. Correct for activity-coefficients. Why is the correction problematic when pH  10?

5.2. Calculate pH and [CO2 

3 ] in a sample with TIC of 1.7, and mHCO 3 of 1.3 mmol/L. (Hint: consider as a first
estimate that TIC  mH2CO*3  mHCO3 , i.e. pH  8.5). Same question for the case that TIC  1.3 and
Alk  1.7 mmol/L

5.3. Calculate the activity and concentration of CO32 at pH  10.5 and pH  6.3, when TIC  2.5 mmol/L
at both pH’s.

5.4. Calculate pH and CO2 species in pure water at a PCO2 of 102 atm.

5.5. Calculate Ca2 concentration when the CO2 pressure is 0.01 atm and calcite dissolves till saturation.

5.6. What CO2 pressure is associated with a Ca2 concentration of 2 mmol/L, assuming equilibrium with cal-
cite, and absence of other reactions?

5.7. Soil-water in equilibrium with calcite has pH  7.6. Calculate the CO2 pressure? (Hint: find HCO
3 as
function of PCO2 with Equations (5.23) and (5.24); same for CO32 with Equation (5.24) and calcite
equilibrium).

5.8. Calculate the time to reach 95% saturation with calcite under open conditions, initial [PCO2]  103.5.
Calculate the distance covered in this time in a limestone fracture of width
 0.05 cm.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 233

5.9. Groundwater-samples from the Veluwe (Netherlands) are listed below (concentrations in mmol/L).
a) Calculate PCO2. b) Which water was in contact with calcite? Did dissolution occur in an open or in a
closed system? c) Do you see effects of contamination?

1. 2. 3. 4.

pH 4.55 7.54 7.86 6.24

Na 0.16 0.41 0.59 0.81
K 0.03 0.02 0.05 0.10
Mg2 0.03 0.20 0.39 0.26
Ca2 0.05 1.55 2.62 1.43
Cl 0.17 0.32 0.51 1.11
HCO3 0.07 3.57 2.75 0.76
SO2
4 0.22 0.01 1.86 0.55
NO3 0.17 .002 0.008 1.47
Fe .001 0.02 0.02 .001
Al 0.14 .001 0.008 .001

5.10. Pouhon is the local name for a spring with Fe2 and CO 2 rich water in Spa, Belgium. An analysis is given
below (concentration in mmol/L):

pH ? Cl .54
Na 1.0 Alk 6.38
K .15 SO 24 .43
Mg2 .54 Fe .34
Ca2 .77 CO2 66.0

a) Calculate pH from HCO 3 /CO2-equilibrium. b) What would be pH of Pouhon-water, if PCO2

decreases to atmospheric level. c) Check the analysis of the Pouhon-water. Alkalinity and CO2 were
titrated at the spring, Fe was analysed in the laboratory, in a non-acidified sample. Explain the error
in the charge balance. d) What would be the change in alkalinity, if all Fe precipitates? e) CO2 in
Pouhon-water might originate from Eifel-volcanism, and is called “juvenile”. What is “juvenile”?
5.11. Write reactions which take place when the following substances are added to a water sample. Does
Alkalinity increase, remain constant, or decrease?
1. CO2 4. MnCl2 7. Na2CO3
2. H2SO4 5. MnCl2  O2 8. NaHCO3
3. FeCl3 6. NaOH 9. Na2SO4
5.12. An analysis of water from a playa (evaporating lake) gave as results:
pH  10.0 SiO2  60 mg/L
What is the contribution of Si to titrated Alkalinity?
Data: Alkalinity  HCO3  2CO32  OH  base – acid.
H4SiO40 ↔ H3SiO 
4  H ; log K  9.1

5.13. Calculate the denudation rate of limestone in mm/yr when PCO2 is 102 atm, precipitation surplus is
300 mm/yr. Density of the limestone is 2.0 g/cm3. Hint: calculate the Ca2 concentration in water, multi-
ply with the water flux.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

234 Carbonates and carbon dioxide

5.14. Calculate the decalcification rate for the unsaturated sand in Figure 5.16. P  0.1 m/yr, temp  10°C,
 0.3, b  1.8 g/cm3.

5.15. Model the profile of Figure 5.16 with PHREEQC. Take decalcification as equilibrium reaction and dedolomi-
tization as kinetic reaction. The pristine sediment contains per L pore water (w  0.1) 0.18 mol calcite and
0.04 mol dolomite (200 times less than in the field). Use 10 cells of 0.1 m each. Discuss the calculated profiles
of dolomite, calcite and of Mg2 and Ca2 after 25, 50 and 75 yrs.

5.16. With PHREEQC, calculate SI for calcite and aragonite of a mixture (0–100% in 10 steps) of Caribbean
seawater with local fresh groundwater (temp. 25°C). Discuss what happens with the solid carbonates
when up to 50% seawater mixes with fresh water, when 50–60% seawater contacts aragonite, and when
the mixture contains !60% seawater. Use the composition (mg/L, except pH):

pH Na K Mg2 Ca2 Cl HCO

3 SO2
4

Seawater 8.18 10000 423 1180 410 17700 180 2610

Groundwater 7.10 14 3.3 7.1 80 34 253 1.8

5.17. From the PWP calcite dissolution rate (Equation 5.43) and the Ca2 / PCO2 relation (Equation 5.24) find
that the overall rate will be r  rfw (1  (cc) ⁄ ).
2
3

5.18. Program the dolomite dissolution rate (Equation 5.50) in PHREEQC and calculate the time to reach 95%
saturation with A/V  1/cm for [PCO2]  101.5 and 103.5.

5.19. Construct a PHREEQC rate for the kinetics of CO2 hydration. Reaction 1:

CO 2  H 2 O → H 2 CO3

with Rfw  k1 [CO2] and Rbw  k1 [H2CO3] .

H2CO3 / H2CO*3  H2CO3 / (CO2  H2CO3)  1 / (630  1).
k1  103 exp(34.69  9252/T )/s (Welch et al., 1969), T is temperature in Kelvin.
Reaction 2:

CO 2  OH → HCO
3

with Rfw  k2 [CO2][OH] and Rbw  k2 [HCO

3 ]. k2  10
6
exp(48.08 – 12720/T)/s (Welch et al.,
1969).
Hint: introduce a new SOLUTION_MASTER_SPECIES Cunh for unhydrated CO2. Let Cunh react to
give H2CO*3 according to:
d [Cunh] / dt  k1[Cunh]  k1[H2CO3] k2[Cunh] [OH]  k2[HCO
3]

a. Calculate the half-life for the reaction when mCunh, 0  1 mM at 10° and 25°C.
b. What is [PCO2] when 1 mM Cunh has hydrated to equilibrium at 25°C?
c. Construct a rate which adds Cunh to the solution and hydrates Cunh until [PCO2]  101.5.
d. Include the kinetics of CO2 hydration in the PWP rate model of calcite dissolution.

5.20. Calculate 14A0 of water in which calcite dissolves closed with respect to CO2 for initial [PCO2]  101.5,
102.5 and 103.5.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 235

5.21.
0
1.0
2.0

ε13CCO2(g) / TIC
3.0
4.0
5.0
6.0
7.0
8.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0
pH

Turner (1982) analyzed enrichment of 13C of CO2(g) with respect to TIC in water (13CCO2(g) / TIC) at 25°C
as a function of pH. Calculate the trend with PHREEQC, use Example 5.10, or in Example 5.11,
CO2(g)/HCO3  exp(24.1e–3 – 9.552/T).

5.22. Model the kinetic dissolution experiment with which we started Section 5.1. After how much time can we
expect   0.98 if we put 5 g calcite in 200 mL water (cf. the laboratory exercise)?

5.23. The cross section through an alluvial landscape shows groundwater flowlines. The CO2 pressure in the water
and the Ca2 concentration depend on the presence of calcite in the rooted soil. Indicate the groundwater
composition (Ca2 concentration and PCO2) in boreholes A and B. Hint: Start with the groundwater flow-
lines; draw the flowline for water that infiltrates at the boundary of the calcite/calcite-free zone. Then start
with borehole B, using concentrations from Table 5.6 for initial PCO2  101.5. Follow the water quality
from borehole B along the flowlines.

PCO2
no calcite
A
calcite

PCO2

Ca2 Ca2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

236 Carbonates and carbon dioxide

REFERENCES
Al, T.A., Martin, C.J. and Blowes, D.W., 2000. Carbonate–mineral/water interactions in sulfide-rich mine tail-
ings. Geochim. Cosmochim. Acta 64, 3933–3948.
Alberic, P. and Lepiller, M., 1998. Oxydation de la matière organique dans un système hydrologique karstique
alimenté par des pertes fluviales (Loiret, France), Water Res. 32, 2051–2064.
Amundson, R., Stern, L. Baisden, T. and Wang, Y., 1998. The isotopic composition of soil and soil-respired
CO2. Geoderma 82, 83–114.
Andersen, M.S., 2001. Geochemical processes at a seawater–freshwater interface. Ph.D. thesis, Lyngby, 145 pp.
Appelo, C.A.J., Beekman, H.E. and Oosterbaan, A.W.A., 1984. Hydrochemistry of springs from dolomite reefs
in the southern Alps of Northern Italy. IAHS Pub. 150, 125–138.
Aravena, R., Wassenaar, L.I. and Plummer, L.N., 1995. Estimating 14C groundwater ages in a methanogenic
aquifer. Water Resour. Res. 31, 2307–2317.
Atkinson, T.C., 1975. Diffuse flow and conduit flow in limestone terrain in the Mendip Hills, Somerset,
England. IAH Congress, Huntsville, Alabama.
Back, W., Hanshaw, B.B., Plummer, L.N., Rahn, P.H., Rightmire, C.T. and Rubin, M., 1983. Process and rate of
dedolomitisation: mass transfer and 14C dating in a regional carbonate aquifer. Geol. Soc. Am. Bull. 94,
1415–1429.
Back, W., Hanshaw, B.B., Herman, J.S., Van Driel, J.N., 1986. Differential dissolution of a Pleistocene reef in
the ground-water mixing zone of coastal Yucatan, Mexico. Geology 14, 137–140.
Baker, A. and Smart, P.L., 1995. Recent flowstone growth rates: field measurements in comparison to theo-
retical predictions. Chem. Geol. 122, 121–128.
Bar-Matthews, M., Ayalon, A., Matthews, A., Sass, E. and Halicz, L., 1996. Carbon and oxygen isotope study
of the active water-carbonate system in a karstic Mediterranean cave: implications for paleoclimate
research in semiarid regions. Geochim. Cosmochim. Acta 60, 337–347.
Bath, A.H., Edmunds, W.M. and Andrews, J.N., 1979. Palaeoclimatic trends deduced from the hydrochemistry
of a Triassic sandstone aquifer, United Kingdom. In Isotope Hydrology II, IAEA-SM-228, 545–568, IAEA,
Vienna.
Batoit, C., Emblanch, C., Blavoux, B., Simler, R. and Daniel, M., 2000. Organic matter in karstic aquifers: a
potential tracer in the carbon cycle. A small scale laboratory model approach. IAHS Pub. 262, 459–463.
Berg, J.M., Tymoczko, J.L. and Stryer, L., 2002. Biochemistry, 5th ed. Freeman, New York, 974 pp.
Berner, R.A., 1975. The role of magnesium in the crystal growth of calcite and aragonite from sea water.
Geochim. Cosmochim. Acta 39, 489–504.
Berner, R.A., Westrich, J.T., Graber, R., Smith, J. and Martens, C.S., 1978. Inhibition of aragonite precipitation
from supersaturated seawater: a laboratory and field study. Am. J. Sci. 278, 816–837.
Bögli, A., 1978. Karsthydrographie und physische Speläologie. Springer, Berlin, 292 pp.
Bolin, B., Degens, E.T., Duvigneaud, P. and Kempe, S., 1979. The global biogeochemical carbon cycle. In
B. Bolin et al. (eds), The global carbon cycle, SCOPE 13, 1–56, Wiley and Sons, New York.
Brook, G.A., Folkoff, M.E. and Box, E.O., 1983. A world model of soil carbon dioxide. Earth Surf. Proc. 8,
79–88.
Brown, J.G. and Glynn, P.D., 2003. Kinetic dissolution of carbonates and Mn oxides in acidic water: measure-
ment of in situ field rates and reactive transport modeling. Appl. Geochem. 18, 1225–1239.
Budd, D.A., 1997. Cenozoic dolomites of carbonate islands: their attributes and origin. Earth Sci. Rev. 42, 1–47.
Busenberg, E. and Plummer, L.N., 1982. The kinetics of dissolution of dolomite in CO2-H2O systems at 1.5 to
65C and 0 to 1 atm PCO2, Am. J. Sci. 282, 45–78.
Busenberg, E. and Plummer, L.N., 1986. A comparative study of the dissolution and crystal growth kinetics of
calcite and aragonite. U.S. Geol. Surv. Bull. 1578, 139–168.
Busenberg, E. and Plummer, L.N., 1989. Thermodynamics of magnesium calcite solid-solutions at 25 and
1 atm total pressure. Geochim. Cosmochim. Acta 53, 1189–1208.
Capaccioni, B., Didero, M., Paletta, C. and Salvadori, P., 2001. Hydrogeochemistry of groundwaters from car-
bonate formations with basal gypsiferous layers: an example from Mt Catria-Mt Nerone ridge (Northern
Appenines, Italy). J. Hydrol. 253, 14–26.
Cardenal, J., Benavente, J. and Cruz-Sanjulián, J.J., 1994. Chemical evolution of groundwater in Triasic
gypsum-bearing carbonate aquifers (Las Alpujarras, southern Spain). J. Hydrol. 161, 3–30.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 237

Cerling, T.E., 1984. The stable isotopic composition of modern soil carbonate and its relationship to climate.
Earth Planet. Sci. Lett. 71, 229–240.
Chapelle, F.H. and Knobel, L.L., 1986. Stable carbon isotopes of HCO 3 in the Aquia aquifer, Maryland: evi-
dence for an isotopically heavy source of CO2. Ground Water 23, 592–599.
Chiodini, G., Cioni, R., Guidi, M. Raco, B. and Marini, L., 1998. Soil CO2 flux measurements in volcanic and
geothermal areas. Appl. Geochem. 13, 543–552.
Chou, L., Garrels, R.M. and Wollast, R., 1989. Comparative study of the kinetics and mechanisms of dissolu-
tion of carbonate minerals. Chem. Geol. 78, 269–282.
Clark, I.D. and Fritz, P., 1997. Environmental isotopes in hydrogeology. CRC Press, Boca Raton, 328 pp.
Davis, K.J., Dove, P.M. and De Yoreo, J.J., 2000. The role of Mg2+ as an impurity in calcite growth. Science 290,
1134–1137.
Deelman, J.C., 2003. Low-temperature formation of dolomite and magnesite.http://users.skynet.be/infolib/
dolomite/bookprospectus.htm
Deer, W.A., Howie, R.A. and Zussman, J., 1966. An introduction to the rock forming minerals. Longman,
London, 528 pp.
Dörr, H. and Münnich, K.O., 1980. Carbon-14 and carbon-13 in soil CO2. Radiocarbon 22, 909–918.
Dowty, E., 1999. ATOMS for Windows and Macintosh, Vs 5. www.shapesoftware.com.
Drake, J.J., 1983. The effects of geomorphology and seasonality on the chemistry of carbonate groundwater.
J. Hydrol. 61, 223–236.
Dreybrodt, W., Buhmann, D., Michaelis, J. and Usdowski, E., 1992. Geochemically controlled calcite precipi-
tation by CO2 outgassing: field measurements of precipiation rates to theroretical predictions. Chem. Geol.
97, 287–296.
Dreybrodt, W., 1981. Kinetics of the dissolution of calcite and its applications to karstification. Chem. Geol. 31,
245–269.
Edmunds, W.M., Bath, A.H. and Miles, D.L., 1982. Hydrochemical evolution of the East Midlands Triassic
sandstone aquifer, England. Geochim. Cosmochim. Acta 46, 2069–2081.
Etiope, G., 1999. Subsoil CO2 and CH4 and their advective transfer from faulted grassland to the atmosphere.
J. Geophys. Res. 104, 16,889–16,894.
Fabricius, I.L., 2003. How burial diagenesis of chalk sediments controls sonic velocity and porosity. AAPG
Bull. 87, 1755–1778.
Farquhar, G.D., Ehleringer, J.R. and Hubick, K.T., 1989. Carbon isotope discrimination and photosynthesis.
Ann. Rev. Plant Physiol. Plant Biol. 40, 503–537.
Ford, D.C. and Williams, P.W., 1989. Karst geomorphology and hydrology. Unwin Hyman, Winchester, MA, 601 pp.
Gabrovs
ek, F. and Dreybrodt, W., 2000. Role of mixing corrosion in calcite-agressive H2O-CO2-CaCO3 solu-
tions in the early evolution of karst aquifers in limestone. Water Resour. Res. 36, 1179–1188.
Garnier, J.-M., 1985. Retardation of dissolved radiocarbon through a carbonated matrix. Geochim. Cosmochim.
Acta 49, 683–693.
Garnier, J.-M., Crampon, N., Préaux, C., Porel, G. and Vreulx, M., 1985. Traçage par 13C, 2H, I et uranine
dans la nappe de la craie Sénonienne en écoulement radial convergent (Béthune, France). J. Hydrol. 78,
379–392.
Garrels, R.M. and Mackenzie, F.T., 1971. Evolution of sedimentary rocks. Norton, New York, 397 pp.
Given, R. and Wilkinson, B.H., 1985. Kinetic control of morphology, composition and mineralogy of abiotic
sedimentary carbonates. J. Sed. Petrol. 55, 109–119.
Gonfiantini, R. and Zuppi, G.M., 2003. Carbon isotope exchange rate of DIC in karst groundwater. Chem.
Geol. 197, 319–336.
Gunn, J., 1981. Hydrological processes in karst depressions. Z. Geomorph. N.F. 25, 313–331.
Gutjahr A., Dabringhaus H. and Lacmann R., 1996. Studies of the growth and dissolution kinetics of the CaCO3
polymorphs calcite and aragonite. II. The influence of divalent cation additives on the growth and dissolu-
tion rates. J. Cryst. Growth 158, 310–315.
Hamada, Y. and Tanaka, T., 2001. Dynamics of carbon dioxide in soil profiles based on long-term field obser-
vation. Hydrol. Proc. 15, 1829–1845.
Hanshaw, B.B. and Back, W., 1979. Major geochemical processes in the evolution of carbonate aquifer systems.
J. Hydrol. 43, 287–312.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

238 Carbonates and carbon dioxide

Hanson, P.J., Edwards, N.T., Garten, C.T. and Andrews, J.A., 2000. Separating root and soil microbial contri-
butions to soil respiration: A review of methods and observations. Biogeochem. 48, 115–146.
Hardie, L.A., 1987. Dolomitization: a critical view of some current views. J. Sed. Petrol. 57, 166–183.
Herman, J.S. and Lorah, M. M., 1988. Calcite precipitation rates in the field: Measurement and prediction for
a travertine—depositing stream. Geochim. Cosmochim. Acta 52, 2347–2355.
Hoogendoorn, J.H., 1983. Hydrochemistry of Eastern Netherlands (in Dutch). DGV-TNO, Delft.
Inskeep, W.P. and Bloom, P.R., 1986. Kinetics of calcite precipitation in the presence of water-soluble organic
ligands. Soil Sci. Soc. Am. J. 50, 1167–1172.
Jakobsen R. and Postma D., 1999. Redox zoning, rates of sulfate reduction and interactions with Fe-reduction
and methanogenesis in a shallow sandy aquifer, Rømø, Denmark. Geochim. Cosmochim. Acta 63, 137–151.
James, A.N., 1981. Solution parameters of carbonate rocks. Bull. Int. Ass. Eng. Geol. 24, 19–25.
James, N.P. and Choquette, P.W., 1984. Limestones – the meteoric diagenetic environment. Geosci. Can. 11,
161–194.
Jurjovec, J., Ptacek, C.J. and Blowes, D.W., 2002. Acid neutralization mechanisms and metal release in mine
tailings: A laboratory column experiment. Geochim. Cosmochim. Acta 66, 1511–1526.
Kalin, R.M., 2000. Radiocarbon dating of groundwater systems. In P.G. Cook and A.L. Herczeg (eds),
Environmental tracers in subsurface hydrology, Chapter 4. Kluwer, Boston.
Keller, C.K. and Bacon, D.H. 1998. Soil respiration and georespiration distinguished by transport analyses of
vadose CO2, 13CO2, and 14CO2. Global Biogeochem. Cycles 12, 361–372.
Kessler, T.B. and Harvey, C.F., 2001. The global flux of carbon dioxide into groundwater. Geophys. Res. Lett.
28, 279–282.
Kern, D.M., 1960. The hydration of carbon dioxide. J. Chem. Educ. 14–23.
Kloppmann, W., Dever, L. and Edmunds, W.M., 1998. Residence time of Chalk groundwaters in the Paris Basin
and the North German Basin: a geochemical approach. Appl. Geochem. 13, 593–606.
Labotka, T.C., Cole, D.R. and Riciputi, L.R., 2000. Diffusion of C and O in calcite at 100 MPa. Am. Mineral.
85, 488–494.
Langmuir, 1971. The geochemistry of some carbonate ground waters in central Pennsylvania, Geochim.
Cosmochim. Acta 35, 1023–1045.
Laursen, S., 1991. On gaseous diffusion of CO2 in the unsaturated zone. J. Hydrol. 122, 61–69.
Lebron, I. and Suarez, D.L., 1996. Calcite nucleation and precipitation kinetics as affected by dissolved organic
matter at 25C and pH > 7.5. Geochim. Cosmochim. Acta 60, 2765–2776.
Lee, R.W., 1997. Effects of carbon dioxide variations in the unsaturated zone on water chemistry in a glacial-
outwash aquifer. Appl. Geochem. 12, 347–366.
López-Chicano, M., Bouamama, M., Vallejos, A. and Pulido-Bosch, A., 2001. Factors which determine the
hydrogeochemical behaviour of karstic springs. A case study from the Betic Cordilleras, Spain. Appl.
Geochem. 16, 1179–1192.
Martin, J.B. and Dean, R.W., 2001. Exchange of water between conduits and matrix in the Floridan aquifer.
Chem. Geol. 179, 145–165.
Maoszewski, P. and Zuber, A., 1991. Influence of matrix diffusion and exchange reactions on radiocarbon ages
in fissured carbonate aquifers. Water Resour. Res. 27, 1937–1945.
Morrow, D.W., 1982. The chemistry of dolomitization and dolomite precipitation. Geosci. Can. 9, 5–13.
Morse, J.W., 1983. The kinetics of calcium carbonate dissolution and precipitation. In R.J. Reeder (ed.), Rev.
Mineral. 11, 227–264.
Morse, J.W. and Arvidson, R.S., 2002. The dissolution kinetics of major sedimentary carbonate minerals. Earth
Sci. Rev. 58, 51–84.
Morse, J.W. and Mackenzie, F.T., 1990. Geochemistry of sedimentary carbonates. Elsevier, Amsterdam, 707 pp.
Mozeto, A.A., Fritz, P. and Reardon, E.J., 1984. Experimental observations on carbon isotope exchange in car-
bonate–water systems. Geochim. Cosmochim. Acta 48, 495–504.
Mucci, A., 1986. Growth kinetics and composition of magnesian calcite overgrowths precipitated from sea-
water: quantitative influence of orthophosphate ions. Geochim. Cosmochim. Acta 50, 217–233.
Mucci, A., 1987. Influence of temperature on the composition of magnesian calcite overgrowths precipitated
from seawater. Geochim. Cosmochim. Acta 51, 1977–1984.
Neretnieks, I., 1981. Age dating of groundwater in fissured rock: influence of water volume in micropores.
Water Resour. Res. 17, 421–422.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 239

Nordstrom, D.K., Plummer, L.N., Langmuir, D., Busenberg, E., May, H.M., Jones, B.F and Parkhurst, D.L.,
1990. Revised chemical equilibrium data for major water–mineral reactions and their limitations.
In D.C Melchior and R.L Bassett (eds), Chemical modeling of aqueous systems II, ACS Symp Ser 416,
398–413.
Parkhurst, D.L., 1997. Geochemical mole-balancing with uncertain data. Water Resour. Res. 33, 1957–1970.
Pitman, J.I., 1978. Carbonate chemistry of groundwater from chalk, Givendale, East Yorkshire. Geochim.
Cosmochim. Acta 42, 1885–1897.
Plummer, L. N. and Busenberg, E., 1987. Thermodynamics of aragonite strontianite solid-solutions: Results
from stoichiometric solubility at 25° and 76°C. Geochim. Cosmochim. Acta 51, 1393–1411.
Plummer, L.N. and MacKenzie, F.T., 1974. Predicting mineral solubility from rate data: application to the dis-
solution of magnesian calcites. Am. J. Sci. 61–83.
Plummer, L.N., Vacher, H.L., Mackenzie, F.T., Bricker, O.P. and Land, L.S., 1976. Hydrogeochemistry of
Bermuda: a case history of groundwater diagenesis of biocalcarenites. Geol. Soc. Am. Bull. 87, 1301–1316.
Plummer, L.N., Wigley, T.M.L. and Parkhurst, D.L., 1978. The kinetics of calcite dissolution in CO2–water
systems at 5° to 60°C and 0.0 to 1.0 atm CO2. Am. J. Sci. 278, 179–216.
Plummer, L.N., Parkhurst, D.L. and Wigley, T.M.L., 1979. Critical review of the kinetics of calcite dissolution
and precipitation. In E.A. Jenne (ed.), Chemical modeling in aqueous systems, ACS Symp. Ser. 93,
537–573.
Plummer, L.N., Busby, J.F., Lee, R.W. and Hanshaw, B.B., 1990. Geochemical modeling in the Madison aquifer
in parts of Montana, Wyoming and South Dakota. Water Resour. Res. 26, 1981–2014.
Plummer, L.N., Prestemon, E.C. and Parkhurst, D.L., 1994. An interactive code (NETPATH) for modeling
net geochemical reactions along a flow path, 2.0. U.S. Geol. Surv. Water Resour. Inv. Rep. 94-4169,
130 pp.
Plummer, L.N., Busenberg, E. and Riggs, A.C., 2000. In-situ growth of calcite at Devils Hole, Nevada: com-
parison of field and laboratory rates to a 500,000 year record of near-equilibrium calcite growth.
Aq. Geochem. 6, 257–274.
Pokrovsky, O.S. and Schott, J., 2001. Kinetics and mechanism of dolomite dissolution in neutral to alkaline
solutions revisited. Am. J. Sci. 301, 597–626.
Raich, J.W. and Potter, C.S., 1995. Global patterns of carbon dioxide emissions from soils. Global Biogeochem.
Cycles 9, 23–36.
Reardon, E.J., Allison, G.B. and Fritz, P., 1979. Seasonal chemical and isotopic variations of soil CO2 at Trout
Creek, Ontario. J. Hydrol. 43, 355–371.
Reardon, E.J., Mozeto, A.A. and Fritz, P., 1980. Recharge in northern clime calcareous sandy soils: soil water
chemical and carbon-14 evolution. Geochim. Cosmochim. Acta 44, 1723–1735.
Reddy, M.M., 1977. Crystallization of calcium carbonate in the presence of trace concentrations of phospho-
rous containing anions. J. Cryst. Growth 41, 287–295.
Reddy, M.M., Plummer, L.N. and Busenberg, E., 1981. Crystal growth of calcite from calcium bicarbonate
solutions at constant PCO2 and 25C: a test of a calcite dissolution model. Geochim. Cosmochim. Acta 45,
1281–1289.
Reeder, R.J., 1983. Crystal chemistry of the rhombohedral carbonates. Rev. Mineral. 11, 1–47.
Rickard, D. and Sjöberg, E.L., 1983. Mixed kinetic control of calcite dissolution rates. Am. J. Sci. 283,
p. 815–830.
Russell, E.W., 1973. Soil conditions and plant growth. Longman, London, 849 pp.
Sacks, L.A., Herman, J.S. and Kauffman, S.J., 1995. Controls on high sulfate concentrations in the Upper
Floridan aquifer in southwest Florida. Water Resour. Res. 31, 2541–2551.
Sanford, W.E., 1997. Correcting for diffusion in Carbon-14 dating of ground water. Ground Water 35, 357–361.
Sanford, W. and Konikow, L.F., 1989. Simulation of calcite dissolution and porosity changes in saltwater mix-
ing zones in coastal aquifers. Water Resour. Res. 25, 655–667.
Saunders, J.A. and Toran, L.E., 1994. Evidence for dedolomitization and mixing in Paleozoic carbonates near
Oak Ridge, Tennessee. Ground Water 32, 207–214.
Sherman, L.A. and Barak, P., 2000. Solubility and dissolution kinetics of dolomite in Ca-Mg-HCO3/CO3 solu-
tions at 25 degrees C and 0.1 MPa carbon dioxide. Soil Sci. Soc. Am. J. 64, 1959–1968.
Schulz, H.D., 1988. Labormessung der Sättigungslänge als Mass für die Lösungskinetik von Karbonaten im
Grundwasser. Geochim. Cosmochim. Acta 52, 2651–2657.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

240 Carbonates and carbon dioxide

Shuster, E.T. and White, W.B., 1972. Source areas and climatic effects in carbonate groundwaters determined
by saturation indices and carbon dioxide pressures. Water Resour. Res. 8, 1067–1073.
Sjöberg, E.L., 1978. Kinetics and mechanism of calcite dissolution in aqueous solutions at low temperatures.
Stockholm Contrib. Geol. XXXII, 1, 1–92.
Smedley, P.L. and Edmunds, W.M., 2002. Redox patterns and trace-element behavior in the East Midlands
Triassic sandstone aquifer, U.K. Ground Water 40, 44–58.
Stoessell, R.K., Moore, Y.H. and Coke, J.G., 1993. The occurrence and effect of sulfate reduction and sulfide
oxidation on coastal limestone dissolution in Yucatan cenotes. Ground Water 31, 566–575.
Stumm, W. and Morgan, J.J., 1996. Aquatic chemistry, 3rd ed. Wiley and Sons, New York, 1022 pp.
Suchet, P.A. and Probst, J.L. 1995. A global model for present-day atmospheric/soil CO2 consumption by chem-
ical erosion of continental rocks. Tellus 47B, 273–280.
Sudicky, E. and Frind, E.O., 1981. Carbon 14 dating of groundwater in confined aquifers: implications of
aquitard diffusion. Water Resour. Res. 17, 1060–1064.
Terjesen, S.G., Erga, O., Thorsen, G. and Ve, A., 1961. Phase boundary processes as rate determining steps in
reactions between solids and liquids. Chem. Eng. Sci. 74, 277–288.
Thorstenson, D.C. and Plummer, L.N., 1977. Equilibrium criteria for two-component solids reacting with fixed
composition in an aqueous phase – example: the magnesian calcites. Am. J. Sci. 277, 1203–1223.
Turner, J.V., 1982. Kinetic fractionation of carbon-13 during calcium carbonate precipitation. Geochim.
Cosmochim. Acta 46, 1183–1191.
Van der Kemp, W., Appelo, C.A.J. and Walraevens, K., 2000. Inverse chemical modeling and radiocarbon dat-
ing of palaeogroundwaters: the Tertiary Ledo-Paniselian aquifer in Flanders, Belgium. Water Resour. Res.
36, 1277–1287.
Van Lith, Y., Vasconcelos, C., Warthmann, R., Martins, J.C.F. and McKenzie, J.A., 2002. Bacterial sulfate
reduction and salinity: two controls on dolomite precipitation in Lagoa Vermelha and Brejo do Espinho
(Brazil). Hydrobiol. 485, 35–49.
Vogel, J.C. 1993. Variability of carbon isotope fractionation during photosynthesis. In J.R. Ehleringer et al.
(eds), Stable isotopes and plant carbon-water relations, 29–46, Academic Press, London.
Walter, L.M. and Hanor, J.S., 1979. Effect of orthophosphate on the dissolution kinetics of biogenic magnesian
calcites. Geochim. Cosmochim. Acta 43, 1377–1385.
Welch, M.J., Lifton, J.F. and Seck, J.A., 1969. Tracer studies with radioactive Oxygen-15 exchange between car-
bon dioxide and water. J. Phys. Chem. 73, 3351–3356.
Weyl, P.K., 1958. Solution kinetics of calcite. J. Geol. 66, 163–176.
Whitaker, F. and Smart, P.L. 1997. Groundwater circulation and geochemistry of a karstified bank-marginal
fracture system, South Andros Island, Bahamas. J. Hydrol. 197, 293–315.
Wigley, T.M.L. and Plummer, L.N., 1976. Mixing of carbonate waters. Geochim. Cosmochim. Acta 40,
989–995.
Yoshimura, K., Nakao, S., Noto, M., Inokura, Y., Urata, K., Chen, M. and Lin, P.-W., 2001. Geochemical and
stable isotope studies on natural water in the Taroko Gorge karst area, Taiwan-chemical weathering of car-
bonate rocks by deep source CO2 and sulfuric acid. Chem. Geol. 177, 415–430.
Zaihua, L., Svensson, U., Dreybrodt, W., Daoxian, Y. and Buhmann, D., 1995. Hydrodynamic control of inor-
ganic calcite precipitation in Huanglong Ravine, China: field measurements and theoretical prediction of
deposition rates. Geochim. Cosmochim. Acta 59, 3087–3097.
Zötl, J.G., 1974. Karsthydrogeologie. Springer, Berlin, 291 pp.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

6

Ion Exchange

Soils and aquifers contain materials like clay minerals, organic matter and metal oxy-hydroxides
which can sorb chemicals. The various processes indicated by the general term sorption are illus-
trated in Figure 6.1. The term adsorption refers to the adherence of a chemical to the surface of the
solid, absorption suggests that the chemical is taken up into the solid, and exchange involves replace-
ment of one chemical for another one at the solid surface.
It is sometimes difficult to separate sorption and other types of reactions involving solids such as
precipitation and dissolution (Sposito, 1984). A major difference is that sorption depends on the
presence of a preexisting solid surface, in contrast to, for example, precipitation. Sorption and ion
exchange have become important topics for hydrologists since these processes regulate the transport
of pollutant chemicals in aquifers and soils. In this chapter we focus on ion exchange while the fol-
lowing chapter will treat sorption processes for heavy metals.
Under steady-state chemical conditions, the composition of a cation exchanger will be in equilib-
rium with the resident groundwater. When the water composition changes as a result of pollution or
acidification, or due to a moving salt/fresh water interface, the cation exchanger readjusts its composi-
tion to the new groundwater concentrations. In this way the exchanger acts as a temporary buffer which
may completely alter the concentrations in water through a process known as ion-chromatography.

A Adsorption: A

A Absorption: A

B B A
B A
A lon exchange:

A B

Figure 6.1. Pictorial definition of different sorption processes.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

242 Ion exchange

A hydrologist who tries to decipher hydrogeological conditions may use the hydrochemical patterns
which are the result of cation exchange. As aquifers become more exploited and polluted, the water
quality along flowlines will change, but due to cation exchange, the water composition may still
reflect aspects of former water quality. This chapter demonstrates the principles of ion exchange in
aquifers and soils. We start with examples of cation exchange as indicators of salinization or desalini-
zation of aquifers. Then we treat exchanger materials, exchange equations, and give examples of salt
water intrusion and chromatographic patterns in aquifers. We also show how exchange and chro-
matography can be modeled with PHREEQC.

6.1. CATION EXCHANGE AT THE SALT/FRESH WATER INTERFACE

The composition of fresh groundwater in coastal areas is often dominated by Ca2 and HCO 3 ions which
result from calcite dissolution. The cation exchanger is then also dominated by adsorbed Ca2. In sea-
water, Na and Cl are the dominant ions, and sediment in contact with seawater will have mostly Na
on the exchanger. When seawater intrudes into a coastal fresh water aquifer, an exchange of cations takes
place:

Na  1⁄2Ca-X2 → Na-X  1⁄2Ca2 (6.1)

where X indicates the soil exchanger. In the reaction Na is taken up by the exchanger, while Ca2
is released. Since the dominant anion Cl remains the same, the water quality changes from a NaCl
to a CaCl2-type of water. The reverse process takes place during freshening, i.e. when fresh water,
with Ca2 and HCO 3 ions, flushes a salt water aquifer:

1
⁄2Ca2  Na-X → ⁄2Ca-X2  Na
1
(6.2)

North Sea
Dunes (Infiltration area)

15
Polders

0
P-CaMix

15 P-CaHCO3

30 P-N
aH P-NaHCO3
CO
3
Elevation (m)

45
B-NaCl*
B-NaCl*
60

B-NaCl
75
B-NaCl
90

S-NaCl
105

120

Figure 6.2. Section through the coastal dunes near Castricum (Netherlands), showing refreshening of a brack-
ish water aquifer. The prefixes P, B and S before the water types indicate fresh, brackish and salt water. Clay lay-
ers are indicated by hatched lines (Stuyfzand, 1985).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Cation exchange at the salt/fresh water interface 243

The sediment now adsorbs Ca2 while Na is released and a NaHCO3-type water results. In this way
the water composition can indicate whether upconing of seawater occurs, or conversely whether fresh
water is flushing salt water from the aquifer.
An example is given in Figure 6.2, showing a section through the Dutch coastal dunes (Stuyfzand,
1985). Groundwater has been pumped from the area for drinking water purposes for a long time. An
alarming rate of upconing of salt water was observed several decades ago, which threatened to salinize
all the pumping wells. To relieve the problem, fresh water from the river Rhine was allowed to infil-
trate into the dunes and then recovered from shallow wells. Deliberately, more water was infiltrated
than abstracted in order to flush the fresh-salt water boundary backwards. This strategy has been suc-
cessful, as indicated by the ‘freshened’ NaHCO3-type water now present in the middle part of the
aquifer (Figure 6.2). The quality pattern shown in Figure 6.2 was constructed based on a detailed
network of sampling wells and allows for an interpretation of flowlines in the aquifer. When Na has
been flushed from the exchanger, the Ca(HCO3)2-type fresh water returns. This water type will
appear sooner when flushing is more rapid, i.e. along the more rapid flowlines, or where the cation
exchange capacity is smaller.
An example of salt water intrusion in the Nile delta, Egypt, is shown in Figure 6.3. The Stiff
diagrams (see also Chapter 1) summarize the groundwater composition and indicate the intrusion of
salt water in boreholes 121 and 129, as demonstrated by the relative excess of Ca2 compared to sea-
water (shown in the top of Figure 6.3). Fresh groundwater in the Nile delta originates almost exclu-
sively from Nile water, which has a Ca(HCO3)2-type composition (shown in Figure 6.3 west of Cairo).

Mediterranean
Sea

sea, 40

Lake Manzala

121, 20

129, 20 231, 10

132
22
111
109
Ismailya canal

50 K2 P3, 2
203
34

meq/L
12 8 4 0 4 8 12
Cl

K  Na
Nile Ca2 HCO3


Cairo Mg2 SO42

Figure 6.3. Stiff diagrams of groundwater in the Nile delta, Egypt, indicating salt water upconing in borehole
121 and 129, and Ca2 / Na-exchange near the Ismailya canal.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

244 Ion exchange

(Ca, Mg) SO4

(Ca, Mg) Cl2

80

80
60

60
Cl

Ca


In t


4
SO

r us

Mg
40

40
io n
20

20
Sea
Fresh Mixing Bm
(Ca, Mg) (HCO3)2 (Na, K)2 SO4
(Na, K) Cl
Mg Bs SO4
Fre
20

20
s

80 80
he
nin
g
40

40
60 60
Na

SO 4
3
O


HC
Mg

40 40
80

80

20 Sea 20
Fresh Bm Bs
Bs Fresh Bm Sea

Ca Na, K HCO3 Cl
80

60

40

20

20

40

60

80

Ca Cl

Figure 6.4. Piper plot showing average compositions of fresh water and seawater, and NaHCO3-sample (Bs)
from Table 6.2. The tail at Bs points to the calculated composition of a conservative mixture Bm.

Groundwater along the Ismailya canal, which runs from Caïro to the Red Sea, shows increasing Cl

concentrations as a result of evapotranspiration of Nile water used for irrigation. The Na concen-
tration increases even more than Cl
, while Ca2 relatively decreases, thus suggesting cation-
exchange of Ca2 for Na in the soil. The sodium originates from Na-containing loess that is blown
in from the desert and settles in the wetter area of the delta, as was observed in the Negev desert
(Nativ et al., 1983).
The effects of cation exchange are particularly evident when analyses are plotted in a Piper dia-
gram (see also Chapter 1), as shown in Figure 6.4. The average compositions of fresh water and sea-
water are shown and a straight line between the two indicates water compositions due to conservative
mixing. If groundwater samples in a coastal area show a surplus of Ca2, compared to the conser-
vative mixture, then this can imply seawater intrusion, while a surplus of Na may indicate freshen-
ing of the aquifer. Typical groundwaters showing these trends are given in Table 6.1.
The chemical reactions occurring during fresh/salt water displacement can be deduced more
precisely by calculating the expected composition based on conservative mixing of salt water and
fresh water, and then comparing the result with the measured concentrations in the water sample.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Cation exchange at the salt/fresh water interface 245

Table 6.1. Groundwaters from the Netherlands showing “only-mixing” water composition, and a composition
influenced by cation exchange (‘CaCl2’ and ‘NaHCO3’-types). Concentrations in mmol/L.

Only mixing CaCl2 water NaHCO3 water

pH 7.5 7.2 6.91 6.6 8.7 8.3

Na 24.53 54.33 341 124 40 2.0
K 0.82 1.41 2.8 2.4
Mg2 2.9 7.1 27.9 30.7 2.8 1.1
Ca2 3.0 7.9 39.6 47.2 0.7 0.65
Cl 27.2 70.8 440 271 25.6 1.4
HCO 3 9.2 15.3 7.0 3.8 14.4 4.0
SO2
4 0.07 0 18.8 4.7 2.7 0.2

The concentration of an ion i, by conservative mixing of seawater and fresh water is:

mi, mix  f sea mi, sea  (1  f sea ) mi, fresh (6.3)

where mi is the concentration of i (mmol/L), fsea the fraction of seawater in the mixed water, and sub-
scripts mix, sea, and fresh indicate a conservative mixture, seawater and fresh water, respectively. Any
change in the concentration mi, react due to reactions then becomes simply:
mi, react  mi, sample  mi, mix (6.4)

where mi,sample is the measured concentration in the sample. The fraction of seawater, fsea, is normally
calculated using the Cl concentration of the sample. Chloride is assumed to be a conservative
parameter, just as it was used in Chapter 2 to find the sea salt contribution in rainwater. The Cl
based fraction of seawater is:
mCl , sample  mCl , fresh
f sea  (6.5)
mCl , sea  mCl , fresh

If seawater is the only source of Cl, then mCl,fresh  0, which simplifies Equation (6.5) to:

f sea  mCl , sample / 566 (6.6)

where the Cl concentration is expressed in mmol/L, and 566 mmol/L is the Cl concentration of
35‰ (grams of salt per kilogram) seawater. While the salinity of seawater may vary, the ratios
among the ions remain constant for all major ions and can be used to calculate the composition of
conservative mixtures.
Table 6.2 illustrates the results of such a calculation for some analyses in Table 6.1. The relative
increases of Na in NaHCO3-water, and of Ca2 in CaCl2-water are evident. Another conspicuous
change is the increase of HCO 
3 in the NaHCO3-type water. This relative HCO3 increase is com-
monly found, and may easily amount to 10 mmol/L or more. When Ca exchanges for Na
2

(Reaction 6.2) the water becomes undersaturated for calcite and dissolution results (Back, 1966;
Chapelle, 1983). Also, the pH of NaHCO3 water often becomes high, above 8, due to calcite disso-
lution. CaCl2 type of water on the other hand, often has a low pH, less than 7. This may be the result
of calcite precipitation, driven by the increase of Ca2 caused by cation exchange (Reaction 6.1).
However, other reactions may also take place. In Table 6.2 the CaCl2 water shows a deficit in sulfate.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

246 Ion exchange

Table 6.2. Recalculated water analyses of Table 6.1, showing actual reaction amounts. Values in mmol/L.

Seawater CaCl2 water NaHCO3 water Fresh water

Sample Mix React Sample Mix React

Na 485.0 341 377 36 40 22 18 0

K 10.6 2.8 8.2 5.4 0
Mg2 55.1 27.9 42.8 14.9 2.8 2.5 0.3 0
Ca2 10.7 39.6 9.0 30.6 0.7 3.3 2.6 3.0
Cl 566 440 440 – 25.6 25.6 – 0
HCO3 2.4 7.0 3.2 3.8 14.4 5.8 8.6 6.0
SO2
4 29.3 18.8 22.8 4.0 2.7 1.3 1.4 0
% seawater 100 78 5 0
Symbol in Figure 6.4: Bs Bm

Sample: given water composition from Table 6.1. Mix: calculated water composition based on conservative
mixing. React: sample – mix, showing effects of cation exchange and other reactions.

Seawater has a high content of sulfate and when it intrudes an anoxic coastal aquifer this may result
in sulfate reduction (Andersen, 2001):

SO 24  2CH 2O → H 2S  2HCO

3 (6.7)
80
80

Ca
CI 

2
60
60


4 

Mg
SO 2

2
40

40
Ex

n
depositio3
ch

CaCO
an

red SO4
g

uct
e
20

20

ion

Mixing
Ex CaCO
ch
solution

an
ge
3
Na


O


3
K

HC


Seawater Salt water intrusion

Fresh water Fresh water intrusion

Figure 6.5. Piper plot showing the composition of groundwater from Zeeland and Western Brabant in
the Netherlands, displaying cation-exchange. Each symbol has a tail that points towards the position on the
mixing line based on Cl concentration of the groundwater sample.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Adsorbents in soils and aquifers 247

Such additional reactions can be visualized in a Piper-diagram by adding a tail going from the
sample point towards the corresponding position on the conservative mixing line (download the
program from www.xs4all.nl/appt).
The NaHCO3-water of Table 6.2 has been entered in Figure 6.4, as point Bs (the actual sample) and
point Bm (the calculated mixture). When only cation exchange takes place, the tail is parallel to the cation
axis of the Piper diamond; when other reactions occur this is not the case (Figure 6.5). The dissolution
of CaCO3 brings the mixture towards the Ca2  HCO 2
3 corner, the reduction of SO4 gives a shift par-
 
allel to the anion-axis, deposition of CaCO3 leads towards the Na and (SO4  Cl ) corner. Figure 6.5
2

shows groundwater analyses from Zeeland and Western Brabant, the Netherlands, where seawater inun-
dations have been common in the past centuries. During freshening of the aquifer (Figure 6.5), the rela-
tive increase of HCO 3 indicates calcite dissolution, while during salt water intrusion not much CaCO3
precipitation appears to take place. It is possible that CaCO3-deposition is masked in the Piper diagram
 
by SO2 4 reduction, since Reaction (6.7) lowers the percentage of (SO4  Cl ) and produces HCO3 .
2

6.2 ADSORBENTS IN SOILS AND AQUIFERS

While in principle all solid surfaces in soils and aquifers can act as adsorbers, solid phases with a
large specific surface area will adsorb most, and the adsorption capacity therefore depends on the
grain size. Solids with a large specific surface area reside in the clay fraction (2 m), but coarser
grains in a sediment are often coated with organic matter and iron oxyhydroxides. The adsorption
capacity is therefore linked to the clay content (fraction 2 m), clay minerals, organic matter (%
C), and oxide or hydroxide content. The current convention is to express the CEC, or Cation
Exchange Capacity of a soil in meq/kg, replacing the unit meq/100 g. In the US soil science litera-
ture, mmolc is often used to indicate “mmol charge” for meq.
An empirical formula which relates the CEC to the percentages of clay (2 m) and organic car-
bon at near neutral pH is (e.g. Breeuwsma et al., 1986):
CEC (meq/kg)  7 (% clay)  35 (% C) (6.8)

Table 6.3 gives the CEC of common soil constituents. Clay minerals show a wide range in CEC
depending upon mineral structure, structural substitutions and the specific surface of the mineral
accessible to water. For iron oxides and organic matter, the CEC is also a function of pH since the
surface oxygens behave as amphoteric acids. Oxyhydroxides may even acquire a positive charge
below pH 8 (discussed in Chapter 7).

Table 6.3. Cation exchange capacities of common soil and sediment materials.

CEC, meq/kg

Kaolinite 30–150
Halloysite 50–100
Montmorillonite 800–1200
Vermiculite 1000–2000
Glauconite 50–400
Illite 200–500
Chlorite 100–400
Allophane up to 1000
Goethite and hematite: up to 1000 (pH ! 8.3, discussed in Chapter 7)
Organic matter (C) 1500–4000 (at pH  8, discussed in Chapter 7)
or, accounting for
pH-dependence: 510  pH – 590  CEC per kg organic carbon (Scheffer and Schachtschabel, 2002).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

248 Ion exchange

EXAMPLE 6.1. Recalculate CEC (meq/kg soil) to concentration (meq/L pore water)
The cation exchange capacity (CEC) of a sediment is expressed in units of meq per kg of dry soil. Express
a CEC of 10 meq/kg of a sediment with porosity   0.2, in meq/L pore water. The sediment consists mainly
of quartz grains with specific weight of 2.65 g/cm3.
ANSWER:
Since quartz has a specific weight of 2.65 g/cm3, 1 kg of dry sediment contains 1000/2.65  377 mL “grains”.
The amount of pore water in 1 kg sediment can be calculated and CEC expressed in units meq/L pore water
(Example 3.7): Total sediment volume is 377/(1  )  472 mL; pore water  472    94 mL.
CEC  10 meq/kg  10/94 meq/mL  106 meq/L pore water.

bulk density b  1000 / 472  2.12 g/cm 3 .

Generally CEC  b / e  meq/kg  kg/L  meq/L

QUESTION:
Express CEC  15 meq/kg in terms of pore water concentration.   0.3, b  1.86 g/cm3.
ANSWER: 93 meq/L

Example 6.1 shows how a CEC of 10 meq/kg, a reasonable value for a sandy aquifer, corresponds to
106 meq/L of pore water. Fresh groundwater normally contains less than 10 meq/L cations, ten times
less than the amount of cations located on the exchanger. The example serves to illustrate that the
amount of adsorbed cations is often very large in comparison to the amount of cations in solution.

6.2.1 Clay minerals

The basic coordination units for clay minerals are tetrahedra and octahedra with oxygens forming
the corners, and a cation residing in the centre (Figure 6.6). In tetrahedral coordination the central
cation is surrounded by four oxygens, in octahedral coordination the cation is surrounded by six oxy-
gens. The oxygens in the coordination units are envisaged as spheres with radii of about 1.4 Å that
should just touch each other. Tetrahedra therefore contain the smaller metal-ions such as Si4 or
Al3, and octahedra the larger metal-ions such as Al3, Mg2, Fe2, Mn2. (Still larger metal-ions
such as Ca2, Na, K, have even larger coordination polyhedra). Within the clay minerals, tetrahe-
dra and octahedra form layers which share oxygens. The way of stacking of these layers determines
the type of clay mineral. In kaolinite the layers of tetrahedra (
) and octahedra () show the repeti-
tion:    . The chemical formula of kaolinite is:

[Al 2 ]vi [Si 2 ]iv O5 (OH)4 (6.9)

where vi and iv indicate six- and four-fold coordination by oxygen. Stacking of a tetrahedral layer of
approximately 3 Å with an octahedral layer of 4 Å gives a repetition of the  structure every 7 Å.
This is the characteristic c-axis spacing (or 001 reflection) observed in X-ray analysis of kaolinite.
The outer oxygens of the octahedra of one layer share protons with the tetrahedra from the next layer.
In mica or soil clay minerals such as smectite or montmorillonite the octahedra are sandwiched
between two layers of tetrahedra, and the structure repeats as:   . The  structure is
about 10 Å thick, and the cohesion of the  layers is mediated by interlayer cations. Another group
of clay minerals, of which chlorite is an example, has its interlayers filled with an octahedral layer.
These structures can be indicated as repetitions of     (of 14 Å each). All these minerals
are found in the finest fraction of soils and aquifers.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Adsorbents in soils and aquifers 249

and  Hydroxyls Aluminums, magnesiums, etc.

Octahedron Octahedral layer

and  Oxygens and  Silicons

Tetrahedron Tetrahedral layer

Figure 6.6. A perspective drawing of tetrahedra and octahedra as coordinating units in minerals (Grim, 1968).

Clay minerals can have a deficit of positive charge which arises from substitutions of cations in the
structure of the crystal. The substitution of Si4 for Al3 reduces the positive charge by one in an
otherwise not much affected structure. The possible substitutions can be understood most easily with
Pauling’s rules for the buildup of crystal structures (Pauling, 1960). The first rule entails that in
filling the sites of cations in the framework of oxygens, the length of the cation-oxygen bond is the
most important. This length is even more important than the charge on the cation. Charge imbalance
can, if needed, be compensated in other units some distance away from the point of imbalance
(although Pauling’s second rule states that the most stable structure is obtained when charges are
balanced nearby).
A charge imbalance is mostly limited to the mica-type clay minerals. The charge imbalance can
be calculated from the structural formula, starting with the mica muscovite:

K xii [Si3Al]iv [Al 2 ]vi O10 (OH)2 (6.10)

The 10 oxygens and 2 hydroxyls give 22 negative charges; four tetrahedral metal sites, and two octa-
hedral metal sites must be filled in this formula. When the charge of the cations adds up to 21, one
single charge is left for the interlayer cation, which is K in muscovite. This K is actually in twelve-
coordination with oxygens from two adjacent tetrahedral layers. When Si4 takes a larger share in
the tetrahedral cation sites, the interlayer charge is diminished. An example is

K 0xii.7 [Si3.3Al0.7 ]iv [Al 2 ]vi O10 (OH)2 (6.11)

which is the typical formula for the clay mineral illite. A further decrease of the interlayer charge is
possible, but reduces the cohesion of the tetrahedral layers around the interlayer cation. The inter-
layer cation may then become hydrated, and the  layers separate as more and more layers of
water are incorporated in the structure.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

250 Ion exchange

OH

OH

OH

Exchangeable cations
nH2O

Oxygens OH Hydroxyls Aluminum, iron, magnesium

and Silicon, occasionally aluminum

Figure 6.7. Perspective drawing of the structure of the clay mineral smectite. (Grim, 1968).

Clay minerals with a mica structure, which expand to 18 Å with Mg-ions as interlayer cations, are
called smectites (Figure 6.7). The hydrated cations in the interlayer space can move freely into solu-
tion to be exchanged for other ions. Smectites (or montmorillonites) typically have an interlayer
charge that ranges from 0.6 to 0.3.
Substitutions also occur in the octahedral layer, for example in vermiculite:
xii [Si Al ]iv [Al Mg ]vi O (OH)
K 0.9 3.8 0.2 1.3 0.7 10 2 (6.12)

The low tetrahedral charge of 0.2 is here surpassed by an octahedral charge of 0.7, and the full
interlayer charge is 0.9. Vermiculites are able to swell to 14 Å with strongly hydrated cations
such as Mg2, despite the large interlayer charge, because the charge imbalance is predominantly
located in the octahedral layer, i.e. at a relatively large distance from the oxygens in the basal
tetrahedral plane to which the interlayer cations are bonded. However, most soil clay minerals with
such a high charge will tend to collapse when ions of suitable size enter the interlayer space, for
example K, NH 4 or ions of similar size and low hydration number. A collapsed structure can only
exchange through solid state diffusion, which is 3–4 orders of magnitude slower than diffusion in
solution.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Exchange equations 251

EXAMPLE 6.2. Structural charge of smectite

A smectite has the chemical formula Na0.6[Si7.6Al0.4]iv[Al3.8Mg0.2]viO20(OH)4 for one unit cell. Calculate
the amount of exchangeable cations per gram of mineral, and calculate the charge density on the basal plane
in eq/m2. The size of the unit cell is a  b  c  5.2  9  10 Å3, and a  b  5.2  9  46.8 Å2 is the
size of the basal plane where the interlayer cations reside.

ANSWER:
The gram-formula weight of the smectite is 733.5 g/mol. One kg contains 1.36 moles of the unit cell for-
mula, hence the exchange capacity is 0.6  1.36  1000  818 meq/kg. The charge of 0.6 is divided over
the two basal planes of a completely separated  layer, or 0.3 qe per 46.8 Å2. This is 6.41017 qe/m2
(qe is the elementary charge). Multiply with the elementary charge 1.61019 C, and obtain 0.1 C/m2; or
divide by Avogadro’s number Na  6  1023/mole, and obtain the charge density as 1.07 eq/m2.

6.3 EXCHANGE EQUATIONS

For ion exchange of Na for K we write the reaction:

Na  ,-X ↔ Na-X  K (6.13)

and the distribution of species is given by the law of mass action:

[ Na-X] [K ]
K Na \ K  (6.14)
[K-X] [ Na ]

The ions in the subscript below the equilibrium constant are written in the order in which they appear
as solute ions in the reaction. (Note that the convention for isotope exchange equations is to start
with the product of the reaction, cf. Chapter 2).
Square brackets in the equations denote activities. The Debye-Hückel theory offers a straightforward
model to relate concentrations and activities in water (Chapter 4). However, for adsorbed cations
there is no unifying theory to calculate activity coefficients and different conventions are in use. Since
the convention has a bearing on the results of exchange calculations, we must discuss this matter.
The standard state, i.e. where the activity of the exchangeable ion is equal to 1, is in all cases an
exchanger which is totally occupied by the same cations (Table 4.1). The activity of each exchange-
able ion is expressed as a fraction of the total, either as molar fraction (as for solutes, Table 4.1), or
as equivalent fraction. Furthermore, the total number can be based on the number of exchange sites,
or on the number of exchangeable cations.
For ion Ii the equivalent fraction I is calculated as:
meq I -X i per kg sediment meq I-X
I   i
(6.15)
CEC ∑ meq I-X
X i
I, J, K,…

where I, J, K, … are the exchangeable cations, with charges i, j, k.

A molar fraction M I is likewise obtained from:

mmol I -X i per kg sediment (meq I-X )/ii

M
I   i
(6.16)
TEC ∑ ( meq I-X )/ i
i
I, J, K,…

where TEC denotes total exchangeable cations, in mmol/kg sediment. The use of fractions always
gives   1. When the activity is calculated with respect to the number of exchangeable cations
this is indicated as [I-Xi], and with respect to the number of exchange sites as [I1/i-X].
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
252 Ion exchange

For homovalent exchange it makes no difference what convention is used, but for heterovalent
exchange the effect is quite notable. For the exchange of Na for Ca2, using the number of
exchangeable cations convention:

Na  1⁄2 Ca-X2 ↔ Na-X  1⁄2 Ca2 (6.17)

with

[ Na-X][Ca 2 ]0.5 Na [Ca 2 ]0.5

K Na\Ca   (6.18)
[Ca-X 2 ]0.5 [ Na ] 0.5 
Ca [ Na ]

When the equivalent fraction of the exchangeable cations is used in Equation (6.18) it conforms to
the Gaines-Thomas convention, after Gaines and Thomas (1953) who were among the first to give a
rigorous definition of a thermodynamic standard state of exchangeable cations. The use of molar
fractions in Equation (6.18) would follow the Vanselow convention (Vanselow, 1932).
If, on the other hand, the activities of the adsorbed ions are expressed as a fraction of the number
of exchange sites (X), then Reaction (6.17) becomes:

Na  Ca0.5-X ↔ Na-X  1⁄2 Ca2 (6.17a)

with

[ Na-X] [Ca 2 ]0.5 Na [Ca 2 ]0.5

K Na\Ca   (6.18a)
[Ca 0.5 -X] [ Na ] Ca [Na ]

Equation (6.17a) agrees with the Gapon convention (Gapon, 1933). In this case the molar and equiv-
alent fractions become identical since both are based on a single exchange site with charge 1.
The differences in Na\Ca exchange behavior resulting from the three different conventions are
shown in Figure 6.8. Both the exchangeable Na and the solute Na are expressed as fractions. In
the calculations KNa\Ca was set to 0.5 for all three equations, which means that [Ca-X2] or [Ca0.5-X]
is twice the [Na-X] when both ions have an activity of 1 in solution. Thus, Ca2 is the preferred, or
selected, ion relative to Na. As is apparent in Figure 6.8, the total solute concentration affects
the selectivity of the exchanger for Ca2 in that the higher charged ion is preferred more strongly
when the total solute concentration decreases. When a clay suspension in a solution containing Na
and Ca2 is diluted, the concentration of adsorbed Ca2 must increase while the solute Ca2
concentration decreases. This effect is a consequence of the exponent that is used in the mass action
equation.
Figure 6.8 shows only small differences among the exchange conventions at low total solute con-
centrations, and one might consider the practical aspects of calculation when making a choice. The
CEC of a sediment is mostly constant, whereas TEC of a heterovalent system varies with the relative
amounts of cations with different charge that neutralize the constant CEC. In most situations the
activities of exchangeable cations are therefore calculated more conveniently (the argument of
Gaines and Thomas, 1953) as equivalent fractions with respect to a fixed CEC. The Gapon conven-
tion is popular among soil scientists. It facilitates the calculation of exchangeable cations in het-
erovalent systems, and forms the basis for calculating the amount of exchangeable Na from the
Sodium Adsorption Ratio (SAR, the ratio of Na over Ca2  Mg2 in water), which is an impor-
tant parameter for estimating irrigation water-quality (Section 6.7). However, the equation does not
perform well when several heterovalent cations are present, which suggests that the Gaines-Thomas
or the Vanselow convention should be preferred (Bolt, 1982; Evangelou and Phillips, 1988). In this
book the Gaines-Thomas convention is generally used.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Exchange equations 253

1
Gaines & Thomas
Gapon
, exchanger
0.8 Vanselow

0.6
[Na-X]  [Ca-X2]
[Na-X]

0.4
1N

0.2

0.01N

0
0 0.2 0.4 0.6 0.8 1
[Na]
, solution
[Na] 2[Ca2]

Figure 6.8. The Na/Ca exchanger composition calculated from solute concentrations using three different
conventions. KNa\Ca  0.5 and at two solution normalities, 1 N and 0.01 N.

EXAMPLE 6.3. Exchange coefficients derived from different conventions as a function of solution normality
Wiklander (1955) determined the molar Ca/K-ratios on several clay minerals and a synthetic resin when in
equilibrium with a solution that contains equivalent amounts of Ca2 and K in Cl-solutions of different
normality. His results are presented below:

CEC (meq/kg) 0.1 N 0.01 N 0.001 N 104 N

Kaolinite 23 – 1.8 5.0 11.1

Illite 162 1.1 3.4 8.1 12.3
Smectite 810 1.5 (10.0) 22.1 38.8
Resin 2500 3.27 10.8 36.0 89.9

Calculate the equivalent and molar fraction of Ca2 and K on the clay, and also the exchange coefficients
according to the conventions of Gaines-Thomas, Vanselow, and Gapon. Assume that solute concentrations
are equal to activities.
ANSWER:
A 10-fold dilution of the solution gives an increase in the
( 
[Ca 2
]) /[K] ratio of
1
0  3.2 in Equation
(6.18). For each 10-fold dilution the Ca /K-ratio should therefore increase by about a factor of 3.2. The
resin (a synthetic ion-exchanger) behaves quite nicely according to this relation. The behavior of the clay
minerals deviates somewhat although the same trend is followed. We are therefore confident that it is mean-
ingful to apply the mass action equation and to calculate the exchange coefficient. The calculation is carried
out only for smectite, leaving the calculation for the other clay minerals as an exercise.
From the measured molar ratio on the clay, we calculate the molar fractions in combination with:

Ca
M   M  1,
K

which for example for 0.1 N gives M

Ca  1.5 K, and  K  1 / 2.5  0.40,  Ca  0.60.
M M M

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

254 Ion exchange

The equivalent fractions are obtained with the help of Equation (6.16):

meq K-X ( meq K-X ) / CEC K

KM   
meq K-X  (meq Ca-X ) / 2 ( meq K-X ) / CEC  (meq Ca-X ) / ( 2 CEC ) K  Ca / 2
2 2

while also in this case

Ca  K  1

which gives K  MK /(MK  2 MCa).

For 0.1 N we obtain K  0.25, and Ca  0.75.
The molal concentrations in solution are for the 0.1 N solution mK 0.05 and mCa2  0.025 mol/kg
H2O, taken equal to activities. The exchange coefficients for the different conventions are

Gaines & Thomas Vanselow Gapon

0.5 [K ] ( M )0.5 [K ]  [K ]
K Ca\K  Ca V
K Ca\K  CaM G
K Ca\K  Ca
K [Ca 2 ]0.5 K [Ca 2 ]0.5 K [Ca 2 ]0.5

Thus, the exchange coefficient for Ca/K sorption on smectite is:

0.1 N 0.01 N 0.001 N 104 N

KCa\K 1.10 (2.05) 1.41 0.78

V
KCa\K 0.61 (1.05) 0.71 0.39
KGCa\K 0.95 (2.00) 1.40 0.78

The Ca\K selectivity of smectite varies somewhat with total solute concentrations, and none of the conven-
tions performs decisively better than the others. In the calculations we assumed that activity coefficients are
equal to 1 for both exchangeable and solute cations, but this may be incorrect. However, if we apply activity
corrections only to the solute ions, the variation increases.

QUESTION:
Calculate KCa\K by applying solute activity corrections for I  0.1 and 104 N?
ANSWER: KCa\K  1.39 and 0.79 for 0.1 and 104 N, respectively.

6.3.1 Values for exchange coefficients

Table 6.4 lists exchange coefficients for various ions following the Gaines-Thomas convention. The
table provides coefficients rather than constants since the values depend upon the type of exchanger
present in the soil, and also on the water composition (Example 6.3). This is due to non-ideal behav-
ior of the exchanger, and because the activity coefficients are assumed to be equal for the solute and
the exchangeable ion. The given ranges represent many measurements from different soils and for
different clay minerals.
The exchange coefficients in Table 6.4 are relative to Na. Exchange coefficients among other
cation-pairs are obtained by combining two reactions. For example, the exchange reaction for Al3
and Ca2 is obtained from:

Na  1⁄2 Ca-X2 ↔ Na-X  1⁄2 Ca2, KNa\Ca  0.4

and

Na  1⁄3 Al-X3 ↔ Na-X  1⁄3 Al3, KNa\Al  0.7

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Exchange equations 255

Table 6.4. Ion exchange coefficients relative to Na following the Gaines-Thomas convention. Based partly
on a compilation by Bruggenwert and Kamphorst, 1982.

Equation: Na  1/i I-Xi ↔ Na-X  1/i Ii

with

[ Na-X] [ I i ]1/i  Na [ I i ]1/i

K Na \ I  
[ I -X i ]1/i [ Na ] 1l /i [ Na ]

Ion I KNa\I Ion I 2 KNa\I Ion I 3 KNa\I

Li 1.2 (0.95–1.2) Mg2 0.50 (0.4–0.6) Al3 0.7 (0.5–0.9)

K 0.20 (0.15–0.25) Ca2 0.40 (0.3–0.6) Fe3 ?
NH 4 0.25 (0.2–0.3) Sr2 0.35 (0.3–0.6)
Rb 0.10 Ba2 0.35 (0.2–0.5)

Cs 0.08 Mn2 0.55
Fe2 0.6
Co2 0.6
Ni2 0.5
Cu2 0.5
Zn2 0.4 (0.3–0.6)
Cd2 0.4 (0.3–0.6)
Pb2 0.3

Subtracting the two reactions, and dividing the two exchange coefficients gives:
1
⁄3Al3  1⁄2Ca-X2 ↔ 1
⁄3Al-X3  1⁄2Ca2, KAl\Ca  0.6

In case of homovalent exchange, the coefficients for the Gapon and Vanselow conventions are iden-
tical to the Gaines-Thomas values. For heterovalent exchange it is possible to derive the coefficients
for the binary case (two cations only, Problem 6.9).
For sediments and soils, the selectivity of cations generally follows the lyotropic series. Cations with
the same charge are more strongly held when their hydration number is smaller, i.e. when the hydration
shell of water molecules is smaller (cf. Section 6.6.1). A similar sequence exists for synthetic, strongly
acid ion-exchange resins (Helfferich, 1959; Rieman and Walton, 1970). In contrast, weakly acid resins
exhibit the reverse selectivity in the high pH range when the resin is fully deprotonated. This may indi-
cate that strong acid cation exchangers generally are the most important in the natural environment.

6.3.2 Calculation of exchanger composition

The composition of the exchanger can be calculated by combining the mass action expressions with
the mass balance for the sum of the exchangeable cations. Analogous to Reactions (6.17) and (6.18)
we can write the general reaction:

1/i I i  1/j J -X j ↔ 1/i I -X i  1/j J j +

which gives:

 Ij/i K Jj \ I [J j ]
J  (6.19)
[ I i ] j/i

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

256 Ion exchange

and also K, L, …, all expressed as a function of I. All fractions are introduced in the mass balance:
 I   J   K  ..  1 (6.20)

to give an equation with one unknown, I. When only monovalent and divalent ions are present, a
quadratic equation results that can be solved (Example 6.4).

EXAMPLE 6.4. Calculate the cation exchange complex in equilibrium with groundwater
Calculate the exchangeable cations in dune sand with a CEC of 10 meq/kg, in equilibrium with dunewater,
Na  1.1, Mg2  0.48, Ca2  1.9 mmol/L. Assume that activity is numerically equal to concentration
in mol/L. Use the exchange coefficients in Table 6.4.

ANSWER:
First express the exchangeable fractions of divalent cations as a fraction of Na. Thus:
[-g 2 ]
Mg  2Na 2
K Na\Mg [ Na ]2

and
[Ca 2 ]
Ca  2Na 2
K Na\Ca [ Na ]2

are introduced into

Mg  Ca   Na  1

Therefore:

 [Mg 2 ] [Ca 2 ] 
2Na  2 
 2     Na  1  0
 K Na\Mg [ Na ] K Na\Ca [ Na ] 
2 2

Substitutions of the coefficients from Table 6.4, and concentrations for solute activities yields:

2Na {0.48103 / (0.5  1.1103 )2  1.9103 / (0.4  1.1103 )2 }   Na  1  0

This quadratic equation is solved to give Na  0.00932, and by back-substitution, Mg  0.138 and
Ca  0.853. With the factor b /   6 kg/L we can recalculate the concentrations from meq/kg to meq/L
pore water (cf. Example 6.1). Thus, the exchangeable cations amount to meqNa-X  Na CEC 6 
0.5 meq/L, meqMg-X2  8.3 meq/L, and meqCa-X2  51.2 meq/L.

The use of the Vanselow convention gives mole fractions. These can be recalculated to equivalent
fractions with (cf. Problem 6.14)
M
I i
I  (6.21)
M
I i   J j  K k 

M M

The Gapon convention makes it particularly easy to calculate exchangeable fractions. All fractions
are linearly related according to:
 I K JG\ I [ J j ]1/ j
J 
[ I i ]1/ i

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Exchange equations 257

When all fractions are entered in the sum   1, some rearrangement gives:

[ I i ]1/ i
I  (6.22)
∑
1/ j
K JG\ I [ J j ]
J  I , J , K ,...

This last equation is also known as the multicomponent Langmuir equation.

6.3.3 Calculation of exchanger composition with PHREEQC

In the database of PHREEQC, cation exchange is defined according to the Gaines-Thomas
convention. The exchanger X is specified under EXCHANGE_MASTER_SPECIES, followed by
exchange half-reactions under keyword EXCHANGE_SPECIES. The exchange master species must
be defined as the first species with log K  0, similarly as for solution species. An excerpt of the
default database is:
EXCHANGE_MASTER_SPECIES
X X-
EXCHANGE_SPECIES
X- = X-
log_k 0.0

Na+ + X- = NaX
log_k 0.0
-gamma 4.0 0.075
Ca+2 + 2X- = CaX2
log_k 0.8
-gamma 5.0 0.165

PHREEQC handles exchange reactions by splitting them into half reactions. For example for Ca2\Na
exchange we may rewrite the reaction in Table 6.4 as:

Ca 2  2 Na-X ↔ Ca-X 2  2 Na (6.23)

with log K  log(1 / KNa\Ca

2
)  log(1 / 0.42)  0.8.
One half reaction needs to be defined as the point of reference, which is:

Na  X ↔ Na-X log K  0.0 (6.24)

If we add Reaction (6.24) twice to Reaction (6.23) (also the log K’s are added), we obtain:

Ca 2  2 X ↔ Ca-X 2 log K  0.8 (6.25)

which is identical to the equation in the database (see above).

PHREEQC uses the same activity coefficient for the exchangeable species as for the aqueous
species. This procedure is based on observations of Na/Ca2 exchange in saline soils. Here, the use
of activity coefficients for the solute ions only leads to an overestimate of Na-X (or, more generally,
of the monovalent ions compared to the divalent ions, cf. Example 6.3). For the calculation of the
exchangeable fractions, the free ion concentrations are used, equal to the total (analytical) concen-
trations minus the aqueous complexes.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

258 Ion exchange

EXAMPLE 6.5. Calculate the exchanger composition in contact with groundwater, using PHREEQC
Repeat the hand calculation of Example 6.4 with PHREEQC. The input file is:

SOLUTION 1
Na 1.1
Mg 0.48
Ca 1.9

EXCHANGE 1
X 0.06 # moles
-equilibrate 1

PRINT
-reset false
-totals true
-exchange true
END

Keyword EXCHANGE defines X  0.06 mol to be in equilibrium with solution 1. Since the solution has
1 kg H2O by default, the CEC is 60 meq/L, in accordance with Example 6.4. The keyword PRINT is used to
reduce the lengthy output. The first line “-reset false” indicates that all printout is to be suppressed, while
“-totals true” and “-exchange true” will print these specific variables. In the output file we find:
----------------------Exchange composition----------------------------------------

X 6.000e-02 mol

Equivalent
Species Moles Equivalents Fraction Log Gamma

CaX2 2.564e-02 5.127e-02 8.545e-01 -0.132

MgX2 4.086e-03 8.172e-03 1.362e-01 -0.130
NaX 5.572e-04 5.572e-04 9.287e-03 -0.033

The results correspond to Example 6.4 within roundoff errors of the logarithms of the K values from Table 6.4.

Other exchange conventions can be implemented also in PHREEQC, as shown by Appelo and
Parkhurst (cf. www.xs4all.nl/appt). For example, the Rothmund-Kornfeld equation has been used
recently in several studies (Bond, 1995; Voegelin et al., 2000; Bloom and Mansell, 2001). The equation
contains an empirical exponent for adapting the activity-ratio of the solute ions and can simulate the
sigmoidal shape often observed in binary isotherms (which reflects that exchangeable cations are pre-
ferred more when present in small concentrations since the sites which favor the cation are filled first).
Thus, for Ca2\K exchange:

Ca 2  2K-X ↔ Ca-X 2  2K

the law of mass action for Brucedale subsoil was (Bloom and Mansell, 2001):

0.618
[Ca-X 2 ]  [K ]2 
   100.281 (6.26)
[K-X]2  [Ca 2 ] 

Example 6.6 illustrates how this equation is introduced in PHREEQC.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Exchange equations 259

EXAMPLE 6.6. Calculate the exchanger composition with PHREEQC, using the Rothmund-Kornfeld equation
For Brucedale subsoil, the exchange species CaX2 is defined as:

0.618Ca2  2KX  CaX2  1.236K

which has the coefficients of mass action Equation (6.26). However, the stoichiometric coefficients for Ca2
and K are incorrect and the equation would generate an error message of PHREEQC, unless “-no_check”
and “mole_balance” are used. These options allow exponents in the mass action law that differ from the stoi-
chiometric coefficients in the reaction equation.
The input file further defines a solution and exchanger which contain K only. Subsequently, by using the
keyword REACTION, Ca2 is added and K is removed and the isotherm is plotted that covers the range of
K from 0–1.
EXCHANGE_SPECIES
0.618Ca+2 + 2KX = CaX2 + 1.236K+ ; -log_k - 0.281
-no_check
-mole_balance CaX2
-gamma 5.0 0.165
SOLUTION 1
K 100 # 100 mmol K/kg H2O
EXCHANGE 1
KX 0.417 # 417 mmol KX, total K in the system = 517 mmol

REACTION
K -1 Ca 0.5 # Remove 517 mmol K+, replace with 258.5 mmol Ca2+
0.517 in 20 steps
USER_GRAPH # Plot isotherm...
-head alpha_K beta_K
-axis_titles "Solute K / (K + 2Ca)" "Exchangeable K / (K + 2Ca)"
-axis_scale x_axis 0 1 0.2 0.1; -axis_scale y_axis 0 1 0.2 0.1
-start
10 graph_x tot("K") / 0.1
20 graph_y mol("KX") / 0.417
-end
END

A plot of the K/Ca2 isotherm is obtained with USER_GRAPH and compared with measured data in
Figure 6.9. The sigmoidal shape of the isotherm indicates that the cation with a relatively low concentration
is favored by the exchanger.

CCa/CT
1.0 0.8 0.6 0.4 0.2 0.0
1.0 0.0

0.8 0.2

0.6 0.4
Ca
K

0.4 0.6

0.2 0.8

0.0 1.0
0.0 0.2 0.4 0.6 0.8 1.0
CK/CT

Figure 6.9. The K/Ca isotherm according to the Rothmund-Kornfeld equation (Bloom and Mansell, 2001).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

260 Ion exchange

Bond (1995) and Bond and Verburg (1997) have shown how results from the Rothmund-Kornfeld
equation can be interpreted in terms of activity coefficients. They obtained activity coefficients
for the exchangeable cations which varied from 0.2 to 1 in Ca2-K systems, and from 1 to 3
in Ca2-Na systems. One remaining problem in their study was that several minerals were
present in the soil, with every mineral having various cation exchange sites, and each site exhibiting
probably different cation preferences. Moreover, Bond and coworkers found that the exchange
coefficients and exponents in the Rothmund-Kornfeld formulas changed with ionic strength,
leading to activity coefficients which vary with salinity. Unfortunately, these variations are poorly
understood although models derived from solid or liquid solutions have been applied (Elprince et al.,
1980; Liu et al., 2004). Thus, the Rothmund-Kornfeld expression may be helpful for detailed
description of accurate laboratory experiments where everything is well known and fixed, but it
may not altogether be applicable in field situations where solution compositions and salinities
vary markedly.

6.3.4 Determination of exchangeable cations

Most methods start with the removal of solute cations by centrifuging a field-moist sample, or by dis-
placing the pore solution with alcohol or a fluid immiscible with water (Figure 6.10). Subsequently, the
sorbed cations are exchanged with an alien cation, while trying to avoid the dissolution of minerals that
can yield false cation contributions. Alternatively, corrections can be made by analyzing the anion
released by the dissolving mineral (alkalinity in the case of carbonates, sulfate in case gypsum is present).

Exchangeable cation determination:

– Displace solute cations with alcohol
or
determine the mass c  εw

– Displace exchangeable cations with

surplus of alien cation

– Prevent or reduce dissolution of calcite,

gypsum, etc.
or
correct via anion from the dissolved mineral

Figure 6.10. Procedure for determining exchangeable cations in soils and sediments.

Many cations have been advocated as the displacing cation, but usually each one yields a different
result since the cations have varying displacing power and induce various side reactions. A popular
method is displacement with 1 M NH4-acetate in which case a single extraction is sufficient to remove
all the exchangeable cations, while not much calcite dissolves if the pH is adjusted to 8.2 (Rhoades,
1982). However, even a small amount of calcite that may dissolve could contribute substantially to
“exchangeable” Ca2 when the exchange capacity is low. Figure 6.11 shows on the left the computed
amounts of calcite that dissolve as a function of the pH of the acetate solution and the CEC (X). On
the right, Figure 6.11 shows the contribution of calcite dissolution to “exchangeable Ca”. For example
when 20 mL NH4-acetate solution is added to 5 g soil with a CEC of 10 meq/kg, then X  2.5 mM.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Exchange equations 261

9 100
8
Calcite dissolved (mmol/L)

% Calcite of total Ca
7 80
6
60 X  2.5 mM
5 0 mM
4 2.5
40 X  12.5 mM
3 12.5
2
X  25 mM 20 X  25 mM
1
0 0
7 7.5 8 8.5 9 7 7.5 8 8.5 9
pH pH

Figure 6.11. The calculated effect of calcite dissolution on determination of Ca-X2 using 1 M NH4-acetate.
Left: the amount of dissolved calcite as a function of pH and X (CEC). Right: the contribution of calcite dis-
solution to “exchangeable Ca”. Calculations included the Ca-acetate complex with an association constant of
101.18 in Martell and Smith (1977).

According to Figure 6.11, 1.9 mM calcite may dissolve at pH  8.5 and X  2.5 mM, which amounts
to (1.9 / (1.9  1.25)) 100%  60% of the total Ca2 in the extract (the sum of calcite and Ca-X2).
The Ca-acetate aqueous complex contributes significantly to calcite dissolution and the Ca2 that orig-
inates from calcite cannot be corrected, since an alkalinity titration of the 1 M acetate solution is useless.
In contrast to the results of batch experiments, the exchange constants for the major cations
derived from column experiments and field injections are more uniform. To obtain constants
from batch experiments that are comparable to flow experiments requires the use of a cation that is
strongly hydrated and only weakly displaces protons, for example Na (Zuur, 1938; Van der Molen,
1958) or Li (Hurz, 2001; Reardon et al., 1983). The recommended procedure is to carry out 2–3
subsequent extractions with 1 M NaCl, followed by an alkalinity titration to correct for dissolving
calcite. Exchangeable Na can be determined separately, in an extraction with 1 M NH4Cl.
Example 6.7 simulates the extraction of exchangeable cations in our dune sand (Example 6.4) with
PHREEQC. The keyword MIX is used to work in absolute units. MIX takes fractions of the actual
amount of solution (you may compare the amounts of water that PHREEQC calculated with the weight
of the solutions in the real-life analytical procedure). The extraction must be repeated once since not all
Ca-X2 is displaced in the first step. The last simulation mixes the two extractions and the total concentra-
tions are calculated in much the same way as would be done when a laboratory experiment is analyzed.
Note that in this sample with a low CEC, the low concentrations can be analyzed only if analytically
pure NaCl is used in the laboratory extraction.

EXAMPLE 6.7. Simulate the analytical measurement of exchangeable cations

Place 5 g (dry) soil in a centrifuge tube and extract twice with 20 mL 1 M NaCl. We neglect the contribution from
the original pore water, but do consider that 2.5 mL of the extractant remains in the centrifuge tube after decant-
ing. As a result, in the 2nd extraction a fraction 2.5 / 20  0.125 of the first extraction is added to the second one.

# Define exchanger and solutions...

SOLUTION 1 # Pore water
Na 1.1; Mg 0.48; Ca 1.9
EXCHANGE 1 # 5 g soil with CEC = 10 meq/kg, or 0.05 mmol X-
-equilibrate 1
X 0.05e-3 # moles

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

262 Ion exchange

SOLUTION 2; Na 1e3; Cl 1e3 # Extractant solution

END

USE exchange 1 # ...put 5 g soil in centrifuge tube

MIX # ...add 20 g 1M NaCl
2 20e-3
SAVE exchange 1; SAVE solution 3 # centrifuge and decant
END

USE exchange 1 # repeat extraction, 2nd time

MIX # 20 g new extractant, 2.5 g of the old extractant
2 20e-3
3 0.125 # 2.5 / 20 part of the old extractant
SAVE solution 4
END

MIX # combine the two centrifuged solutions...

3 0.875 # 17.5 / 20 part decanted after centrifuging
4 0.8888 # 20 / 22.5 idem
USER_PRINT
-start
10 print "CaX2 = ", (tot("Ca") * tot("water") * 4e5), "meq/kg"
20 print "MgX2 = ", (tot("Mg") * tot("water") * 4e5), "meq/kg"
-end
END

The last simulation combines the two decanted solutions and prints the exchangeable cations. In line 10 of
USER_PRINT, tot(“Ca”) gives the molality of total Ca2, tot(“water”) provides the kg of water and the
product of the two, the moles of Ca2  . When multiplied with 1000 (g/kg)/(5 g soil)  2000
(meq/mol)  4  105, it yields meq/kg soil. The result is:
------------------------------------------------------User print------------------------------------------------
CaX2 = 8.4129e+00 meq/kg
MgX2 = 1.3418e+00 meq/kg

which is nearly equal to CEC  the exchangeable fraction, calculated in Example 6.4 (there is still some
trace of Ca2 and Mg2 left on the exchanger in the centrifuge tube). Exchangeable Na must be deter-
mined by means of a separate extraction with NH4Cl.

QUESTION:
Perform the extraction with 1 N NH4Cl. Look in PHREEQC.DAT how NH 4 is defined (a special case,
because oxidation to N2 and NO3 is to be avoided). Also assume that 5 g soil now contains 1% pore water.

6.4 CHROMATOGRAPHY OF CATION EXCHANGE

The principles of ion exchange are routinely used in the laboratory in ion-chromatography, a method
to separate and analyze ions from a complex mixture. The chemical analyst uses a column packed with
beads of an ion-exchanger. The pores within and between the beads are filled with fluid, and the total
amount of fluid in the column is termed the column pore volume, or just the pore volume (V0, m3). The
column is percolated with different solutions, and as a result of cation exchange a concentration pattern
develops at the end of the column. Strongly selected cations will displace other ions from the exchanger
and be transported at a relatively low velocity. In displacement chromatography, the cation in the
injected solution is sorbed more strongly than the resident cations, which are flushed in an ordered
manner from the column. The ion with the lowest selectivity is leached first, then the next favored, etc.,
and finally the cation arrives that is injected in the column. If, on the other hand, the cation in the
incoming solution has a lower affinity for the exchanger, the process is called elution chromatography.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 263

In this case, the concentrations at the outlet of the column change more gradually. Sometimes the total
ion concentration of the injected fluid is different from the resident solution. This will induce a salinity
front, or an anion-jump when one pore volume has been displaced from the column.

6.4.1 Field examples of freshening

Displacement chromatography applies to field situations where fresh water, containing mainly Ca2,
displaces seawater dominated by Na and Mg2. Ca2 is favored by natural ion exchangers over
Mg2 and Na. It is possible to obtain some idea of the timescales involved for flow and cation
exchange, if we simplify the fresh water to contain only Ca2 and HCO 3 . Example 6.8 shows how
such a calculation is performed.

EXAMPLE 6.8. Flushing of an exchange complex

Cultivated fields on islands in the Dutch Wadden Sea often show brackish patches in an otherwise fresh
groundwater area. The brackish water is a remnant of transgressions during the last century. They occur
where clay layers at the surface obstruct the downward percolation of fresh water. Estimate the time needed
for refreshening and flushing of the exchange complex for the situation depicted in Figure 6.12. The
CEC of the sediment is 10 meq/kg, bulk density b  2.0 g/cm3 and porosity   0.30. The salt water
has the composition of seawater, the fresh water is pure Ca2- and HCO3-water in equilibrium with
[PCO2]  0.01 atm.

P  300 mm/yr

285 mm/yr Fresh 285 mm/yr

 

15 mm/yr Saline 15 mm/yr

2m
CEC  10 meq/kg

Figure 6.12. Flushing of salt water below a clay layer: The 2 m. thick aquifer has a porosity 0.3 and
therefore contains 0.6 m water. The precipitation surplus is 300 mm/yr, of which 285 mm/yr flows directly
into the ditches, while 15 mm/yr percolates downward into the aquifer.

ANSWER:
First we calculate the number of years needed to flush the pore volume with a rainfall surplus of 15 mm/yr.
The seawater chloride is flushed after one pore volume, and the concentrations drop to the fresh HCO 3-
level. Then, the exchange complex is flushed. Initially the fresh water equilibrates with the original seawater
exchange complex. This continues until all the Na has been flushed. Next comes Mg2, which in turn is
displaced by Ca2. The flushing time is in all cases the ratio of the amount of exchangeable cation over the
amount of fresh water Ca2, multiplied with the flushing time for one pore volume.
The aquifer contains initially 0.6 m salt water, which is displaced in 40 years with 15 mm/yr. The amount
of exchangeable cations in the sediment can be calculated on a pore water basis: 2 g sediment has a vol-
ume of 1 cm3, and has 0.02 meq exchangeable cations, and 0.3 mL water. A sediment column of 2  1  1 m3
contains 2  2  106  4  106 g sediment, with 4 104 meq exchangeable cations, and 0.6 m3 water.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

264 Ion exchange

Percolation gives 0.015  1  1  1 m3  0.015 m3/yr. The fraction of exchangeable Na in equilibrium
with seawater is calculated as in Example 6.4:

2Na {[Ca 2 ] / ( K Na\Ca [ Na ])2  [Mg 2 ] / ( K Na\Mg [Na ])2 }   Na − 1  0

Assume that activity equals molal concentration i.e. [Na]  0.485, [Mg2]  0.055 and [Ca2]  0.011, and
use the exchange constants from Table 6.4, KNa\Ca  0.4, and KNa\Mg  0.5, to find Na  0.583, and subse-
quently Mg  0.318, and Ca  0.099. The sediment column contains 0.583  4  104  23316 meq Na-X.
Fresh water in equilibrium with calcite at PCO2  0.01 atm contains: Ca2  102.1  [PCO2] ⁄3 
1

1.7 mmol/L (cf. Chapter 5). The percolating volume thus brings 0.015  1700  2  51 meq Ca2 each year
into the soil. The time needed to flush all Na-X is then 23316 / 51  457 years. Similarly we find the time needed
for flushing of Mg2, using the assumption that Mg-X2 flushing first starts after all exchangeable Na-X has been
removed, i.e. after 249 years (this time for Mg removal is only approximate). The results are summarized below.
Years needed for removal

Salt (high Cl ) 40
Na (poor soil structure) 457
Mg (hydrological tracer) 249
Total: 746

4 Seawater

3
meq/L

2
NaHCO3 Mg(HCO3)2 Ca(HCO3)2
1

0 200 400 600 800

years

Figure 6.13. Water compositions at the end of a flowline as seawater is displaced by fresh water.

Figure 6.13 shows how the water composition changes over time at the end of a flowline. We will discuss the
bad soil structure associated with NaHCO3 type water in later sections of this Chapter, but note already that
gypsum is used to accelerate the reclamation of saline soils. Gypsum dissolution will increase the Ca2 con-
centration in the percolating water and thereby speed up flushing of Na (Šimůnek and Suarez, 1997). In
these (simple) calculations, the effects of dispersion and geological heterogeneity were neglected. These
tend to accelerate the initial breakthrough of Ca2, but slow down the complete removal of Na and Mg2.

QUESTIONS:
What are the flushing times for Na and Mg2 if the CEC were twice higher?
ANSWER: twice longer
What are the flushing times when the Ca2 concentration is 10 times higher because gypsum is added to
the soil?
ANSWER: 10 times shorter, 45.7 year for NaHCO3 type water.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 265

Figure 6.13 shows how the water composition changes with time at the end of a flowline as calcu-
lated in Example 6.8. Alternatively the flushing times can be calculated using the retardation for-
mula for sharp fronts (cf. Example 3.6):
R  1  q / c (3.33)

which for cation exchange becomes:

R  1  CEC  / c (6.27)

Since the CEC is constant, only the fraction  of the exchangeable ions varies. At the front where
Na is being flushed, the initial Na is 0.583 and with CEC  66.7 meq/L pore water, this corre-
sponds to qNa  0.583  66.7  38.9 meq/L. After the front, solute Na concentration must balance
mHCO3  2  mCa2, therefore cNa  2  1.7  3.4 meq/L. Once all Na has been flushed from the
column, qNa  0 and cNa  0. Hence:

( qNa )1  ( qNa )2 38.9  0

R1  1   1  1  11.4  12.4 (6.28)
(c Na )1  ( cNa )2 3.4  0

The Na-front is 12.4 times retarded compared to the conservative front, which corresponds to
12.4  40  496 years from the onset of freshening.
The second front where Mg2 is flushed is obtained similarly, and also more accurately than in
Example 6.8. The injected Ca2 has not yet arrived at the end of the flowline (it exchanges with
Mg2), so that Ca2 remains unchanged, i.e. Ca  0.099. After Na has been flushed the remain-
der is Mg-X2 and Mg  1  0.099  0.901 with qMg  60.1 meq/L. Given the exchangeable
fractions of Ca2 and Mg2, the solution composition must obey:

Mg [Ca 2 ]
 0.4 
2


 
Ca [Mg ]
2  0.5 

or [Ca2]  0.0703 [Mg2]. Since the sum of the two cations is 3.4 meq/L,
cCa  cMg  (0.0703  1) cMg  3.4.
Hence cMg  3.18 meq/L.
The Mg2-front arrives with a retardation (R2) of:

( qMg )2  ( qMg )3 60.1  0

R2  1   1  1  18.9  19.9 (6.29)
( cMg )2  ( cMg )3 3.18  0

which amounts to 19.9  40  796 years. This is longer than calculated in Example 6.8 since the elution
of Ca2 together with Mg2 is now duly accounted for (for an exact answer, see Appelo et al., 1993).
We have calculated the retardation relative to flushing one pore volume through the column. If
the flowline is twice as long, twice as much time is needed for all the transitions and fronts. In other
words, we have found the characteristic velocity v  dx / dt of a front. This is illustrated in a
time / distance diagram (Figure 6.14), where the transitions appear as straight lines.
The freshening chromatographic pattern calculated in Example 6.8 has been observed in the injec-
tion experiment of Valocchi et al. (1981). Fresh water from a sewage plant with tertiary treatment (i.e.
giving near drinking water quality) was injected in a brackish-water aquifer, shown in Figure 6.15. The
native brackish water contained large amounts of Na and Mg2, and the fresh water had Ca2 as the
major cation; the cation exchange capacity of the silty aquifer sediment was as high as 100 meq/kg.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

266 Ion exchange

3
Ca(HCO3)2 Mg(HCO3)2

NaHCO3
1

Seawater
(Salinity front)

0 1 2 3 X

Figure 6.14. Front propagation and characteristic velocity v  dx / dt as calculated in Example 6.8.

Figure 6.16 shows the Ca2 and Mg2 concentrations in an observation well 16 m downstream from
the point of injection. The initial Ca2 and Mg2 concentrations are those in the undisturbed aquifer.
After injection of about 200 m3, the concentrations start to decrease and at 500 m3, the fresh water
has replaced all of the brackish water. The Ca2 (and Mg2) concentrations become now lower than
in the injected water, and remain low until about 5000 m3 have been injected. This stage corresponds
to the dilution step after the salinity front where Ca2 (and Mg2) are exchanged for Na. When
exchangeable Na becomes exhausted, the Ca2 and Mg2 concentrations increase again. Ca2
increases towards the concentration in the injection water. However, Mg2 substantially surpasses
the injected concentration because exchangeable Mg2 is displaced. The water changes quality vari-
ations were modeled with a numerical model using exchange constants obtained from batch experi-
ments with the same aquifer material, obviously with very good results (Figure 6.16).

Distance from injection well

160 40 30 20 10 0 16 43
Sand and gravel

Alternating layers sand and clay

5 Silt and sand

Meters

Clay
10
S23
(16 m)
Depth

(11 m)
15 (20 m) I1
(40 m)
(160 m) P5
P6
P7
S 22
20 (43 m) S24

S22 P7 P6 P5 I1 S23 S24

Figure 6.15. Aquifer outline for the Valocchi experiment, showing the fresh water injection well I1 and the
observation wells (Valocchi et al., 1981. Copyright by the Am. Geophys. Union).

The injection experiment by Valocchi et al. (1981) displayed the cation exchange phenomena especially
clearly since the CEC was high and the conditions in the aquifer before injection were uniform.
This is not always true for field studies where freshening and salinization may have occurred repeatedly
in the same aquifer. A reversal of flow does not restore the original quality conditions since the
cation exchange processes are non-linear. The chromatographic pattern may then become blurred.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 267

103 103
Computer Computer
5 simulation 5
simulation
Field data Field data

Mg2 (mg/L)
Ca2(mg/L)

2 2
102 Injection water 102

5 5

2 2
Injection water
10 10
102 2 5 103 2 5 104 2 5 105 102 2 5 103 2 5 104 2 5 105
m3 Injected m3 Injected

Figure 6.16. Breakthrough data for Ca2 and Mg2 in boring S23 of Figure 6.15 during injection of fresh water
into a brackish aquifer. The drawn lines represent simulated water qualities (Valocchi et al., 1981. Copyright by the
Am. Geophys. Union).

However, freshening quality patterns have been observed in studies by Lawrence et al. (1976),
Chapelle and Knobel (1983), Stuyfzand (1985, 1993, cf. Figure 6.2), Beekman (1991), Walraevens
and Cardenal, (1994), and in a diffusion profile by Manzano et al. (1992) and Xu et al. (1999).
The freshening chromatographic pattern was observed in exemplary fashion in the Aquia aquifer
(Chapelle and Knobel, 1983; Appelo, 1994). This aquifer consists of a Paleocene, glauconite rich
sand, confined by thick clay layers (Figure 6.17). The vertical exaggeration in the figure is mislead-
ing; the aquifer is 70 m thick and extends over 90 km. On a laboratory scale, this corresponds to a
hair-thin chromatographic column. The area was uplifted and emerged in the Pleistocene. As a
result, fresh CaHCO3-water infiltrated and displaced the saline pore water. While Cl has been
flushed almost entirely from the aquifer, cation exchange still reveals the previously more saline con-
ditions (Figure 6.18). The increased alkalinity in the downstream end is due to calcite dissolution as
Ca2 exchanges for Na, and NaHCO3 type of water is formed. Behind the NaHCO3 type water follows
first a KHCO3, then a MgHCO3 type of water and finally the fresh CaHCO3 infiltration water quality.

Recharge
area Chesapeake Bay

300
feet

0 Pleistocene

Piney point aquifer

0 5 10 miles
Marlboro clay
Aquia aquifer

Figure 6.17. Cross section showing the outline of the Aquia aquifer, Maryland, USA (Appelo, 1994).

Chromatographic patterns allow a time-frame to be associated with the succession of

various water qualities. The calculated profiles in Figure 6.18 were obtained when the first 50 km of
the aquifer had been flushed 8 times. Using 14C dating, Purdy et al. (1992) found an age of 12 ka
for water at a distance of 50 km from the recharge area. The chromatographic pattern required there-
fore 8  12,000  96,000 years to become established under present day flow conditions.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

268 Ion exchange

8 1

0.8
6
mmol/L

mmol/L
0.6
Alkalinity
4 Mg2ⴙ Kⴙ
0.4
2
0.2
Naⴙ
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60

pH
8
pH
2
7
mmol/L

1
Ca2ⴙ
0
0 10 20 30 40 50 60
Distance along flowpath (miles)

Figure 6.18. Water quality patterns in the Aquia aquifer; symbols indicate observed concentrations, lines are
modeled (Appelo, 1994).

Compared to the duration of the Pleistocene this seems a short timespan. Probably flow in the
aquifer was enhanced when the confining clay layers were sliced up by glacier tongues from
Pleistocene icesheets, thus creating an upward outlet for water from the aquifer. An almost identical
water quality pattern has been found in the Ledopaniselian aquifer at the Dutch / Belgian border
(Walraevens and Cardenal, 1994). Also in this case, flow times were established based on 14C meas-
urements corrected for the dissolution of calcite (Van der Kemp et al., 2000).

6.4.2 Salinity effects in cation exchange

Both in the Valocchi experiment (Figure 6.16) and in the Aquia aquifer (Figure 6.18) the salinity
jumped when fresh water displaced the salt water. The anions traveled unhindered through the
aquifer and determined the speed of the salinity front, while the injected cations exchanged with the
sediment and were retarded. Consequently, directly after the salinity front, the water attained equi-
librium with the original seawater exchanger. Example 6.9 calculates the composition of NaHCO3
water which has lost Ca2, and has equilibrated with the seawater exchange complex.

EXAMPLE 6.9. The water composition after passage of a salinity front during freshening
Calculate the water quality when fresh water with 3.4 meq anions/L displaces seawater (Example 6.8)

ANSWER:
The exchangeable fractions in equilibrium with seawater are Na  0.583, Mg  0.318, Ca  0.099. The
anion concentration is 3.4 meq HCO 3 /L, which limits the total cations to 3.4 meq/L. Hence:

mNa  2 mMg2  2 mCa 2  0.0034 eq/L (6.30)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 269

We furthermore have from exchange equilibrium

Ca (K Na\Ca [ Na ])2

[Ca 2 ]  (6.31)
2Na

and

Mg (K Na\Mg [Na ])2

[Mg 2 ]  (6.32)
2Na

Again, we set [I i]  mI i, and obtain on combining Equations (6.30), (6.31) and (6.32):

 2 K2 2 Mg K Na\Mg
2 
2    mNa  0.0034  0
Ca Na\Ca
mNa 
 2Na 2Na 

This quadratic equation can be readily solved with the fractions I in Example 6.8, and with KNa\Ca  0.4,
and KNa\Mg  0.5, to give mNa  3.39103 mol/L. Further back-substitution in Equations (6.31) and
(6.32) gives mMg2  2  106, and mCa2  0.5106 mol/L. This shows that our assumption that Na is
the only cation of importance after the salinity jump, was indeed justified.

During dilution, divalent ions are preferentially adsorbed in comparison to monovalent Na (Figure 6.8).
When fresh water displaces saltwater, dilution takes place and Na is therefore desorbed from the
exchanger (Example 6.9). For the exchange reaction of Ca2 with Na:
1
⁄2 Ca2  Na-X ↔ 1
⁄2 Ca-X2  Na

the law of mass action yields:

[ Na ] [Ca-X 2 ]
 KCa\Na
[Ca 2 ] [ Na-X]

If Na is diluted 10 times, then Ca2 must be diluted 100 times to maintain equilibrium with the same
exchangeable activities [Na-X] and [Ca-X2]. Similarly for Al3 / Na exchange, if Na is diluted
10 times, Al3 must be diluted 1000 times. For ions with the same charge the dilution is the same.
The dilution factor f can be used to calculate (more easily than in Example 6.9) the composition
behind a salinity front from the solute concentrations ahead of the front:

( mNa )0 2( mMg2  mCa 2 )0

   (i mI i )1 (6.33)
f f2

where the subscripts 0 and 1 indicate concentrations before and after the salinity front, and f is the
dilution factor. For Example 6.9 this yields:

485 2  (55  11)

  3.4 (6.34)
f f2

which is a quadratic equation that can be solved to yield f  142.9 (hence, mNa  3.39 mmol/L, as
before).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

270 Ion exchange

QUESTION:
Calculate the concentrations of Ca2 and Mg2 (mg/L) after the salinity jump in the injection experiment
of Valocchi et al. (Figure 6.16). The original concentrations in the aquifer were: (mNa)0  86.5,(mMg2)0
 18.2, (mCa2)0  11.1 mmol/L. Anions in the injected water sum up to  (i mIi)1  14.7 meq/L.
ANSWER: f  6.5, Ca2  10.5 mg/L, Mg2  10.5 mg/L, the same as in Figure 6.16.

The salinity effect is particularly clear when the concentrations decrease behind a salinity front since
the exchange complex will control the water composition longer when water is more dilute. But,
sometimes the effects are unmistakable when the salinity increases. Ceazan et al. (1989) injected
NH4Br solution in an aquifer with a relatively low total anion concentration of 0.5 meq/L.

Injection well Observation well

Lake

1080
1070
1.5 m Microequivalents per liter 1060
Ca2  Mg2

NH4 Br
12 300
2.8 250 570

10
Percent of injectate

Percent of injectate

2.4 200
concentration

concentration

150 398
2.0 8
100
1.6 Br 6 50
1.2 0

4 150 NH
4K

0.8 NH4
2 100
0.4 413
50
0 0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Time after injection (hr) Time after injection (hr)

Figure 6.19. Concentrations of ions in an observation well 1.5 m downstream of the point of injection of
NH4Br (Ceazan et al., 1989).

The passage of the injected fluid was monitored in an observation well located 1.5 m downstream
from the injection well (Figure 6.19). The arrival of Br after 2 hours was accompanied by an increase
in the total anion concentration to 1.8 meq/L. The NH 
4 concentration did not follow Br since NH4

2 2
was adsorbed upstream. Instead, the divalent cations Ca and Mg showed a marked increase
(Figure 6.19). The ratio (cCa2Mg 2) / cNa was 0.67 (meq/meq) in the native water (Table 6.5) and
changed to 1.54 at the peak of the Br concentration. This is again due to the salinity effect, and, using
Equation (6.28) we can calculate that f  0.43, cNa  0.7 meq/L and cCa2Mg 2  1.08 meq/L in the
peak. These numbers are in excellent agreement with the observations of Ceazan et al. (Figure 6.19).

Table 6.5. Concentrations (meq/L) in the injection experiment of Ceazan et al., 1989.

Time (h) Br anions c c2 c2 / c ratio

1.4 0 0.50 0.3 0.2 0.67

2.08 1.28 1.78 0.7 1.08 1.54

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 271

QUESTION:
The Cd2 concentration in resident groundwater in the experiment of Ceazan is estimated to be 1 g/L.
Calculate the concentration at the Br peak.
ANSWER: 1 / f 2  1 / 0.432  5.4 g/L.

6.4.3 Quality patterns with salinization

The development of a chromatographic pattern depends on the ratio of exchangeable and solute
concentration, which gives the velocity of a front (Equation 6.27). When the ratio is small, because
the CEC of the aquifer is low or the solute concentrations are high, the succession of the fronts will
be rapid. In case of seawater intrusion, the high concentrations accelerate the transitions of the
water types and reduce their lateral extent in the aquifer. This may be the reason why a chromato-
graphic sequence of salinization has not been observed at the field scale, even though CaCl2-
water type is an ubiquitous indicator of salt water upconing in coastal areas (Section 6.1).
Information on the sequence of compositions with salt water intrusions must therefore be obtained
from column or small-scale field experiments (Appelo et al., 1990; Beekman, 1991; Bjerg et al.,
1993; Van Breukelen et al., 1998; Gomis-Yagnes et al., 2000; Andersen, 2001), or from models as
discussed in the next section.
Beekman and Appelo (1990) performed column experiments, displacing fresh water with sea-
water diluted 1:1 with distilled water (Figure 6.20). When about 75 mL had been leached from the
column, the brackish water had displaced the fresh water and both Ca2 and Mg2 increased because
of the exchange with injected Na and the salinity effect, where cations must balance the increased
Cl. The amount of adsorbed Ca2 was small and therefore Ca2 rapidly dropped again towards the
Ca2 concentration in the brackish water. The Mg2 concentration showed a small decrease after the
initial salinity jump and then an increase towards the brackish water Mg2 content. The high Ca2
concentrations resulting from the salinity jump, may, in combination with high SO2
4 concentrations,
lead to gypsum precipitation (Gomis et al., 2000).

50
CEC  1.40 meq/100 g

40
Concentration (mmol/kg H2O)

2  Mg2

30

CI/10
20

Na/10

10 2  Ca2

NH
4 K

0
0 50 100 150 200 250
Effluent volume (mL)

Figure 6.20. Column experiment with once diluted seawater displacing fresh water (Beekman and
Appelo, 1990).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

272 Ion exchange

The simulated lines in Figure 6.20 were obtained with the Gapon exchange model, in which
Ca2 / Mg2 exchange is written as:

0.5Mg 2  Ca 0.5 X ↔ Mg0.5 X  0.5Ca 2 (6.35)

which has the mass action equation:

[Ca 0.5 -X][Mg 2 ]0.5

 KCa\Mg (6.36)
[Mg0.5 -X][Ca 2 ]0.5

The square roots for the solute activities in this equation induce the particular effect of favoring
sorption of small concentrations of Ca2 or Mg2. Gomis et al. (1996) have shown later that the
Mg2 concentration in the experiment is modeled more accurately when all species are given an
equal exponent of 1:

[Ca 0.5 -X][Mg 2 ]

 KCa\Mg (6.37)
[Mg0.5 -X][Ca 2 ]

(cf. Problem 6.15).

6.4.4 Fronts and chromatographic sequences

Chromatographic sequences in heterovalent, multicomponent systems are difficult to calculate, and
models that discretize the Advection-Reaction-Dispersion equation may have problems because
small concentration variations disappear by numerical dispersion. Van Veldhuizen et al.
(1998) developed a model that integrates the matrix of the retardations {dq / dc}, calculates the posi-
tion of the fronts and is dispersion-free. However, the program solves only for uniform initial condi-
tions and one input solution. For such systems a few rules can be formulated (Helfferich and
Klein, 1970):

– In a system with n ions, a chromatographic sequence develops with n constant compositions

(“plateaux”) from the initial to the final composition.
– The initial solution composition may change in response to a salinity change.
– Transitions between the plateaux may be sharp (a shock), or smooth (a wave).
– The n  1 transitions involve the sequential reaction of two ions. The first to react are the ions
with the lowest affinity; they are not much retarded because dq / dc is small.
– The transition is a shock if the ion with the higher affinity of the two increases in concentration,
and a wave if the concentration decreases. All ions with still higher affinity will similarly increase
or decrease in concentration.
– The ion with the lower affinity of the two reacts counter to the higher affinity ion
(decreases in concentration in a shock, increases in a wave). All ions with still lower affinity react
likewise.
– A concentration increase in the injected solution (compared to the initial resident solution) of the
ion with the higher affinity of the two gives a shock, whereas concentration decrease in the
injected solution normally produces a wave.

Table 6.6 gives for a system with 6 ions, Li, Na, K, Mg2, Ca2 and Al3, the initial concentra-
tion in a column and the composition of the injected solution. The ions are listed in the order of
increasing affinity for the exchanger.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 273

Table 6.6. Initial concentrations (mol/L) in a model column and the composition of the injected solution.
CEC  0.1 eq/L. The resulting chromatographic sequence is illustrated in Figure 6.21.

Li Na K Mg2 Ca2 Al3

Initial 0.015 0.005 0.011 0.004 0.008 0.0015

Injected 0.008 0.015 0.004 0.01 0.006 0.003
Transition 1, shock 2, wave 3, shock 4, wave 5, shock

The chromatographic sequence after a transport distance of 9 m is shown in Figure 6.21. Six
plateaux are visible where the concentrations remain constant, while five transitions exist where all
the concentrations change. At transition 1, Na displaces Li and, because the concentration of Na
is higher in the injected solution than initially in the column this leads to a sharp front. Next K
exchanges for Na (transition 2), and since the injected concentration of K is lower than in the ini-
tial column solution, this produces a wave. In this fashion, the rules stated above will become clear
when the traces of the individual ions are followed in Figure 6.21. Basically it is the retardation equa-
tion which controls the velocity of the fronts, depending on the relative masses of the adsorbed and
solute ions. If the change in sorbed mass is relatively high, then more time is necessary for trans-
porting it via a relatively small change in solute concentration.
Manual calculation of the position of sharp fronts was already demonstrated in Section 6.4.1 and
is elaborated in Appelo (1994b) and Appelo et al. (1993). Broadening fronts in ion exchange can be
solved with a modification of the retardation Equation (6.27):

CEC d CEC d
R  1  1 (6.38)
dc A0 d

Here dc  A0d where A0 is the total ion concentration (meq/L) and  the equivalent fraction in
solution. By integrating Equation 6.38 it is possible to define the isotherm when the dispersivity in
the column is small and the flow velocity is low (DeVault, 1943; Bürgisser et al., 1993; Appelo,
1996, cf. Example 6.10).

20
Transition: 5 4 3 2 1 0
18
16
Na
14
12 K
mmol/L

10 Ca
8 Li
6
Mg
4
2 Al
0
0 2 4 6 8 10
Distance (m)

Figure 6.21. A chromatographic sequence along a flowline for a system with heterovalent cations, specified
in Table 6.6. The conservative front has traveled 9 m. The transitions were calculated with the model MIE of Van
Veldhuizen et al., 1998.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

274 Ion exchange

EXAMPLE 6.10. Flushing of K from a column

Van Eijkeren and Loch (1984) packed a laboratory column with cation exchange resin-beads. The column
was equilibrated with 0.3 meq/L KCl, and then flushed with 0.3 meq/L NaCl. The column length
L  6.9 cm; water flux  56.7 cm/day; dispersion coefficient DL  11.3 cm2/day. Experimental data of the
fraction of K in the effluent, K  mK / (mNa  mK), are shown in Figure 6.22. Let us model these data
with the retardation Equation (6.38).
A small dispersion coefficient yields a variance 2  2Dt of the concentrations around the midpoint break-
through. The time needed to flush one pore volume t  6.9/56.7  0.12 days. Hence  
   1.67 cm.
(2Dt)
Or, scaled to column length,   1.67 / L  1.67 / 6.9  0.24. The breakthrough curve for a conservative
element can now be constructed following Example 3.13, with c / ci  0.84 at V  (1  0.24)V0; c / ci  0.5
at V  V0; c / ci  0.16 at V  (1  0.24)V0; etc.; V  volume injected in the column. The conservative
breakthrough is plotted in Figure 6.22 (line V0,).
The next step is to obtain the CEC of the column by integrating the area between the conservative break-
through and the K-experimental points. This gives CEC  5.66 A0, where A0 is the total ion concentration
in the effluent (0.3 meq/L). The exchange equation for K & Na is reworked with [K]  KA0 and
[Na]  (1  K)A0 to:

K K\Na K
K 
1  K  K K\Na ,

and the derivative is:

d , K K\Na

dK
( )
2
1  K  K K\Na ,

From Equation (6.38), R  1  CEC / A0 (d / d)  V*, where V* is the number of pore volumes that an
exchanged element arrives later than the same fraction of a conservative ion. V* can be read for a few frac-
tions from the experimental data to solve for KK\Na. The average is KK\Na  2.7. All variables have now been
estimated and the elution curve can be constructed as shown in Figure 6.22. (Rather than multiplying V0,
with R, we added R  1 to V0,, since dispersion for K is mostly determined by the exchange isotherm).
Van Eijkeren and Loch modeled the experimental data with a more intricate model, incorporating mobile
and immobile fractions. They used a lower KK\Na  1.7 obtained from batch experiments. It is interesting that
a similar discrepancy between K’s from batch experiments and a flow experiment was found by Rainwater
et al. (1987), who used essentially the same theory as given here. Rainwater et al. modeled a laboratory sand
box with injection and withdrawal wells.

αK
1.0
Experimental points,
68% of conservative breakthrough

Van Eijkeren and Loch, 1984

K
VK  5.66  V0,α
within V0  and 24%

(1  αK · K  αK)2
βK · αNa βK · (1 αK)
K   2.7
αK · βNa αK · (1 βK)
0.5

V0,α

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 pore volumes, V
1 2 days

Figure 6.22. The exchange of K by Na in a column packed with cation exchange resin-beads.
Dots: relative K concentrations from the experiment (Van Eijkeren and Loch, 1984). Line: modeled results.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 275

QUESTION:
Find d / d for K  0.5 and K  0.1, and calculate the retardation for these fractions.
ANSWER: for K  0.5: d / d  0.79, R  1  5.66  0.79  5.47. For K  0.1:
d / d  1.97, R  12.2.

6.4.5 Modeling chromatographic sequences with PHREEQC

PHREEQC calculates ion chromatographic sequences via the ARD Equation (3.66). The three
components of this equation, Advection, Reaction and Dispersion, are calculated sequentially for a
1D column or flowline. Figure 6.23 illustrates the discretization of the flowline in a number of cells.
One timestep, or a shift using PHREEQC terminology, moves the mobile cell contents into the next
cell. Subsequently, the reactions between the immobile entities (exchangers, minerals, etc.) and the
solution are calculated. Then for each timestep, dispersion is calculated by mixing the contents of
adjacent cells. Once more this is followed by calculating reactions between the mobile and immobile
entities. During the next timestep, everything is repeated. The mathematical details of the model are
discussed in Chapter 11; here we use the model to investigate the combined effects of transport and
ion-exchange.

After 3 shifts: PHREEQC Dispersion

Water volumes
o + + +
4 3 2 1

+ + + o o o
1 2 3 4 5 6

Cells

EXCHANGE 1–3 EXCHANGE 4–6

KINETICS 1–3 ----
EQUILIBRIUM_PHASES 4–6

Figure 6.23. PHREEQC in a 1D flowline.

Consider an 8 cm long laboratory column, filled with very coarse sand, CEC  1.1 meq/L pore
water. The initial solution is 1 mM NaNO3, and the column is flushed with 0.6 mM CaCl2 solution.
The pore water flow velocity is 3.17106 m/s (100 m/yr), the dispersivity  is 0.2 mm. First, we
investigate whether the reaction front will be broadening or sharpening for these conditions. The fol-
lowing input file provides the isotherm, plotted in Figure 6.24.
SOLUTION 1
Na 1
EXCHANGE 1
X 1.1e-3; -equil 1
REACTION 1
Na -1 Ca 0.5; 2.1e-3 in 20 steps

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

276 Ion exchange

USER_GRAPH
-head a_Ca b_Ca
-start
10 graph_x 2*tot(“Ca”)/1e-3
20 graph_y 2*mol(“CaX2”)/1.1e-3
-end
END

1
Exchanger Ca/(Na  Ca)

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
Solute Ca/(Na  Ca)

Figure 6.24. Ca/Na exchange isotherm for CEC  1.1 meq/L.

The isotherm is extremely curved, with a steep slope (high retardation R  1  dq / dc) for small
Ca2-concentrations and a small slope (low retardation) for high Ca2 concentrations. Therefore the
front in the column experiment will be self-sharpening.
What results do we expect from the column experiment? The Cl front arrives at the outlet after
1 pore volume (PV) (Figure 6.25A). The total anion concentration then increases from 1 to 1.2 meq/L
and Na must increase as well to 1.2 meq/L. The Ca2 ion arrives later since it is taken up by the
exchanger. After one PV, all the Na in the effluent comes from the exchanger, until it is exhausted
after q / c  1.1 / 1.2  0.917 PV. The retardation of the Na front is therefore 1.917. At the
same time Ca2 appears at the column outlet.

QUESTIONS:
What is the retardation of the Na / Ca2 front if the CEC  2.4 meq/L?
ANSWER: 1  2.4 / 1.2  3 PV
What is the retardation if the CEC  2.2 meq/L and 1.2 mM CaCl2 is injected?
ANSWER: 1  2.2 / (2 1.2)  1.917 PV

The same calculations can be done with PHREEQC using the input file:

# Define column... # right hand: change required for 40 cell column

SOLUTION 1-10 # SOLUTION 1-40
Na 1; N(5) 1
EXCHANGE 1-10 # EXCHANGE 1-40
X 1.1e-3; -equilibrate 1
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 277

1.4
A
1.2

1 Na
mmol/L

0.8

0.6 Cl Ca

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
Pore volumes

1.4
B
1.2

1 Na
mmol/L

0.8
0.6 Ca
Cl
0.4
0.2

0
0 0.5 1 1.5 2 2.5 3
Pore volumes

Figure 6.25. A column containing 1 mM NaNO3 solution and 1.1 mM exchangeable Na-X is eluted with a 0.6 mM
CaCl2 solution; A: without dispersion, B: with dispersion. Symbols are for a 10 cell model, lines for a 40 cell model.

# Define injected solution...

SOLUTION 0
Ca 0.6; Cl 1.2

TRANSPORT
-cells 10 # 40
-length 8e-3 # 2e-3 meter
-shifts 30 # 120
-time_step 2523 # 631 seconds
-punch 10 # 40 cell number that is graphed
-flow_direction forward
-boundary_conditions flux flux
-diffusion_coefficient 0.0e-9 # effective D_e, m2/s
-dispersivity 0 # meter

USER_GRAPH
-headings PV Na Cl Ca
-plot_concentration_vs time
-start
10 graph_x (step_no + 0.5) / cell_no
20 graph_y tot("Na")*1e3, tot("Cl")*1e3, tot("Ca")*1e3
-end
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

278 Ion exchange

In the PHREEQC file, SOLUTION 1–10 and EXCHANGE 1–10 define solution and exchanger for
10 cells in the column. SOLUTION 0 is the injected solution. Keyword TRANSPORT defines the
column discretization and the transport steps. The column consists of 10 cells, each with length
x  8 mm. Solution 0 is transported 30 times (“shifts”) to the next cell. A total of 30 shifts for 10
cells entails that 3 pore volumes are injected. The time step for one shift is t  2523 seconds
and is the same for all cells. The pore water flow velocity is v  x / t  8  103 / 2523 
3.17106 m/s. To obtain the concentrations in the effluent of the column, the last cell 10 is
punched, “-punch 10”. The boundary conditions for the column end are of the “flux” type where
a given mass enters the column per unit time, without a diffusive contribution (cf. Section 3.6.1).
This boundary condition applies to laboratory columns where the in- and outlet tubings have a
much smaller diameter than the column and the concentration gradient at the end of the column
is therefore zero. The dispersion coefficient, DL  De  v, consists of the effective diffusion
coefficient, De, and the product of flow velocity and dispersivity, , and both are zero in the
input file.
PHREEQC calculates concentrations midway in each cell (cell-centered). Since the midpoint
concentration arrives half a shift later at the cell boundary, the pore volume eluted (or injected) is the
number of (shifts  0.5) divided by the number of cells in the column:
PV  (step_no  0.5)/cell_no
Here “step_no” is a special BASIC word for shift number and “cell_no” stands for the cell number
that is processed. In this case only the last cell is graphed (“punch 10” in TRANSPORT).
The results of this input file are displayed in Figure 6.25A and comply with our manual
predictions. Also PHREEQC calculates Ca2 breakthrough at 1.917 PV. The best check on the
behavior of a numerical model is a comparison with an analytical solution of the problem. Another,
easier, but approximate check is to refine the grid in the model and see if the results remain the same.
We can change the number of cells to 40 (as indicated on the right side in the input file), with a cell
length of 2 mm and 120 shifts, and now punch cell 40. The results of the 40 cell model are identical
to the 10 cell model, and clearly, the model behaves well. When we include a dispersivity of
2  103 m in the input file, Cl breakthrough becomes sigmoidal (Figure 6.25B), but Ca2 break-
through is not much affected because the sharpening front counteracts dispersion.
Next, we add 0.2 mM KNO3 to the column solution of 1 mM NaNO3. With KNa\K  0.2 in the data-
base, the concentrations of Na-X and K-X become the same. Again, we expect Cl to arrive after
one PV at the end of the column, while Ca2 is retarded because it displaces Na and K. In the mean-
time, the effluent maintains a Na / K ratio of 5 in equilibrium with the initial exchanger, while on the
exchanger the Na / K ratio is 1. Therefore, the exchanger is emptied of Na before K. The elution
of Na ends with the retardation R1  1  (0.55  0) / (1  0)  1.55 PV. Next the K concentration
increases to 1.2 mM to compensate the anion charge. When K is the only cation in solution, K is also
the only ion on the exchanger and K-X increases to the exchange capacity of 1.1 mM. K is depleted
with the retardation R2  1  (1.1  0) / (1.2  0)  1.917 PV. Run the file, after having added
graphic output for K on line 20 of USER_GRAPH, and observe the result (Figure 6.26).
On average, the modeled front positions agree well with the calculated retardations. After the
depletion of Na, K increases but does not attain 1.2 mM because dispersion smears the front and
K-X is exhausted beforehand. The K / Ca2 front is steep compared to the Na / K transition
since the slope of the Ca2 / K isotherm for the final concentration is much smaller than of
K / Na. When comparing the 10 and 40 cell model results, the modeled Na / K transition is
seen to be more affected by coarse discretization than the K / Ca2 front. A coarse discretization
may lead to numerical dispersion, particularly for an indifferent front, but this can be checked by
refining the grid. A sharpening front will counteract numerical dispersion, while a broadening front
will overrule the numerical dispersion.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 279

1.4

1.2 Cl
1 Na
K
mmol/L
0.8

0.6 Ca

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
Pore volumes

Figure 6.26. A column with 1 mM NaNO3 and 0.2 mM KNO3 in solution, and 0.55 mM Na-X and 0.55 mM
K-X is eluted with a 0.6 mM CaCl2 solution. Points are for a model with 10 cells, lines for 40 cells.

QUESTION:
Rerun the file with 1.198 mM Na and 1 M Mg2 instead of K. In this case the amounts of Na-X and
Mg-X2 are approximately equal. Discuss the shape of the Mg2 / Na and Ca2 / Mg2 fronts.

EXAMPLE 6.11. PHREEQC model for Valocchi’s field injection experiment

Table 6.7 gives the water compositions of Valocchi’s experiment discussed earlier (Figure 6.16). Model the
data for well S23 at 16 m from the injection well. The pore volume V0  295 m3 (a conservative tracer
arrives at S23 when 295 m3 have been injected), the dispersivity is 1 m.

Table 6.7. Water compositions in the fresh water injection experiment (Valocchi et al., 1981).
KNa\i is for the reaction 1/i I-Xi  Na ↔ 1/i Ii  Na-X. Concentrations in mmol/L.

Na Mg2 Ca2 Cl X

Native 86.5 18.2 11.1 160 750

Injected 9.4 0.5 2.13 14.7
KNa\i 1.0 0.54 0.41

ANSWER:
In the input file we first define the initial aquifer with exchangeable cations. Since the exchange coefficients
used by Valocchi are slightly different from those in Table 6.4, the log K values are redefined under
EXCHANGE_SPECIES following Equations (6.23–6.25). Next, the injected solution and the transport
parameters are defined. We use three cells, with well S23 located at the end of the third cell, but add a fourth
cell to mimick an “infinite” aquifer. Each cell has the same volume, and because we have radial flow,
the lengths decrease with the square root of distance. The first cell has the length:

length(l)  total_length / ntot (6.39)

and successive cells have

length(n)  length(l) ( n  (n  1) ) (6.40)

–
In our case, length(1)  16 /
3  9.24 m.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

280 Ion exchange

SOLUTION 1-4 Well S23 # define aquifer...

Na 86.5; Mg 18.2; Ca 11.2
Cl 160.0
EXCHANGE_SPECIES
Mg+2 + 2X- = MgX2; -log_k 0.54; -gamma 5.5 0.2
Ca+2 + 2X- = CaX2; -log_k 0.78; -gamma 5.0 0.165
EXCHANGE 1-4
X 0.7500; -equilibrate 1
PRINT; -reset false # ... note to reduce printout
END

SOLUTION 0 # define injected solution and xpt...

Na 9.4; Mg 0.5; Ca 2.13
Cl 14.66
TRANSPORT
-cells 4
-shifts 400
-length 9.24 3.83 2.94 2.48
-dispersivity 1.0
-punch 3
USER_GRAPH
-headings Injected Na Cl Mg Ca
-axis_scale x_axis 100 100000 auto auto true # use logarithmic axes
-axis_scale y_axis 0.1 160 auto auto true
-plot_concentration_vs time
-start
10 V0 = 295 # m3 from injection well to S23
20 graph_x V0 * (step_no + 0.5)/cell_no
30 graph_y 1e3 * tot("Na"), 1e3 * tot("Cl"), 1e3 * tot("Mg"), 1e3 * tot("Ca")
-end
END

The model results are shown in Figure 6.27 and match the concentrations observed by Valocchi et al. well.
Note that the length of the first three cells of the model sum to 16 m. The role of the fourth cell is to allow
dispersion beyond this point, as happens in the aquifer. When the grid is refined twice, the fronts sharpen
only slightly which indicates that numerical dispersion is negligible. However, now two extra cells are
needed to describe dispersion beyond the model path correctly, the total length of the additional cells should
at least sum up to the dispersivity.

1000

100
mmol/L

Cl
10 Na

Ca
1
Mg

0.1
100 1000 10000 100000
m3 injected

Figure 6.27. Model lines and observed concentrations for the Valocchi case.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chromatography of cation exchange 281

QUESTIONS:
Refine the grid twice.
Include equilibrium with Calcite and a CO2 pressure of 102 atm in the model; make sure to adapt pH and
alkalinity in solution 0 and the aquifer solutions so that the Ca2 concentration remains equal.
Invert the quality pattern, i.e. inject brackish water in a fresh aquifer.

Lastly in this section, we compare a model of seawater intrusion by advective transport and by dif-
fusion. The following PHREEQC input file models 20 years advective transport of a twice diluted
seawater into a fresh aquifer with a CEC of 0.1 eq/L, and calcite and CO2 as mineral buffers.
SOLUTION 1-30 # Fresh pore water and sediment...
pH 7.27
Na 0.485; Mg 0.8; Ca 2.0
Cl 0.566; C(4) 3.5
EQUILIBRIUM_PHASES 1-30
Calcite; CO2(g) -2.0
SAVE solution 1-30
PRINT; -reset false
END # Note the END (see text)
EXCHANGE 1-30
X 0.1; -equilibrate 1
END

SOLUTION 0 # 1/3 seawater enters the aquifer...

pH 8.3
Na 162; Mg 18.4; Ca 3.6
Cl 189; Alkalinity 2.4
TRANSPORT # 20 yr flow, 5 m/yr, in 150 m...
-cells 30; -length 5
-time_step 3.15e7
-flow_direction forward; -shifts 20
-dispersivity 1; -punch_frequency 20
USER_GRAPH
-heading dist Na Mg*10 Ca Cl Alk*10
-init false # -init is shorthand for initial_solutions
-plot_concentration_vs x
-start
10 graph_x dist
20 graph_y tot("Na")*1e3, tot("Mg")*1e4, tot("Ca")*1e3,\
tot("Cl")*1e3, Alk*1e4 # Note line continuation with \
-end
END

First, the pore water is brought to equilibrium with calcite and a [PCO2]  102. The solution is saved
and following “END”, the next simulation equilibrates the exchanger with the solution. The order of
the calculations is important here. If END were left out, then the solution would first equilibrate with
the exchanger, after which the mineral reactions would be calculated. The exchanger would then pro-
vide buffering. CO2 is included to balance the pH decrease during calcite precipitation. Normally,
there are proton sources and sinks by sorption on mineral surfaces and organic matter, which can be
included in the model with SURFACE, as will be discussed in Chapter 7.
The profile for seawater intrusion in Figure 6.28 displays overall the same pattern as Figure 6.20.
As seawater intrudes, Na is taken up by the exchanger, releasing Ca2 and Mg2 which display
maxima due to the salinity effect. Calcite precipitation is a secondary result and reflected by a
decrease in the alkalinity.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

282 Ion exchange

200
180 Cl
160
140 Na
120 Mg  10

mmol/L
100
80
60 Alk  10
40
Ca
20
0
0 50 100 150
Distance (m)

Figure 6.28. PHREEQC simulation of salt water intrusion by advective transport into a fresh water aquifer.

To model diffusion with PHREEQC, only the TRANSPORT keyword needs to be adapted in the pre-
vious file.
TRANSPORT
-cells 30; -length 0.2
-time_step 3.15e9;
-flow_direction diffusion; -shifts 1
-boundary_conditions constant closed
-diffusion_coefficient 0.3e-9

The model is now for 100 years diffusion into 6 m sediment. The “flow_direction” is diffusion, and
the boundary conditions are “constant” for constant concentrations at the first cell of the column and
“closed” at the last cell. Shifts define now the number of time steps that are calculated for diffusion
(here 1  3.15109 give 100 years total time). Figure 6.29 shows the results of the modified file,
with concentration patterns that follow more the trace of Cl than in the previous case. For example,
the Mg2 peak is completely absent.
When seawater diffuses into a stagnant fresh groundwater, part of the exchanged ions from the
sediment diffuse away at the same rate as the displacing ions arrive. The difference in diffusion flux
of a conservative element and an exchangeable element is partially determined by the CEC. When
the CEC is not too high, the effect of cation exchange on the water composition is modest and less
conspicuous than when seawater intrudes the aquifer by flow. In case of diffusive transport the
resulting water compositions will resemble those of a simple mixture of fresh and salt water.

200
180
160 Cl
140
120
mmol/L

100
80 Na
60
40 Alk  10

20 Ca Mg  10
0
0 1 2 3 4 5 6
Distance (m)

Figure 6.29. PHREEQC simulation of salt water intrusion by diffusion (initial concentrations are identical to
Figure 6.28).
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Physical non-equilibrium 283

6.5 PHYSICAL NON-EQUILIBRIUM

Flow conditions in a porous medium are often inhomogeneous with water flowing in certain parts
of the medium and being stagnant within particle aggregates, cemented clusters or organic clods.
A diffusive exchange of ions will take place between the stagnant zones and the mobile flow region.
Figure 6.30 visualizes the situation. First, there is film diffusion through a boundary layer of stagnant
water that surrounds the particles. Next intraparticle diffusion may occur inside particle aggregates
which is a much slower process. Lastly, there are the chemical reaction rates which vary from instan-
taneous for ion exchange to slow for silicate mineral reactions.

1. Bulk (fast) 1
Transport and reaction

2. Film (slow) 2

3. Particle
Processes

(slow) 4
4 3b
3a
4. Chemical
reaction
4

Figure 6.30. Rate determining processes for reactions in a porous medium (Weber and Smith, 1987).

When diffusion into the stagnant zones is unable to equalize the concentration gradients brought
about by advective flow, the system is in physical non-equilibrium. An estimate of the pore size that
can be homogenized by diffusion is obtained by using the formula (Section 3.5.2):
 2  2De t (3.60)

where 2 is the variance of the diffusion (Gaussian) curve, De is the effective diffusion coefficient
and t is time. The distance covered by diffusion can be compared with the distance traveled by advec-
tive flow during the same period of time, as illustrated in Figure 6.31. The lines are drawn for the
equality   x, where  
 2 Det (diffusion) and x  vt (advective transport).

ag
Region attained by diffusion (mm)

gr
eg
10 ph ate
di ys s
ffu
sio ica too
n l n la
ca on rg
n
f -e e f
eq ill s qu or
ui m ilib dif
lib al riu fus
1 riu l a m io
m gg n:
re
ga
te
s:

De  1011 m2/s De  109 m2/s

0.1
1 10 100 1000
Flow velocity (m/yr)

Figure 6.31. Approximate size of stagnant zones that can be homogenized by diffusion as a function of the
pore water flow velocity for two diffusion coefficients.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

284 Ion exchange

For example, when De  109 m2/s,   1 mm enclosing 68% of the chemical, is reached in
t  (103)2 / 2  109  500 s. The distance of 1 mm is covered by advective flow in the same
time when v  x / t  103 / 500  2  106 m/s, or 63 m/yr. Clearly, the size of a stagnant zone
that will be homogenized with a normal groundwater flow velocity of around 20–200 m/yr is small.
For a homogeneous sand the size of the stagnant zone corresponds to the pore size, which is about
equal to the particle size below which 10% of the particles exist (Perkins and Johnston, 1963).
The size of stagnant zones is also about equal to the dispersivity and an idea of flow velocity required
for full equilibrium in columns can be obtained from the dispersion of the breakthrough curve of a
conservative tracer. Figure 6.31 shows that column dispersivities of a few mm require flow veloci-
ties below about 50 m/yr. Field dispersivities of several meters, on the other hand, indicate stagnant
zones that will never attain full equilibrium when the concentrations vary in water that flows
around them.
A mobile fraction of the porosity can be defined as

m  em / ew (6.39)

where w is the total water filled pore volume and m is the mobile part.
Figure 6.32 shows how the fraction of mobile water may affect the shape of tritium breakthrough
curves. In the upper graph the aggregates are smaller than 2 mm and for the applied flow velocity, all of
the pore water is mobile. The lower graph shows results for a column packed with aggregates up to 6 mm.

1.0
3H O
0.8
2 2 mm

0.6
c/c0

T0  0.512
0.4
Φm  0.94
T0  2.33
0.2

0.8 6 mm

0.6
c/c0

0.4 Φm  0.53
T0  0.512
T0  3.16
0.2

0
0 1 2 3 4 5 6
T

Figure 6.32. Calculated and observed effluent curves for tritium movement through a clay loam for small
aggregates 2 mm (upper) and larger aggregates 6 mm (lower graph). T  number of pore volumes; T0 is
injected volume of tritiated water or the change to unlabeled water (Van Genuchten and Cleary, 1982).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Physical non-equilibrium 285

In this case the fraction of mobile water m is 0.53, as obtained by curve-fitting the concentrations in the
column effluent. The tailing of the effluent concentrations in the lower graph is due to diffusion into or
out of the dead-end pores. The breakthrough of tritium also occurs earlier (Figure 6.32) because the
velocity of the liquid in the mobile region increases by a factor (m)1 when the volume of the stagnant
zone is inaccessible.

6.5.1 Modeling stagnant zones

We assume that the mobile and immobile zones consist of two homogeneous boxes which exchange
mass depending upon the concentration difference:
dM im / dt  Vim Rim dcim / dt   (cm  cim ) (6.40)

where the subscript m indicates mobile and im indicates immobile area. Mim are moles of chemical
in the immobile zone, Vim is the volume of the immobile part (m3), Rim is the retardation
(Rim  1  (dq / dc)im), cm is the concentration in the mobile area, and cim the concentration in the
immobile part (both in mol/L), and  is the exchange factor or mass transfer coefficient (s1). The
concentration cim is an average for the entire stagnant zone, even though in reality the concentration
can be expected to vary with distance from the interface.
A mass increase in the immobile zone is balanced by mass decrease in the mobile zone and vice
versa:

M im  Vim Rim cim  M m  Vm Rm cm (6.41)

Thus, the concentration changes depend on the relative size of the boxes (Vm, Vim), the concentration
changes in the solid part (qm, qim in the retardation equation), and the exchange factor . Instead
of the volumes Vm and Vim, the porosities m and im as fractions of the total volume may be used in
the formulas.
We can integrate Equation (6.40) with the initial concentrations cm  cm0 and cim  cim0 at t  0,
and the mass balance (6.41) to obtain:
cim  im f cm0  (1  im f ) cim0 (6.42a)

where
im  Rm em / (Rm em  Rim eim ) (6.43a)

and

 t 
f  1  exp   (6.44)
 im Rim eim 

The concentration in the mobile box is found similarly:

cm  m f cim0  (1  m f ) cm0 (6.42b)

with

m  Rim eim / (Rm em  Rim eim ) (6.43b)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

286 Ion exchange

Since  is constant and f is also constant for a given simulation time, Equation (6.42) describes a mix-
ing process in which a fraction of the mobile cell mixes into the immobile cell and vice versa. For large
 and t (or large distance), the exponential term in (6.44) becomes negligible, f → 1, and mixing is
complete. In that case, the solute concentrations in the two boxes become the same, as you may check.
The exchange factor  is a function of the contact surface of the two boxes and can be related to
the shape of the immobile zone (Van Genuchten, 1985):
De eim
 (6.45)
(af s →1 )2

where De is the diffusion coefficient in the stagnant zone, a is the size of the immobile area (an effec-
tive diffusion length), and fs→1 is a shape factor. The shape factors given by Van Genuchten (1985)
are listed in the PHREEQC manual (Table 1, p. 53). For example for spheres, a is the radius of the
sphere, and fs→1  0.21.
The effects of stagnant zones can be calculated with PHREEQC, using the option “-stagnant” in
keyword TRANSPORT. A two-box model can be simulated by adding a stagnant cell to each mobile
cell in the column. Let us calculate again the column in which 1.1 mM Na-X is exchanged by inject-
ing 0.6 mM CaCl2 solution, but now we increase the flow velocity 50 times to 1.59104 m/s
(5 km/yr). We subdivide the porosity in a mobile (m  0.2) and a stagnant part (im  0.1). The
stagnant part consists of exchanger beads with a diameter 2a  2 mm (equal to the dispersivity) and
a diffusion coefficient De  0.3109 m2/s. With the shape factor for spheres, fs→1  0.21, the
exchange factor   6.8104/s. We compare the outflow profiles when the exchange sites are dis-
tributed homogeneously over the mobile and immobile cells and when the exchanger is confined to
the immobile parts of the column.
SOLUTION 1–81 # mobile 1-40, immobile 42-81
Na 1
EXCHANGE 1-81 # Distribute exchanger homogeneously...
X 1.1e-3; -equilibrate 1
# Distribute X heterogeneously, _im = 0.1 _m = 0.2
# All X in mobile, X_m = X * (_im + _m)/_m...
#EXCHANGE 1-40
# X 1.65e-3; -equil 1
# All X in immobile, X_im = X * (_im + _m)/_im ...
#EXCHANGE 42-81
# X 3.3e-3; -equil 1
END

SOLUTION 0; Ca 0.6; Cl 1.2

TRANSPORT
-cells 40; -length 2e-3; -shifts 176
-time_step 12.6 # note smaller timestep: higher flow velocity
-punch 40
# 1 stagnant layer,  = 6.8e-4/s, _m = 0.2, _im = 0.1
-stagnant 1 6.8e-4 0.2 0.1
-disp 0.002
USER_GRAPH
-head PV Na Cl Ca
-plot t
-start
10 graph_x (step_no + 0.5) / (cell_no * 1.5)
20 graph_y tot(“Na”)*1e3, tot(“Cl”)*1e3, tot(“Ca”)*1e3
-end
PRINT
-reset false
END
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Physical non-equilibrium 287

1.4
A
1.2

1 Na
mmol/L
0.8
Cl
0.6 Ca
0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
Pore volumes

1.4
B
1.2
1
mmol/L

0.8
0.6 Na
Cl Ca
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3
Pore volumes

Figure 6.33. A column with stagnant zones and exchangeable Na-X is displaced with a CaCl2 solution. In
6.33A, the thin lines are the concentrations for a column without stagnant zones (from Figure 6.25B), the thick
lines are for Na-X distributed homogeneously over the mobile and immobile zones. In 6.33B, the exchange sites
are confined to the immobile zone.

In the input file, the mobile cells are numbered 1–40, and the immobile cells are 42–81. Immobile
cell 42 mixes with cell 1, cell 43 mixes with cell 2, etc. The exchanger can be assigned to any of these
cells and corrected for the volumetric contribution of the cell (stagnant or mobile) to maintain equal
total exchange capacity in the column (cf. the input file). Only the solutions of the mobile cells are
displaced, so that the number of shifts must be increased to obtain displacement of the total pore vol-
ume of mobile plus immobile water. For this same reason, the formula for the x-axis in the plot is
modified since the total pore volume is 1.5 times larger than the mobile volume.
Run the file, and compare the results shown in Figure 6.33A with the column without stagnant
zones (Figure 6.25B). The Cl curve has become more disperse because the ion enters the stagnant
zones which initially are free of Cl. The Ca2 / Na front arrives earlier because only 2⁄3 of the total
cation exchange capacity is readily accessible. Next we confine the exchange sites to the stagnant
zone only (Figure 6.33B). The cations follow now a more gradual curve. A small part of Ca2 arrives
together with Cl at the end of the column because exchange with the stagnant zone is too slow to
take away all the Ca2 from the mobile water.

QUESTIONS:
Will the pattern for Ca2 / Na be different when exchange is limited to the mobile zone?
What happens with the Cl concentrations if the flow velocity is doubled to 10 km/yr? And how are the
Ca2 / Na concentrations affected? (Run the files)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

288 Ion exchange

0.8

0.6
c /c0

0.4

0.2

0
0 2 4 6 8 10 12
Pore volumes

Figure 6.34. Effects of non-equilibrium during flow-interruption. The flow was interrupted after 4 pore
volumes (Brusseau et al., 1989).

Dispersion, and chemical and physical non-equilibrium may produce similar effluent curves, and
therefore experiments can often be simulated equally well with different models (Nkedi-Kizza et al.
1984). For example, the Cl breakthrough in Figure 6.33 for the column with stagnant zones, can be
matched almost perfectly by assuming a homogeneous column and a dispersivity of 8 mm (instead of
2 mm). Similarly in Figure 6.32, the tritium curve in the column with stagnant aggregates of 6 mm,
can be simulated by increasing the dispersivity to 6 mm. On the other hand, the effluent curves for
exchanging or sorbing species are more unique and less easily fitted by changing the dispersivity.
The effect of non-equilibrium in column experiments or aquifer clean-up schemes can be
assessed by stop-flow or flow-interruption (Brusseau et al., 1989; Koch and Flühler, 1993). In
Figure 6.34 the flow was stopped after 4 pore volumes. When flow was reactivated, the concentra-
tion had decreased because of diffusion into the immobile zone. Similarly during decreasing con-
centrations, when the chemical seems leached out, a temporary flow stop may cause an increase in
concentration when the flow is restarted because of diffusion out of the immobile zone. During
aquifer clean-up an increase in concentration after pumping is temporarily halted may be due to this
effect (Bahr, 1989).

QUESTIONS:
Consider the effect of dispersion on Cl breakthrough. Make the column of Figure 6.33 homogeneous and
change the dispersivity to 8 mm.
Estimate the equilibration time for the column of Figure 6.33. (Hint: in Equation 6.44, make t /
(imRimim) ! 4.6)
ANSWER: t ! 406 s.
Calculate the effect of flow-interruption for the experiment depicted in Figure 6.33B. Stop the flow for the
estimated equilibration time after 1.5 PV’s have been injected, then resume the flow. repeat the experiment, but
allow for a 10 times longer equilibration time. Is the estimated equilibration time correct?

6.6 THE GOUY-CHAPMAN THEORY OF THE DOUBLE LAYER

We have considered macroscopic equations that relate concentrations in solution to those on the
exchanger surface. But if we want to describe what occurs mechanistically when a charged surface

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The gouy-chapman theory of the double layer 289

attracts or repels ions from a solution, we must go down to the molecular level (Moore, 1972; Feynman
et al., 1989). In our case, we could consider the cations and anions as point charges which can move in
an electric field. A bulk solution in which the number of cations and anions are equal, has zero electric
potential. A negatively charged surface has a negative potential that attracts positively charged ions and
repels negative ones. However, diffusion will smear out concentration-steps, and cause the transitions
to become more gradual. The net effect (Figure 6.35) is a gradual change in concentrations in the so-
called Gouy-Chapman diffuse double layer (DDL) which surrounds the charged surfaces in a solution.
The electric potential will not change abruptly at the solid-solution interface, but gradually
lessen as distance from the interface increases. The relative concentrations in the electric field are
given by Boltzmann’s law:

(
mi, a / mi, b  exp zi Fa / RT ) / exp (zi Fb / RT ) (6.46)

where mi is the concentration of ion i (kmol/m3; formally the activity should be used) with charge
number zi, F is Faraday’s constant (C/mol),  is potential (Volt), R is the gas constant (J/mol/K), and
T absolute temperature (K). Subscripts a and b indicate the position in the electric field.
In the electrically neutral solution at infinite distance from the surface, the potential %  0.
Therefore:

(
mi / mi,∞  exp zi F / RT ) (6.47)

Any concentration difference between anions and cations will create a charge density  (C/m3):

  F  zi mi  F  zi mi,∞ exp zi F / RT ( ) (6.48)

Clay surface

0 X 0 X 0 X

25

20 a b c
Concentration

15 15 15

10 10 10

5 5 5

0 0 0
0 X 0 X 0 X

Figure 6.35. Distribution of counterions. a) Minimal energy and maximal electrostatic attraction; b) maximal
concentration distribution and entropy; c) compromise: the distribution in a diffuse double layer (Bolt and
Bruggenwert, 1978).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

290 Ion exchange

Let us calculate, for a planar surface, how the concentrations change in the x-direction. The electric
field strength changes with the charge density as:
dE / dx  x / e (6.49)

where  is the dielectric constant of water, 6.951010 F/m at 25°C.

A stronger electric field strength conforms to a steeper potential gradient:
E  d / dx (6.50)

Combining the three equations gives the Poisson-Boltzmann equation:

d 2  / dx 2  (F / e) zi mi,∞ exp(zi F / RT ) (6.51)

This equation can be integrated once (first multiply both sides with 2(d / dx), then integrate over x):

(
(d / dx )2  2(RT / e) mi,∞ exp zi F / RT  constant ) (6.52)

In the free solution, d/ dx  0 and   0, and the constant is:

constant  2(RT / e) mi,∞ (6.53)

Therefore, the fundamental double layer equation for a planar surface is:

(d / dx )2  2(RT / e) mi,∞ {exp (zi F / RT )  1} (6.54)

The next step is to integrate once more for simplified conditions, or to use a numerical integrator.
For a simple mono-monovalent solution with zi  1, the double layer equation is:

(d / dx )2  2(RT / e) mi,∞ {exp (F / RT )  exp ( F / RT )  2} (6.55)

or

{
d / dx  (2(RT / e) mi,∞ ) 2 cosh(F / RT )  2 } (6.56)

or

d / dx  (8(RT / e) mi,∞ ) sinh(F / 2RT ) (6.57)

With   0 at x  0, the latter integrates to:

1   bax  exp(b0 )  1   
 ln  tanh   1⁄2 ln   (6.58)
b   2  exp(b0 )  1   

where b  F / (2RT) and a 

 8(RT / ) 
m i,%.
The potential can now be calculated for any x, and subsequently the concentrations of the ions in
the double layer follow from the Boltzmann equation. An example is provided in Figure 6.36. At the
negatively charged surface, the cations are counter-ions and attracted to the surface, while the anions
are co-ions and repelled from the surface. Accordingly, the cations increase in concentration towards
the surface, while the anions decrease in concentration. The surface charge is balanced both by the
counter-ion excess, and by the co-ion deficit or co-ion exclusion.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The gouy-chapman theory of the double layer 291

100

101 C  0.1 N

Concentration (N)
Cations C  0.01
102

Anions
3 C  0.001
10

104
0 40 80 120
Distance from pore wall (Å)

Figure 6.36. Distribution of cations (full lines) and anions (dotted lines) in the double layer on a negatively
charged surface at 3 concentration levels in the free solution (Nielsen et al., 1986. Copyright by the Am.
Geophys. Union).

The potential at the surface can also be calculated simply for a mono-monovalent solution. The
charge of the double layer, DL, is equal and opposite to the charge of the surface:

 DL  F  D / As  F  / As , (6.59)

where DL is in C/m2 and D and  indicate the charge in moles of the double layer and on the sur-
face, respectively.
The double layer charge equals the integrated charge density:
∞ ∞
d 2
 DL  ∫0  dx  e ∫ dx 2
dx (6.60)
0

After integration we obtain:

 DL   e(d / dx ) x0 (6.61)

With (6.57) and   0 at x  0 we obtain the Gouy-Chapman double layer equation:

 DL  (8(RT / e) mi,∞ ) sinh(F0 / 2RT ) (6.62)

The above derivation is for a mono-monovalent solution, but conventionally, the equation is applied
also for an equivalent ionic strength I:
I  1⁄2 mi z 2i (6.63)

Thus,

 DL  0.1174 I 0.5 sinh(F0 / 2RT ) (6.64)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

292 Ion exchange

EXAMPLE 6.12. Calculate the surface potential for 1.07106 eq/m2 surface charge on montmorillonite
in a 0.1 M NaCl and in a 0.025 M MgSO4 solution.
Use the identities sinh(x)  (ex  ex) / 2 and arcsinh(y)  ln(y  (y2  1) ⁄2).
1

ANSWER:
For both solutions, I  0.1 mol/L. Thus, F   964851.07106  0.103 C/m2. Divide by 0.117 
(0.1)0.5 and find 2.78. Hence F0 / 2RT  ln(2.78  (2.782  1)0.5)  1.75, and 0  89.7 mV.
However, for a di-divalent solution, the correct equation is (note the disappearance of the factor 2):
 DL  0.117 mi0.5
,∞ sinh(F0 / RT ) (6.65)

For the MgSO4 solution we divide 0.103 C/m2 by 0.117  (0.025)0.5, and find 5.56. Hence, F0 / RT  2.42,
and 0  62.1 mV.

For small arguments, exp(x)
(1  x). Therefore, when   25 mV, sinh(F / 2RT)
F / 2RT,
and we obtain the simple formula (at 25°C):
 DL  2.29 I 0.5 0 (6.66)

Already at an earlier stage, the approximation for small  can be applied to the Poisson-Boltzmann
Equation (6.52) for a homovalent solution:
d 2  / dx 2  2(F 2 /eRT ) zi2 mi,∞  (6.67)

which can be simplified using

2  2(F 2 /eRT ) zi2 mi, % (6.68)

to:

d 2  / dx 2  2  (6.69)

This equation is much easier to integrate, and yields (check by differentiating!):

  0 exp(x ). (6.70)

In essence the potential decays exponentially from the charged surface, and so will the extent of a
diffuse double layer, given the exponential concentration changes of cations and anions.
The units of 2 are m2, as can be verified from Equation (6.68). The parameter 1 is termed
the Debye length; this parameter also appears in the Debye-Hückel theory for calculating ion activ-
ity coefficients. The ionic strength can be incorporated in the Debye length to give:

e RT 1
1  (6.71)
2(N a qe )2 I

With   6.951010 F/m, R  8.314 J/K/mol, T  298 K, Na  6.0221023/mol, qe  1.61019C,

and I expressed in mol/dm3, Equation (6.71) becomes:

1  3.09/ I (Å) (6.72)

It is customary to consider the point where   0 / e, i.e. where x  1, as the thickness of the
DDL. The influence of the ionic strength on the thickness of the DDL is shown in Figure 6.37.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The gouy-chapman theory of the double layer 293

“Thickness” DL (ψ  ψ0&e)
120

80

40

Fresh water
0
3 2 1
log l

Figure 6.37. The extension of the double layer (“thickness”) as a function of the ionic strength I.

6.6.1. Numerical integration of the double layer equations

The double layer equations have been used to estimate selectivities among mono- and divalent
ions (Eriksson, 1952; Bolt, 1967). Equation (6.55) can also be integrated numerically, which
allows to include the closest approach distance of the ions to the surface (Neal and Cooper, 1983).
Ions with a large hydration shell (given by the ion-size parameter in the Debye-Hückel equation,
Table 4.2) cannot come close to the surface and will thus exhibit a relatively small selectivity. In the
following we use the solver for ordinary differential equations (ode’s) provided in Matlab.
The instructions for Matlab are written in the two socalled “m” files given in Table 6.8. The cal-
culation is for a system with Na, Ca2 and Cl. In the first m-file some basic definitions of the
physical constants, Debye-Hückel parameters and concentrations in the free solution are provided.
Subsequently, the solver is called with:
[ psi, x]  ode45(“x_psidot”, 0 ,  x , x0 , tolerance);

Matlab will give (x) in the form of two vectors from 0 at x0 up to % at x. In the m-file in Table 6.8,
0 is 0.1 V, x0 is 3.5 Å (which is the Debye-Hückel å parameter for Cl, the smallest of the three
ions), % is 11011, and x is obtained from the integration.
The statement “x_psidot” in ode45 invokes the m-file with the differential equation. In this file,
the sum of the ion-concentrations is calculated (i.e. the right hand side of Equation (6.55)) and returned
in inverted form, (dx / d). Thus, the integration is along , and the tolerance of 11017 applies to
x (m). We could integrate along x, going piecewise from the smallest to the largest Debye-Hückel
value and account explicitly for the singularities at the closest ion-approach distances. However, we
do not know exactly the distance where the potential is so small that it can be considered zero
(
11011) which may cause the integration to fluctuate if it is continued too far along x. The
Runge-Kutta procedure of ode45 is quite stable and accounts for the singularities by a step-size
decrease at those points.
After integration of the double layer equation, the program calculates the concentrations as a
function of x with the Boltzmann equation, and plots them in a graph similar to Figure 6.38. The
example calculation is for a solution with 8 mM Na, 1 mM Ca2, and 10 mM Cl. Several things
can be noted in Figure 6.38. First, the cations increase very rapidly towards the surface. For exam-
ple, the Ca2 concentration at 5 Å from the surface (1.5 Å from the point where the potential
is 0.1 V) is about 1.6 mol/L. Second, the amount of Ca2 in the double layer is much larger than of
Na, despite the equivalent ratio of 1 : 4 in solution. Third, the concentration of Na descends more
quickly when Ca2 enters the double layer, as can be seen in the break of the Na curve at 5 Å.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

294 Ion exchange

0.8 Ca/2

0.6
mol/L

0.4 Na

0.2
Cl  30

0
0 0.5 1 1.5 2 2.5 3
Distance from charged surface (nm)

Figure 6.38. Concentrations of Na, Ca2 and Cl in a double layer in which the approach to the charged sur-
face is limited by the Debye-Hückel å parameter.

The high proportion of Ca2 is due to the quadratic weight of charge in the Boltzmann term (exp(z)).
The increased descent of the Na concentration is related to a more rapid decrease in the charge
density when the divalent Ca2 ion is also present.

Table 6.8. Matlab program to calculate concentrations in the double layer at a charged surface, with varying
closest approach distance for different ions. Note: the % sign indicates a comment.

file: dl.m
function dl % (% = Comment) integrates double layer

% Uses : x_psidot.m

global Na Ca Cl Na_0 Ca_0 Cl_0 dha_Na dha_Ca dha_Cl;

global RT F epw x psi;

% J/kmol C/kmol e_w e_0

RT = 2477.57e3; F = 96485.e3; epw = 78.5 * 8.854e-12;

% Closest distance to surface = Debye-Huckel A (m) ...

dha_Na = 4.0e-10; dha_Ca = 5.0e-10; dha_Cl = 3.5e-10;

% conc’s in kmol/m3 at x = inf, psi = 0 ...

Na_0 = 0.008; Ca_0 = 0.001; Cl_0 = 0.01;

% Call Matlab ODE solver ...

% psi_0 psi_end x_0 tol
[psi, x] = ode45(‘x_psidot’, -0.100, -1.e-11, dha_Cl, 1e-17);

% Calculate c = c_0 * exp(-zFpsi/RT) ...

lx = length(x); Na = zeros(lx,1); Ca = zeros(lx,1); Cl = zeros(lx,1);
d = exp(-F*psi/RT);
Na = Na_0 * d; Ca = Ca_0 * d.*d; Cl = Cl_0 ./ d;
% c = 0 for x < DH A param ...
for i = 1:lx, if x(i) < 5.1e-10
if x(i) < dha_Na, Na(i) = 0; end; if x(i) < dha_Ca, Ca(i) = 0; end;
if x(i) < dha_Cl, Cl(i) = 0; end, else break, end, end;

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The gouy-chapman theory of the double layer 295

Table 6.8. (Contd)

figure(1);
clf; hold on; axis([0, 30e-10, 0, 1]);
plot(x, 30*Cl, 'y:'); plot(x, Na, 'm-'); plot(x, 0.5*Ca, 'c- -');

file:x_psidot.m
function ans = x_psidot(psi, x);

% derivative for ode ...

% dx/d(psi) = (2RT/epw * SUM c_i*[exp(-z_i F psi / RT) - 1])^-.5

global Na_0 Ca_0 Cl_0 dha_Na dha_Ca dha_Cl RT F epw;

d = exp(-F*psi/RT);

if x >= dha_Na, d1 = Na_0 * (d - 1) ; else d1 = 0; end;

if x >= dha_Ca, d2 = Ca_0 * (d*d - 1); else d2 = 0; end;
if x >= dha_Cl, d3 = Cl_0 * (1/d - 1); else d3 = 0; end;
dall = d1 + d2 + d3; if dall == 0, dall = 1e-21; end;

ans = (2 * RT / epw * abs(dall))^(-0.5);

The program is easily extended to more ions. Figure 6.39 shows a multicomponent example
with Na, K, Mg2 and Ca2 for the solution composition in Table 6.9. As before, the figure
shows peaks for the various ions at the closest approach distance. The traces for Ca2 / 5 and
Mg2 in this figure are equal since the concentration of Ca2 in the free solution is 5 times
higher than of Mg2. Otherwise, the relations among the approach distances, head-end poten-
tials and the relative enrichment of an ion in the double layer are not simple to generalize.
However, some feeling of the diffuse double layer’s behavior can be gained by calculating exchange
constants KNa\I and comparing with empirical cation exchange constants noted in Table 6.4.

0.5

Ca/5
0.4

0.3 Mg
mol/L

0.2

Na
0.1
K*5

0
0 0.2 0.4 0.6 0.8 1 1.5 2
Distance from charged surface (nm)

Figure 6.39. Concentrations of Na, K, Mg2 and Ca2 in the double layer, with variable closest approach
to the surface for the different ions.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

296 Ion exchange

Table 6.9. Apparent cation exchange selectivities with respect to Na from double layer accumulation at a
surface with charge 0.1 C/m2 (cf. Example 6.1). The 0.01 M Cl solution contains (in mmol/L): Na 3.7; K
0.3; Mg2 0.5; Ca2 2.5. The 0.1 M Cl solution is 10 times more concentrated. Debye-Hückel å parameters are
4, 3.5, 5, 5.5 and 3.5 Å for Na, K, Mg2, Ca2 and Cl, respectively.

Cl (mol/L) 0 (V) KNa\K KNa\Mg KNa\Ca Cl (%)

0.01 0.091 0.91 1.13 0.98 3.5

0.1 0.063 0.87 0.82 0.71 10.6
Table 6.4: 0.2 0.5 0.4

The integrated amounts in the double layer give I-X, which can be used to calculate selectivities
according to:
Na +  1/ i I -X i ↔ Na-X  1 & i I i; K Na\I

Results for cation concentrations which are near to the (present day) average Rhine water in Holland,
and for a 10 times more concentrated water, are given in Table 6.9.
The K values in Table 6.9 are all close to 1.0, both for K and the divalent cations. When con-
centrations increase in the free solution, the K value for the divalent cations decreases, indicating
more enrichment for the divalent ions. The decrease in K is less than was calculated by Bolt (1982)
with approximative analytical solutions. The relative constancy of the double layer K indicates that
exchange formulas can be used to estimate double layer contents. Furthermore, the observed
exchange constants (from Table 6.4) are significantly smaller than the estimated double layer coun-
terparts. In other words, the greater observed selectivity for ions other than Na is not captured by
the double layer theory, at least not with the standard Debye-Hückel parameters. For example, to
obtain the observed KNa\K  0.2 with double layer theory, the approach distance for K must be
diminished to zero and 0 must be lowered to 0.2 V. For the divalent ions, the observed selectivity
can be obtained by diminishing the approach distance to 0 and 0.5 Å for Ca2 and Mg2, respec-
tively. This suggests that inner-sphere bonds are contributing to ion exchange of divalent ions.
In Table 6.9 the charge compensation by anion exclusion is given as a percentage of the total
charge of 0.1 C/m2, amounting to 3 and 10% for the 0.01 and 0.1 M Cl solutions, respectively.
Actually, most of the structural charge of clay minerals such as montmorillonite is balanced in the
interlayer region, which probably contains only water and cations (Karaborni et al., 1996).
The double layer theory was developed here for a planar surface, which permits an easy first inte-
gration step. All other shapes, even spheres, require integration of the second order Poisson-
Boltzmann equation, for example with a central differences algorithm as is used for calculating
diffusion (cf. Chapter 11). Alternatively, for spherical shapes, the polynomial approximations
derived by Bartschat et al. (1992) can be used.

6.6.2 Practical aspects of double layer theory

The thickness of the double layer (Figure 6.37) indicates how far a negative potential from a clay min-
eral surface extends into solution. Clay particles are repelled from each other when the negative poten-
tials from opposite clay particles overlap. A suspension of clay particles becomes stable when the double
layer is so thick that the clay particles repel each other: the colloidal solution remains then peptized. A
thin double layer on the other hand, may induce flocculation, the suspension is separated in a clear super-
natant solution and a flocculated clay. The double layer thickness increases with decreasing ionic
strength and with decreasing charge of the counterions. If a CaSO4 solution is replaced by a NaCl solu-
tion of equal molality, the ionic strength decreases by a factor of 4, and the Debye length increases by 2.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 The gouy-chapman theory of the double layer 297

100

conductivity (percent)
0.01 N 0.01 N

Relative hydraulic
80

60
Netanya H2O
40
Nahal Oz
H2O
20

0
0 5 10 20 0 5 10 20
Sodium adsorption ratio (SAR )

Figure 6.40. Hydraulic conductivity of a sandy loam (Netanya) and a silty loam (Nahal Oz) soil as a
function of the SAR and the concentration of Nacl in the leaching solutions. From Shainberg and Oster, 1978.

Swelling of clay in the soil may therefore be associated with low ionic strength and monovalent cations
in the soil solution. Swelling of clay may induce transport of clay particles, leading to a clogging of
pores, and to a decrease of the hydraulic conductivity (Quirk and Schofield, 1955; Suarez et al., 1984).
Figure 6.40 illustrates the decrease of the hydraulic conductivity as a function of SAR, i.e. the Na&
Ca2

ratio in solution, and of total concentration (normality).

Swelling has even more outspoken effects when fresh water displaces seawater in an aquifer.
Initially, in seawater, the concentrations are so high that the clay remains flocculated, irrespective of the
high Na concentration in seawater. Displacement and dilution by fresh water will decrease the ionic
strength and increase the Na/
 Ca2 ratio in solution. Both factors promote the movement of clay par-

ticles. Goldenberg (1985) used dune sand mixed with smectite clay and found a marked decrease in
hydraulic conductivity when seawater was replaced by fresh water. In later experiments he could actu-
ally observe the migration of clay particles (Goldenberg and Mandel, 1988). Experiments by Seaman
et al. (1995), Fauré et al. (1996) and Grolimund et al. (1996) showed similar results. Transport of clay
particles also influences other properties of an aquifer, such as porosity and dispersivity (Mehnert and
Jennings, 1985; Beekman and Appelo, 1990), and can lead to clogging of wells when fresh water is
injected into brackish water aquifers (Brown and Signor, 1974). Also, transport with colloids may facil-
itate the transport of hazardous chemicals, radioactive elements and strongly sorbing organic chemi-
cals, which would otherwise be immobile because of their large tendency for sorption and
consequent high retardation (Van der Lee et al., 1992; Puls and Powell, 1992; Saiers and Hornberger,
1996). For describing colloid mobility, filtration theory is often used (Yao et al., 1971) which con-
siders solution composition effects only implicitly (see Kretzschmar et al., 1999, for a review).
The extent of the double layer determines whether pores are electrostatically “free”. Pores which
are so small that the double layers overlap are completely filled with a negative potential. Anions are
then unable to pass the pores. Passage of cations is hindered as well since the outflowing solution
must remain neutral. A consequence is that only water molecules and uncharged species (H4SiO4,
uncharged complexes) can pass the pore. This process is termed hyperfiltration or membrane filtra-
tion (Fritz, 1986). Figure 6.37 shows that the double layer extends to about 0.01 m when the ionic
strength is 0.001 mol/L. Some effect may therefore be expected in clay soils (particles and
pores 1 m), or compacted clays. The process is self-regulating since the double layer will shrink
as concentrations and ionic strength increase when a residual brine forms, which allows the passage
of more and more solute ions. The double layer approaches the size of a single layer of ions
if concentrations are higher than 1 mol/L, and hyperfiltration would be reduced in that case.
The process of membrane filtration has nevertheless been invoked for explaining anomalous ion-
ratios in deep brines (Kharaka and Berry, 1973; Demir, 1988) and calcite precipitation (Fritz and
Eady, 1985). A warning with sampling of turbid waters is also appropriate here. Hyperfiltration may
take place when water samples are filtered over membrane-filters (pore size 0.45 m, or sometimes
even 0.1 m) with too high pressure and against a too large resistance of the filtered solids.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
298 Ion exchange

1.6

Cl (mmol/L)
1.2

0.8
CECi  0.042 eq/L
0.4 Cli  0.96 mmol/L
f  0.10
0
0 0.5 1 V

Figure 6.41. Concentrations in pressure filtrate from a Na-illite suspension. Data points from Bolt (1961), line
calculated by Appelo (1977). V is the fraction of water remaining in the suspension, CECi is the initial (CEC),
Cli is the intial Cl concentration in the suspension, f is the ratio of activity coefficients inside the suspension
and in the outer solution.

A related phenomenon takes place when water is squeezed from compacting sediments or
clay suspensions. The pressure filtrate may then have higher ion concentrations than the original
pore water (Bolt, 1961), and, consequently, the concentrations in the remaining pore water will
decrease when water is pressed out the suspension (Von Engelhardt and Gaida, 1963). These effects
are difficult to trace in stratigraphic sequences in the field since other water-rock interaction
processes may be more dominating. However, the effects are of importance when sediment samples
are centrifuged or squeezed to obtain interstitial water for chemical analyses (De Lange, 1984;
Saager et al., 1990).
The pressure filtration effects are related to compaction of the diffuse double layer, and can
be quantitatively described with the Donnan theory of suspensions (Appelo, 1977). Two examples
are shown in Figures 6.41 and 6.42. Figure 6.41 gives the composition of water squeezed from a
Na-illite suspension as a function of V, the relative volume of water that remains in the suspension.

CECi  0.67 eq/L

Cli  1.07 eq/L
1.0 f  0.90
Cl (mol/L)

CECi  0.67 eq/L

Cli  0.576 eq/L
0.5 f  0.83

CECi  0.72 eq/L

Cli  0.157 eq/L
f  0.55
0
1 2 3 4
Void ratio

Figure 6.42. Concentrations in pore water of a compacted clay suspension. Solid line indicates data from Von
Engelhardt and Gaida (1963), dotted line (shown where deviating from the measured data) is calculated
by Appelo (1977). Void ratio is the volume of water divided by the volume of solids. For explanation of other
symbols see the caption with Figure 6.41.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Irrigation water quality 299

Initially the salt concentration is 0.96 mmol Na / L, while the clay content amounts to 42 meq
CEC/L. The figure shows that the salt concentration in the pressure filtrate increases sharply when
about half of the water has been pressed from the suspension. Another example, in Figure 6.42,
shows the concentrations in the pore solutions of a clay as a function of the void ratio, defined as the
volume of water divided by the volume of solids. Three different salt solutions in the initial pore
water, varying from 0.157 to 1.07 eq/L, show marked decreases in concentration when water is
squeezed from the clay suspension. The effect is quite notable due to the high CEC of 0.67 or
0.72 eq/L in the initial suspension.
Another aspect of the relative anion-deficit in the DDL is anion-exclusion (Bresler, 1973).
Whenever pores are so small that double layers overlap, or when waterfilms become, in unsaturated
soils, thinner than the DDL, the anions are hesitant to enter pores with a negative potential. The path
of anions is then confined to only a part of the porespace, which is easily measured by means of col-
umn experiments. An anion, affected by exclusion, will show an early breakthrough compared to
water flow (found, for example, with tritium); the faster flow may easily be as much as 10% in soils
(Bond et al., 1982; James and Rubin, 1986), or even 50% in compacted clay (Demir, 1988). A field
experiment by Gvirtzmann and Gorelick (1991) showed similar effects.
We have seen (Figure 6.36) that the concentration of the anions will decrease towards a nega-
tively charged surface. The relative deficit of anions (or more generally, of the co-ions which have
the same valence as the adsorbing solid) balances part of the charge of the solid, and is thus rightly
termed negative adsorption (Van Olphen, 1977). Negative adsorption increases with higher salt con-
centrations on solids with a fixed charge, such as clay minerals. A higher anion deficit means that
less cations are adsorbed, and the CEC (defined from cations alone) may be expected to decrease
when solute concentrations increase. This effect may lead to an about 10% lower apparent CEC of
soils in a 0.1N Cl solution (Table 6.9). Since other adsorption mechanisms may play a role, such as
competition with protons, or the adsorption of complexes (Sposito et al. 1983; Griffioen and Appelo,
1993), the net effect is very difficult to detect experimentally in a sediment or soil which generally
consists of a mixture of clay minerals, organic matter and oxides which all exchange somewhat
differently.

6.7 IRRIGATION WATER QUALITY

Soils containing a high concentration of Na in the soil solution are notorious for having a poor
soil structure. The permeability is reduced (Figure 6.40), and heavy agricultural machinery is not
supported. The effects are related to a high degree of swelling of clays caused by extended double
layers surrounding the clay particles. Diffuse double layers increase in thickness when monovalent
ions make up a high proportion of the exchangeable cations. Problems with soil structure arise
when Na forms about 15% of the exchangeable cations, and the ionic strength is less than about
0.015 (i.e. fresh water). With Na, Mg2, and Ca2 as the dominant cations in natural water, the
fraction of exchangeable Na is a function of the ratio of Na over the square root of Ca2 and
Mg2 concentrations. The fraction of Na can be calculated with the Gaines-Thomas convention
(Example 6.4):

2Na {[Ca 2 ] / ( K Na\Ca [Na ])2  [Mg 2 ] / (K Na\Mg [Na ])2}  Na  1  0

However, in irrigation studies the Gapon convention is more popular, and offers an easier relation-
ship between exchanger-fraction and solute-concentrations. From the reaction:
Na  Ca0.5-X ↔ Na-X  1⁄2Ca2 (6.17a)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

300 Ion exchange

we have:

[Na-X][Ca 2 ]0.5  [Ca 2 ]0.5

G
K Na\Ca   Na (6.18a)
[Ca 0.5 -X][ Na ] Ca [ Na ]

or

Na [Na ]
 K Na\Ca
G (6.73)
Ca [Ca 2 ]0.5

The value of the Gapon constant K GNa\Ca  0.5. A further simplification is to assume that Mg2 and
Ca2 are equally selected by the soil, i.e. K GNa\Ca  K GNa\Mg, and to add up the Ca2 and Mg2 con-
centrations. Since furthermore MgCa  1 – Na in a soil with only Na, Mg2 and Ca2, we
obtain:
Na [Na ]
 K Na
G
\ Ca (6.74)
1  Na [Ca 2 ]  [Mg 2 ]

where the quotient Na/(1 – Na) is termed the Exchangeable Sodium Ratio (ESR). The ESR is,
according to Equation (6.74), a function of the activity of [Na] divided by the square root of the
sum of Ca2 and Mg2 activities.

EXAMPLE 6.13. Calculation of the Exchangeable Sodium Ratio (ESR).

Calculate the ratio of Na /

Ca in water where the limit of 15% Na in the exchange complex is reached.

ANSWER:
Equation (6.74) gives:

0.15 [Na ]
 0.5
1  0.15 [Ca 2 ]

—
and the limiting ratio Na/
Ca is 0.35 (moles/L).5 (assuming concentrations equal to activities).

Soil scientists often use the sodium adsorption ratio (SAR) of water as a measure of ESR to estimate the
suitability of irrigation water. In the formula for SAR, concentrations in mmol/L are used, and the Mg2
concentration is added to the Ca2 concentration. With this convention Equation (6.74) becomes:

Na 1000 mNa

 0.5 (6.74a)
1  Na 1000 mCa  mMg

where mi is now the concentration of i in mmol/L. In this equation

mNa
SAR 
mCa  mMg

so that Equation (6.74a) may be written as:

ESR  0.0158 SAR (6.75)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Irrigation water quality 301

The critical ESR  0.15, is reached when SAR  10. The equation as obtained from exchange the-
ory can be compared with the empirical relationship (US Soil Salinity Lab, 1954):

ESR  0.013  0.015 SAR (6.76)

which is the relationship found for a number of soils (Figure 6.43). The figure also includes a plot of
ESR calculated with the Gaines-Thomas convention, with KNa\Ca  0.37. The value of KG has been
found to decrease when soil organic matter increases (Gheyi and Van Bladel, 1976; Curtin et al.,
1995), indicating specific binding (complexation) of Ca2 to organic matter.
The quality of irrigation water is not only determined by the SAR in water at the time of
application. The ratio of Na over the square root of Ca2  Mg2 concentrations of water may
change by evapotranspiration and chemical reactions in the soil. Most important are the concentrat-
ing effect of evapotranspiration and the precipitation of calcite, which both increase SAR.
Evapotranspiration has this effect because of the square root of the Ca2 concentration in the for-
mula for SAR, while the precipitation of calcite simply decreases the Ca2 concentration in water. An
adjusted SAR can be calculated in which calcite precipitation is taken into account (Ayers and
Westcot, 1985; Example 6.14).

1.00

Gaines-Thomas
Gapon

.75
Exchangeable sodium ratio (ESR)

.50

.25

0
0 10 20 30 40 50 60
Sodium adsorption ratio (SAR)

Figure 6.43. Relationship between SAR and ESR in a number of soils; full line gives the empirical relationship
(agrees with the Gapon convention), dotted line gives Gaines-Thomas convention with KNa\Ca  0.37. Adapted
from US Soil Salinity Lab (1954).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

302 Ion exchange

EXAMPLE 6.14. Calculation of SAR adjusted for calcite precipitation

An analysis of groundwater provided the results:

pH Na K Mg2 Ca2 Cl HCO

3 SO2
4

8.5 62 0.1 9.0 81 168 183 9.0 mg/L

– Is the analysis reliable?

– Is the water quality suitable for irrigation when irrigation return flow is 10%?

ANSWER:
Recalculating the analysis in meq/L gives:

Na K Mg2 Ca2  Cl HCO

3 SO2
4 
2.7 0.002 0.74 4.05 7.49 4.74 3.00 0.19 7.93

The difference is 0.44, or 3% of (cations–anions), and the analysis is correct. SAR of the water is 1.7, and
it offers excellent irrigation quality. However, concentrations increase upon evapotranspiration of 90% of
the water, and this gives, by itself, an increase of SAR. Moreover, one can expect that calcite precipitates, so
that the Ca2 concentration might decrease despite the increase in concentration due to evapotranspiration.
The Ca2 concentration in equilibrium with calcite can be calculated with the equations derived in Chapter 5.
It is easiest to assume a constant CO2-pressure:

Ca 2  2HCO
3 ↔ CaCO3  CO 2  H 2O; K  106

With a CO2-pressure of 0.01 atm, then at equilibrium with calcite:

[Ca 2 ][HCO
3 ]  10
2 8 (6.77)

Note that HCO 3 as used here, is the actually measured concentration which may have resulted from other
reactions than calcite dissolution: we cannot assume HCO3  2Ca2, as was did in Chapter 5.
Neglecting activity coefficients and complexes allows to use concentrations instead of activities. When
the cations are 10 times concentrated through evapotranspiration, the concentration of Ca2 increases to
20.2 mmol/L and of HCO 3 to 30 mmol/L. or

(0.02)(0.03)2  104.74

An amount  will precipitate, which can be calculated with Equation (6.77):

(0.02  )(0.03  2)2  108

Trial and error gives   0.0143 mol/L, and the final concentrations are:

Na Mg2 Ca2 HCO

3

27 3.7 5.9 1.4 mmol/L

In the resulting water SAR  27 /

(3.7  5.9)  8.7, which is still on the safe side.
We note once more, that water types with more equivalents Ca2  than HCO3 offer relatively safe irrigation
water qualities from the SAR point of view.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 303

QUESTION:
Repeat the calculation with PHREEQC. Evaporate water with keyword REACTION.
ANSWER:   13.7 mmol Calcite/L, SAR  8.5.

PROBLEMS
Laboratory simulation of the exchange reaction:
Add a CaCl2 solution to an exchanger in Na form. Ca2 displaces part of the Na from the exchanger sites and
hence both the Ca2 and Na concentrations change. We calculate the equilibrium constant KNa\Ca. The resin
DOWEX 50 with an exchange capacity CEC  5 meq/g is used as a substitute for the soil.

You receive:

– 300 mL 2.5 mM CaCl2 solution. Determine Ca2, Cl (titrations)

– 200 mg Na-X (DOWEX 50) in a 500 mL erlenmeyer flask

Add 200 mL of the CaCl2 solution to the exchanger in the erlenmeyer flask. Stir for 5 minutes, and analyze the
solution for Ca2  and Cl (titrate), and Na (flame-photometer). Estimate the Na concentration beforehand,
and determine the correct dilution factor to be in the analysis range of the flamephotometer.

Analyze:
in the CaCl2 solution:
Ca2  ………. mmol/L Cl  ………. mmol/L
after reaction with Na-X:
Ca2  ………. mmol/L Na  ………. mmol/L   ………. meq/L (correct?)

Calculate:
Ca-X2  ………. meq/L Na-X  ………. meq/L
Ca  ………. Na  ………. KNa\Ca  ……….

The distribution-coefficient is for Na ………. L/kg, and for Ca2 ………. L/kg.
(Use the analyzed concentrations, when calculating the distribution coefficients).

6.1. Calculate the structural charge per gram of illite, K0.7[Si3.4Al0.6]iv[Al1.9Fe20.1]viO10(OH)2, as well as the
charge density. The  layers are stacked 20 units deep; the unit cell size is as of smectite, one unit cell
contains 24 oxygens.

6.2. Also for vermiculite, Mg0.4[Si3.8Al0.2]iv[Mg0.6Al1.4]viO10(OH)2; unit cell size is as of smectite, the layers
are completely separated.

6.3. Repeat the calculation of Example 6.4 with all concentrations in water 10 times higher (note how the rel-
ative proportion of exchangeable Na-X changes). Do the same for seawater using the concentrations listed
in Table 6.2.

6.4. Determine the exchange coefficient KNa\Cd on smectite from the following data of Garcia-Miragaya and
Page (1976). The amount of smectite is 2.05 meq/L.

NaClO4 (N) Cdtotal (M) adsCd (%)

0.03 0.40 91.1

0.80 87.8

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

304 Ion exchange

6.5. In the dunes of the island Ameland two boreholes (A and B) have been drilled. Water from three filters
(1–3), and the nearby sea (4) has the following composition (concentrations in mmol/L):

1 2 3 4

pH 6.7 6.8 7.0 8.2

Na 0.86 18.4 112 485
K 0.04 0.1 3.9 10.6
Mg2 0.09 0.45 10 55.1
Ca2 3.0 0.9 31.3 10.7
Cl 1.0 3.0 200 566
HCO 3 6.1 18.0 0.3 2.4
SO2
4 0.02   29.3

A B

4
1
Sea Sea

2 3 Salt/fresh interface

a. Explain the composition of these water samples.

b. Is the salt / fresh-water interface moving upward or downward at A, B?
c. Calculate the exchangeable fractions Na, K, Mg and Ca on the sediment with the Gaines-Thomas
convention for sample 1 and 4 (assume activity  molality).
d. Recalculate the fractions of exchangeable cations in meq/L pore water, given CEC  10 meq/kg;
b  1.8 g/cm3; w  0.3.
6.6. In the southern part of the Nile-delta the composition of the groundwater is determined by:
– composition of Nile-water, used as irrigation-water
– processes during infiltration of Nile-water in the soil
– processes in the soil
a. What are the processes during infiltration? Data (concentrations in mmol/L):

Na K Mg2 Ca2 HCO

3 Cl SO2
4 pH

Nile-water 0.5 0.1 0.4 0.7 2.2 0.5 0.1

Groundwater 8.0 0.1 0.2 0.7 2.7 4.5 1.0 8.1

b. In the northern part of the delta, near the sea, water is pumped with a composition of:

Na K Mg2 Ca2 HCO

3 Cl SO2
4

180 2.5 35 15 7.5 270 1.5 (borehole 121)

– What processes influenced this composition?

– If more water is pumped up, will water become more brackish or fresher?
c. Calculate TDS (Total Dissolved Solids) of the analysis in b.
6.7. Native groundwater in the injection test of Valocchi et al. (1981) had the composition Na  86.5,
Mg2  18.2 and Ca2  11.1 mmol/L (Figures 6.16, 6.27). Injected water has 14.66 meq Cl/L.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 305

Selectivity coefficients were (Gaines and Thomas convention, activity  molal concentration)
KNa\Mg  0.54 and KNa\Ca  0.41. Sediment CEC  750 meq/L pore water.
Calculate
a. the composition of the exchange complex,
b. the composition of water after the salinity jump, in equilibrium with the original exchange complex
(compare with Figure 6.16),
c. the exchange complex in equilibrium with injected water (cf. Example 6.11),
d. the number of porevolumes that the exchange complex can maintain the high Na concentration
found under b) (When the outcome is multiplied with V0  295 m3, the porevolume for observation
well S23, the rise of Mg2 must be found: compare with Figure 6.16).

6.8. The fresh groundwater resident in the column shown in Figure 6.20 has Na  5.6, Mg2  1.9, and
Ca2  2.0 mmol/L. Injected is water with 235 meq anions/L. The CEC of the sediment is 60 meq/L pore
water. KNa\Mg  0.55, KNa\Ca  0.35 (Gaines and Thomas convention, solute activities  molal concen-
trations).
Calculate
a. the composition of the exchange complex,
b. the composition of water after the salinity jump, in equilibrium with the original exchange complex,
c. the number of porevolumes that the exchange complex can maintain the high Ca2concentration
found under b).

6.9. Derive equations for heterovalent exchange of divalent ions and Na which relate the Gaines-Thomas K
to the Vanselow K and the Gapon K.

6.10. Derive the equation for the thickness of the double layer, Equation (6.71).

6.11. Calculate the potential at a smectite surface in a solution of 0.01 M NaCl, and estimate the concentration
of Na and Cl at 0, 10, 50 and 100Å distance from the smectite surface. The smectite has a surface
charge of 0.09 C/m2.

6.12. Calculate SAR in the irrigation return flow water of Example 6.14 without taking calcite precipitation into
account. Compare with SAR of the irrigation water.

6.13. Calculate the exchange coefficients for Ca / K exchange on illite and ion-exchange resin from the data of
Wiklander (1955), as given in Example 6.3.

6.14. Derive a general formula to calculate equivalent fractions from molar fractions, and vice versa. Hint: use
I CEC  i I TEC, and find two relations for TEC / CEC in which only M’s or ’s appear.

6.15. Calculate a Ca2 / Mg2 isotherm with PHREEQC, using Gapon’s convention:

0.5 Ca2  Mg1⁄2-X ↔ Ca1⁄2-X  0.5Mg2

and also for

Ca2  Mg1⁄2-X ↔ Ca1⁄2-X  Mg2

Derive KG’s from Table 6.4. Use “-no_check” and “-mole_balance” for the 2nd case. Consult Figure 6.24
for obtaining an isotherm.

6.16. Make a plot of the K / Na exchange-isotherm (K vs K) with PHREEQC for KK/Na  2.7, used in
Example 6.10.

6.17. Calculate the retardation of Na in the aquifer below the Vejen waste site, Figure 3.21. mCa  1 mM, mNa
varies from 1 M to 10 mM, CEC  50 meq/L, KCa\Na  2.85.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

306 Ion exchange

REFERENCES
Andersen, M.S., 2001. Geochemical processes at a seawater-freshwater interface. Ph.D. thesis, Lyngby, 145 pp.
Appelo, C.A.J., 1977. Chemistry of water expelled from compacting clay layers: a model based on Donnan
equilibrium. Chem. Geol. 19, 91–98.
Appelo, C.A.J., 1994. Cation and proton exchange, pH variations, and carbonate reactions in a freshening
aquifer. Water Resour. Res. 30, 2793–2805.
Appelo, C.A.J., 1994b. Some calculations on multicomponent transport with cation exchange in aquifers.
Ground Water 32, 968–975.
Appelo, C.A.J., 1996. Multicomponent ion exchange and chromatography in natural systems. In P.C. Lichtner,
C.I. Steefel and E.H. Oelkers (eds), Rev. Mineral. 34, 193–227.
Appelo, C.A.J., Willemsen, A., Beekman, H.E. and Griffioen, J., 1990. Geochemical calculations and observa-
tions on salt water intrusions. II. Validation of a geochemical model with laboratory experiments. J. Hydrol.
120, 225–250.
Appelo, C.A.J., Hendriks, J.A. and Van Veldhuizen, M., 1993. Flushing factors and a sharp front solution for
solute transport with multicomponent ion exchange. J. Hydrol. 146, 89–113.
Ayers, R.S. and Westcot, D.W., 1985. Water quality for agriculture. FAO Irr. Drain. Pap. 29, 174 pp.
Back, W., 1966. Hydrochemical facies and groundwater flow patterns in northern part of Atlantic coastal plain.
U.S. Geol. Surv. Prof. Paper 498-A, 42 pp.
Bahr, J.M., 1989. Analysis of nonequilibrium desorption of volatile organics during field test of aquifer decon-
tamination. J. Contam. Hydrol. 4, 205–222.
Bartschat, B.M., Cabaniss, S.E. and Morel, F.M.M., 1992. Oligoelectrolyte model for cation binding by humic
substances. Env. Sci. Technol. 26, 264–294.
Beekman, H.E., 1991. Ion chromatography of fresh and salt water intrusions. Ph.D. thesis, Free University,
Amsterdam, 198 pp.
Beekman, H.E. and Appelo, C.A.J., 1990. Ion chromatography of fresh- and salt-water displacement: laboratory
experiments and multicomponent transport modelling. J. Contam. Hydrol. 7, 21–37.
Bjerg, P.L., Ammentorp, H.C. and Christensen, T.H., 1993. Model simulations of a field experiment on cation
exchange-affected multicomponent solute transport in a sandy aquifer. J. Contam. Hydrol. 12, 291–311.
Bloom, S.A. and Mansell, R.S., 2001. An algorithm for generating cation exchange isotherms from binary
selectivity coefficients. Soil Sci. Soc. Am. J. 65, 1426–1429.
Bolt, G.H., 1961. The pressure filtrate of colloidal suspensions, II. experimental data on homoionic clays.
Kolloid-Z. 175, 144–150.
Bolt, G.H., 1967. Cation exchange equations used in soil science – a review. Neth. J. Agric. Sci. 15, 81–103.
Bolt, G.H. (ed.), 1982. Soil chemistry, B. Physico-chemical models. Elsevier, Amsterdam, 527 pp.
Bolt, G.H. and Bruggenwert, M.G.M. (eds), 1978. Soil chemistry, A. Basic elements. Elsevier, Amsterdam, 281 pp.
Bond, W.J., 1995. On the Rothmund-Kornfeld description of cation exchange. Soil Sci. Soc. Am. J. 59, 436–443.
Bond, W.J. and Verburg, K., 1998. Comparison of methods for predicting ternary exchange from binary
isotherms. Soil Sci. Soc. Am. J. 61, 444–454.
Bond, W.J., Gardiner, B.N. and Smiles, D.E., 1982. Constant flux absorption of a tritiated calcium chloride solu-
tion by a clay soil with anion exclusion. Soil Sci. Soc. Am. J. 46, 1133–1137.
Breeuwsma, A., Wösten, J.H.M., Vleeshouwer, J.J., Van Slobbe, A.M. and Bouma, J. 1986. Derivation of land
qualities to assess environmental problems from soil surveys. Soil Sci. Soc. Am. J. 50, 186–190.
Bresler, E., 1973. Anion exclusions and coupling effects in nonsteady transport through unsaturated soils: I.
Theory. Soil Sci. Soc. Am. Proc. 37, 663–669.
Brown, R.F. and Signor, D.C., 1974. Artificial recharge – state of the art. Ground Water 12, 152–160.
Bruggenwert, M.G.M. and Kamphorst, A., 1982. Survey of experimental information on cation exchange in
soil systems. In G.H. Bolt, 1982, op. cit., 141–203.
Brusseau, M.L., Rao, P.S.C., Jessup, R.E. and Davidson, J.M., 1989. Flow interruption: a method for investi-
gating sorption nonequilibrium. J. Contam. Hydrol. 4, 223–240.
Bürgisser, C., Černík, M., Borkovec, M. and Sticher, H., 1993. Determination of nonlinear adsorption
isotherms from column experiments: an alternative to batch studies. Env. Sci. Technol. 27, 943–948.
Ceazan, M.L., Thurman, E.M. and Smith, R.L., 1989. Retardation of ammonium and potassium transport through
a contaminated sand and gravel aquifer: the role of cation exchange. Env. Sci. Technol. 23, 1402–1408.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 307

Chapelle, F.H., 1983. Groundwater geochemistry and calcite cementation of the Aquia aquifer in Southern
Maryland. Water Resour. Res. 19, 545–558.
Chapelle, F.H. and Knobel, L.L., 1983. Aqueous geochemistry and the exchangeable cation composition of
glauconite in the Aquia aquifer, Maryland. Ground Water 21, 343–352.
Curtin, D., Selles, F. and Steppuhn, H., 1995. Sodium-calcium exchange selectivity as influenced by soil prop-
erties and method of determination. Soil Sci. 159, 176–184.
De Lange, G.J., 1984. Shipboard pressure-filtration system for interstitial water extraction. Med. Rijks Geol.
Dienst (Haarlem, the Netherlands) 38, 209–214.
Demir, I., 1988. The interrelation of hydraulic and electrical conductivities, streaming potential and salt fitra-
tion during the flow of chloride brines through a smectite layer at elevated pressures. J. Hydrol. 98, 31–52.
DeVault, D., 1943. The theory of chromatography. J. Am. Chem. Soc. 65, 532–540.
Elprince, A.M., Vanselow, A.P. and Sposito, G., 1980. Heterovalent, ternary cation exchange equilibria: NH4+-
Ba2+-La3+ exchange on montmorillonite. Soil Sci. Soc. Am. J. 44, 964–969.
Eriksson, E., 1952. Cation-exchange equilibria on clay minerals. Soil Sci. 74, 103–113.
Evangelou, V.P. and R.E. Phillips, 1988. Comparison between the Gapon and Vanselow exchange selectivity
coefficients. Soil Sci. Soc. Am. J. 52, 379–382.
Fauré, M.-H., Sardin, M. and Vitorge, P., 1996. Transport of clay particles and radioelements in a salinity gra-
dient: experiments and simulations. J. Contam. Hydrol. 21, 255–267.
Feynman, R., Leighton, R.B. and Sands, M.L., 1989. The Feynman lectures on physics, vol. 2. California Inst.
Technol.
Fritz, S.J., 1986. Ideality of clay membranes in osmotic processes: a review. Clays Clay Min. 34, 214–223.
Fritz, S.J. and Eady, C.D., 1985. Hyperfiltration-induced precpitation of calcite. Geochim. Cosmochim. Acta 49,
761–768.
Gaines, G.L. and Thomas, H.C., 1953. Adsorption studies on clay minerals. II. A formulation of the thermody-
namics of exchange adsorption. J. Chem. Phys. 21, 714–718.
Gapon, E.N., 1933. Theory of exchange adsorption V. J. Gen. Chem. (USSR) 3, 667–669, (Chem. Abstr. 28,
4516, 1934)
Garcia-Miragaya, J. and Page, A.L., 1976. Influence of ionic strength and inorganic complex formation on the
sorption of trace amounts of Cd by montmorillonite. Soil Sci. Soc. Am. J. 40, 658–663.
Gheyi, H.R. and Van Bladel, R., 1976. Calcium-sodium exchange in some calcareous soils and a montmoril-
lonite clay as compared with predictions based on double layer threory. Geoderma 16, 159–169.
Goldenberg, L.C., 1985. Decrease of hydraulic conductivity in sand at the interface between seawater and dilute
clay suspensions. J. Hydrol. 78, 183–199.
Goldenberg, L.C. and Mandel, S., 1988. Some processes in the sea-water/fresh-water interface, as influenced
by the presence of gases. Natuurwet. Tijdschr. 70, 288–299, (SWIM Conf. Ghent, Belgium).
Gomis, V., Boluda, N. and Ruiz, F., 1996. Application of a model for simulating transport of reactive multispecies
components to the study of the hydrochemistry of salt water intrusions. J. Contam. Hydrol. 22, 67–81.
Gomis-Yagues, V., Boluda-Botella, N. and Ruiz-Bevia, F. 2000. Gypsum precipitation/dissolution as an expla-
nation of the decrease of sulphate concentration during seawater intrusion. J. Hydrol. 228, 48–55.
Griffioen, J. and Appelo, C.A.J., 1993. Adsorption of calcium and its complexes by two sediments in calcium-
hydrogen-chlorine-carbon dioxide systems. Soil Sci. Soc. Am. J. 57, 716–722.
Grim, R.E., 1968. Clay mineralogy. McGraw-Hill, New York, 596 pp.
Grolimund, D., Borkovec, M., Barnettler, M. and Sticher, H., 1996. Colloid-facilitated transport of strongly
sorbing contaminants in natural porous media. Env. Sci. Technol. 30, 3118–3123.
Gvirtzman, H. and Gorelick, S.M., 1991. Dispersion and advection in unsaturated porous media enhanced by
anion exclusion. Nature 352, 793–795.
Helfferich, F., 1959. Ionenaustauscher Band 1. VCH, Weinheim, 520 pp.
Helfferich, F. and Klein, G., 1970. Multicomponent chromatography. M. Dekker, New York, 419 pp.
Hurz, G., 2001. Lithium chloride solution as an extraction agent for soils. J. Plant Nutr. Soil Sci. 164, 71–75.
James, R.V. and Rubin, J., 1986. Transport of chloride ion in a water-unsaturated soil exhibiting anion-exclusion.
Soil Sci. Soc. Am. J. 50, 1142–1149.
Karaborni, S., Smit, B., Heidug, W., Urai, J. and Van Oort, E., 1996. The swelling of clays: molecular simula-
tions of the hydration of montmorillonite. Science 271, 1102–1104.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

308 Ion exchange

Kharaka, Y.K. and Berry, F.A.F., 1973. Simultaneous flow of water and solutes through geological membranes.
I. Experimental investigation. Geochim. Cosmochim. Acta 37, 2577–2603.
Koch, S. and Flühler, H., 1993. Non-reactive solute transport with micropore diffusion in aggregated porous
media determined by a flow-interruption method. J. Contam. Hydrol. 14, 39–54.
Kretzschmar, R., Borkovec, M., Grolimund, D. and Elimelech, M., 1999. Mobile subsurface colloids and their
role in contaminant transport. Adv. Agron. 66, 121–193.
Lawrence, A.R., Lloyd, J.W. and Marsh, J.M., 1976. Hydrochemistry and groundwater mixing in part of the
Lincolnshire Limestone aquifer, England. Ground Water 14, 12–20.
Liu, C., Zachara, J.M. and Smith, S.C., 2004. A cation exchange model to describe Cs+ sorption at high ionic
strength in subsurface sediments at Hanford site, USA. J. Contam. Hydrol. 68, 217–238.
Manzano, M., Custodio, E. and Carrera, J., 1992. Fresh and salt water in the Llobregat delta aquitard. Proc. 12th
Saltwater Intrusion Meeting, 207–238, CIMNE, Barcelona.
Martell, A.E. and Smith, R.M., 1977. Critical stability constants, vol. 3. Plenum Press, New York.
Mehnert, E. and Jennings, A.A., 1985. The effect of salinity-dependent hydraulic conductivity on saltwater
intrusion episodes. J. Hydrol. 80, 283–297.
Moore, W.J., 1972. Physical chemistry, 5th ed. Longman, London, 977 pp.
Nativ, R., Issar, I. and Rutledge, J., 1983. Chemical composition of rainwater and floodwaters in the Negev
desert, Israel. J. Hydrol. 62, 201–223.
Neal, C. and Cooper, D.M., 1983. Extended version of Gouy-Chapman electrostatic theory as applied to the
exchange behavior of clay in natural waters. Clays Clay Miner. 31, 367–376.
Nielsen, D.R., Van Genuchten, M.Th. and Biggar, J.W., 1986. Water flow and solute transport processes in the
saturated zone. Water Resour. Res. 22, 89S-108S.
Nkedi-Kizza, P., Biggar, J.W., Selim, H.M., Van Genuchten, M.Th., Wierenga, P.J., Davidson, J.M. and Nielsen,
D.R., 1984. On the equivalence of two conceptual models for describing ion exchange during transport
through an aggregated oxisol. Water Resour. Res. 20, 1123–1130.
Pauling, L., 1960. The nature of the chemical bond. Cornell Univ. Press, 644 pp.
Perkins, T.K. and Johnston, O.C., 1963. A review of diffusion and dispersion in porous media. Soc. Petrol.
Engin. J., march 1963, 70–84.
Puls, R.W. and Powell, R.M., 1992. Transport of inorganic colloids through natural aquifer material: implica-
tions for contaminant transport. Env. Sci. Technol. 26, 614–621.
Purdy, C.B., Burr, G.S., Rubin, M., Helz, G.R. and Mignerey, A.C., 1992. Dissolved organic and inorganic 14C
concentrations and ages for coastal plain aquifers in southern Maryland. Radiocarbon 34, 654–663.
Quirk, J.P. and Schofield, R.K., 1955. The effect of electrolyte concentration on soil permeability. J. Soil Sci. 6,
163–178.
Rainwater, K.A., Wise, W.R. and Charbeneau, R.J., 1987. Parameter estimation through groundwater tracer
tests. Water Resour. Res. 23, 1901–1910.
Reardon, E.J., Dance, J.T. and Lolcama, J.L., 1983. Field determination of cation exchange properties of cal-
careous sand. Ground Water 21, 421–428.
Rhoades, J.D., 1982. Cation exchange capacity. In A.L. Page et al. (eds), Methods of soil analysis, Agron.
Monogr. 9, 149–157. Soil Sci. Soc. Am., Madison.
Rieman, W. and Walton, H.F., 1970. Ion exchange in analytical chemistry. Pergamon Press, Oxford, 295 pp.
Saager, P.M., Sweerts, J.-P. and Ellermeijer, H.J., 1990. A simple pore-water sampler for coarse, sandy sedi-
ments of low porosity. Limnol. Oceanogr. 35, 747–751.
Saiers, J.E. and Hornberger, G.M., 1996. The role of colloidal kaolinite in the transport of cesium through
laboratory sand columns. Water Resour. Res. 32, 33–41.
Scheffer, F. and Schachtschabel, P., 2002. Lehrbuch der Bodenkunde, 15th Aufl. Elsevier, Amsterdam,
607 pp.
Seaman, J.C., Bertsch, P.M. and Miller, W.P., 1995. Chemical controls on colloid generation and transport in a
sandy aquifer. Env. Sci. Technol. 29, 1808–1815.
Shainberg, I. and Oster, J.D., 1978. Quality of irrigation water. Int. Irr. Inf. Cent. Bet Dagan, Israel.
Šimůnek, J. and Suarez, D.L., 1997. Sodic soil reclamation using multicomponent transport modeling. J. Irr.
Drain. Eng. 123, 367–376.
Sposito, G., 1984. The surface chemistry of soils. Oxford Univ. Press, New York, 234 pp.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 309

Sposito, G., Holtzclaw, K.M., Charlet, L., Jouany, C. and Page, A.L., 1983. Na-Ca and Na-Mg exchange in
Wyoming bentonite in ClO 
4 and Cl background ionic media. Soil Sci. Soc. Am. J. 47, 51–56.
Stuyfzand, P.J., 1985. Hydrochemistry and hydrology of the dune area between Egmond and Wijk aan Zee (in
Dutch). KIWA, SWE, 85.012, Nieuwegein.
Stuyfzand, P.J., 1993. Hydrochemistry and hydrology of the coastal dune area of the Western Netherlands. Ph.D.
Thesis, Free University, Amsterdam, 366 pp.
Suarez, D.L., Rhoades, J.D., Lavado, R. and Grieve, C.M., 1984. Effect of pH on saturated hydraulic conduc-
tivity and soil dispersion. Soil Sci. Soc. Am. J. 48, 50–55.
US Soil Salinity Lab. Staff, 1954. Diagnosis and improvement of saline and alkali soils. USDA Handb. 60, US
Gov. Print. Office, Washington DC.
Valocchi, A.J., Street, R.L. and Roberts, P.V., 1981. Transport of ion-exchanging solutes in groundwater: chro-
matographic theory and field simulation. Water Resour. Res. 17, 1517–1527.
Van Breukelen, B.M., Appelo, C.A.J. and Olsthoorn, T.N., 1998. Hydrogeochemical transport modelling of 24
years of Rhine water infiltration in the dunes of the Amsterdam water supply. J. Hydrol. 209, 281–296.
Van der Kemp, W., Appelo, C.A.J., Walraevens, K., 2000. Inverse chemical modeling and radiocarbon dating of
palaeogroundwaters: the Tertiary Ledo-Paniselian aquifer in Flanders, Belgium. Water Resour. Res. 36,
1277–1287.
Van der Lee, J., Ledoux, E. and Demarsily, G., 1992. Modeling of colloidal uranium transport in a fractured
medium. J. Hydrol. 139, 135–158.
Van der Molen, W.H., 1958. The exchangeable cations in soils flooded with seawater. Staatsdrukkerij, Den
Haag, 167 pp.
Van Eijkeren, J.C.H., annd Loch, J.P.G., 1984. Transport of cationic solutes in sorbing porous media. Water
Resour. Res. 20, 714–718.
Van Genuchten, M. Th., 1985. A general approach for modeling solute transport in structured soils. IAH Mem.
17, 513–525.
Van Genuchten, M. Th. and Cleary, R.W., 1982. Movement of solutes in soil: computer-simulated and labora-
tory results. In G.H. Bolt, 1982, op. cit., 349–386.
Van Olphen, H., 1977. An introduction to clay colloid chemistry, 2nd ed. Wiley and Sons, New York, 318 pp.
Van Veldhuizen, M., Hendriks, J.A. and Appelo, C.A.J., 1998. Numerical computation in heterovalent chro-
matography. Appl. Numer. Mathem. 28, 69–89.
Vanselow, A.P., 1932. Equilibria of the base-exchange reactions of bentonites, permutites, soil colloids and zeo-
lites. Soil Sci. 33, 95–113.
Voegelin, A., Vulava, V.M., Kuhnen, F. and Kretzschmar, R., 2000. Multicomponent transport of major cations
predicted from binary adsorption experiments, J. Contam. Hydrol. 46, 319–338.
Von Engelhardt, W. and Gaida, K.H., 1963. Concentration changes of pore solutions during the compaction of
clay sediments. J. Sed. Petrol. 33, 919–930.
Walraevens, K. and Cardenal, J., 1994. Aquifer recharge and exchangeable cations in a Tertiary clay layer
(Bartonian clay, Flanders-Belgium). Min. Mag. 58A, 955–956.
Weber, W.J. and Smith, E.H., 1987. Simulation and design models for adsorption processes. Env. Sci. Technol.
21, 1040–1050.
Wiklander, L., 1955. Cation and anion exchange phenomena. In F.E. Bear (ed.), Chemistry of the soil, Chapter
4. Wiley and Sons, New York.
Xu, T., Samper, J., Ayora, C., Manzano, M. and Custodio, E. 1999. Modeling of non-isothermal multi-component
reactive transport in field scale porous media flow systems. J. Hydrol. 214, 144–164.
Yao, K.M., Habibian, M.T. and O’Melia, C.R., 1971. Water and waste water filtration: concepts and applica-
tions. Env. Sci. Technol. 5, 1105–1112.
Zuur, A.J., 1938. Trans. 2nd Comm. and Alkali-Subcomm., Int. Congr. Soil Sci., Helsinki, B 66–67.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

7

Sorption of Trace Metals

Heavy metals like Cd, Pb and Ni, and also As, are strongly poisonous and therefore the highest
admissible concentration for these substances is very low (Table 1.1). Enhanced heavy metal con-
centrations constitute a serious problem for water supplies around the world. Once released to
groundwater, the mobility of trace metals is to a large extent controlled by sorption processes. Part
of trace metal sorption is due to ion exchange and can be estimated with the formulas from Chapter 6.
However, most sorption is connected with specific binding of heavy metals to the variable charge
surfaces of oxides and organic matter.
Variable charge solids sorb ions from solution without releasing other ions in equivalent propor-
tion. Their surface charge can be positive or negative depending on the pH and the solution compo-
sition. Variable charge solids are important in regulating the mobility of both positively charged
heavy metals such as Pb2 and Cd2, and of oxyanions such as HAsO42 and H2PO 4 . Modeling of
sorption on variable charge solids is more complex than ion exchange, and generally requires com-
puter programs for an exact answer. Essentially, the surfaces of oxides and hydroxides gain a pH
dependent charge due to sorption of protons and other ions from solution. The surface charge creates
a potential difference between the solution and the surface. This potential will influence the
approach of ions towards the surface, and thus couples back to the charge development on the sur-
face. However, some ions are bound so strongly that they even remain fixed on a surface that has the
same charge and repels the ion electrostatically.

7.1 THE ORIGIN AND OCCURRENCE OF HEAVY METALS IN GROUNDWATER

Groundwater is often heavily polluted near mines of sulfide minerals like pyrite (FeS2), galena
(PbS), sphalerite (ZnS) or arsenopyrite (FeAsS). Waste materials, containing residual sulfide miner-
als, are piled up in tailings and may become oxidized, producing sulfuric acid:


FeS2  15⁄4O2  7⁄2H2O l Fe(OH)3  2SO2
4  4H (7.1)

Concomitantly, large amounts of heavy metals are released which remain soluble in the acid mine
drainage (Stollenwerk, 1994; Appleyard and Blowes, 1994). Figure 7.1 shows the pore water com-
position of a mine tailing. In the upper part, the pH is less than 2.5, and the water contains over
20,000 mg Zn/L and over 100 mg Cu/L. Downward the pH increases to above 4 and the Cu concen-
tration drops to 1 mg/L, while Zn remains high and only decreases in the saturated part of the tail-
ings. Apparently, the pH has a major influence on the mobility of the heavy metals, but the effect is
different for the various metals. Acid mine drainage is globally a major environmental issue. There
are thousands and thousands of metal sulfide mines, many of them abandoned, that continuously
release low pH and heavy-metal loaded water into the environment.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

312 Sorption of trace metals

pH mV mg/L
1 2 3 4 5 6 7 8 0 250 500 750 1000 10000 100000
0.0 0.0 0.0

2.0 2.0 2.0

Depth (m)
4.0 4.0 4.0

6.0 6.0 6.0

pH Eh SO4
8.0 8.0 8.0
mg/L mg/L mg/L
10 1000 100000 1 100 10000 0.01 1.0 1000
0.0 0.0 0.0

2.0 2.0 2.0

Depth (m)

4.0 4.0 4.0

6.0 6.0 6.0

Fe Zn Cu
8.0 8.0 8.0

Figure 7.1. Pore water geochemical profiles for pH, Eh and selected elements in a mine tailings impound-
ment. Note the logarithmic scale for SO4, Fe, Zn and Cu. .– indicates the water table at time of sampling
(Appleyard and Blowes, 1994).

Lowering of the water table by pumping may give access of atmospheric oxygen to sulfides which were
once submerged and protected against oxidation (Kinniburgh et al., 1994; Larsen and Postma, 1997).
Figure 7.2 shows the Ni2 distribution in the zone of drawdown in the Beder aquifer (Larsen and
Postma, 1997). The high nickel concentration is accompanied by a high sulfate concentration which
suggests that pyrite oxidation is taking place. Nickel has not moved far in the aquifer since pumping
started, apparently because sorption on iron oxyhydroxide and manganese oxides retards its movement.

N S
40 RB 3
RB 12
30
Pond RB 11 Clay
RB 13
RB 14 1967
20

1994
10
Elevation (m)

10

20
Sand

30

0 1 2 3 4 5 Clay
40
Ni (µmol/L)
0 50 75 135 300 m

Figure 7.2. Distribution of dissolved nickel in the unsaturated and saturated zone of the Beder aquifer,
Denmark. The water table at the time of sampling in 1994 and the pre-production water table in 1967 is indicated
(Larsen and Postma, 1997).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The origin and occurrence of heavy metals in groundwater 313

NO3 (mg/L) SO4 (mg/L) Fe (mg/L)

0
5
10
Depth (m.b.s)

15
20
25
30
35
40
45
0 100 200 300 0 200 400 0 5 10 15 20

As (µg/L) Co (µg/L) Ni (µg/L) Zn (µg/L)

0
5
10
Depth (m.b.s)

15
20
25
30
35
40
45
0 30 60 0 200 400 600 0 500 1000 0 500 1000

Figure 7.3. Groundwater composition in a drinking water protection area near Vierlingsbeek, The Netherlands.
Pyrite oxidation is coupled to nitrate reduction and leading to the release of trace metals to the groundwater
(Van Beek and Van der Jagt, 1996).

However, the SO2 2

4 concentration has already increased in the pumping wells, because SO4 is more
mobile and not sorbed at the groundwater pH of 7.
Another oxidant for pyrite is nitrate, leached from increased agricultural application of fertilizer
and manure. The reaction is in this case:

FeS2  3NO 2
3  2H 2 O → Fe(OH)3  1.5N 2  2SO 4  H

(7.2)
and is further discussed in Chapter 9. Figure 7.3 shows the groundwater composition of a
drinking water protection area at Vierlingsbeek. From 15 to 20 m below surface there is a decrease
of 150 mg NO 2
3 /L accompanied by an increase of 155 mg SO4 /L, in accordance with Reaction
(7.2). Together with sulfate, trace components As, Co, Ni and Zn are released and reach concentra-
tions many times higher than permissible in drinking water. Again, the pumping wells in the
area showed an increase of sulfate, but not of the trace metals. The trace metals are sorbed and
thus retarded with respect to groundwater flow, although eventually they will also arrive at the
production well.
The chemical weathering of detrital silicates like amphiboles and biotite, and of clay minerals
and oxyhydroxides may also release trace metals (Edmunds et al., 1992). Atmospheric input can be
important as well. At circum-neutral pH, the heavy metals remain fixed by sorption or solid equili-
bria, but when water turns acid they can be mobilized. Figure 7.4 shows the water chemistry in a shal-
low aquifer consisting of unreactive siliceous sand. The upper groundwater has become acid mainly
due to acid rain, and has a pH of 4.5 and increased concentrations of heavy metals like Ni, Co and
Be (Kjøller et al., 2004). Further down, the pH increases to around 6.5 and the concentrations of
heavy metals are much reduced.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

314 Sorption of trace metals

mol/L mol/L pH
0.0 0.2 0.4 0.00 0.04 0.08 3 4 5 6 7
43
41
Co Ni
39 Be
Elevation (m)

37
35
33
31
29
27
25

Figure 7.4. The mobilization of trace metals in an aquifer acidified by acid rain (Kjøller et al., 2004).

The contamination of groundwater by arsenic is a serious problem in Bangladesh and West Bengal
(Smedley and Kinniburgh, 2002) and Vietnam (Berg et al., 2001). Figure 7.5 shows the arsenic distri-
bution in well waters of Bangladesh. In the upper 100 m, arsenic is as high as 1000 g/L, while the
highest admissible WHO concentration is 10 g/L (Table 1.1). Arsenic may substitute for sulfide in
pyrite and the oxidation of pyrite can be one of the causes for the high arsenic contents of groundwater
(Smedley and Kinniburgh, 2002). Another hypothesis is that arsenic adsorbed on Fe-oxide particles
had been transported by the Ganges and other large rivers down to the delta. Here the sediment was
deposited together with organic material. Subsequent development of anoxic conditions caused the
Fe-oxides to become reduced, resulting in the release of the arsenic to the groundwater (Ravenscroft
et al., 2001). Recently, Appelo et al. (2002) have proposed that the displacement of adsorbed arsenic by
dissolved carbonate could be the cause of the mobilization of arsenic (cf. Chapter 11).

As (µg/L)
0 500 1000 1500
0

100 As (µg/L)
0 200 400 600 800 1000
Well depth (m)

20
Well depth (m)

200
40

60
300 80

100

400

Figure 7.5. Arsenic concentration of groundwater in wells of Bangladesh. The insert shows that most arsenic
is found in the uppermost aquifer (Kinniburgh and Smedley, 2001).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption isotherms and distribution coefficients 315

100

80

% Sorbed 60
Cu2 Ni2
Cr3 Pb2
40
Zn2
Cd2 Ca2
20

0
3 4 5 6 7 8
pH

Figure 7.6. Adsorption of heavy metals on the surface of ferrihydrite as a function of pH (Stumm, 1992, based
on the compilation of Dzombak and Morel, 1990).

In Chapter 6 the capacity of sediment particles to sorb major ions like Na and Ca2 was given by
the cation exchange capacity, the CEC. The CEC was a constant related to the clay mineral and
organic carbon content. The exchange concept works well for major cations in general and for trace
metals when the pH is constant, but trace metals behave differently when pH varies. Figure 7.6
shows the sorption of heavy metals on amorphous iron oxyhydroxide (ferrihydrite or hydrous
ferric oxide) as a function of pH. All metals show zero sorption at low pH and increased sorption as
pH increases. Apparently, H ions compete with the heavy metals for the sorption sites. However,
the pH where 50% of the total amount of metal is sorbed varies for different metal ions. One of
the challenges of hydrogeochemistry is to quantify this sorption behavior, and to assess its influence
on the mobility of heavy metals and to predict arrival times of increasing concentrations of these
elements.

7.2 SORPTION ISOTHERMS AND DISTRIBUTION COEFFICIENTS

The relation between sorbed and dissolved solute concentrations at a fixed temperature is called a
sorption isotherm. Two equations are often employed to describe the relation, the Langmuir and the
Freundlich isotherm. The Freundlich isotherm has the form:

sI  K F c nI (7.4)

where sI is the sorbed concentration (mol/kg, g/g, etc.), cI is the solute concentration of chemical I
(mol/L, g/mL, the same chemical mass units as for s), and KF and n are adjustable coefficients.
Usually, n is smaller than 1, so that the increase of sorbed concentration lessens as the solute con-
centration increases. For fitting of experimental data, the Freundlich equation can be linearized by
a log transform to:

log sI  log K F  n log cI (7.5)

which enables the constants to be derived by linear regression. Alternatively, a non-linear

regression can be performed directly on Equation (7.4), with the advantage that the errors can be
weighted as desired (Kinniburgh, 1986). The procedure is easy using a spreadsheet program
(Example 7.1).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

316 Sorption of trace metals

EXAMPLE 7.1. Freundlich sorption isotherm for Cd2 on loamy sand

Christensen (1984) determined sorption of Cd2 on loamy sand in 1 mM CaCl2, pH  6.0, as tabulated in
Table 7.1. Find the coefficients of the Freundlich sorption isotherm with logarithmic weighting of the data
(log sest  log sobs)2 and compare with absolute weighting (sest  sobs)2, where s is sorbed concentration in
g/g soil. Use a non-linear solver, for example from Excel.

Table 7.1. Dissolved and sorbed concentrations of Cd2

on loamy sand.

Dissolved cCd (g/L) Sorbed sCd (g/g)

3.1 0.86
6.1 1.12
5.9 1.71
7.1 1.68
8.1 2.03
9.9 2.46
12.3 2.85
13.0 3.36
13.6 3.22
16.0 3.25
19.1 3.56
24.1 3.76
25.8 4.17
27.6 4.58
33.2 4.82
36.4 5.19

ANSWER:
The data are entered in columns A2..17 and B2..17 of the spreadsheet. Guesses for KF and n are typed in E2
and D2, respectively. Calculate the estimated sorbed concentration in C2 using the formula E$2
* A2 ^ D$2 and copy the formula into C3..17 for the other datapoints. Calculate in column D2..17 the
squared difference of the values in B and C,
(C2  B2)^2 etc. for absolute weighting of the errors, or the
log of the numbers,
(log(C2)  log(B2))^2 etc. for logarithmic weighting, and sum the column in D19.

4
sCd (µg/g soil)

0
0 10 20 30 40
cCd (µg/L)

Figure 7.7. Freundlich isotherms for Cd2 sorbed on loamy sand. Data points from Christensen (1984),
grey and black lines are best-fits using absolute and logarithmic weighting of errors, respectively.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption isotherms and distribution coefficients 317

This sum is to be minimized by adjusting the values of KF and n in E2 and D2, for example in Excel. Use
Tools, Solver and follow the instructions. The result will be,

with absolute weights: sCd  0.568 (cCd)0.617

with logarithmic weights: sCd  0.431 (cCd)0.713

The isotherms are slightly different (Figure 7.7). In general, a logarithmic or relative weighting may be
preferable as it weights more highly the small values in the fit which may be the most accurately determined
and the most important for extrapolation.

QUESTION:
Derive the Freundlich parameters with relative weighting of the errors (
((C2  B2) / C2)^2, etc.).
ANSWER: sCd  0.468 (cCd)0.690

With the Freundlich equation, sorption extends infinitely as concentrations increase, which is unre-
alistic since the number of sorption sites can be expected to be limited. Also, it is generally observed
that the distribution coefficient, i.e. the ratio K d  sI / cI, becomes a constant when concentrations
are small, but in the Freundlich equation the value of

K d′  K F c n1 (7.6)

increases indefinitely as concentrations decrease (for n  1).

The Langmuir equation has a better theoretical background and can be derived from the law of
mass action for a sorption reaction:
[sI ]
s  I ↔ sI ; Ks  (7.7)
I [ s][ I ]

With the mass balance for sorption sites:

s  s I  stot (7.8)

it gives:

[stot ]K s [I ]
[s I ]  I
(7.9)
1  K s [I ]
I

where stot is the total concentration of sorption sites. The activity of the sorbed species can be
expressed as a fractional surface coverage by dividing by stot. Langmuir used a kinetic approach to
derive the equation for gas adsorption on a surface and obtained:
smax cI
sI  (7.10)
K L  cI

It can be checked that the equations are identical if stot  smax and KL  1 / Ks . I

The Langmuir equation shows that the sorbed concentration sI increases linearly with solute con-
centration cI, if cI  KL. When the concentration of I is very high and cI !! KL, the surface
becomes saturated, and sI  stot.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

318 Sorption of trace metals

In the linear part of the curve, the ratio of sorbed and solute concentrations is constant:
K d′  sI / cI  stot / K L (7.11)

If sI is expressed in mass/kg solid and cI in mass/L water, K d has the dimensions L/kg. For the expe-
riment with loamy sand and Cd2 from Example 7.1, a Langmuir equation can be fitted (using loga-
rithmic weighting of the errors) with stot  9.50 g/g soil and KL  30.9 g/L. Hence, K d  307 L/kg.
It is of interest to extrapolate these numbers to field conditions, assuming b /w  6 kg/L. The retar-
dation R  1  K d b /w  1845 will slow down the movement of Cd2 appreciably, but the sorp-
tion maximum of 0.084 mmol/kg is fairly small and must be considered when calculating the mass that
is retained by the soil.

7.2.1 Distribution coefficients from ion exchange

We can use ion exchange equations to calculate a priori values for distribution coefficients of heavy
metals. For a heavy metal I i in exchange equilibrium with Ca2, we have:
1 / i I i  1⁄2Ca-X2 i 1 / i I-Xi  1⁄2Ca2 (7.12)
with
[I-X i ]1/i [Ca 2 ]0.5
K I \ Ca  (7.13)
[Ca-X 2 ]0.5 [I i ]1/i

from which we obtain:

i
[I -X i ] I 0.5  (7.14)
i 
 i 
 K
 I \Ca 2
Ca
 0 . 5 
[I ] [I ]  [Ca ] 

The equivalent fraction I is multiplied by the CEC to give the exchangeable I-Xi in meq/kg; the
activity [I i] is multiplied by the charge i and 1000 to give (numerically) the aqueous concentration
in meq/L. Thus, we get the distribution coefficient in L/kg:
i
CEC  0.5 
K d′   K I \Ca 2
Ca
 0 . 5  (7.15)
1000 i  [Ca ] 

The K d can subsequently be made non-dimensional, by multiplying by the sediment bulk density
b (kg/L) and dividing by the porosity w.
In fresh water Ca approaches 1 (cf. Example 6.4) and the Ca2-concentration can be estimated
for a given CO2-pressure and equilibrium with calcite (Chapter 5). With an estimate of the CEC, the
distribution coefficient is obtained (Examples 7.2 and 7.3).

EXAMPLE 7.2. Estimate the distribution coefficient of Cd2

The loamy sand of Example 7.1 contained 6.2% clay (2 m) and 0.35% organic carbon. Estimate the
distribution coefficient in the experiment with mCa  1 mM, for Cd2 at trace quantities, and express the
2

distribution coefficient in dimensionless units assuming w  0.3.

ANSWER:
Estimate the CEC with Equation (6.8): CEC  7  6.2  35  0.35  55.7 meq/kg (Christensen (1984)
analyzed 75 meq/kg). The exchange coefficient KCd\Ca  KNa\Ca / KNa\Cd  1 (from Table 6.4). With

Equation (7.15) the distribution coefficient is: K d  55.7 / 2000  (1  1 /
103 2
)  27.8 L/kg.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption isotherms and distribution coefficients 319

Bulk density b  2.65  (1  )  1.86 kg/L (cf. Example 6.1). Hence the dimensionless distribution
coefficient is 27.8  b /   172.

QUESTIONS:
What is the effect of doubling the Ca2 concentration?
ANSWER: Kd halves
What is the effect of doubling the CEC?
ANSWER: Kd doubles
Calculate Kd with PHREEQC
ANSWER: Kd  19 L/kg, note the effect of activity corrections for CdX2, Problem 7.20.

The distribution coefficient, estimated in Example 7.2, is about a factor 10 smaller than calculated
with the Langmuir fit (K d  307 L/kg). Also, the maximal sorption capacity for Cd2 would be the
CEC / 2  27.8 mmol/kg which is 330 times higher than was estimated with the Langmuir equation
(stot  9.50 g/g  0.085 mmol/kg). Apparently, the soil contains a small amount of sorbent that binds
Cd2 very strongly in a process not captured by the ion exchange formulae. Christensen (1984) and
many others, e.g Boekhold et al. (1991), noted a strong pH dependency of the distribution coefficient of
Cd2 in soils that may be related to sorption on variable charge surfaces discussed later in this chapter.
Major elements are normally neglected in the calculation of the distribution coefficient from exper-
imental data, but they can be incorporated in the multicomponent ion exchange formulae from Chapter
6. If the charge of the trace element and major ions is equal, a simple linear equation is obtained from
Equation (6.22). For example for Sr2, as a trace element, with respect to the major ions Mg2 and Ca2:
[Sr 2 ]
Sr  (7.16)
[Sr 2 ]  2
KCa\Sr [Ca 2 ]  K Mg\Sr
2 [Mg 2 ]

EXAMPLE 7.3. Distribution coefficient for Sr 2

Let us consider the applicability of the multicomponent relation (7.16) for data provided by Johnston et al.
(1985) who carefully determined Sr 2 distribution coefficients for sediments near a radioactive waste site. The
distribution coefficients were obtained for sands and tills at varying Sr 2 concentrations in groundwater and
synthetic solutions. A typical result is shown in Figure 7.8 for two samples of weathered till. Synthetic
groundwater composition, and CEC of the two samples are given in Table 7.2.

Table 7.2. Water compositions and sediment CEC in an experiment for determining the Sr2 distribution
coefficient (Johnston et al., 1985). Aqueous concentrations in mmol/L.
Sample CEC (meq/kg) Ca2 Mg2 Na K Sr2

A4 53 .75 .80 .09 .08 10850

A8 34 .75 .33 .3 .43 10850

The groundwater used for sample A4 contains mainly Ca2 and Mg2 in solution, which allows the experi-
mental results to be modeled as a cation exchange reaction with respect to only the divalent ions (Mg2 and
Ca2). The distribution coefficient is in our exchange model:
CEC Sr
K d′ 
2000 mSr 2

where mSr is molality of Sr2 (mol/kg H2O). By combining the reactions listed in Table 6.4 we obtain the
exchange coefficients KCa\Sr2  0.77, and KMg\Sr2  0.49. We furthermore assume [I]  mI and find

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

320 Sorption of trace metals

in combination with Equation (7.16):

CEC /2000
K d′  (7.17)
cSr 2 /87600  0.77 mCa 2  0.49 mMg2

where cSr is concentration of Sr2 in mg/L (we use mg/L to allow for a direct comparison with experimen-
tal data from Johnston et al., 1985). The conditions for sample A4 thus lead to:
K d′ (A4)  0.0265 / (cSr / 87600  9.7104 ),

and for sample A8:

K d′ (A8)  0.017 / (cSr / 87600  7.4103 ).

Lines corresponding to these equations are plotted in Figure 7.8, and are in good agreement with experi-
mental values.

100
Sample A4
Sample A8
Distribution coefficient (mL /g)

A4
A8

10

1
106 104 102 100 102 104
Sr in equilibrium solution (mg/L)

Figure 7.8. Strontium distribution curves for two soil samples. Experimental data from Johnston et al.,
1985. Lines are estimated here.

EXAMPLE 7.4. Exchange coefficient of Cd 2 vs Na on montmorillonite

Garcia-Miragaya and Page (1976) determined adsorption percentages of Cd2 from NaClO4 solutions
of different normality on 2.05 meq/L montmorillonite. A selection of their data is presented in the Table
below.

NaClO4 (N) Cdtotal (M) adsCd (%) NaClO4 (N) Cdtotal (M) adsCd (%)

0.01 0.40 97.1 0.05 0.40 73.6

0.80 96.3 0.80 69.2

Determine the exchange coefficient KNa/Cd for the Gaines and Thomas convention.

ANSWER:
Inspection of the data shows that the percentage of Cd2 adsorbed (adsCd) appears to be independent of
the amount of Cd2 added to the solution. In other words, the ratio of sorbed Cd2 over solute Cd2 is

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption isotherms and distribution coefficients 321

independent of Cdtotal, and the distribution coefficient is a constant. The concentration of Na has a marked
effect, however, and we expect that cation exchange is operative. The reaction is:

Na  0.5Cd-X 2 ↔ Na-X  0.5Cd 2

with

 Na [Cd 2 ]0.5
K Na\Cd  (7.18)
[Na ] 0.5
Cd

We have X  exchange capacity of the montmorillonite, 2.05103 M,

[Na]  mNa  solution normality N,
Cd X / 2  Cdtotal adsCd / 100, or Cd  Cdtotal adsCd / (50 X),
[Cd2]  mCd 2  Cdtotal(1  adsCd / 100),
and assume Na  1. This gives in (7.18):

1 (1  adsCd / 100)
K Na\Cd 
N adsCd / 50X
and we obtain: for N  0.01: KNa\Cd  0.55 and 0.63, and for N  0.05: KNa\Cd  0.38 and 0.43. Compare
with Table 6.4.

The distribution coefficient formulae generally work quite well for metal ions such as Ni2, Zn2,
Cd2, Sr2 and alkaline metals on montmorillonites (Garcia-Miragaya and Page, 1976;
Shiao et al., 1979; Baeyens and Bradbury, 1997), but deviations occur. For example, in Figure 7.9, dis-
tribution coefficients are plotted for various metal ions at trace concentration in exchange with Na
on montmorillonite. We replace (Ca / [Ca2])0.5 with (Na / [Na]) in Equation (7.15) and expect to
find, in a plot of log K d versus log mNa, a slope of 2 for Sr2 and Co2, and of 1 for Cs.

100000

10000
Distribution coefficient (mL/g)

1000

100

Sr pH 5
10 Eu pH 5
Am pH 6.5
Co pH 5
Cs pH 5
1
0.01 0.1 1 10
Na concentration (M)

Figure 7.9. Distribution coefficients of various radionuclides as a function of the Na concentration, on a Na
smectite (Shiao et al., 1979).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

322 Sorption of trace metals

However, the experimental data in Figure 7.9 indicate that the distribution coefficient for Co2 has a
slope of 1, similar to Cs, while americium is apparently so strongly sorbed that an increase of the
Na concentration has no effect at all. The slope for Eu3 is 3 at low concentrations of Na, but dis-
plays an unexpected minimum at higher concentrations. Clearly, the high Na concentrations are
beyond the usually encountered range in fresh ground water, but the anomalies are important for
assessing the behavior of radionuclides at waste sites. We have already calculated that in a natural soil,
specific sorbers may be present which bind heavy metals very strongly (Examples 7.1, 7.2). Likely
candidates for these solids are the edges of clay minerals, oxyhydroxides and organic matter that we
will discuss in the following sections.

7.3 VARIABLE CHARGE SURFACES

Goethite (-FeOOH) is a ubiquitous mineral in soils and aquifers and is probably the most often
used mineral for studying sorption behavior of trace metals because it can be easily synthesized in
the laboratory. The primary structural unit is an octahedron of 6 oxygens which coordinates to the
central Fe3 ion (Figure 6.6 displays an octahedron). In the crystal structure of goethite, the octa-
hedra share oxygens to form rows of two octahedra as illustrated in Figure 7.10. The rows are inter-
connected at the corners, and give the impression of being surrounded by open channels which are,
however, occupied by the protons. On the outside of the crystal, the channels become “grooves” in
the (010) and (01/0) faces. These can act as a trap for ions and may be responsible for the distinct
sorption behavior of goethite (Russell et al., 1974).
The tendency for oxygens to bind or to loose protons can be calculated with Pauling’s bond
valence concept and is related to the number of bonds to Fe3 atoms (Hiemstra et al., 1989; Venema
et al., 1996a). One Fe3 ion donates 3 / 6  0.5 bond valence to each of the 6 surrounding oxy-
gens. An oxygen connected to 3 Fe3 has a formal charge of 2  3  0.5  0.5. It needs half a
proton for charge balance and thus, will share a proton with another oxygen. All the internal oxygens
in the goethite structure are triply coordinated to iron, but at the surface of the crystal, doubly or
singly coordinated oxygens exist which are especially greedy for charge compensation by heavy
metal ions. The singly coordinated oxygens are indicated with “A” in Figure 7.10, the hydroxyls as
“B”, and the doubly coordinated oxygens as “C”. The singly coordinated oxygens are present on all
the faces which parallel the c-axis, but are most abundant on the {010} form.
The reactivity of goethite thus depends on the crystal morphology, which will change depending
on the growth stage of the crystal. The planes of the {110} form grow rapidly and are found on small
crystals in the laboratory and in soils (Cornell and Schwertmann, 2003). However, quickly develop-
ing faces fade away during extended growth and the two planes of the {010} form, which possess the
deep groove, become dominant on larger, natural crystals. Most experimental binding constants
apply to the surface properties of small, needle-like crystals, but it can be expected that the (010) and
(01/0) faces of larger goethite crystals provide more sites per given surface area and will also assert
a stronger influence on binding.

7.3.1 Titration curves with suspended oxide particles

The surfaces of Fe-oxides, clays, other minerals or organics may become protonated or deprotonated
and obtain different surface properties as a function of pH. In many ways surfaces behave similarly
to amphoteric solutes as illustrated in the following. First we titrate a volume of H2O with NaOH or
HCl. Electroneutrality requires that:
mH  mNa  mOH  mCl (7.19)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Variable charge surfaces 323

Corner sharing on (100) (Cd–Fe  3.77 Å) Corner sharing on (110) (Cd–Fe  3.77 Å)
C
DC
SC
DC B

SC

A
TC
SE

(100)
DE
Edge sharing on (021)
(Cd–Fe  3.3 Å)
b c
Corner sharing on (010) (110)
(Cd–Fe  3.77 Å) (010)
a
(021)

Figure 7.10. Atoms in the structure of goethite and, at the same orientation, morphology of the lath shaped
crystals. Cadmium adsorbs mainly in the form of a corner sharing octahedron with “A” type oxygens (see text).
The sorption coordination units are indicated by SC (single corner), DC (double corner), TC (triple corner),
SE (single edge) (Randall et al., 1999).

From the H2O dissociation constant we derive mOH  1014 / mH, neglecting the difference between
activity and molal concentration. The relation between [H] or mH and added base mNa is then:

mNa  1014 / mH  mH  mCl (7.20)

and we calculate the titration curve of pure water (Figure 7.11).

Next we add an amphoteric solute AH, i.e. a substance which can act both as acid and as base,
following the reactions:

AH 
2 ↔ AH  H ; K a1 (7.21)

and
AH ↔ A  H; Ka 2 (7.22)

where Ka1 and Ka2 are the acid dissociation constants. The species AH 
2 and A are included in the
electroneutrality equation, which becomes:
mNa  1014 / mH  mH  mCl  mA  m  (7.23)
AH2

where the species AH  

2 and A are obtained as function of [H ] with the speciation formula
(Equation 5.13). Again, we calculate the titration curve (Figure 7.11).
The titration curve for pure water (Figure 7.11) shows no buffering and the pH is a simple loga-
rithmic function of the amount of NaOH or HCl added. The amphoteric acid titration curve
shows two steps, corresponding to Reactions (7.21) and (7.22). AH 2 is the dominant species below
pH 7.29, AH is predominant between pH 7.29 and 8.93 and A above pH 8.93.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

324 Sorption of trace metals

11


(AH2  AH  A)
9

pH
7

Goethite
5

H2O
3
2 1 0 1 2
NaOH added (mM)

Figure 7.11. Titration-curves, showing the change of pH with addition or subtraction of NaOH (subtraction
means addition of HCl) to pure H2O, 1 mM AH, and a suspension of goethite. The amphoteric acid AH and
goethite have both log Ka1  7.29 and log Ka2  8.93 as intrinsic acidity constants. Background electrolyte
is 0.01 N NaCl.

The protonation/deprotonation reactions at the surface of goethite are:

˜ FeOH 
2 ↔ ˜ FeOH  H ; ′
Ka1 (7.24)

and
˜ FeOH ↔ ˜ FeO  H; ′
Ka2 (7.25)
where Ka1 and Ka2 are apparent equilibrium constants. For comparison, the titration-curve of goethite
is shown in Figure 7.11 using the same values of Ka1  and Ka2 as for AH and an equivalent amount of
acid neutralizing capacity. The titration curve of goethite shows a more gradual variation without the
stepwise changes of AH.
In solution, the individual AH molecule is not affected by dissociation of other AH molecules.
On the surface of goethite, however, the adsorption sites are positioned closely together. If one group
has lost a H, it will be more difficult to desorb a H from the neighbor groups since the increased
negative charge holds the remaining H more strongly. When part of the ˜FeOH surface sites has
dissociated into ˜FeO and H due to increasing pH, the apparent dissociation constant Ka2  for the
remaining ˜FeOH will decrease. Likewise for the association of ˜FeOH with H, as more
˜FeOH 2 has formed in response to decreasing pH, the more difficult is the attachment of addi-
tional H, and the apparent Ka1  will increase. The gradual change in the goethite titration curve is
therefore due to the development of electrostatic charge on the goethite surface. When studying the
association of ions with mineral surfaces one needs to consider both the chemical binding of the ion
at the surface and the charge development on the surface (Dzombak and Morel, 1990).

7.3.2 Surface charge and point of zero charge, PZC

The surface charge of the solid is the sum of structural deficits, unbalanced bonds at the crystal
surface and charge generated by the adsorbed ions:
  0  i  0  H   M   A (7.26)
 is the specific charge of the mineral (eq/kg) and the sum of 0, the structural charge due to crystal
structure substitutions, and i which is due to the various complexed ions as specified below.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Variable charge surfaces 325

H is the charge due to protons:

{
H  ˜SOH
2  ˜SO

} { } (7.27)

For cation sorption the surface charge contribution is given by M:

{
 M  (m  1) ˜SOM (m1) } (7.28)
From sorption of anion Aa, the surface charge is:

 A  (a  1) ˜SA(a1) { } (7.29)

The anions also comprise oxyanions such as PO3 4 .
The sign of the surface charge can be measured using the electrophoretic mobility of suspended
particles in an electric field. An example is shown in Figure 7.12 with rutile (TiO2) particles placed in
solutions of different composition (Fuerstenau et al., 1981). At the lowest Ca(NO3)2 concentration
(Figure 7.12a) it is protonation that determines the surface charge. At low pH the rutile surface is pos-
itively charged and the particles move towards the negative electrode, at high pH the surface is nega-
tively charged and the particles migrate to the positive electrode. The point where the surface charge is
zero in these measurements, is called the iso-electric point (IEP). For the rutile used, it was pH 6.5.

4 4
Electrophoretic mobility (µm/s per volt/cm)

Ba
1.67103 M

0.33103 M Sr
2 2 Ca
Mg
4
1.6710 M

0.33104 M
0 0

1.67105 M

2 2
Rutile 0.33105 M
Rutile
0.33103 M Nitrate salt
Na
Ca(NO3)2
4 6 8 10 4 6 8 10
(a) pH (b) pH

Figure 7.12. Dependence of the electrophoretic mobility of rutile particles on pH and electrolyte concentra-
tion of NO3-salts of alkaline-earth ions. (a) Mobility at various concentrations of Ca(NO3)2. (b) Mobility at an
electrolyte concentration of 0.33 mM of alkaline-earths. From Fuerstenau et al. (1981).

As the Ca(NO3)2 concentration increases, more Ca2 will adsorb on the rutile surface and diminish
the negative surface charge. As the result, the electrophoretic mobility decreases and it can even become
reversed in the sense that the particles move towards the negative electrode. Thus, simple alkaline-earth
ions, which are normally considered as inert background-electrolytes, can reverse the surface potential
of oxides (Parks, 1990). In the low pH range the electrophoretic mobility of the rutile particles was
not affected by the Ca(NO3)2 concentration and clearly nitrate does not change the surface charge of
the rutile particles.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

326 Sorption of trace metals

Figure 7.12b compares the ability of different ions to change the surface charge of the rutile parti-
cles. At the one end, the Na ion seems to have little influence on the surface charge while at the
other end, Ba2 changes the surface charge to strongly positive even at the pH corresponding to the
iso-electric point. Apparently, ions have different abilities to displace protons from the surface.
Alkaline-earth ions like Ca2 and Mg2 act mainly as background electrolytes as long as the proto-
nated surface is positive or neutral and it is only when their concentration becomes high that they
may become potential determining ions. Heavy metals such as Cu2, Pb2, Cd2, but also Ba2, on
the other hand, have a very strong ability to displace protons from the surface and that is the reason
why heavy metal adsorption is strongly pH dependent (Figure 7.6).
The charge of the solid is compensated in the diffuse double layer (Section 6.6):
   D  0, (7.30)

where D is the charge of the double layer (eq/kg).

The distribution of ions at the particle surface is visualized in Figure 7.13. Alkali metals, includ-
ing Na, and acid anions such as Cl keep a shell of hydration water and bind via outer-sphere com-
plexes to the surface. Protons, heavy metals, and most oxyanions are found closer to the surface
where they bind via inner-sphere complexes to the structural oxygens. In Figure 7.13b a Cu2 ion
has displaced a H on a structural oxygen atom. Note also the inner-sphere, bidentate surface-com-
plex of phosphate. The binding mechanism and the environment around the molecules can be
deduced from electron resonance spectra, using various means of excitation (Hawthorne, 1988;
Atkins and de Paula, 2002). Lastly, at some distance away from the surface, the diffuse layer extends
into the solution, with counter-ions at higher and co-ions at lower concentration than in the free solu-
tion (Section 6.6).

(a) (b) Oxygen

Central ion

OH
Diffuse ion
P
Outer-sphere O
complex
Inner-sphere
complex F

Cu


H2 Cl

H Water
molecule
 
Na

s a  d

Figure 7.13. Sposito’s visualization of a charged surface showing inner- and outer-sphere bonding and ions in
the diffuse double layer. Figure 7.13b shows planes associated with the various types of bonding: “s” for sur-
face hydroxyl groups, “a” for inner-sphere complexes, “” for outer-sphere complexes and “d” for ions in the
diffuse layer (Stumm, 1992).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Variable charge surfaces 327

At a given pH, the proton charge will compensate all other charge, and   0. This pH is called the
point of zero charge, PZC, or pHPZC:
  0 at pHPZC . (7.31)

At the PZC, the surface has also zero potential, 0  0, and therefore PZC equals the iso-electric
point IEP when specific adsorption is absent. More precise definitions of PZC exist which account
for the various processes that determine the charge (Sposito, 1984). The Point of Zero Net Proton
Charge recognizes that the mineral may have a permanent structural charge, but at the pHPZNPC the
contributions of surface protonation and deprotonation reactions to the total charge are balanced:
H  0 at pHPZNPC . (7.32)

The Pristine Point of Zero Charge, pHPPZC, is the PZC for a mineral without structural charge, and
without specific adsorption of other ions than protons:
  H  0 at pHPPZC . (7.33)

Lastly, the Point of Zero Salt Effect, pHPZSE, indicates the intersection point of titration curves at
different ionic strengths:
d / dI  0 at pHPZSE . (7.34)
Table 7.3 lists PZC’s for a number of minerals. Iron oxides have PZC’s ranging from 8.5 to 9.3 and will
at most groundwater pH values be neutral or weakly positive. Birnessite,
-MnO2, has a PZC of 2.2 and
is negatively charged at the pH of most groundwaters. Minerals have a general capacity for anion
exchange (in the double layer) when the pH is below the PZC, and a cation exchange capacity when the
pH is above PZC. The capacity for exchange depends on the difference between the PZC and the pH of
the solution. Figure 7.14 shows adsorption of Na and Cl on the oxides SnO2 and ZrO2 as a function
of pH. The PZC of SnO2 is found at pH  4.8, and is slightly lower than the PZC of ZrO2 at 6.8. When
the pH is higher than the PZC the oxide surface adsorbs Na from solution, when the pH is below the
PZC, Cl is adsorbed. Both ions are sorbed in small, equal quantities at the PZC of the oxide.

Table 7.3. The Point of Zero Charge, pHPZC , of clays and common soil
oxides and hydroxides (Parks, 1967; Stumm and Morgan, 1996; Davis and
Kent, 1990; Venema et al., 1996a).

pHPZC

Kaolinite 4.6 (Parks)

Montmorillonite 2.5 (Parks)
Corundum, -Al2O3 9.1 (Stumm and Morgan)
-Al2O3 8.5 (Stumm and Morgan)
alpha-Al(OH)3 5.0 (Stumm and Morgan)
Hematite, -Fe2O3 8.5 (Davis and Kent)
Goethite, -FeOOH 9.3 (Venema et al.)
Fe(OH)3 8.5 (Stumm and Morgan)
Birnessite,
-MnO2 2.2 (Davis and Kent)
Rutile, TiO2 5.8 (Davis and Kent)
Quartz, SiO2 2.9 (Davis and Kent)
Calcite, CaCO3 9.5 (Parks)
Hydroxyapatite, Ca5OH(PO4)3 7.6 (Davis and Kent)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

328 Sorption of trace metals

Uptake of ions (mol/kg solid)

1.0
ZrO2

SnO2
SnO2

0.5 Cl
Na
ZrO2

0
0 1 2 3 4 5 6 7 8 9 10 11 12
pH

Figure 7.14. pH-dependent sorption on SnO2 and ZrO2: Cl is adsorbed at low pH, Na at high pH (Kraus
and Phillips, 1956).

7.3.3 Sorption edges

The pH dependency of metal sorption on oxides can be depicted in a plot of sorbed fraction versus
the pH. Figure 7.15 shows adsorption curves for three divalent metals on rutile. The pH where 50%
of the total metal in the suspension is sorbed, is called the pH50 or sorption edge. In Figure 7.15 the
pH50 is at pH  5.5 for Cd2, at 3.7 for Cu2, and at pH  2.0 for Pb2. The sorption edge of all these
cations moves to a lower pH when the concentration of rutile particles increases.

100

Cu2
80 Cd2
Pb2

60
% Sorbed

 % Sorbed
40

20

 pH
0
0 1 2 3 4 5 6 7 8 9
pH

Figure 7.15. Adsorption of some divalent metal ions on TiO2 (rutile). Comparison of pH-values at 10 and at
90% adsorption allows the reaction mechanism to be determined. Modified from Schindler and Stumm (1987).

Rutile has its PZC at ca. pH  5.8 and in contrast to the preceding example (Figure 7.14), the heavy
metals are sorbing on the rutile surface at a pH below the PZC, i.e. in the pH range where the oxide
surface has a positive charge. The reaction where a heavy metal (M2) competes with a proton for
the surface site is:
˜SOH  M m ↔ ˜SOM (m1)  H; KSOM (7.35)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Variable charge surfaces 329

This is called a monodentate ligand formation. The heavy metal with a divalent charge may also bind
to two surface sites, and the reaction becomes:
2 ˜SOH  M 2 ↔ (˜SO)2 M  2H; K ( ˜ SO) (7.36)
2M

The association of a metal ion with two surface sites is often termed bidentate ligand formation.
However, the binding of the metal ion to two oxygens at the mineral surface is not an actual biden-
tate bond. Because the oxygen atoms at the mineral surface are positioned closely together, they act
as one unit in the law of mass action and Reaction (7.36) is normally written as:

( ˜SOH)2  M 2 ↔ (˜SO)2 M  2H; K ( ˜ SO) (7.37)

2M

— —
If we calculate the ratio M / M, where M is the adsorbed and M the solute concentration, for a high
and a low metal sorption ratio, e.g. of 10 / 1 and of 1 / 10, we find for Equation (7.35):
[H ]high sorption 1 [˜S
SOH]high sorption
 (7.35a)
[H ]low sorption 100 [˜SOH]low sorption

and for Equation (7.36):

[H ]high sorption 1 [˜SOH]high sorption
 (7.36a)
[H ]low sorption 10 [˜SOH]low sorption
—
If M is present in trace quantities only, [˜SOH]high sorption  [˜SOH]low sorption, and the increase of
the adsorption ratio from 1 / 10 to 10 / 1 requires a pH-increase of either 2 units (for Equation 7.35)
or of 1 unit (for Equation 7.36). This pH-increase corresponds to an adsorption increase from 9.1 to
90.9%. Inspection of Figure 7.15 shows that pH is 2.0 for Cd2, and about 1.6 for Cu2 and Pb2.
It suggests that Equation (7.35) is appropriate for Cd2 on rutile, while a combination of mono- and
bidentate bonds could be surmised for Cu2 and Pb2.
However, the situation is more complicated for Cu2 and Pb2. The sorption edge for these met-
als is located at a low pH where [˜SOH 2 ] is the dominant surface species. When the pH increases,
[˜SOH] will increase and the assumption [˜SOH]high sorption  [˜SOH]low sorption is no longer
valid. Instead, in Equations (7.35a) and (7.36a), the ratio [˜SOH]high sorption / [˜SOH]low sorption
increases and therefore pH becomes smaller. According to (7.35), sorption of a divalent metal ion
will increase the positive surface charge and surface potential which will also favor [˜SOH] as
compared to [˜SOH2 ]. In conclusion, a steep sorption edge can result from, either bidentate (7.36a)
versus monodentate (7.35a) binding near the PZC where [˜SOH] is the dominant surface species,
or at a low pH because of competition between [˜SOH] and [˜SOH 2 ].

Sorption of arsenate and arsenite

Oxyanions will be attracted to positively charged surfaces and repelled from negative surfaces and
so will exhibit high sorption at low pH and low sorption at high pH. We may use arsenic as an exam-
ple. Arsenic is found in groundwaters as arsenate, As(5), (Smedley et al., 2002) and arsenite, As(3)
(Mukherjee et al., 2000). Both As(5) and As(3) form protolytes which may release protons stepwise
in a similar way to carbonic acid. For As(5), the first dissociation constant of H3AsO4, is
log K1  2.24 and this species is therefore rarely important. The second dissociation reaction is:
H 2 AsO 2 
4 ↔ HAsO 4  H ; log K  6.9 (7.38)
At pH  6.9, H2AsO 4 is the dominant form and at pH ! 6.9 HAsO4 predominates. These two
2

As(5) species will be most abundant in groundwater.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

330 Sorption of trace metals

(a)  Oxygen
100  Iron
0.27 mM As(3) As–Fe  3.38 Å
 Arsenic
80  Hydrogen
As adsorbed (mmol/kg)

60 y
z
As–Fe  3.57 Å
40 0.13 mM As(3)
(b)
As–Fe  3.38 Å
20 0.27 mM
As(5)
0
2 4 6 8 10 12 x
pH y
z
As–Fe  3.57 Å

Figure 7.16. Left; pH dependent sorption of As(3) and As(5) on goethite (2.5 g/L). Right; section through
goethite along the (001) plane showing the position of adsorbed As(3) on the (110) plane (Manning et al., 1998).

For As(3) the first dissociation reaction is:

H3AsO3 ↔ H 2 AsO 
3 H ; log K  9.2 (7.38a)
Therefore at pH  9.2, the uncharged H3AsO3 is the predominant species.
Figure 7.16 shows the sorption of As(3) and As(5) to goethite. For As(5) there is a steady decrease
in sorption as pH increases corresponding to the decreasing positive surface charge and increasing
importance of the HAsO42 species. For As(3) there is a sorption maximum at pH 4–9.2 with sorption
decreasing both at higher and lower pH. Apparently, H3AsO3 is outcompeted from the surface by pro-
tons when the pH is less than 4. Essentially all of the arsenite is sorbed on goethite in the range
7  pH  10 from the 0.133 mM As(3) solution, corresponding to 53 mmol/kg adsorbed As (Figure
7.16). In the 0.267 mM As(3) solution, up to 84% of the total concentration of 107 mmol/kg is adsorbed.
At near neutral pH the adsorption of uncharged As(3) is larger than of negatively charged As(5) (but this
depends on the concentration of As, cf. Problem 7.6). The right side of Figure 7.16 shows a section
through goethite along the (001) plane with the double row of Fe-centered octahedra. The As atoms
adsorb on the (110) plane, sharing oxygens with the Fe3 octahedra in the mineral structure.

Sorption on Mn-oxide, clay and calcite

Manganese oxides are highy effective in sorbing heavy metals and the low PZC of birnessite (Table 7.3)
indicates that the surface will be negatively charged under most conditions. Sorption of various trace
metals on the manganese oxide birnessite,
-MnO2, is illustrated in Figure 7.17 and shows strong
sorption even at a low pH. For example, Pb2 is 100% sorbed at pH 5. Sorption of transition metal
ions on birnessite is only partially reversible (Murray, 1975; McKenzie, 1980). The structure of bir-
nessite allows an easy uptake of metal ions in the octahedral layer when the atomic radii fit. Such
behavior is transitionary between adsorption processes and solid solution behavior. However, for
alkaline-earths and Mn2, adsorption is reversible. The sorption site on the birnessite surface is cen-
tered on 3 oxygens surrounding a vacancy in the octahedral layer with a charge of 2. The surface

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Variable charge surfaces 331

complexation model based on this diprotic site contains the following reactions (Appelo and
Postma, 1999):
˜ MnO3H 2  H ↔ ˜ MnO3H
3 (7.39)

˜ MnO3H 2 ↔ ˜ MnO3H  H (7.39a)

˜ MnO3H ↔ ˜ MnO32  H (7.39b)

˜ MnO32  M m ↔ ˜ MnO3 M ( m2) (7.39c)

˜ MnO3H  M m ↔ ˜ MnO3HM ( m1) (7.39d)

The metals M m bind via inner-sphere complexes to the three oxygens. The association reactions
(7.39c and d) are similar to complexation of metal ions to diprotic acids such as H2CO3.
The diprotic sorption site has a distinctive broadening effect on the sorption edge. While for most
oxyhydroxides an increase in the adsorption ratio from 1 / 10 to 10 / 1 was covered in 1–2 pH units
(Figures 7.6 and 7.15), this needs 5 pH units or more for birnessite (Figure 7.17). As the pH
increases, more and more negative charge accumulates on the surface and desorption of additional
H becomes increasingly more difficult. Because the diprotic site is divalent, the potential effect
increases more with pH than for other oxides.

2.0

1.8 Pb

1.6 Cu
Mn
Co
Sorbed (mmol/g)

1.4
Zn
1.2 Ni

1.0

0.8

0.6

0.4

0.2

0
1 2 3 4 5 6 7 8
pH

Figure 7.17. pH dependent sorption of trace metals to birnessite (
-MnO2). Sorption is expressed as mmol
trace metal per gram of birnessite. The total amount of trace metal added is 2 mmol trace metal per 2 g of
birnessite (McKenzie, 1980).

The oxygen groups on surfaces of clay minerals may be protonated and deprotonated in a similar
way and display pH dependent sorption in addition to the permanent negative charge which accounts
for most of their cation exchange properties (Chapter 6). Figure 7.18 shows sorption
of a trace concentration of Ni2 on the surface of montmorillonite. The background electrolyte in
the experiments is 0.33 mM Ca(NO3)2 and since the affinity of Ca2 and Ni2 for ion exchange
sites is almost the same (Table 6.4), Ca2 fills most of the constant charge exchange sites.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

332 Sorption of trace metals

Nickel is only present in trace amounts and shows pH dependent sorption somewhat like on oxides.
However, at pH  5 about 10% of total Ni2 remains sorbed to the constant charge exchange sites
instead of going down to zero as on ferrihydrite (cf. Problem 7.19). Accordingly, clay minerals show
additive sorption behavior towards heavy metals, featuring both pH dependent specific adsorption
and pH independent ion exchange (Bayens and Bradbury, 1997; Kraepiel et al., 1999). Note in
Figure 7.18 that sorption diminishes at pH ! 10 due to complex formation in solution.

100

80
% Ni sorbed

60

40

20

0
3 4 5 6 7 8 9 10
pH

Figure 7.18. Sorption of trace amounts of nickel on montmorillonite. The background electrolyte concentra-
tion is 3103 M Ca(NO3)2 and the total Ni2 concentration around 108 M (Bradbury and Baeyens, 1999).

Also calcite which is ubiquitous in aquifers, may be important as a sorbent. When trace metal con-
centrations are very low, sorption on calcite may govern the aqueous concentration, while at higher
concentrations solid solutions may become controlling. Zachara et al. (1991) determined sorption of
various cations on calcite as function of pH (Figure 7.19). A 107 M solution of the divalent cation
was added after the calcite had equilibrated at a given pH, and the loss from solution was recorded as
percent adsorbed. Zachara et al. (1991) interpreted the sorption behavior as the result
of competition between Ca2 and the foreign ion for surface sites. Sorption increases with pH (Figure
7.19), because the Ca2 concentration decreases towards higher pH to maintain calcite equilibrium.

100 CO2  10
3.46

2 7
Me  10 M
25 g/L CaCO3 Cd
80 Zn
% Adsorbed

60 Mn Co
Ni

40

20 Ba

Sr
0
7 8 9
pH

Figure 7.19. Trace metal sorption on calcite in a system brought in equilibrium with calcite at varying pH
(Zachara et al., 1991).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Variable charge surfaces 333

At a given pH, sorption decreases in the order Cd2 ! Zn2 ! Mn2 ! Co2 ! Ni2 ! Ba2, Sr2
(Figure 7.19). Since sorption of an alien cation on the crystal sites impedes the dissolution and growth
of calcite, this order of binding strength is expected to reflect the inhibitory drive of the cations.
Comparison with Figure 5.35, which illustrates the growth inhibition effect of heavy metals on calcite
and aragonite, indeed gives the same sequence: Zn2 ! (Co2 ( Mn2 ( Ni2) ! Sr2, Ba2.
Sorption of protons and other species on calcite and siderite has also been measured with a rapid
titration technique that prevents calcite dissolution by Charlet et al. (1990) and subsequently on the
much slower dissolving magnesite by Pokrovsky et al. (1999). These data were modeled with a sur-
face complexation model while accounting for electrostatic effects (Van Cappellen et al., 1993).
However, the results indicate that electrostatic effects are minor on the surface of carbonates.

7.3.4 Sorption, absorption, and coprecipitation

Before continuing with the modeling of sorption, we will have a brief look at the complexities of the nat-
ural processes. Sorption comprises a whole suite of reactions ranging from adsorption to solid solution
formation. There is a continuum going from one stage to the other and the processes may be sequential.
Often there is an initial fast adsorption step followed by a slow step where the adsorbed species are
incorporated into the crystal structure to form a solid solution. This is for example observed for Cd2 on
the calcite surface (Davis et al., 1987; Stipp et al., 1992). For Mn-oxide it was already men-
tioned that adsorbed heavy metal ions can diffuse into octahedral positions and build a solid
solution. Experimentally these phenomena are noted by the irreversible adsorption of ions on mineral
surfaces.

100
90
80
70
% Adsorbed

60
50
72 hours
40
30
20
Desorption
2 hours
10 Adsorption
0
4 5 6 7 8 9
pH

Figure 7.20. Adsorption/desorption of 105 M Zn2 on 103 M Fe-ferrihydrite, I  0.1. Desorption was ini-
tiated after holding the system at pH 9 for respectively 2 and 72 hours (modified from Schultz et al., 1987).

Figure 7.20 shows experimental results for adsorption and desorption of Zn2 on ferrihydrite.
Desorption was initiated by lowering the pH of suspensions that were aged at pH 9. A short aging
period of two hours produced a desorption curve almost similar to the sorption edge (Figure 7.20).
However, when the aging period was increased to 72 hours, only about 60% of the Zn2 could be

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

334 Sorption of trace metals

desorbed and the remainder stayed bound in the ferrihydrite structure. Similar time-dependent sorp-
tion phenomena have been noted by e.g. McKenzie (1980), Brümmer et al. (1988), Barrow et al.
(1989), and Gerth et al. (1993).
Fuller et al. (1993) investigated both the adsorption of arsenate on ferrihydrite, and coprecipita-
tion of arsenate in ferrihydrite. After 24 hours of reaction approximately twice as much arsenate was
removed from solution by precipitation as compared to adsorption, while the As / Fe molar ratio in
the precipitate became as high as 0.2. During adsorption a continued slow uptake of As(5) was
observed. However, during coprecipitation some of the arsenate was released over time due to
recrystallization of the ferrihydrite.
Irreversible sorption phenomena are difficult to deal with in field and modeling studies. Usually,
an emphasis is given to reversible surface complexation models, as is done in the remainder of this
chapter, but it should be noted that we are dealing with relatively low concentrations of trace metals.
In general, these models are not as definitive and robust as the model for cation exchange of major
ions discussed in Chapter 6.

7.4 SURFACE COMPLEXATION

Sorption involves two effects; a chemical bond between the ion and the surface atoms, and an elec-
trostatic effect that depends on the surface charge. These yield two terms in the equation for the
Gibbs free energy of surface complexation:

G des  G chem  G coul (7.40)

The Coulombic term (here for a desorption reaction) reflects the electrical work required to move
ions away from a charged surface. Gcoul corresponds to the difference in energy of the state where
one mole of an ion resides at the surface with the potential 0, and the state where the ion resides in
the bulk of the solution with potential   0 (Volt). For desorption it is:

G coul  G 0  G 0  zF (0  0 )  zF0 (7.41)

where z is the charge of the ion, and F is the Faraday constant (96,485 C/mol).
The Gibbs free energy is related to the mass action constant by the equation (Chapter 4):

RT ln Ka  G des (7.42)

Substitution in Equation (7.40) yields:

RT ln Ka  G chem  G coul  RT ln K int  zF0 (7.43)

or:

zF0
log Ka  log K int  (7.44)
RT ln 10
Here Gdes results in an apparent dissociation constant Ka that can be measured experimentally.
Gchem gives Kint, the intrinsic dissociation constant describing the chemical binding. While Kint is a
constant, Ka must vary with the charge of the surface.
For example for dissociation of a surface proton:

˜SOH
2 ↔ ˜SOH  H (7.45)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Surface complexation 335

the apparent dissociation constant is:

[˜SOH][H ]
K a1  (7.46)
[˜SOH2]

The symbol Ka1 is used to indicate the first of two possible deprotonation steps of the surface.
The apparent constant for dissociation of ˜SOH into ˜SO is denoted as Ka2. It is difficult to
relate the concentration of a surface species to its activity. As discussed by Sposito (1983, 1984),
the surface complex has an activity of 1 when fully covering the surface, but in a charge-free envi-
ronment. Thus, surface activity includes the Coulombic term, but this term is not included in the
standard state.
The surface potential 0 can be calculated using the constant capacitance model (Schindler and
Stumm, 1987). This model considers the charged surface to be balanced by a parallel layer of counter
ions by analogy with a parallel plate condenser. It also assumes a linear relation between surface
charge and surface potential:
  e 0 (7.47)

where  is the specific surface charge in Coulombs/m2, and  is the specific capacitance in
Farads/m2. The form of Equation (7.47) can be derived from the simplified double layer equation for
small potentials:

 DL  2.29 I 0.5 0 (6.66)

The specific capacitance  depends both on the solution composition and the properties of the surface.
If only the surface species ˜SOH 2 is present, the surface charge is:

  F ˜SOH {
2 / As } (7.48)

where As is the specific surface in m2/kg, and {˜SOH 2 } is expressed in moles per kg adsorbing
solid. In the constant capacitance model, the surface potential is (Equation 7.47):

0 
F {˜SOH2 } (7.49)
e As

We can combine Equations (7.49) and (7.44), and obtain for the first dissociation constant, with
z  1 for H:

log K a1  log K int 

F2
e As RT ln 10
˜SOH
2{ } (7.50)

Part of the factor before {˜SOH

2 } is customarily lumped together as:

F2

e As RT

so that we write Equation (7.50) more simply as:


log K a1  log K int 
ln 10
{
˜SOH
2 } (7.51)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

336 Sorption of trace metals

12
1
2
11 3 4
5
10

9
0.31 g Al2O3/200 mL
pH
8
(1) 1 mM NaCl
7
(2) 5 mM NaCl

6 (3) 10 mM NaCl
5 (4) 50 mM NaCl
5 4
3 (5) 100 mM NaCl
21
4
1.0 0.8 0.6 0.4 0.2 0.2 0.4 0.6 0.8 1.0
Acid f1 Base f2

Figure 7.21. Effect of different concentrations of background electrolyte on titration curves of -Al2O3. f is
equivalent fraction of titrant added. Modified from Stumm and Morgan, 1981 (Aquatic chemistry, 2nd ed.).

and similarly for Ka2:

2
log K a2  log K int 
ln 10
{
˜SO } (7.52)

The factor  (and 2) can be obtained from the slope of a plot of the surface charge (i.e. of ˜SOH2 ,
or ˜SO) vs the apparent log Ka, and the intercept provides an estimate of Kint (cf. Stumm and
Morgan, 1996). The factors  and 2 should be identical for Ka1 and Ka2, as follows from the
derivation, but this is often not the case (Schindler and Stumm, 1987).
The factor  (and 2) must be determined for each background electrolyte concentration. The
constant capacitance model does not describe the effects of changes in concentration of the back-
ground electrolyte. A background electrolyte such as NaCl has the effect of equalizing the slope of
the titration curve over a large pH-interval, as illustrated in Figure 7.21. This equalization indicates
that surface hydroxyls dissociate more readily in the basic branch, while protons associate easier in
the acid limb. In other words, the effect of the electrostatic term decreases with increasing concen-
tration of NaCl. This shows (cf. Equations 7.47 and 6.66) that the surface capacitance () increases
as the concentration of the background electrolyte increases.

EXAMPLE 7.5. Calculate the specific capacitance (F/kg) of a -Al2O3 suspension for I  0.1 M NaCl
from the equation

{
pK a  7.2  14.5 ˜ AlOH
2 . }
Find the ratio {˜AlOH} / {˜AlOH 2 } at pH  7 and 6 for this ionic strength. The total number of sites is
0.03 mol/kg Al2O3. Also estimate the specific capacitance for I  0.001 M NaCl from the titration plot
shown in Figure 7.21.

ANSWER:
The capacitance As is contained in the slope of pKa vs {˜AlOH
2 } (Equation 7.50). In this case,

e As  F 2 /(14.5RT ln10)  1.1105 F/kg.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Surface complexation 337

From Equation (7.51) we have:

log K a1  log ({˜AlOH} / {˜AlOH })  pH  7.2  14.5 {˜AlOH }


2

2 (7.53)

At pH  7, we can neglect {˜AlO}, and therefore {˜AlOH

2 }  {˜AlOH}  0.03 mol/kg. Set
{˜AlOH2 }  x, and fill in Equation (7.53):

log((0.03  x ) / x )  0.2  14.5x,

which can be solved by trial and error to give x  {˜AlOH 2 }  0.0148 mol/kg. Thus, the ratio {˜AlOH} /
{˜AlOH 2 }  0.0148/0.0152  1. (Without electrostatic contribution, their ratio would be 0.63).
Similarly, at pH  6, {˜AlOH 2 }  0.026 mol/kg, and the ratio {˜AlOH} / {˜AlOH2 }  0.15.


(Without electrostatics 0.064). Dividing the calculated {˜AlOH2 } by 0.03 mol/kg yields the fraction of
sites occupied by protons, 0.5 and 0.87 for pH 7 and 6, respectively, which can also be found from curve 5
in Figure 7.21.
For I  0.001 M we read from Figure 7.21, at pH  7, {˜AlOH 2 }  0.15  0.03  0.0045 mol/kg solid.
Hence {˜AlOH} / {˜AlOH 2 }  5.67, and from Equation (7.51) we obtain  / ln10  212. Similarly for
pH  6, {˜AlOH 2 }  0.35  0.03  0.01 mol/kg solid, and  / ln 10  140. We take the average, and
find the capacitance As  F2 / (176 RT ln10)  9.3103 F/kg. Thus, the capacitance has decreased by
about a factor of 10 for a 100 times dilution of the background concentration.

QUESTIONS:
Why is the ratio {˜AlOH}/{˜AlOH2} smaller when electrostatics are excluded?
ANSWER: the positive surface potential repels positive species.
Is the ratio also smaller (without electrostatics) when pH ! 8.3?
ANSWER: no, the surface carries a negative charge, so the ratio will be larger.

Example 7.5. shows that the capacitance of Al2O3 increases with the square root of the ionic
strength, in excellent agreement with the simplified double layer equation, Equation (6.66). For
larger potentials,  ! 25 mV, the full double layer equation should be used:

 DL  0.1174 I 0.5 sinh(F0 / 2 RT ) (6.64)

which also relates charge and potential approximately by the square root of the ionic strength. This
relation is used in various computer models for calculating surface complexation.

EXAMPLE 7.6. Calculate the surface potential for 2.6107 meq/m2 on -Al2O3 in 0.1 M NaCl solution.
Use the identities sinh(x)  (ex  ex) / 2 and arcsinh( y)  ln( y  ( y2  1) ⁄2).
1

ANSWER:
We have F   96485  2.6107  0.0251 C/m2. Divide by 0.1174  (0.1)0.5 and find 0.676. Hence
F0 / 2RT  ln(0.676  (0.6762  1)0.5)  0.633, and 0  32.5 mV.

QUESTION:
Calculate the potential using the simplified double layer equation?
ANSWER: 0  34.7 mV.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

338 Sorption of trace metals

Experimentally, the surface characteristics are determined by titration at different concentrations of

a background electrolyte (Kinniburgh et al., 1995). The procedure is to start at a low electrolyte con-
centration, titrate forward (and backward if desired) to a given endpoint, and then add more elec-
trolyte. Adding background electrolyte at the acid endpoint will increase pH, while at the basic
endpoint it will decrease pH (cf. Figure 7.21). The amount of titrant which must be added to return
to the endpoint pH before salt was added, gives the starting point for the next titration in the graph.
The titrations will intersect at the point where the surface has zero charge, as in Figure 7.21.
Formally, this PZC is the point of zero salt effect, PZSE.
If a single intersection point is not obtained, other processes are operative. For example, sorption
of CO2 from the laboratory atmosphere is notorious for giving different forward and backward titra-
tion curves and CO2 must be removed by purging the suspension for extended time with nitrogen gas
under acid conditions (Evans et al., 1979). Sorption of CO2 is also known to lower the PZSE of
goethite from about 9 to 8.5 while the IEP is not affected (Zeltner and Anderson, 1988), whereas for
hematite both PZSE and IEP decrease from 9.3 to 8.5 (Lenhart and Honeyman, 1999). Modeling of
CO2 sorption experiments indicates that the neutral species is dominant on the surface (Appelo et al.,
2002). Thus, the IEP of goethite is not much affected, while the PZC takes into account the buffer-
ing by sorbed CO2 during the titration.
Figure 7.21 shows how adding a salt, such as KCl, will shift the pH of the suspension towards the
PZC. Usually, the pH of a soil in 1 M KCl solution is lower than the pH in distilled water, which indi-
cates that PZC is lower than the pH of soil water. This is typical for most soils in temperate climates
with clay minerals such as illite and smectite which have a PZC  3. However, tropical soils with
iron oxide as the main charging mineral often have a higher pH in 1 M KCl, which we can relate now
to the higher PZC of the oxides that are present in the soil.

7.4.1 Surface complexation models

Two types of models are commonly used for surface complexation: the two-layer model and the
triple layer model. In the first type, the diffuse double layer starts immediately at the charged
surface, and the surface potential is directly connected to the surface charge via the capacitance
(Equation 7.47). The capacitance is user-definable in the constant capacitance model, and defined
by the ionic strength in the double layer model via the Gouy-Chapman relation (Equation 6.64).
In the second type, three different layers with different capacitances are assumed, one layer at the
surface called the Stern layer, another layer starting at a closest approach distance (for hydrated
ions, but the distance is assumed equal for all), and one more corresponding with the diffuse double
layer. Three capacitances must be defined for the three layers, but only the one for the Stern layer
is usually adjusted (Davis et al., 1978; Davis and Kent, 1990). The capacitance for the inter-
mediate layer is usually fixed at 0.2 F/m (Davis and Kent, 1990), but varied by others (Hiemstra
and Van Riemsdijk, 1991). The capacitance of the diffuse double layer is obtained from the
Gouy-Chapman relation (Equation 6.64). Figure 7.22 shows the essential features of the two types
of models.
Surface binding can be purely electrical, as for ions in the diffuse double layer, or it can involve
a combination of complexation and electrical work. However, both model types consider only
complexed ions and completely ignore the composition of the diffuse double layer. In the two-layer
models, the potential in the Boltzmann term is identical for all the ions which complex to the sur-
face. In the triple layer model, ions can be assigned to bind directly to the surface (the zero plane),
or be separated from it by a hydration shell (the beta plane). This gives a choice of two potentials for
the modeling (Davis and Kent, 1990). In addition, the charge of the surface-bound ion can also be
distributed over the zero plane and the  plane which influences the distribution of potential because
the capacitances of the layers are different (Hiemstra and Van Riemsdijk, 1996).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Surface complexation 339

Diffuse layer
of positive Compact or
Localized negative counter ion Stern layer
surface charge charge o  d

 
   
   
Solid Solid Distribution of charge
 Solution 
Solution
 
 
o
d d
Gouy–Chapman model o  Stern–Grahame model

Positive
0 d 0 Distribution of potential
Negative 
o o

o  d  0 o    d  0

zeo zed
d  0.1174 l sinh d  0.1174 l sinh
2kT 2kT

o
o  d o   
C1
o   d
  d  
C2 C2

Figure 7.22. The two-layer and the triple layer models for calculating the electrostatic contribution to surface
complexation. In two-layer model, the diffuse double layer starts at the charged surface while in the triple layer
model, the diffuse double layer start two layers away from the surface.

Most chemical observations such as sorption-edges, proton-metal release ratios and ionic strength
effects can be very well described by all the models (Westall and Hohl, 1980; Stumm, 1992; cf. also
Venema et al., 1996b). The triple layer model has more options to account for mechanistic details of
the sorption process (Hiemstra and Van Riemsdijk, 1996) and probably also for the physical behavior
of the particles in an electric field (Davis and Kent, 1990). However, Dzombak and Morel (1990) have
derived a comprehensive database for sorption on ferrihydrite with the double layer model which fits
many observations. Thus, the double layer model has become the de-facto standard for heavy metal
sorption modeling in natural environments, while the greater flexibility of the triple layer model can
be helpful for modeling detailed laboratory experiments (Venema et al., 1996b) and for incorporating
the charge distributions of oxyanions (Rietra et al., 1999). The double layer model is included in
PHREEQC, and the triple layer model can be computed with MINTEQ (Allison et al., 1991; a
Windows version can be found at www.lwr.kth.se/english/OurSoftware/Vminteq/index.htm).

EXAMPLE 7.7. Calculate the equivalent ionic strength for a constant capacity model
An arsenic sorption experiment on goethite reported by Manning and Goldberg, 1996, in
0.1 M NaCl, was modeled with the constant capacity relation: DL  1.06 . What is the ionic strength?

ANSWER:
The theory is in Section 6.6. For small potentials 25 mV, we use Equation (6.66), DL  2.29
I   1.06 , which gives I  1.06 / 2.29  0.463, and I  0.21 mol/L. For larger potentials we use the double
layer Equation (6.64), DL  0.1174 I sinh(F / 2RT )  1.06  which is solved by trial and error for I. For small
, we find again I  0.21 mol/L. For higher , I varies: I  0.1 mol/L for   80 mV, and I  0.069 mol/L
for   150 mV. Manning and Goldberg’s (1996) results can be well modeled with I  0.2 mol/L.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

340 Sorption of trace metals

Sorption site densities

For calculating the surface charge, the sorption density must be derived from sorption data (mol/kg)
and surface area of the solid (m2/kg), giving the specific number of sorption sites (sites/nm2,
mol/m2, or mol/mol metal-oxide). The number of sorption sites can also be estimated geometrically
from the crystal structure and morphology, but it is usually determined from acid-base titration for
protons, or from sorption plateaux for specific elements (Davis and Kent, 1990). The number of
sorption sites may be different for various elements.

Table 7.4. Measured sorption densities for various elements on goethite, after Davis and Kent, 1990.

Ion H F SeO2
3 PO34 Pb2
Sites/nm2 2.6–4.0 5.2–7.3 1.5 0.8 2.6–7

In Table 7.4, F has a higher sorption density on goethite than any of the other elements because it
exchanges with structural oxygens besides being sorbed on the positive goethite surface (Davis and
Kent, 1990). In principle, it is possible to calculate sorption for various sites, each with its own sorp-
tion density and set of competing ions, but most models consider no more than two types of sites.
The number of sites per nm2 can be recalculated to moles per liter, as shown in Example 7.8.

EXAMPLE 7.8. Recalculate sites/nm2 to mol/L for a suspension with 0.2 mol goethite/L (17.8 g/L). The
goethite has 2 sites/nm2 and a surface area of 40 m2/g.

ANSWER:
First recalculate sites/nm2 to mol/m2. There are 2 sites/nm2  21018 nm2/m2/(Na  6.0221023 sites/mol) 
3.32 mol/m2. Multiply by the surface area of goethite: 3.32  40 m2/g  133 mol/g  88.85 g/mol
goethite  11.8 mmol/mol goethite. Find the site concentration in mol/L: 11.8  0.2 mol/L  2.36 mmol/L.

7.4.2 The ferrihydrite (Fe(OH)3) database

Dzombak and Morel (1990) have obtained a coherent database for surface complexation on hydrous
ferric oxide (Hfo, or ferrihydrite) by fitting the results of numerous laboratory experiments. This
database is also available in PHREEQC. Dzombak and Morel’s model defines complexation reac-
tions for two sites on hydrous ferric oxide (Table 7.5), a strong site Hfo_sOH and a weak site
Hfo_wOH, while the Gouy-Chapman double layer equation is used for defining the surface poten-
tial as function of surface charge and ionic strength.
The database included in PHREEQC is obviously not complete, but constants for other elements
can be estimated with linear free energy relations (LFER). The LFER relations listed in Table 7.5 are
based on the regression of optimized surface complexation K’s with the KMOH for the first hydroly-
sis constant of the metal in water, or the second dissociation constant of the acid anion. For example,
the constant for surface complexation of carbonate:
Hfo_wOH  CO32  H  Hfo_wCO 3  H2O (7.54)
is estimated using:
HCO  2
3  H  CO3 ; log K a 2  10.33 (7.54a)
For reaction (7.54) the LFER relation (Table 7.5) gives log K2  6.384  0.724 
(10.33)  13.86. However, this value is much higher than the log K2  12.56, which was
optimized using measurements of CO2 sorption on ferrihydrite (Appelo et al., 2002).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Surface complexation 341

Table 7.5. Properties of Hfo (ferrihydrite) in Dzombak and Morel’s (1990) database, and sorption-reaction
equations for various ions. LFER stands for linear free energy relations and are used to estimate unknown values
by analogy to aquesous complexation reactions.

Weak sites Strong sites Surface area Mol wt PZC

Hfo_w, mol/mol Fe Hfo_s, mol/mol Fe m2/g m2/mol g/mol pH
0.2 0.005 600 5.33104 88.85 8.11

...... Reactions ......

Protons on strong sites:
Hfo_sOH2  Hfo_sOH  H log Ka1 7.29
Hfo_sOH  Hfo_sO  H log Ka2 8.93
Protons on weak sites:
Hfo_wOH 2  Hfo_wOH  H

log Ka1 7.29
Hfo_wOH  Hfo_wO  H log Ka2 8.93
Transition metals:
Hfo_sOH  Mm  Hfo_sOM(m1) H K1
LFER: log K1  4.374  1.166 log KMOH1
Hfo_wOH  Mm  Hfo_wOM(m1) H K2
LFER: log K2  7.893  1.299 log KMOH
Alkaline earth cations:
Hfo_sOH  M2  Hfo_sOHM2 K1
Hfo_wOH  M2  Hfo_wOM  H K2
Hfo_wOH  M2  H2O  Hfo_wOMOH  2H K3
Di- and monovalent anions:
Hfo_wOH  Aa  H  Hfo_wA(a1)  H2O K2
LFER: log K2  6.384  0.724 log Ka22
Hfo_wOH  Aa  Hfo_wOHAa K3
LFER: log K3  0.485  0.668 log Ka2
1)
KMOH is for: Mm  OH  MOH(m1). 2) Ka2 is for: HA(a1)  H  Aa.

In the optimization an uncharged complex appeared to be the dominant species, which in Dzombak
and Morel’s database is reserved for the trivalent anions only. The overall sorption of CO2 is there-
fore overestimated by the LFER as illustrated by a comparison of the optimized and LFER-derived
sorption envelopes (Figure 7.23).

1
Fet  8.7 mM
Fraction sorbed of CO2

0.8

0.6

0.4

0.2
Fet  0.87 mM

0
2 4 6 8 10
pH

Figure 7.23. Sorption of 4.6106 M CO2 on Hfo (0.87 and 8.7 mmol Fe/L) calculated with the double layer
model. Full lines are for optimized complexation constants and species, dotted lines are for Hfo_wCO
3 with a
complexation constant estimated using LFER. Datapoints from Zachara et al., 1987.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

342 Sorption of trace metals

Let us return to Examples 7.1 and 7.2 dealing with Cd2 sorption on sand. Can we explain the meas-
ured value of the distribution coefficient Kd  307 L/kg as due to sorption on iron oxyhydroxide in
the sand?

EXAMPLE 7.9. Sorption of Cd2 to iron oxyhydroxide in loamy sand

To measure the content of iron oxyhydroxides, Christensen (1984) extracted the sand with dithionite, and
found 2790 ppm Fe. Derive the Kd (L/kg) for Cd2.

ANSWER:
The iron oxyhydroxides in the soil probably consist of goethite, which is more crystalline than
ferrihydrite and has an about 10 times smaller surface area (for goethite prepared in the laboratory).
We assume therefore 10 times fewer sites as compared to ferrihydrite, i.e. 0.02 mol weak sites/mol Fe, and
5  104 mol strong sites/mol Fe, and also a 10 times lower specific surface area (60 m2/g) and grams of
“Ferrihydrite” in the PHREEQC input file.

SOLUTION 1
pH 6.0; pe 14 O2(g) -0.68
Ca 1; Cl 2; Cd 1e-6

SURFACE 1
# 2790 ppm Fe / 55.85 = 50 mmol/kg * 89 = 4.45 g ''Ferrihydrite''/kg
Hfo_w 1e-3 60 4.45 # 1e-3 mol weak sites, 60 m2/g specif. surf., 4.45 g ferrihydrite
Hfo_s 0.025e-3 # 0.025e-3 mol strong sites
-equil 1

USER_PRINT
-start
10 print ''K_d (L/kg) = '', (mol(''Hfo_wOCd+'')+mol(“Hfo_sOCd+''))/ tot(''Cd'')
-end
END

which gives in the output file:

--------------------------------------------------------User print----------------------------------------------
K_d (L/kg) = 4.2411e-01

QUESTIONS:
What is the Kd when there are 10 times more surface sites?
ANSWER: 10 times higher
What is the Kd at pH  7, 8 and 9?
ANSWER: Kd  23.5 at pH  7, Kd  151 at pH  8, Kd  864 at pH  9

The value of Kd  0.42 L/kg calculated for sorption of Cd2 onto iron oxyhydroxides in the soil
(Example 7.9) is much lower than the measured value of 307 L/kg. Therefore, iron oxyhydroxides
contribute very little to Cd2 sorption on sand at pH  6. This corroborates experiments of Zachara
et al. (1992) who determined sorption on soil separates and noted that the oxides may even block the
access of Cd2 to sorption sites on clay minerals. In near neutral or slightly acidic water, sorption to
iron oxyhydroxides is probably mostly important for Pb2 (Lion et al., 1982) and Cu2 (inspect the
PHREEQC database), and for oxyanions such as As and Se.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Surface complexation 343

7.4.3 Diffuse double layer concentrations in surface complexation models

Neither the double layer model, nor the triple layer model considers the composition of the diffuse dou-
ble layer. The surface charge is simply balanced by ions in solution, whereas in reality the double layer is
attached to the surface for charge balance. When a model run separates the solution and the solid, with-
out considering the double layer, then both will carry a net charge, which is of course impossible.
Let us imagine an experiment where 89 mg neutral ferrihydrite is weighed in 3 centrifuge tubes.
Next, 20 ml 0.01 M HCl is added to the first tube. After equilibration, the tube is centri-
fuged, and the fluid is poured into the next tube containing ferrihydrite, and so on. What is then the
Cl concentration in the third tube? The PHREEQC input file for this experiment is given in Table
7.6. Note that SURFACE 1 is defined with the “-diffuse_layer” option, for calculating the composi-
tion of the diffuse double layer.
The results in Table 7.6 indicate that the Cl concentration decreases from 10 to 2.16 mmol/L in
the first tube and to 0.009 mmol/L in the third tube. The Hfo has become protonated by the HCl solu-
tion and acquired a positive charge. The charge is balanced by an excess of anions in the diffuse layer
that remain with the surface when solution and solid are separated, and causes a decrease in the Cl
concentration. When the diffuse layer calculation is switched off (the default option in PHREEQC),
the Cl concentration remains 10 mmol/L. The positive charge of the surface is now counterbal-
anced by a negative charge balance in solution. The loss of protons is higher than when diffuse_layer
option is considered and the pH is calculated (wrongly!) to increase to 8.1 in the third tube.
The diffuse_layer option in PHREEQC is computer intensive and usually the results are insensi-
tive to its setting, for example, because the pH is determined by other equilibria. Therefore, modeling
of variable charge surfaces may often proceed with the double layer option off, but it should be ascer-
tained whether the effects are indeed minor for the problem at hand (cf. Problems 7.4 and 7.5).

Table 7.6. PHREEQC input file for modeling the reaction of Hfo with HCl in 3 centrifuge tubes, and
calculation results.

# Put 89 mg neutral Hfo in 3 centrifuge tubes...

SURFACE 1-3
Hfo_wOH 2e-4 600 0.089 # 0.2 mmol weak sites, 600 m2/g specif. surf., 89 mg Hfo
-diffuse_layer
END

# Take 1st tube, add 20 g of 10 mM HCl, react and decant in tube 2...
USE surface 1
SOLUTION 1; pH 2.0 charge; Cl 10; water 0.02
SAVE solution 2
END

# 2nd tube, react and decant in tube 3...

USE surface 2; USE solution 2
SAVE solution 3
END

# 3rd tube...
USE surface 3; USE solution 3
END

The output gives: Initial solution tube 1 2 3

Cl 10.0 2.16 0.17 0.009 mmol/L
pH 2.04 2.69 3.78 5.04

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

344 Sorption of trace metals

7.5 COMPLEXATION TO HUMIC ACIDS

Organic material is commonly extracted from soils and sediments with 0.5 M NaOH. When the result-
ing, dark-colored solution is brought to pH  1 with HCl, part of the organic matter precipitates.
The fraction which precipitates is called humic acid, while the fraction that remains soluble at
pH  1 is called fulvic acid. The organic extracts contain macromolecules with 60–10,000 atoms of
C, H, O, N and S and other elements, and molecular weights ranging from 500 to 1500 for fulvic
acid, and from 1000 to 85,000 for humic acid. Average elemental compositions of soil extracted
organic matter are given in Table 7.7.

Table 7.7. Average elemental composition of humic and fulvic acid extracted from soil. Data in Schnitzer and
Khan (1978), recalculated to 100% for the given elements.

C (mol %) H (mol %) O (mol %) N (mol %) S (mol %) Ash (g/kg)

Humic 40.2 38.8 19.9 1.0 0.08 30

Fulvic 35.1 39.5 24.7 0.6 0.1 15

In soil water the concentration of dissolved organic carbon (DOC) ranges from 0.1 to 3 mM, in
groundwater from 0.01 to 1 mM, and in rivers draining swamps the concentration may be as high as
5 mM. Small size organic acids and saccharides make up about 75% of the DOC in ground-
water and only about 25% consists of fulvic and humic acid (Thurman, 1985; Routh et al., 2001).
However, the contribution of humic acid may increase to over 90% in high pH, high DOC (!10 mM
C) groundwater (Artinger et al., 1999).
The charge on humic and fulvic acids can be determined by titrating with acid or base and, as
shown in Figure 7.24, it is strongly pH dependent. The charge usually also depends on the ionic
strength and an increase in ionic strength, by adding neutral salt, will tend to decrease the pH.

10
meq charge/(g fulvic or humic A)

Fulvic acid
8

Humic acid
4

2
3 5 7 9 11
pH
Figure 7.24. Titration curves for fulvic acid extracted from groundwater in  0.1,  0.03, 0.01 and 
0.003 M NaCl (upper curves) and for average humic acid (lower curve) in  0.1 M NaCl. Lines modeled
according to Tipping and Hurley’s model (Higgo et al., 1993).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 345

Aliphatic
dicarboxylic
acid
Hydroxy acid
OH OH
CO2H
CO2H
CH
CH
HO2C CH2 CO2H
Phenol
group
HO2C CH2 O
CH OH
CH2
Aromatic CH2 CH2
dicarboxylic OH
acid
OCH3 CH2 CH2 CO2H

CO2H Aliphatic
acid
Aromatic
acid

Figure 7.25. Various types of carboxylic acids and phenolic acid on humic acids (Thurman, 1985).

However, the ionic strength effects are irregular and within the measurement error in Figure 7.24.
The fulvic acid shown in Figure 7.24 has a total charge of 11.4 meq/g, while the humic acid has a
total charge of 7.1 meq/g. Not even at pH  11 are all sites deprotonated, and likewise, at pH  3
not all sites are protonated.
The charge buildup is attributed to two main groups of acids, carboxylic (R-COOH groups) with
pKa  5, and phenolic with pKa ! 8. The acidity of the functional groups is determined by their posi-
tion within the humic molecule, as illustrated in Figure 7.25. For example, the (linear) aliphatic acid has
a pKa of 4.8, while the two carboxylic groups on the aromatic ring have pKa’s of 2.9 and 4.4 (Thurman,
1985). One may expect a more or less Gaussian distribution of the acidity constants, akin to the variation
for simple organic acids shown in Figure 7.26. Furthermore, the apparent acidity constants will change
as charge builds up on the molecule, just as for oxide surfaces. Therefore, the approach used for model-
ing the deprotonation and complexation behavior of humics is similar to that for variable charge oxide
surfaces, in that a general electrostatic contribution is combined with an intrinsic surface complexation
K for the various ions. However, humic molecules have small, spherical or cylindrical shapes without a
fixed structure which complicates the electrostatic calculations, and the variation in the acidic groups
and their orientation within the molecule is much greater than for an oxide, for which at least some struc-
tural order exists. Example 7.10 shows how to estimate the size and charge density for humic particles.

EXAMPLE 7.10. Estimate the spherical surface area of a fulvic acid of 1000 g/mol, density 1 g/cm3. Also
estimate the surface charge density for fulvic acid at pH  7 (7.5 meq/g according to Figure 7.24).

ANSWER:
The volume of 1 molecule of fulvic acid is 1 cm3/1 mmol/6.0221020 (molecules/mmol)  1024 (Å/cm)3 
1661 Å3. The radius of the molecule is 7.3 Å, and the surface area is 4.081010 cm2/mol fulvic acid. The cal-
culated radius agrees well with physical determinations (Aiken and Malcolm, 1987; Buffle, 1990; Clapp
and Hayes, 1999) and estimates from electrostatic calculations (De Wit et al., 1993a). The molecule expands
with decreasing ionic strengths below 0.1 and with increasing charge or pH (Avena et al., 1999).
The humic acid has a charge of 7.5 eq/kg at pH  7, hence the charge density is
7.5  96485 C/mol/4.081010  17.7 C/cm2. Note that the charge density is high compared to oxide
surfaces, and that it will further increase 1:1 with the molecular weight for spherical particles if charge
increases 1:1 with the molecular weight (Bartschat et al., 1992).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

346 Sorption of trace metals

10 Other acidic groups (63)

0
2 4 6 8 10 12 14

20
Phenolic groups (95)
10

Number of tabulated pKa values

0
2 4 6 8 10 12 14

70
Carboxyl groups (409)
60

50

40

30

20

10

0
2 4 6 8 10 12 14
pKa

Figure 7.26. The acidity constants of simple organic acids show an approximately Gaussian distribution
(Perdue et al., 1984).

Compared to oxide minerals, the ionic strength has a relatively small effect on the pH titration curves
of humics, but a greater bearing on sorption of metals at a fixed pH (Bartschat et al., 1992).
Basically, the small effect on titration curves is due to the wide range of acidity constants in humics:
when the salt concentration increases at a given pHx, the proton dissociation increases (according to
electrostatic theory) for the group with pKa
pHx, but the protons are taken up by the next, more acid
group. The large effect on metal sorption indicates that fewer sites are responsible for metal sorption
(K’s are more restricted), or that ion exchange with the cation from the background electrolyte is
important. Electrostatics by itself, does not offer a sufficient correction to obtain the smooth and
gradual pH titration curves if only one type of site is available with two pKa’s as on oxide surfaces.
The necessary larger variation in intrinsic K’s is accounted for by taking various discrete sites
(Tipping and Hurley, 1992) or by assuming a continuous distribution of K’s (Perdue and Lytle, 1983;
De Wit et al., 1993b; Kinniburgh et al., 1999). The behavior of humics is accordingly dominated by
site heterogeneity rather than electrostatics. The electrostatic contribution is bypassed altogether by
some authors and proton release is modeled as an ion exchange process of H versus the cation of
the background electrolyte (Westall et al., 1995).

7.5.1 The ion association model

When present, the electrolyte effect can be explained as due to an ion association of the background
electrolyte with a charged fulvic acid (˜F). For example, Na with ˜F gives the reaction:
˜FNa ↔ Na  ˜F; K ˜FNa (7.55)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 347

with

[Na ][ ˜F ] [Na ]  ˜F

K ˜FNa   (7.56)
[ ˜FNa]  ˜FNa

where F.. is the fraction of the surface species on fulvic acid. The apparent acid dissociation
constant is calculated from experimental data:

[H ] f [H ] f
K a2   (7.57)
(1  f )  ˜FH

Here, f is the fraction which has dissociated, as found from a titration experiment, and is equal to:

f   ˜F   ˜FNa (7.58)

Combining with Equation (7.56) gives:

f   ˜F (1  [Na ] / K ˜FNa ) (7.59)

We introduce this relation in Equation (7.57). The quotient [H] F / FH has an apparent con-
stant defined by Equation (7.52) and therefore:

2  [Na ] 
log Ka2  log K int 
ln 10
{ }
˜F  log  1 


K ˜ FNa 
(7.60)

This equation shows that Ka2 decreases with increasing dissociation of the surface groups (increas-
ing pH), and increases when the salt concentration increases. If the dissociation constant K˜FNa
is small, the increase of Ka2 would be about ten-fold for every ten-fold increase in concentration.
This is indeed observed in titrations of polymer solutions (Marinsky, 1987). Titrations of humic and
fulvic acid follow a similar trend (Figure 7.27).

0.001 M NaNO3
84 ppm FA
66
6 118
0.010 M NaNO3
163 ppm FA
242
Apparent pKa

55
5 0.100 M NaNO3
1834 ppm FA
188
752

3
pKint
0 0.5 f 1.0

Figure 7.27. The variation of pKa of fulvic acid with degree of dissociation at three different ionic strength
levels. The ionic medium is NaNO3. Lines are drawn according to the constant capacitance/ion association
model. The fulvic acid is from Armadale Bh soil horizon. Experimental data after Marinsky (1987).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

348 Sorption of trace metals

Assume that the electrostatic term for a fulvic acid may be approximated by:
2
 ˜F   f f (7.61)
ln 10
so that Equation (7.60) can be rewritten as (with log K  p K):

 [Na ] 
pK a  pK int   f f  log  1   (7.62)
 K ˜FNa 

where the constants for the fulvic acid ˜FH follow the notation used earlier for a solid surface
˜SOH. Figure 7.27 shows an approximately linear increase of pKa with f, and also a decrease as the
concentration of NaNO3 increases. The value of K˜FNa can be estimated from two points in the plot,
using:

K ˜FNa  mNa,1  10 pKa,1 pKa,2 (K ˜FNa  mNa,2 ) (7.63)

The lines in Figure 7.27 were drawn with pKint  2.8, f  3.7, and K˜FNa  0.03.

7.5.2 Tipping and Hurley’s discrete site model “WHAM”

Tipping and Hurley (1992) and Tipping (1998) have formulated the model WHAM for complexing
of heavy metals by humic acid. The acronym WHAM stands for “Windermere humic acid model”,
for Lake Windermere in the English Lake District. In the model, the carboxylic (“HA”) and the phe-
nolic (“HB”) acid groups are each assigned 4 sites. Each site in turn is given a charge nHA(i) or nHB(i)
(meq/g), a pKa(i) for proton dissociation and a pKMHA or pKMHB for metal dissociation. The parame-
ters for the A and B sites are determined with the algorithm:
for i = 1 to 4:
nHA(i) = (1-fpr) x nHA/4; {...fixed charge on each site}
pKa(i) = pKHA + (2i - 5)/6 x pKHA; {...varying pKa}
pKMHA(i) = pKMHA; {...fixed metal complexation constant}
end.

and similar for HB.

The values for the various parameters are given in Table 7.8. Tipping and Hurley suggest that some
of the sites are so close together that they form bidentate bonds, controlled by a proximity factor fpr.
These bidentate sites have two pKa’s equal to the pKa of the constituent monoprotic sites, and a
pKMH2AB equal to the sum of the two individual pKM’s ((pKMHA  pKMHA) or (pKMHA  pKMHB)).

Table 7.8. Basic parameters for Tipping and Hurley’s (1992) model for complexation of protons and metals to
average humic acid.

nHAa nHBa fpr pKHA pKHB pKHA pKHB P

4.73 nHA/2 0.4 3.26 9.64 3.34 5.52 103

Metals Mg2 Ca2 Ni2 Cu2 Zn2 Cd2 Pb2

pKMHA 3.3d 3.2d 1.4 0.63 1.7c 1.52 0.81

pKMHB 7.1b 7.0b 4.5b 3.75 4.9b 5.57 3.04
a
nH in meq/g humic acid. b Estimated with the relation: pKMHB  2.57  1.38pKMHA.
c
Christensen and Christensen, 1999. d Lofts and Tipping, 2000.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 349

(Note that Table 2 of Tipping and Hurley, 1992, incorrectly assigns (pKMHA  pKMHB) to all sites,
Tipping, pers. comm.). Tipping and Hurley (1992) estimate a proximity factor fpr  0.4 which
allocates 40% of the total charge to 12 bidentate bonds. Perhaps, this can be seen more clearly by
inspecting the PHREEQC input file shown in Table 7.9.

The empirical electrostatic term

Tipping and Hurley (1992) mimicked the electrostatic Boltzmann factor, exp(zF/ RT), by an
empirical relation (for z  1):
exp(F0 / RT )  exp(2P log(I ) Z )  TH , (7.64)

where P is an adjustable coefficient (P  103 for average humic acid), and Z is the charge on the
humic acid, eq/g. The factor TH in Equation (7.64) varies with I and Z, and you may check that with
Z varying from 3 to 6 meq/g, for I  0.3, TH will change from 0.72 to 0.52 (low to high charge
Z). For I  0.003, TH changes from 0.21 to 0.044 (low to high charge). At higher charge TH
becomes smaller, and the cations are more attracted to the humic acid. Also, at a lower ionic strength,
TH decreases, because the capacitance around the charged particles decreases. This increases the
negative potential for a given charge (  / ), and draws cations stronger to the humic acid.
The double layer theory for a planar surface also links charge and ionic strength with 0, and we may
try to connect formula (7.64) with the double layer Equation (6.64). The relation between potential and
charge for a monovalent electrolyte according to double layer theory is (cf. Examples 6.12 and 7.6):

F0 / 2RT  ln(y  (y2  1) ⁄2)

1

–
where y  H / (0.1174
I ), and H  ZF / A is the specific charge on the humic particle (C/m2).
Thus, the equivalent double layer DL  exp(F0 / RT)  (y  (y2  1) ⁄2)2. For the desired con-
1

nection of TH  DL, we may now adapt A (in Example 7.7 we linked the constant capacity model
with the double layer model by adapting I, since that parameter was then the unique variable for the
double layer model). With a least squares fit we obtain:

A  159300  220800/ I 0.09  91260/ I 0.18 (7.65)

The equation implies an increase of the specific surface as the ionic strength decreases. Thus, the
calculated charge density becomes smaller at lower ionic strength, which reduces the Boltzmann
factor for humic acid. This is in agreement with the smaller effect of ionic strength on a spherical
surface (of humic acid) than on a planar surface (of ferrihydrite).

The Donnan term

The humic molecule carries a negative charge which radiates out to form a double layer, in which
cations are enriched in addition to the complex bound amounts. Tipping and Hurley (1992) and others
(Kinniburgh et al., 1999; Kraepiel et al., 1999) calculate the cations in the double layer with a so-called
Donnan model. In the Donnan model a single, averaged potential is assumed. The cations are enriched
in the Donnan double layer according to:
mi,DL  mi,∞ exp(zi FDL / RT ) (7.66)

where mDL is the concentration in the Donnan phase, where the potential is DL.
The ratio of, for example, Na and Ca2 in the two phases is:
mNa,DL / mNa,%  (mCa,DL / mCa,%) ⁄2
1
(7.67)
and the ions sum up to counterbalance the negative charge of the humic acid.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

350 Sorption of trace metals

Table 7.9. A PHREEQC data file for calculating metal complexation to humic acid according to Tipping and
Hurley (1992).

# download database T&H.DAT from www.xs4all.nl/~appt

DATABASE T&H.DAT
SURFACE_SPECIES
# Cd (Example in fig. 10 of Tipping and Hurley, 1992)
H_aH + Cd+2 = H_aCd+ + H+; log_k -1.500
H_bH + Cd+2 = H_bCd+ + H+; log_k -1.500
H_cH + Cd+2 = H_cCd+ + H+; log_k -1.500
H_dH + Cd+2 = H_dCd+ + H+; log_k -1.500

H_eH + Cd+2 = H_eCd+ + H+; log_k -4.640

H_fH + Cd+2 = H_fCd+ + H+; log_k -4.640
H_gH + Cd+2 = H_gCd+ + H+; log_k -4.640
H_hH + Cd+2 = H_hCd+ + H+; log_k -4.640

H_abH2 + Cd+2 = H_abCd + 2H+; log_k -3.000

H_adH2 + Cd+2 = H_adCd + 2H+; log_k -3.000
H_afH2 + Cd+2 = H_afCd + 2H+; log_k -6.140
H_ahH2 + Cd+2 = H_ahCd + 2H+; log_k -6.140
H_bcH2 + Cd+2 = H_bcCd + 2H+; log_k -3.000
H_beH2 + Cd+2 = H_beCd + 2H+; log_k -6.140
H_bgH2 + Cd+2 = H_bgCd + 2H+; log_k -6.140
H_cdH2 + Cd+2 = H_cdCd + 2H+; log_k -3.000
H_cfH2 + Cd+2 = H_cfCd + 2H+; log_k -6.140
H_chH2 + Cd+2 = H_chCd + 2H+; log_k -6.140
H_deH2 + Cd+2 = H_deCd + 2H+; log_k -6.140
H_dgH2 + Cd+2 = H_dgCd + 2H+; log_k -6.140
SURFACE 1
# For Psi vs I (= ionic strength) dependence, adapt specific surface A in PHRC:
# A = 159300 - 220800/(I)^0.09 + 91260/(I)^0.18
# Example: A = 34170 m2/g for I = 0.01 mol/L
# 1 g humic acid
H_a 7.10e-4 34.17e3 1
H_b 7.10e-4; H_c 7.10e-4; H_d 7.10e-4

H_e 3.55e-4; H_f 3.55e-4; H_g 3.55e-4; H_h 3.55e-4

H_ab 1.18e-4; H_ad 1.18e-4; H_af 1.18e-4; H_ah 1.18e-4

H_bc 1.18e-4; H_be 1.18e-4; H_bg 1.18e-4; H_cd 1.18e-4
H_cf 1.18e-4; H_ch 1.18e-4; H_de 1.18e-4; H_dg 1.18e-4
-equil 1
-diffuse_layer # diffuse layer calculations included
SOLUTION 1
pH 7.0
Cd 1e-6 Fix_cd -9.0
Na 10; N(5) 2 charge
PHASES
Fix_cd; Cd+2 = Cd+2; -log_k 0.0
SELECTED_OUTPUT
-file fig_10.prn; -reset false
-act Cd+2
-mol H_aCd+ H_bCd+ H_cCd+ H_dCd+ H_eCd+ H_fCd+ H_gCd+ H_hCd+\
H_abCd H_adCd H_afCd H_ahCd H_bcCd H_beCd H_bgCd H_cdCd H_cfCd H_chCd H_deCd
H_dgCd
USE solution none
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 351

Equation (7.67) has the form of an exchange formula, with K(7.67)  1. However, when the different
standard state of exchangeable ions is taken into account, it appears that K(7.67)  KCa\Na (2 CEC) ⁄2,
1

where CEC is the charge of the humic acid in eq/L. In Section 6.6.1 we calculated that the enrich-
ment of cations in a double layer follows exchange theory with KI\Na
1, and the simple Donnan
model may not be correct. Generally, the Donnan model underestimates the contribution of the
higher-charged ions (Problem 7.22).

An example calculation
A PHREEQC input file for Tipping and Hurley’s model is given in Table 7.9. The 4 carboxylic sites
are defined as H_a, H_b, H_c and H_d, and the 4 phenolic sites are H_e..H_h. The bidentate sites are
entered separately with letters noting the corresponding monoprotic bonds (H_ah is the combination
of sites H_a and H_h). The input file is for calculating the distribution of Cd2 over the various sites
of average humic acid and the double layer and uses the database T&H.DAT that can be downloaded
from www.xs4all.nl/appt. Diffuse layer calculations are included, which diminishes the surface
potential and reduces sorption of Cd2. Output in file “fig_10.prn” closely matches the results
depicted in Figure 10 in Tipping and Hurley (1992).
Again, we return to Example 7.1, since we still want to know whether the organic carbon content
can explain the Kd  307 L/kg for sorption of Cd2 on sand.

EXAMPLE 7.11. Sorption of Cd2 to organic matter in loamy sand

Christensen (1984) analyzed 0.7% OM (Organic Matter,
0.35% OC) in the sand. Find the Kd (L/kg) for Cd2.

ANSWER:
Let’s make an input file for the actual experiment, 1 g soil or 7 mg OM reacts with 50 mL solution with a
trace concentration of Cd2. The data for Tipping and Hurley’s model can be downloaded from
www.xs4all.nl/appt and will not be repeated here, only the actual calculation of the distribution coeffi-
cient is presented:
# Cd sorption on OC in Christensen's xpt
DATABASE T&H.DAT # from www.xs4all.nl/~appt
PRINT; -reset false; -user_print true

SURFACE 1
# 1 g loamy sand = 0.007 g HA, multiply site conc's for 1 g HA with 7e-3
# Specific surface = 46.5e3 m2/g for 1 mM CaCl2 (I = 0.003 mol/L)
H_a 4.97e-6 46.5e3 0.007
H_b 4.97e-6; H_c 4.97e-6; H_d 4.97e-6

H_e 2.483e-6; H_f 2.483e-6; H_g 2.483e-6; H_h 2.483e-6

H_ab 8.278e-7; H_ad 8.278e-7; H_af 8.278e-7; H_ah 8.278e-7

H_bc 8.278e-7; H_be 8.278e-7; H_bg 8.278e-7; H_cd 8.278e-7
H_cf 8.278e-7; H_ch 8.278e-7; H_de 8.278e-7; H_dg 8.278e-7
-equil 1
-diffuse_layer

SOLUTION 1
pH 6.0
Ca 1; Cd 1e-6; Cl 2 charge
-water 0.05

USER_PRINT
-start
10 H_Cd = mol("H_aCd+") + mol("H_bCd+") + mol("H_cCd+") + mol("H_dCd+")
20 H_Cd = H_Cd + mol("H_eCd+") + mol("H_fCd+") + mol("H_gCd+") + mol("H_hCd+")

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

352 Sorption of trace metals

30
H_Cd = H_Cd + mol("H_abCd") + mol("H_adCd") + mol("H_afCd") + mol("H_ahCd")
40
H_Cd = H_Cd + mol("H_bcCd") + mol("H_beCd") + mol("H_bgCd") + mol("H_cdCd")
50
H_Cd = H_Cd + mol("H_cfCd") + mol("H_chCd") + mol("H_deCd") + mol("H_dgCd")
60
print " ug Cd/g soil =", H_Cd * 112.4e6 * 0.05,\
" ug Cd/L =", tot("Cd") * 112.4e6,\
" Kd (L/kg) =", H_Cd / tot("Cd”) * 50
70 print " ug Cd/g soil in DL =", edl("Cd") * 112.4e6, chr$(13)
80 dl_Ca = edl("Ca")*2 - edl("water")*tot("Ca") * 2
90 dl_Cl = edl("Cl") - edl("water")*tot("Cl")
100 print "Excess eq Ca in DL =", dl_Ca
110 print "Excess eq Cl in DL =", dl_Cl, “Ca - Cl charge in DL =", dl_Ca-dl_Cl
120 print "Surface charge =", edl("Charge")
-end
END

In keyword USER_PRINT the concentrations of surface complexes are summed up in mol/kg H2O and con-
verted to g Cd/g by multiplying with the atomic weight in g/mol, and with 0.05 (kg H2O/g soil). The
distribution coefficient is likewise found by multiplying the quotient of complexed and solute concentra-
tions by 50 kg H2O/kg soil. The concentration of Cd2 in the double layer and the charge due to excess Ca2
and Cl in the double layer are also printed. These quantities are given in moles by PHREEQC, not in molal-
ity, and they are not multiplied with 0.05. The result is:

------------------------------------------------------User print-----------------------------------------------------------------------------

ug Cd/g soil = 1.5080e-01

ug Cd/L = 1.1240e-01 Kd (L/kg) = 1.3416e+03
ug Cd/g soil in DL = 1.5122e-03

Excess eq Ca in DL = 2.2492e-05
Excess eq Cl in DL = -3.2998e-06 Ca - Cl charge in DL = 2.5792e-05
Surface charge = -2.5796e-05

The calculated distribution coefficient Kd  1342 L/kg is about 5 times larger than determined. However, we
assumed that all the organic matter is humic acid whereas a large part probably consists of more unreactive
lignin. The contribution of the double layer to the overall sorbed Cd2 is about 1%. The total (possible) charge
of the OM is 0.007 g OM  7.095 meq/g HA  5105 eq. Approximately half the sites are dissociated at
pH  6 (Surface charge  2.58105 mol). About 90% of this charge is compensated by an excess Ca2 and
10% by a deficit of Cl. The double layer content for Cd2 has been neglected which is reasonable given the
small amounts Cd2. On the other hand, the double layer contains 5 times more Ca2 than is complex-bound
and clearly the complex constants for a weakly bound cation are difficult to obtain if only traditional wet-
chemical analyses are used.

QUESTIONS:
Obtain a printout of total complexed Ca2?
ANSWER: complexed mol Ca/g soil  2.8367106
Calculate Kd for pH  7.0 and 8.0?
ANSWER: Kd  1749 L/kg at pH  7, Kd  2663 L/kg at pH  8.
Calculate Kd when I increases to 0.01 mol/L by adding NaCl (use Equation (7.65) to adapt the
specific surface A)?
ANSWER: Kd  888 L/kg
Calculate Kd when I increases to 0.01 mol/L by adding CaCl2?
ANSWER: Kd  384 L/kg

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 353

100
Sediment B  1.0 g/L
90 CdT  107 M
80 I  0.6 (NaNO3)
T  25°C
70
60

% Sorbed
50
Sediment treatment
40
Unaltered
30 MgCl2 extracted

20 NaOAc & NH2OH–HCl

extracted
10 H2O2 extracted
0
3 4 5 6 7 8 9 10
pH

Figure 7.28. Cadmium adsorption onto sequentially extracted estuarine salt marsh sediment as a function of
pH (Lion et al., 1982).

Clearly, sorption of Cd2 to the loamy sand of Christensen (1984) is likely to be located mainly in
the organic fraction. This corroborates the experiments of Lion et al. (1982) who extracted oxides
and organic matter from an estuarine sediment and determined sorption edges for Cd2 and Pb2 on
the residues and also on the intact sediment (Figure 7.28). Treatment with acetate and hydroxyl-
amine/hydrochloride extractant showed no effects for Cd2 (this extract removes manganese oxides
and mainly amorphous iron oxyhydroxides), but the removal of organic carbon with H2O2 shifted the
sorption edge from pH  5.9 to pH  6.5, indicating that a major sorbent had been eliminated. For
Pb2, however, the sorption edge shifted 0.7 unit higher pH when the oxides were removed, and
remained at that pH when organic carbon subsequently was extracted.
We can calculate the contributions of the various fractions in the sand to the sorption
isotherm of Example 7.1 and compare the total with the observed distribution (Figure 7.29).
The amount of organic carbon was reduced by 7 (the amounts of ‘H_’ in Example 7.11 were
divided by 7) to obtain the observed distribution coefficient for small concentrations of Cd2.

6
Total
5
s Cd (µg/g soil)

4
Cd-OM

1 CdX2

0
0 10 20 30 40
cCd (µg/L)

Figure 7.29. Cd2 sorption on sandy loam, modeled as the sum of sorption on iron oxyhydroxide (negligible
and omitted from the graph), ion exchange (Cd-X2) and complexation on organic matter (Cd-OM).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

354 Sorption of trace metals

The amount sorbed on iron oxyhydroxide is negligible, exchangeable Cd-X2 on clay contributes about
10%, while organic matter provides the dominant contribution to the total isotherm. Note that the
modeled sorption isotherm is less curved than the data, indicating that the number of strongly sorbing
sites in the model is too small. Similar comparisons of a deterministic model for heavy metal sorption
have been presented by Cowan et al. (1992) and Lofts and Tipping (2000). Generally, modeling seems
able to describe various data to within a factor of ten.

7.5.3 Distribution models

In Section 7.3.3 we found that the stoichiometry of the exchange reaction determined the steepness
of the sorption edge. To generalize this finding we write the sorption Reaction (7.7) as:
s  q I ↔ sI ; Ks (7.68)
I

Here q is the stoichiometric reaction coefficient. Charges have been left out for simplicity, but you
may think of sorption of neutral molecules, gases or organics. Most of this theory was originally
developed for gas-adsorption, and subsequently applied to sorption of solute ions (Van Riemsdijk
et al., 1986). The mass action equation for Reaction (7.68) yields a modified form of the Langmuir
equation:
[stot ] K s [I ]q
[sI ]  I
(7.69)
1  K s [I ]q
I

This equation can be rewritten as:

K s [I ]q

 I
(7.70)
1  K s [I ]q
I

where
 sI / stot is the fraction of surface sites covered by I. When q equals one, then Equation
(7.70) is identical to the Langmuir equation (7.9). If [I] is small, Equation (7.70) approaches the
Freundlich equation (7.4).
When
is plotted vs log [I], a sigmoidal curve results, which broadens as q decreases. Sips (1948)
has shown that this curve closely approaches an integrated probability curve when the log K’s of the
sorption sites are normally distributed around an average log Kav and with a standard deviation :
log Ki   log K  log K 2 
1

i  ∫ exp     d log K
av
(7.71)
 2$    2  
∞ 

where
i are all the sites with binding strength up to log Ki. The relation among the Sips parameters
and those characterizing the normal distribution is:
log K s  q log K av (7.72a)
I

and

q  1/ ( 2 ) (7.72b)
Figure 7.30 shows sorption of protons as a function of pH in case log Kav  4, for two values of
–
  1 /
2 and   1.7 (corresponding to q  1 and q  0.416, respectively). There is a close
correspondence but the Sips curve is slightly broader in the tails.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 355

1
Probability
0.9
  1 / 2 Sips
0.8
0.7
Fraction of DOM

0.6
0.5
0.4
0.3
0.2
  1.7
0.1
0
2 3 4 5 6 7 8 9
pH

Figure 7.30. Sorption of protons on a surface with probability-distributed binding strength. Comparison of the
Sips curve with a probability curve, log Kav  4, q  1 and 0.416.

Equation (7.70) can be generalized for a multicomponent system in the form of the so-called
Henderson-Hasselbalch (HH) equation:
K s [ I ]q

 I
(7.73)
1  ∑ (K s [ I ]q )
I I

However, the deviation with respect to the normal distribution increases because of tailing of the dis-
tribution. Van Riemsdijk et al. (1986) therefore derived what is called the multicomponent Langmuir-
Freundlich (LF) equation:

( ∑ K [ I ])
p
Ks [I ] I sI

I  I
(7.74)
∑ I Ks I
[I ]
1  ( ∑ K [ I ])
p
I sI

where p indicates the heterogeneity among the sites, with the same function as q has for the individ-
ual ions. The integration of the normal distribution of log K’s (Equation 7.71) is an option in the
geochemical model MINTEQ (Allison et al., 1991) for calculating sorption of heavy metals to
humic acids. The comparison of the Sips distribution (Equation 7.72) and MINTEQ is excellent for
a single component (similar to Figure 7.30). In multicomponent solutions, the different tailings of
the distributions give notable effects. Figure 7.31 compares competitive sorption of protons and
Ca2 on humic acid, at variable pH and [Ca2]  0.001, calculated by MINTEQ and the HH (7.73)
and LF (7.74) approximations. The LF equation approximates the Gaussian distribution quite well,
while the HH curves are too broad.
The multicomponent Sips’ equations are used in the non-ideal competitive adsorption (NICA)
model developed by Koopal and coworkers (Koopal et al., 1994; Kinniburgh et al., 1998, 1999):

{∑ ( K [I ] ) }
p
qI
qI K s [ I ]qI I sI

I  I
(7.75)
∑I ( Ks [I ] 1  {∑ ( K [ I ] ) }
qI p
) qI
I I sI

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

356 Sorption of trace metals

0.9
MINTEQ
0.8 H-DOM
HH
0.7 LF

0.6
Fraction of DOM

0.5

0.4
Ca-DOM
0.3

0.2

0.1

0
2 3 4 5 6 7 8 9
pH

Figure 7.31. Comparison of two-component sorption of protons and Ca2 by three equations: MINTEQ is a
numerical integration of the normal distribution of log K’s; HH and LF correspond to Equations (7.73) and
(7.74), respectively. Data: log Km,H  3.78, log Km,Ca  2.9,   1.7 for both, [Ca2]  0.001.

The NICA model considers two types of sites (carboxylic and phenolic, similar to the Tipping and
Hurley model), each with a parameter p which reflects the heterogeneity within the sites. Basically,
sorption of protons is determined by titrations at various ionic strength. These data are corrected to
a master curve with an electrostatic model for spherical or cylindrical particles (De Wit et al., 1993a, b).
For the master curve, values of qH and p are derived by curve fitting (Buffle, 1990; Rusch et al.,
1997). Parameters for heavy metal sorption are likewise determined from sorption data at a fixed pH
(Benedetti et al., 1995). More recently, the parameters are estimated with non-linear estimation pro-
grams (Milne et al., 2001, 2003).

7.5.4 Humic acids as carriers of trace elements

Most of the experimental work for deriving sorption parameters has been done with an extracted
and subsequently purified humic acid. But how applicable are the obtained parameters for
groundwater DOC? Christensen and Christensen (1999) equilibrated solutions with groundwater DOC,
heavy metals and cation exchange resins or aquifer sands. DOC was derived from a waste site percolate
and added in variable concentrations. Equilibrium will be achieved between the resin and the free metal
in solution. When increased DOC complexes more metal, the concentration of free metal decreases. The
total solute concentration of Ni2, Zn2 and Cd2 increases with DOC as a result of increasing com-
plexation in solution, and desorption from the resin or aquifer materials (Figure 7.32).
Also shown in Figure 7.32 are model lines for the concentration ratio, given by Tipping and
Hurley’s ‘WHAM’ model, and by MINTEQ. Clearly, the modeled lines are not exactly matching the
observations. The WHAM model predicts too much complexation for sample L2, and too little for
sample L1. The MINTEQ model tends to underestimate complexation to DOC, but still matches the
data within a factor of 3. Thus, at least an idea of the importance of humic acids as carriers of
heavy metals can be derived with the models. In the experiments of Figure 7.32, the solubility is
enhanced by a factor of 2 with about 100 mg C/L for Cd2 and Zn2, and with 40 mg C/L for Ni2.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Complexation to humic acids 357

(M  MDOC)/M (M  MDOC)/M
10 10
8 L1 Cadmium 8 L2 Cadmium
6 6
5 5
4 4
3 3
2 2

1 1
0 50 100 150 200 250 0 50 100 150 200 250
10 10
8 L1 Nickel 8 L2 Nickel
6 6
5 5
4 4
3 3
2 2

1 1
0 50 100 150 200 250 0 50 100 150 200 250
DOC (mg C/L) 100
L2 Zinc
50
Resin experiments
20
Aquifer material
Aquifer material 10
WHAM default proton parameters 5
WHAM specific proton parameters
2
MINTEQA2 default proton parameters
1
0 50 100 150 200 250
DOC (mg C/L)

Figure 7.32. Solute concentration of Ni2, Zn2 and Cd2 as function of DOC in water in equilibrium
with exchange resin or aquifer materials, relative to water without DOC (M is free, MDOC is complexed to
DOC) (Christensen and Christensen, 1999). L1 and L2 are two different water samples, L1 with a higher Ca2
concentration.

These DOC concentrations are much higher than encountered in average groundwater. Therefore,
enhanced transport of heavy metals by dissolved humic acids in groundwater takes place only in
dark, tea colored waters and it seems that it may be ignored in most other cases.

QUESTION:
Obtain the distribution coefficient Kd (L/kg) for Cd2 on OC in water L2 for 70 mg DOC/L from the
experiment in Figure 7.32. Water L2 has pH  6.6, Ca2  1.3 mM, I  0.013 mM. Calculate Kd as in
Example 7.11.
ANSWER: Kd
104 L/kg in the experiment. The WHAM model estimates:
Kd  1.9104 L/kg

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

358 Sorption of trace metals

7.6 KINETICS OF SURFACE COMPLEXATION

Sorption of trace metals on humic acid or oxides proceeds quickly, with half-lives of minutes or less
(Bunzl and Schimmack, 1991; Hachiya et al., 1984). However, longer reaction times can enhance sorp-
tion and shift the sorption edge to lower pH (Gerth et al., 1993). Desorption generally takes more time
than adsorption, and the time needed for desorption increases further when the duration of the adsorp-
tion experiment is longer (Section 7.3.4). This suggests that several mechanisms operate in the sorp-
tion process, for example sorption on weak and strong sites on oxides, with slower kinetics for the
strong sites. Another explanation may be that metals slowly diffuse into the oxide structure during
adsorption, and that outward migration during desorption is slower because inward diffusion continues
as long as the internal structure has not been saturated. We can try to model a kinetics experiment, and
establish the various processes by following the time dependent variation in the species concentrations.
The pressure jump technique is used to study rapid kinetics (Sparks, 1989). A high pressure of
about 100 atm is applied to the sample and suddenly released. The return to normal pressure within
less than 100 ms is accompanied by changes in the solution composition that can be recorded with
an accurate conductivity meter. Hayes and Leckie (1986) have applied this technique to measure the
sorption kinetics of lead on goethite and used the triple layer model to interpret the data. Equilibrium
sorption of lead followed the reaction:
˜ FeOH  Pb2 ↔ ˜ FeOPb  H (7.76)
The rate data were modeled with the forward rate:

r f  k1 m˜ FeOH [Pb2 ]s (7.77)

where rf is the forward rate in mol/L/s, k1 the rate constant (1/s), mi the concentration in mol/L, and
[Pb2]s  [Pb2] exp (2F0 / RT) is the (solute) activity of Pb2 at the surface. The backward
reaction was likewise:

rb  k1 m˜ FeOPb [H ]s (7.78)

The rate constants are linked by detailed balancing. At equilibrium the net rate is zero, rf  rb, and
therefore k1 / k1  Kint exp(F0 / RT), where Kint is the intrinsic equilibrium constant for the sorption
reaction, Equation (7.76). Hayes and Leckie (1986) found k1
2.5105/s at 100 atm pressure. The
value of k1 decreased when pressure was increased, suggesting that stronger sorption sites gained
importance at greater pressure. A fast and a slow relaxation was also observed for sorption of Cu2
on goethite (Grossl et al., 1994).
The kinetic reaction can be linearized for small perturbations which permits an easy evaluation
of the effect of various solution parameters on the rate (Hayes and Leckie, 1986; Zhang and Sparks,
1990). For larger changes, e.g. modeling sorption from zero metal concentration onwards, the lin-
earization is incorrect. However, the full rate equations can be included in PHREEQC, thus permit-
ting a modeling approach which was hitherto not possible. The sorption edge of Pb2 can be
reasonably modeled with the strong and weak sites of Hfo, but log K must be adapted for the strong
sites (Figure 7.33). Note in Figure 7.33 that the model appears to shift the sorption edge too much
for changes in ionic strength.
For modeling kinetic sorption, we need the surface potential 0. This can be obtained by calcu-
lating sorption on a dummy surface with a very small concentration. The complexation model is not
electrically neutral and requires a double layer calculation for charge compensation. However, the
surface is positively charged at all times in this experiment, and we can, more simply, neutralize the
surface reaction with NO 3 . Part of the input file for PHREEQC which calculates kinetic sorption of
Pb to ferrihydrite is shown in Table 7.10 (download the complete file from www.xs4all.nl/appt).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of surface complexation 359

100
1.0 M
90
0.3 M
80 0.1 M
0.01 M
70 Model, 0.1 M
60 Model, 1 M
% Sorbed

50

40

30

20

10

0
3 4 5 6 7
pH

Figure 7.33. Sorption edge for Pb2 on goethite. Data from Hayes and Leckie (1986). Double layer model
with log K changed to 3.7 for the strong sites.

The file starts with the definition of the basic setup of the experiment. Imagine a titrator which maintains
the pH at 5.0 in a beaker with 10 mM NaCl solution and some ferrihydrite. We will explore time-depend-
ent sorption as 2 mM Pb(NO3)2 is added to the beaker and use RATES and KINETICS to define the
reactions and not do so with keyword SURFACE. However, we do need the surface potential of ferrihy-
drite which we obtain using the SURFACE keyword. The concentration of surface sites is given an
insignificant concentration, say 108 smaller than actual. The potential at the surface is independent of
the concentration of ferrihydrite. Furthermore, when a surface species is multiplied by 108, it will give
the sorbed equilibrium concentration.
The rates are calculated according to Equations (7.77) and (7.78) for all the surface species. A
few specific points may bev noted when studying the input file in Table 7.10:

– The Boltzmann factor is used in various rate equations, and is therefore stored in temporary mem-
ory (“put(Boltzm, 99)”) in the first rate, for later recall with “Boltzm  get(99)”.
– Line 20 in the first rate defines the forward rate for the species. Note that “kin(“kin_surfs_oh”)”
provides the moles of the Hfo_sOH species.
– Line 30 defines the backward rate. The moles of the species is given by m, and is divided by Kint
of the reaction (Kint  103.7 for sorption of Hfo_sOPb).
– Line 40 calculates the moles of reaction by multiplying by k1  parm(1) and by “time”. The rate
equations have very different rates and the fastest reactions may need very small timesteps for
reducing oscillations. These “stiff ” equations should be solved with the ordinary differential
equation solver implemented in PHREEQC-2.9.
– Line 50 stores the moles of reaction in temporary memory (“put(moles, 1)”) for calculating mass and
charge balance for the sorption sites, and line 60 “saves” the reaction, finishing the rate definition.
– The rate “kin_surfs_oh” is for species “Hfo_sOH”. This species is defined by mass balance on the
strongly sorbing surface sites. The same applies to the corresponding species on the weak sites.
– The rate “kin_surf_ntr” sums the charge of all the kinetics reactions, and balances it by adding or
subtracting NO 3 . In this way, pH effects which result from balancing the electrical neutrality in

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

360 Sorption of trace metals

Table 7.10. Partial PHREEQC input file for calculating kinetic sorption of Pb to weak and strong sites on
ferrihydrite.

# Rates for calculating kinetic sorption of Pb

# This is a partial file. Download complete file sk_pb_t.ode from www.xs4all.nl/~appt

PHASES; Fix_ph; H+ = H+; -log_k 0.0

EQUILIBRIUM_PHASES 1; Fix_ph -5.0 HNO3

SOLUTION 0-2
pH 5.0; Na 10; N(5) 10 charge

# dummy surface for calculating  and limiting conc’s...

SURFACE 1
Hfo_w 18.14e-11 600 7.981e-8 # ...site conc’n * 1e-8
Hfo_s 0.45e-11
-equil 0
# Rate equations...
RATES # Strong surface sites...
kin_surfs_Pb
-start
10 Boltzm = exp(-38.924 * edl("Psi"))
12 put(Boltzm, 99)
20 r_f = kin("kin_surfs_oh") * act("Pb+2") * (Boltzm*Boltzm)
30 r_b = m / (10^3) * act("H+") * Boltzm
40 moles = parm(1) * (r_f - r_b) * time
50 put(moles, 1); 60 save -moles
-end

# Similar rates are defined for the other species on strong surface sites
# kin_surfs_h, kin_surfs_o

# kin_surfs_OH species is obtained by mass balance on sites...

kin_surfs_oh
-start
10 r = get(1) + get(2) - get(3)
20 moles = m - r
30 save m - moles
-end

# And also for all the species on the weak sites...

# kin_surfw_Pb, kin_surfw_h, kin_surfw_o, kin_surfw_oh

# charge balance, sum the reactions...

kin_surf_ntr
-start
5 s = 0
10 for i = 1 to 6
20 if exists(i) then s = s + get(i)
30 next
40 save -s
-end
END

KINETICS 1
kin_surfs_Pb; -formula Pb 1. H -1.; -m0 0.0; -parm 2.5e5
# also other species on strong sites...
kin_surfs_oh; -formula OH 0. ; -m0 2.880e-4
(Contd)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of surface complexation 361

Table 7.10. (Contd)

# also other species on weak sites...

kin_surf_ntr; -formula NO3 1. ; -m0 1.0
-step 1 3 11 25 40 80 100 240 780 2000 4000 10000 20000 49120
-cvode true
# Adsorption in pH-stat...
USE equilibrium_phases 1
SOLUTION 1
pH 5.0; Na 10; N(5) 10 charge; Pb 2.0 # Add Pb to solution
USE surface 1
# etc. for print

the solution are avoided. For a negative surface, or for humic acids, one would counterbalance
with cations from solution, as in the Donnan approach discussed earlier. It is possible to calculate
neutralization by a mixture of ions with the exchange formulae given in Chapter 6, “put” the
amounts for the different ions in storage, and then “get” the amount in a rate which contains the
reactant formula for the specific ion only.
– The initial amount (“m0”) (only exemplified for a few species in Table 7.10) of each kinetic
reactant equals the initial surface composition. It can be obtained by multiplying the equilibrium
surface species concentration from the first simulation by 108.
This sorption model does consider weak and strong sites, and illustrates an interesting feature noted
by Morel and Hering (1993) for the sorption of aqueous complexes with different complexation
strengths. Sorption on the weaker sites is usually faster because more sites are present. This results
in temporary over-loading of the weak sites. The overshoot retards sorption on the strong sites since
it must take place via redistribution and desorption from the weak sites.
Figure 7.34 shows adsorption of Pb2 on the weak sites to go through a maximum at about
10  500 s, while sorption on the strong sites increases steadily with a half-life of 103.4 s
45 min.
2.7

Without weak sites, the half-life is shorter by a factor of 2.5. The half-life decreases for high concentra-
tions of Pb2 as is illustrated in Figure 7.35, which compares sorption isotherms for 45 minutes kinet-
ics and at equilibrium. Equilibrium sorption on the weak sites is very nearly linear for the concentration
range from 0–2 mM Pb2, and 45 minutes are sufficient to reach near equilibrium for the weak sites.

2.0

mPb
1.5 q_weakPb
mmol/L

1.0

0.5
q_strongPb
0
0 1 2 3 4 5
log (seconds)

Figure 7.34. Adsorption of Pb2 on strong and weak sites of ferrihydrite, modeled as a kinetic process. Note
that the redistribution of sorbed complexes continues after 102.7  500 s, even when the solute concentration of
Pb2 is already near the final value.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

362 Sorption of trace metals

1.8

1.6
qtot
1.4

1.2
Sorbed (mmol/L)

qweak
1.0

0.8

0.6

0.4 qstrong
0.2

0
0 0.1 0.2 0.3 0.4 0.5
Solute (mmol/L)

Figure 7.35. Adsorption isotherms of Pb2 at 45 min. (kinetics, dotted lines) and at equilibrium (continuous
lines), weak and strong sites of ferrihydrite.

The initial part of the overall equilibrium sorption isotherm is curved because the strong sites are filled
up almost completely with Pb2 (the concentration of the weak sites is 18 mM, the concentration of
strong sites is 0.45 mM). The kinetic sorption isotherm for strong sites lies below the equilibrium
isotherm, but approaches it at higher concentrations. This suggests that the experimental determination
of sorption isotherms may require more equilibration time for low than for high concentrations.
When adsorption is stopped after 45 minutes, by replacing the fluid with a NaCl solution with-
out Pb2, the redistribution of sorbed complexes is not yet complete. The concentration of Pb on
the strong sites therefore continues to increase during the overall desorption process. The solute
concentration may even become higher due to redistribution among the species than in the final
equilibrium solution, as is shown in Figure 7.36. When adsorption is extended over 1 day, the
system approaches equilibrium. Desorption occurs then almost exclusively from the weak sites.

2.0 2.0
1 day adsorption 45 min. adsorption

1.5 1.5

q_weak
mmol/L

mmol/L

q_weak
1.0 1.0

0.5 q_strong 0.5 q_strong

mPb
mPb
0 0
0 1 2 3 4 5 0 1 2 3 4 5
log (seconds) log (seconds)

Figure 7.36. Time curves for desorption of Pb2 from ferrihydrite after 45 minutes and 1 day of adsorption.
Note that Pb2 adsorption on the strong sites increases over short times even though there is overall desorption.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Field applications 363

Desorption is modeled with PHREEQC by defining a new KINETICS block with initial concentra-
tions m0 equal to the concentrations m at the end of the adsorption step, as shown in the complete,
downloaded input file.
Lastly, we may note that the final solute equilibrium for the 45 minutes experiment is slightly less
than for one day adsorption time, due to the kinetic readjustment on the strong and the weak sites.
The observation that desorption is less when the adsorption time has been in the order of several
days, therefore cannot be explained by kinetic rearrangements among strong and weak sorption
sites. Rather, it must be related to diffusion of the metal ions into the mineral structure.

7.6.1 Extrapolation of adsorption kinetics for other metal ions

Surface complexation of heavy metals on oxides involves an inner-sphere reaction where the water
molecules from the hydration shell of the ion are removed and replaced by direct coordination to the
surface oxygens of the solid. Generally, the rate for the water exchange rate is 4–5 orders of magni-
tude greater than for the inner-sphere adsorption reaction (Wehrli et al., 1990; Figure 7.37) and can
be used for estimating adsorption rates. For example, the rate for water exchange on Zn2 is 107.5/s
(Figure 7.37). This is 500 times slower than for Pb2, and accordingly, we estimate the half-life for
Zn2 sorption on the weak sites to be around 500  90 s (1⁄2 day), while for sorption on the strong
sites it is estimated to be 500  45 min (15 days).

5
Pb2
4
Cu2
3
Zn2
log kads (int) (1/M/s)

2
Mn2
1 Co2

0
VO(OH)
1

2 VO2
3 Fe3
Al3 log kads (int)  4.16  9.52 log k(H2O)

1 2 3 4 5 6 7 8 9 10
log k_w (1/s)

Figure 7.37. Rate constants for adsorption of metal ions on oxide surfaces (kads) are related to exchange rates
of water in the hydration shell (kw) (Stumm, 1992).

7.7 FIELD APPLICATIONS

Surface complexation theory is useful to obtain a conceptual understanding of heavy metal sorption
behavior on mineral surfaces. However, the theory has been developed from experimental data using
a single mineral, ferrihydrite or goethite, and the application to natural sediments poses compli-
cations. Grains in aquifer sediment are normally coated by mixtures of iron oxyhydroxides, clay
minerals and organics. Often the coating consists of sandwiches of organics and clays with
predominantly negative surface charge, and iron oxyhydroxides with predominantly positive surface

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

364 Sorption of trace metals

charges. It would be quite unlikely that the Dzombak and Morel (1990) database for ferrihydrite is
completely adequate for a natural sediment. There are, however, some environmental settings where
it is reasonable to apply the Dzombak and Morel (1990) database. For example as a model substance
for heavy metal behavior in acid mine drainage, or for an iron oxide filter bed in a drinking water
plant. As an illustration, the concentration changes in a reduced groundwater which contains heavy
metals, and is oxidized in a drinking water plant, can be calculated as shown in Example 7.12.

EXAMPLE 7.12. Calculate heavy metal removal from groundwater with Fe2 oxidation during
aeration and filtration
The groundwater composition is (mg/L):

pH Ca2 Alk Cl Fe2 As Cd2 Ni2

6.5 100 120 100 5 10103 10103 30103

ANSWER:
We make an input file for PHREEQC, define the groundwater (SOLUTION) and calculate the composition
after oxidation (EQUILIBRIUM_PHASES) with surface complexation (SURFACE). The surface complexes
of CO2 on Hfo are copied from the database (SURFACE_SPECIES). NaOH or HCl is added to increase or
decrease pH (REACTION). Output is directly in g/L for the heavy metals (USER_ PRINT). The file must
be run with the wateq4f.dat database that includes the elements As and Ni.

DATABASE wateq4f.dat
SOLUTION 1
pH 6.5; -units mg/l; Ca 100; Alkalinity 120; Cl 100; Fe(2) 5
As 10e-3; Cd 10e-3; Ni 30e-3

PHASES; Ferrihydrite; Fe(OH)3 + 3H+ = Fe+3 + 3H2O; -log_k 2.0

EQUILIBRIUM_PHASES 1
O2(g) -0.68 # P_O2 = 0.21 atm (log 0.21 = -0.68)
Ferrihydrite 0.0 0.0 # SI = 0.0, initial amount = 0.0
# CO2(g) -3.0 # For bubbling out CO2

SURFACE 1
Hfo_w Ferrihydrite 0.2 5.33e4 # weak sites, 0.2 mol/mol, surface area 5.33e4 m2/mol
Hfo_s Ferrihydrite 0.5e-2 # strong sites 2.5% of weak sites
-equil 1
SURFACE_SPECIES
Hfo_wOH + CO3-2 + H+ = Hfo_wCO3- + H2O; log_k 12.56
Hfo_wOH + CO3-2 + 2H+ = Hfo_wHCO3 + H2O; log_k 20.62

#REACTION 1
# NaOH 1; 1.e-3 # Increase pH for increased metal removal
# HCl 1; 2.e-3 # Reduce alkalinity and bubble out CO2

USER_PRINT
-start
10 print ' pH As_tot Cd Ni '
20 print -la("H+"), tot("As") * 74.92e6, tot("Cd") * 112.4e6, tot("Ni") * 58.71e6
-end
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Field applications 365

Results are given in Table 7.11 for five cases, 1) oxidation and sorption onto the precipitated Hfo;
2) oxidation and sorption, with CO2 partially removed; 3) with 1 mmol NaOH/L added to increase pH for
possibly increased metal removal; 4) with 2 mmol/L HCl added to decrease carbonate complexation for
nickel by bubbling out CO2, and 5) the effect of removing CO2 surface complexes in case 1.

Table 7.11. Effects of operational conditions on heavy metal concentrations in groundwater after aeration
and precipitation of Fe(OH)3. Concentrations in g/L.

pH Astot Cd2 Ni2

Groundwater 6.50 10 10 30
Oxidized 6.42 1.3 9.6 29.5
Aerated, PCO2  103 8.10 0.13 4.5 28.7
NaOH added 8.25 0.14 3.8 29.1
HCl added 7.11 0.01 8.2 25.9
Oxidized, no CO2 cplx 6.42 0.003 9.6 29.5

The results of Example 7.12 indicate that aeration, with precipitation of ferrihydrite and sorption on
its surface, is removing most As, but not much of Cd2 and Ni2. During aeration, part of the CO2 will
be degassing, which is modeled by imposing CO2 equilibrium with a partial pressure of 103 atm. The
pH increases to 8.10 due to the loss of CO2, and more than half of the Cd2 is now sorbed. Increasing
the pH further will enhance sorption of the heavy metal cations, as is shown by adding 1 mmol
NaOH/L. Adding more NaOH will increase total CO2 when the CO2 pressure is fixed. This leads to car-
bonate complexes in solution which reduce the free metal ion activity and thereby diminish sorption.
Apparently, sorption of Ni2 on Hfo is not very strong. Lastly, a comparison of case 1 and 5, with and
without CO2 complexes, shows that sorption of As is much reduced by competition with CO2.
In the case of acid mine drainage, streamwaters often have a low pH and contain high
amounts of dissolved iron and heavy metals. When base is added to the streamwater, iron oxyhy-
droxides are expected to precipitate, removing by adsorption and coprecipitation some of the dis-
solved heavy metals. Figure 7.38 shows the copper content of an acid mine drainage stream.

4.4 24 m 70 m
Dissolved Total
Observed
Simulated

3.6
Copper (µM)

2.8

2.0

8 10 12 14 16 18 8 10 12 14 16 18
Time (hr) Time (hr)

Figure 7.38. The variation in copper concentration in streamwater, during an experiment where a pulse of Na2CO3
was added to an acid mine drainage stream at Leadville, Colorado, USA. Shown is the copper concentration over
time at 24 and 70 m downstream from the point of Na2CO3 injection. Closed symbols give the total Cu concentra-
tion and open symbols the Cu concentration in samples passed through 0.1 m filters. The solid lines are model pre-
dictions based on the Dzombak and Morel (1990) surface complexation model for ferrihydrite (Runkel et al., 1999).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

366 Sorption of trace metals

The upstream pH value is about 3.4, but in the pulse, where a strong Na2CO3 solution was added,
the pH increased to above 5.6. The high pH water arrived at 12.00 hr at the sampling point at
24 m and the dissolved copper concentration decreased markedly (open symbols), while the total
(unfiltered) concentration did not change much. This was attributed to increased sorption of Cu2 on
the surface of iron oxyhydroxide, suspended in water and sedimented on the streambed. After the
high pH spike had passed, the Cu2 concentration increased temporarily above the background
concentration due to desorption from the iron oxyhydroxide in the streambed. The lines give
the model results based on a surface complexation model using the ferrihydrite database of
Dzombak and Morel (1990) and generally match the field data well. Recently, Swedlund and
Webster (2001) found that in sulfate-rich water, such as acid mine drainage, a ternary sulfate com-
plex, like ˜FeOHCuSO4, also needs to be taken into consideration.
As mentioned above, sandy aquifer sediments will have other sorption properties than
ferrihydrite because the surface coatings consist of mixtures with clay. Wang et al. (1997) measured
trace metal sorption on a wide range of natural sediments and their results for copper are
shown in Figure 7.39. The results show a wide variation in the sorption properties of the various
sediments. The pH where 50% is adsorbed varies from about 4.7 to 6.7. A surface complex-
ation model was fitted to the data with variable complexation constants. These constants were
found to be proportional to the first hydrolysis constant of the metal ion in water and also
related to the relative amounts of organic carbon and various oxide minerals in the
sediment. Possibly, the variation can be explained by assigning each sediment the properties
of its individual sorbing components in the correct proportion, as we did for Cd2 in a sand
(Figure 7.29).

100.0
HLa1
NNa1
80.0
LNa1
Cu2 sorbed (%)

HGa3
60.0 HIa1
HNa1
40.0 CGa4
QTa1
MNa1
20.0 BEa1
ZHa1
0.0
2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5
pH

Figure 7.39. pH-dependent sorption of copper on a range of natural sediments (Wang et al., 1997).

The spatial heterogeneity of the adsorbers in aquifer sediments deserves also attention. Mn-oxides
are extremely effective scavengers of heavy metals and often observed in distinct layers which
may become enriched in trace elements. Figure 7.40 shows the speciation of Ni2 in a sediment
core with separate bands of Mn-oxide and Fe-oxide. The two plots to the left in Figure 7.40 show
results of sediment extractions using a mild reductant, hydroxylamine (HA). Just below 2 m depth,
HA extracts a large amount of Mn from Mn-oxide and Ni2 is clearly associated with the Mn-oxide.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 367

mmol/kg µmol/kg mmol/kg µmol/kg

0 10 20 30 40 50 0 200 400 0 100 200 0 200 400
0

0.5

Ni
Depth (m)

1.5

2 Mn Ni
Fe
2.5 Mn
Fe
3
HA HA HCl HCl
3.5

Figure 7.40. Sequential extraction of Mn, Fe and Ni2 from a sandy sediment with a distinct Mn-oxide band
over a more diffuse Fe-oxide band. HA: extraction with hydroxylamine, HCl: with 6 N HCl (Larsen and
Postma, 1997).

Extraction with HCl releases a lot of Fe from the sediment, corresponding to the Fe-oxide band, but
there is little Ni2 associated with this layer. If the Mn-oxides become reduced, for example by reac-
tion with Fe2 (Postma and Appelo, 2000), large amounts of Ni2 may be mobilized (Stollenwerk,
1994; Larsen and Postma, 1997).

PROBLEMS
7.1. Find the specific capacitance (F/kg) of -Al2O3 at pH  5 in 0.001 M NaCl from Figure 7.21.
7.2. Find the ratio {˜AlOH} / {˜AlOH  2 } for -Al2O3 at pH  5.4 in 0.1 M NaCl, using Equation 7.53.
There are 0.03 mol sites/kg. Is this correct in Figure 7.21?
7.3. The plot of pKapp vs f for a humic acid at 3 different concentrations of Al(NO)3 is presented in Figure 7.41.
Estimate pKint,  / ln 10, and Kh-Na. Assume that Al3 forms the complex:

˜ hOAl 2 ↔ Al3  ˜ hO; K h-Al

5
pKa

0 0.2 0.4 0.6 0.8 1.0

f

Figure 7.41. Apparent acid dissociation constants of humic acid plotted against degree of dissociation at
concentrations of 0.001, 0.01, and 0.1 M of background electrolyte. From Tipping et al. (1988).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

368 Sorption of trace metals

7.4. Seaman et al. (1996) measured the retardation of Br on a quartz sediment coated with iron oxides as a
function of the Br concentration, see table below. The sediment contained 3.6% Fe, b  1.8 g/cm3,
0.3.
Use PHREEQC to find the retardation for Br in the double layer of Hfo at pH  5, 8.5 mM weak sites,
600 m2/g, 3.8 g Hfo, for these KBr concentrations. Explain the trend with concentration. How much of the
analyzed iron is active as Hfo?
mM KBr Retardation
1 2.15
10 1.35
100 1.08
7.5. Model a column experiment of Br retardation, Figure 7.42. Experiment D has the same iron
oxyhydroxide parameters as Problem 7.4. Column length 10 cm. Initial solution: pH  5, 4 mM K
and Cl balances charge in solution. Influent solution: 1 mM KBr, pH  5, K balances charge.
Find the dispersivity from Experiment A.

1.2
A
1.0

0.8
A
Br (C/Co)

0.6
B
0.4 C
D
0.2

0.0
0 1 2 3 4 5
Pore volume

Figure 7.42. Bromide breakthrough in column experiments with quartz sands coated with variable
amounts of iron oxide (Seaman et al., 1996).

7.6. Compare sorption of As(3) and As(5) on ferrihydrite at concentrations of 0.267 mM and 0.001 mM, both
in 0.1 M NaCl solution with pH  7. For comparison with Figure 7.16: assume that goethite has 10 times
less sorption sites and surface area than ferrihydrite, but with the same chemical properties. Explain the
relative increase of As(5) sorption (arsenate anion) at the low concentration.
7.7. Wehrli et al. (1990) have shown that the rate constants for surface complexation are related to the water
exchange rate of the metal ion (cf. Figure 7.37). If the exchange rate for Pb2 sorption on ferrihydrite
decreases from 2.5105/s to 2.5103/s, how will the half-life of the sorption change?
7.8. The water exchange rate for Zn2 is approximately 500 times smaller than for Pb2 (cf. Figure 7.37).
Modify the PHREEQC input file for kinetic sorption of Pb2 (Table 7.10) to calculate the kinetics of
Zn2 sorption on ferrihydrite.
7.9. Determine the half-life for sorption of 0.1 mM Pb2 on the strong sites of ferrihydrite with the
PHREEQC input file given in Table 7.10.
7.10. Calculate the equilibrium sorbed concentrations of Pb2 on Hfo_w  18.14103 M and
Hfo_sOPb  0.451011 M, 7.981 g Hfo. Solution composition: fixed at 2 mM PbCl2 in 10 mM NaNO3
solution, pH  5.
7.11. Modify the PHREEQC input file from Table 7.10 and compare equilibrium and 45 min. kinetic sorption
isotherms for weak and strong sites over the range 0–2.0 mM Pb. Hint: Set INCREMENTAL_REAC-
TIONS false, and use keyword REACTION to add PbCl2.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 369

7.12. Modify the PHREEQC input file in Table 7.10 to calculate desorption after 45 min. adsorption. Hint: the
initial masses m0 of kinetic reactants must be changed for desorption.
7.13. Check that a decrease of the rate constant by a factor 100 will increase the reaction half-life by 100. Hint:
change parm(1) in the PHREEQC input file.
7.14. Find the Langmuir parameters for sorption of Cd2 on loamy sand, data in Example 7.1. Define the con-
centration range where the Langmuir isotherm is linear to within 10%. Also the concentration where
qI  0.75 qmax.
7.15. Calculate the distribution coefficient of 0.2 M Zn2 for a sediment with CEC  7 meq/kg, b  2.1 g/cm3,
w  0.2, mCa  2 mmol/L, using PHREEQC. What is the effect of b) doubling the Zn concentration, c)
doubling the Ca2 concentration, d) doubling the CEC, e) changing the pH from 7 to 5?
7.16. Calculate the distribution coefficient of Zn2 for a solution with 88 mg ferrihydrite/L at pH  7,
mMg2  2 mmol/L, using PHREEQC. What is the effect of b) doubling the Zn concentration, c) doubling
the Mg2 concentration, d) doubling the amount of ferrihydrite, e) changing the pH from 7 to 5?
7.17. Discuss the effects of doubling the Mg2 concentration in Problem 7.16 in terms of surface potential
changes due to increased sorption of Mg2 and increased ionic strength. Compare with the effect of
Ca2, when Mg2 is replaced by Ca2.
7.18. Use PHREEQC to study sorption isotherms for Pb2, Zn2 and Ca2 on ferrihydrite, pH range 3–9,
Hfo_w 20103 600 8.8, Hfo_s 0.5103 , 0.1 M NaNO3, 2 mM M(NO3)2. What is the effect of halving
the concentration of Hfo; decreasing the concentration of NaNO3 to 0.01 M; decreasing the concentration
of the metal ion to 0.2 mM.
7.19. Estimate the percentage Ni2 at trace (108 M) concentration sorbed on smectite in 0.33 mM Ca2 solu-
tion, 1 g smectite/L, CEC  8.7 meq/g. Compare with Figure 7.18, pH  5. Note to include NiX2
(cf. Table 6.4) and HX with log_k 4.5.
b. Include 8.8 g ferrihydrite, vary log_k for sorption of Ni2 on the strong sites?
7.20. In Example 7.2, include the activity correction for CdX2 (cf. Section 6.3.3) and recalculate K d with
PHREEQC.
7.21. Estimate the retardation of 14C due to sorption of HCO 3 on ferrihydrite in the Sherwood sandstone
aquifer (cf. Chapter 5). Smedley and Edmunds (2002) found 400 ppm Fe in 0.2 M oxalate extractions,
equivalent to 42 mM ferrihydrite/L pore water. Hint: Use water in equilibrium with calcite and dolomite
at [PCO 2 ]  102.5, and equilibrate with the surface.
7.22. Calculate the counterions which balance the charge of 1 g HA (humic acid)/L at pH 7, 1 mM NaCl, 1 mM
CaCl2 according to the Donnan model (Hint: use Example 6.4) and according to the double layer com-
position (Hint: cf. Example 7.11). Calculate KCa\Na for the double layer composition.
7.23. Lets make a pill with FeSO4 and NaClO4 as purificants for water with a harmful content of As and F.
NaClO4 oxidizes bacteria and also Fe2, which precipitates as ferrihydrite and sorbs As and F. This reaction
is acid, and if we add calcite, the Ca2 concentration will increase and CaF2 may precipitate. Determine the
composition of the pill to bring the As and F concentrations down to below the drinking water limits in a
water with:
pH 7.67 and Na 2.93, Ca2 0.59, Alk 3.85, F 0.26 and As 1.3103 mmol/L (4.9 mg F/L and 97 g As/L).

REFERENCES
Aiken, G.R. and Malcolm, R.L., 1987. Molecular weight of aquatic fulvic acids by vapor pressure osmometry.
Geochim. Cosmochim. Acta 51, 2177–2184.
Allison, J.D., Brown, D.S. and Novo-Gradac, K.J., 1991. MINTEQA2, a geochemical assessment data base and
test cases for environmental systems. Report EPA/600/3-91/-21. US EPA, Athens, GA.
http://www.epa.gov/ceampubl/mmedia/minteq/index.htm

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

370 Sorption of trace metals

Appelo, C.A.J. and Postma, D., 1999. A consistent model for surface complexation on birnessite (d-MnO2) and
its application to a column experiment. Geochim. Cosmochim. Acta 63, 3039–3048.
Appelo, C.A.J., Van der Weiden, M.J.J., Tournassat, C. and Charlet, L., 2002. Surface complexation of ferrous
iron and carbonate on ferrihydrite and the mobilization of arsenic: Env. Sci. Technol. 36, 3096–3103.
Appleyard, E.C. and Blowes, D.W., 1994. Applications of mass-balance calculations to weathered sulfide mine
tailings. In C.N. Alpers and D.W. Blowes (eds), Environmental geochemistry of sulfide oxidation. ACS
Symp. Ser. 550, 516–534.
Artinger, R., Buckau, G., Geyer, S., Fritz, P., Wolf, M. and Kim, J.I., 1999. Characterization of groundwater
humic substances: influences of sedimentary organic carbon. Appl. Geochem. 15, 97–116.
Atkins, P.W. and de Paula, J, 2002. Atkins’ Physical Chemistry, 7th ed. Oxford Univ. Press, 1149 pp.
Avena, M.J., Vermeer, A.W.P. and Koopal, L.K., 1999. Volume and structure of humic acids studied by viscom-
etry pH and electrolyte concentration effects. Coll. Surf. A 151, 213–224.
Baeyens, B. and Bradbury, M.H., 1997. A mechanistic description of Ni- and Zn sorption on Na-montmoril-
lonite. J. Contam. Hydrol. 27, 199–248.
Barrow, N.J., Gerth, J. and Brümmer, G.W., 1989. Reaction—kinetics of the adsorption and desorption of
nickel, zinc and cadmium by goethite. 2. Modeling the extent and rate of reaction. J. Soil Sci. 40, 437–450.
Bartschat, B.M., Cabaniss, S.E. and Morel, F.M.M., 1992. Oligoelectrolyte model for cation binding by humic
substances. Env. Sci. Technol. 26, 264–294.
Benedetti, M.F., Milne, C.J., Kinniburgh, D.G., Van Riemsdijk, W.H. and Koopal, L.K., 1995. Metal ion binding to
humic substances: application of the non-ideal competitve adsorption model. Env. Sci. Technol. 29, 446–457.
Berg, M., Tran, H.C., Nguyen, T.C., Pham, H.V., Schertenleib, R. and Giger, W., 2001. Arsenic contamination
of ground and drinking water in Vietnam: a human health threat. Env. Sci. Technol. 35, 2621–2626.
Boekhold, A.E., Van der Zee, S.E.A.T.M. and De Haan, F.A.M., 1991. Spatial patterns of cadmium contents
related to soil heterogeneity. Water Air Soil Poll. 57, 479–488.
Bradbury, M.H. and Baeyens, B., 1999. Modelling the sorption of Zn and Ni on Ca-montmorillonite. Geochim.
Cosmochim. Acta 63, 325–336.
Brümmer, G.W., Gerth, J., Tiller, K.G., 1988. Reaction kinetics of the adsorption and desorption of nickel, zinc
and cadmium by goethite. I. Adsorption and diffusion of metals. J. Soil Sci. 39, 37–51.
Buffle, J. 1990. Complexation reactions in aquatic systems: an analytical approach. Ellis Horwood, Chichester,
692 pp.
Bunzl, K. and Schimmack, W., 1991. Kinetics of ion sorption on humic substances. In D.L. Sparks and D.L.
Suarez (eds), Rates of soil chemical processes. Soil Sci. Soc. Am. Spec. Pub. 27, 119–134.
Charlet, L., Wersin, P. and Stumm, W., 1990. Surface charge of MnCO3 and FeCO3. Geochim. Cosmochim. Acta
54, 2329–2336.
Christensen, T.H., 1984. Cadmium soil sorption at low concentrations: 1. Effect of time, cadmium load, pH, and
calcium. Water Air Soil Poll. 21, 105–114.
Christensen, J.B. and Christensen, T.H., 1999. Complexation of Cd, Ni, and Zn by DOC in polluted groundwa-
ter: a comparison of approaches using resin exchange, aquifer material sorption, and computer speciation
models (WHAM and MINTEQA2). Env. Sci. Technol. 33, 3857–3863.
Clapp, C.E. and Hayes, M.H.B., 1999. Sizes and shapes of humic substances. Soil Sci. 164, 777–789.
Cornell, R.M. and Schwertmann, U., 2003. The Iron Oxides, 2nd ed. Wiley-VCH, Weinheim, 664 pp.
Cowan, C.E., Zachara, J.M., Smith, S.C. and Resch, C.T., 1992. Individual sorbent contributions to cadmium
sorption on ultisols of mixed mineralogy. Soil Sci. Soc. Am. J. 56, 1084–1094.
Davis, J.A. and Kent, D.B., 1990. Surface complexation modeling in aqueous geochemistry. Rev. Mineral. 23,
177–260.
Davis, J.A., James, R.O. and Leckie, J.O., 1978. Surface ionization and complexation at the oxide/water interface.
1. Computation of electrical double layer properties in simple electrolytes. J. Coll. Interf. Sci. 63, 480–499.
Davis, J.A., Fuller, C.C. and Cook, A.D., 1987. A model for trace metal sorption processes at the calcite surface:
Adsorption of Cd2+ and subsequent solid solution formation. Geochim. Cosmochim. Acta 51, 1477–1490.
De Wit, J.C.M., Van Riemsdijk, W.H. and Koopal, L.K., 1993a. Proton binding to humic substances. 1.
Electrostatic effects. Env. Sci. Technol. 27, 2005–2014.
De Wit, J.C.M., Van Riemsdijk, W.H. and Koopal, L.K., 1993b. Proton binding to humic substances. 2.
Chemical heterogeneity and adsorption models. Env. Sci. Technol. 27, 2015–2022.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 371

Dzombak, D.A. and Morel, F.M.M., 1990. Surface complexation modeling: hydrous ferric oxide. Wiley and
Sons, New York, 393 pp.
Edmunds, W.M., Kinniburgh, D.G. and Moss, P.D., 1992. Trace metals in interstitial waters from sandstones:
acidic inputs to shallow groundwaters. Env. Poll. 77, 129–141.
Evans, T.D., Leal, J.R. and Arnold, P.W., 1979. The interfacial electrochemistry of goethite (a-FeOOH), espe-
cially the effect of CO2 contamination. J. Electroanal. Chem. 105, 161–167.
Fuerstenau, D.W., Manmohan, D. and Raghavan, S., 1981, The adsorption of alkaline-earth metal ions at the
rutile/aqueous interface. In P.H. Tewari (ed.), Adsorption from aqueous solutions, 93–117, Plenum Press,
New York.
Fuller, C.C., Davis, J.A. and Waychunas, G.A., 1993. Surface chemistry of ferrihydrite: Pt. 2. Kinetics of arse-
nate adsorption and coprecipitation. Geochim. Cosmochim. Acta 57, 2271–2282.
Garcia-Miragaya, J. and Page, A.L., 1976. Influence of ionic strength and inorganic complex formation on the
sorption of trace amounts of Cd by montmorillonite. Soil Sci. Soc. Am. J. 40, 658–663.
Gerth, J., Brümmer, G.W. and Tiller, K.G., 1993. Retention of Ni, Zn and Cd by Si-associated goethite. Z.
Pflanzenern. Bodenk. 156, 123–129.
Grossl, P.R., Sparks, D.L. and Ainsworth, C.C., 1994. Rapid kinetics of Cu(II) adsorption/desorption on
goethite. Env. Sci. Technol. 28, 1322–1429.
Hachiya, K., Sasaki, M., Saruta, Y., Mikami, N. and Yasunaga, T., 1984. Static and kinetic studies of adsorption-
desorption of metal ions on the ?-Al2O3 surface. J. Phys. Chem. 88, 23–27, 27–31.
Hawthorne, F.C. (ed.) 1988. Spectroscopic methods in mineralogy and geology. Rev. Mineral. 18, 798 pp.
Hayes, K.F. and Leckie, J.O., 1986. Mechanism of lead ion adsorption at the goethite-water interface. ACS
Symp. Ser. 323, 114–141.
Hiemstra, T. and Van Riemsdijk, W.H., 1991. Physical-chemical interpretation of primary charging behavior of
metal (hydr)oxides. Coll. Surf. 59, 7–25.
Hiemstra, T. and Van Riemsdijk, W.H., 1996. A surface structural approach to ion adsorption: the charge distri-
bution (CD) model. J. Coll. Interf. Sci. 179, 488–508.
Hiemstra, T., Van Riemsdijk, W.H. and Bolt, G.H., 1989. Multisite proton adsorption modeling at the solid/
solution interface of (hydr)oxides: a new approach, 1. Model description and evaluation. J. Coll. Interf. Sci.
133, 91–104.
Higgo, J.J.W., Kinniburgh, D., Smith, B. and Tipping, E., 1993. Complexation of Co2+, Ni2+, UO22+ and Ca2+ by
humic substances in groundwaters. Radiochim. Acta 61, 91–103.
Johnston, H.M., Gillham, R.W. and Cherry, J.A., 1985. Distribution coefficients for strontium and cesium at a
storage area for low-level radioactive waste. Can. Geotechn. J. 22, 6–16.
Kinniburgh, D.G., 1986. General purpose adsorption isotherms. Env. Sci. Technol. 20, 195–204.
Kinniburgh, D.G. and Smedley, P.L. (eds), 2001. Arsenic contamination of groundwater in Bangladesh. BGS
Techn. Rep. WC/00/19, British Geological Survey.
Kinniburgh, D.G., Gale, I.N., Smedley, P.L., Darling W.G., West, J.M., Kimblin, R.T., Parker, A., Rae, J.E.,
Aldous, P.J. and O’Shea, M.J., 1994. The effect of historic abstraction of groundwater from the London
basin aquifers on groundwater quality. Appl. Geochem. 9, 175–195.
Kinniburgh, D.G., Milne, C.J. and Venema, P., 1995. Design and construction of a personal-computer-based
automatic titrator. Soil Sci. Soc. Am. J. 59, 417–422.
Kinniburgh, D.G., Van Riemsdijk, W.H., Koopal, L.K. and Benedetti, M.F., 1998. Ion binding to humic sub-
stances. In E.A. Jenne (ed.) Adsorption of metals by geomedia, 483–520, Academic Press, San Diego.
Kinniburgh, D.G., Van Riemsdijk, W.H., Koopal, L.K., Borkovec, M., Benedetti, M.F. and Avena, M.J., 1999.
Ion binding to natural organic matter: competition, heterogeneity, stoichiometry and thermodynamic con-
sistency. Coll. Surf. A 151, 147–166.
Kjøller, C., Postma, D. and Larsen, F., 2004. Groundwater acidification and the mobilization of trace metals in
a sandy aquifer. Env. Sci. Technol. 38, 2829–2835.
Koopal, L.K., Van Riemsdijk, W.H., De Wit, J.C.M. and Benedetti, M.F., 1994. Analytical isotherm equations
for multicomponent adsorption to heterogeneous surfaces. J. Coll. Interf. Sci. 166, 51–60.
Kraepiel, A.M.L., Keller, K. and Morel, F.M.M., 1999. A model for metal adsorption on montmorillonite.
J. Coll. Interf. Sci. 210, 43–54.
Kraus, K.A. and Phillips, H.O., 1956. J. Am. Chem. Soc. 78, 249.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

372 Sorption of trace metals

Larsen, F. and Postma, D., 1997. Nickel mobilization in a groundwater well field: release by pyrite oxidation
and desorption from manganese oxides. Env. Sci. Technol. 31, 2589–2595.
Lenhart, J.J. and Honeyman, B.D., 1999. Uranium(VI) sorption to hematite in the presence of humic acid.
Geochim. Cosmochim. Acta 63, 2891–2901.
Lion, L.W., Altmann, R.S. and Leckie, J.O., 1982. Trace metal adsorption characteristics of estuarine
particulate matter: evaluation of Fe/Mn oxide and organic surface coatings. Env. Sci. Technol. 16,
660–666.
Lofts, S. and Tipping, E., 2000. Solid-solution metal partitioning in the Humber rivers: application of WHAM
and SCAMP. Sci. Total Env. 251, 381–399.
Manning, B.A., Fendorf, S.E. and Goldberg, S., 1998. Surface structures and stability of arsenic (III) on
goethite: spectroscopic evidence for inner-sphere complexes. Env. Sci. Technol. 32, 2383–2388.
Manning, B.A. and Goldberg, S., 1996. Modeling competitive adsorption of arsenite and molybdate on oxide
minerals. Soil Sci. Soc. Am. J. 60, 121–131.
Marinsky, J.A., 1987, A two-phase model for the interpretation of proton and metal ion interaction with charged
polyelectrolyte gels and their linear analogs. In W. Stumm (ed.), Aquatic surface chemistry, 49–81, Wiley
and Sons, New York.
McKenzie, R., 1980. The adsorption of lead and other heavy metals on oxides of Mn and Fe. Aust. J. Soil Res.
18, 61–73.
Milne, C.J., Kinniburgh, D.G. and Tipping, E., 2001. Generic NICA-Donnan model parameters for proton bind-
ing by humic substances. Env. Sci. Technol. 35, 2049–2059.
Milne, C.J., Kinniburgh, D.G., Van Riemsdijk, W.H. and Tipping, E., 2003. Generic NICA-Donnan model
parameters for metal-ion binding by humic substances. Env. Sci. Technol. 37, 958–971.
Morel, F.M.M. and Hering, J.G., 1993. Principles and applications of aquatic chemistry. Wiley and Sons, New
York, 374 pp.
Mukherjee, M., Sahu, S.J., Jana, J., Bhattacharya, R., Chatterjee, D., De Dadal, S.S., Bhattacharya, P. and Jacks,
G., 2000. The governing geochemical processes responsible for mobilisation of arsenic in sedimentary
aquifer of Bengal Delta Plain. In P.L. Bjerg, P. Engesgaard and Th.D. Krom (eds), Groundwater 2000,
201–202, Balkema, Rotterdam.
Murray, J.W., 1975. The interaction of metal ions at the manganese dioxide-solution interface. Geochim.
Cosmochim. Acta 39, 505–519.
Parks, G.A., 1967. Surface chemistry of oxides in aqueous systems. In W. Stumm (ed.), Equilibrium concepts
in aqueous systems. Adv. Chem. Ser. 67, 121–160.
Parks, G.A., 1990. Surface energy and adsorption at mineral-water interfaces: an introduction. Rev. Mineral. 23,
133–175.
Perdue, E.M. and Lytle, C.R., 1983. Distribution model for binding of protons and metal ions by humic sub-
stances. Env. Sci. Technol. 17, 654–660.
Perdue, E.M., Reuter, J.H. and Parrish, R.S., 1984. A statistical model of proton binding by humus. Geochim.
Cosmochim. Acta 48, 1257–1263.
Pokrovsky, O.S., Schott, J. and Thomas, F., 1999. Processes at the magnesium-bearing carbonates/solution inter-
face. 1. A surface speciation model for magnesite. Geochim. Cosmochim. Acta 63, 863–880.
Postma, D. and Appelo, C.A.J., 2000. Reduction of Mn-oxides by ferrous iron in a flow system: Column exper-
iment and reactive transport modeling. Geochim. Cosmochim. Acta 64, 1237–1247.
Randall, S.R., Sherman, D.M., Ragnarsdottir, K.V. and Collins, C.R., 1999. The mechanism of cadmium sur-
face complexation on iron oxyhydroxide minerals. Geochim. Cosmochim. Acta 63, 2971–2987.
Ravenscroft, P., McArthur, J.M. and Hoque, B.A., 2001. Geochemical and palaeohydrological controls on pol-
lution of groundwater by arsenic. In W.R. Chappell et al. (eds), Arsenic exposure and health effects IV,
Elsevier, Amsterdam.
Rietra, R.P.J.J., Hiemstra, T. and Van Riemsdijk, W.H., 1999. The relationship between molecular structure and
ion adsorption on variable charge minerals. Geochim. Cosmochim. Acta 63, 3009–3015.
Routh, J., Grossman, E.L., Murphy, E.M. and Benner, R., 2001. Characterization and origin of dissolved
organic carbon in Yegua ground water in Brazos County, Texas. Ground Water 39, 760–767.
Runkel, R.L., Kimball, B.A., McKnight, D.M. and Bencala, K.E., 1999. Reactive solute transport in streams: a
surface complexation approach for trace metal sorption. Water Resour. Res. 35, 3829–3840.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 373

Rusch, U., Borkovec, M., Daicic, J. and Van Riemsdijk, W.H., 1997. Interpretation of competitive adsorption
isotherms in terms of affinity distributions. J. Coll. Interf. Sci. 191, 247–255.
Russell, J.D., Parfitt, R.L., Fraser, A.R. and Farmer, V.C., 1974. Surface—structures of gibbsite, goethite and
phosphated goethite. Nature 248, 220–221.
Schindler, P.W. and Stumm, W., 1987. The surface chemistry of oxides, hydroxides and oxide minerals. In W.
Stumm (ed.), Aquatic surface chemistry, 83–110, Wiley and Sons, New York.
Schnitzer, M. and Kahn, S.U., 1978. Soil organic matter. Elsevier, Amsterdam, 319 pp.
Schultz, M.F., Benjamin, M.M. and Ferguson, J.F., 1987. Adsorption and desorption of metals on ferrihy-
drite: Reversibility of reaction and sorption properties of the regenerated solid. Env. Sci. Technol. 21,
663–669.
Seaman, J.C., Bertsch, P.M., Korom, S.F. and Miller, W.P., 1996. Physicochemical controls on nonconservative
anion migration in coarse-textured alluvial sediments. Ground Water 34, 778–783.
Shiao, S.Y., Rafferty, P., Meyer, R.E. and Rogers, W.J., 1979. Ion exchange equilibriums between montmoril-
lonite and solutions of moderate to high ionic strength. ACS Symp. Ser. 100, 297–324.
Sips, R., 1948. On the structure of a catalyst surface. J. Chem. Phys. 16, 490–495.
Smedley, P.L. and Edmunds, W.M., 2002. Redox patterns and trace-element behavior in the East Midlands
Triassic sandstone aquifer, U.K. Ground Water 40, 44–58.
Smedley, P.L. and Kinniburgh, D.G., 2002. A review of the source, behaviour and distribution of arsenic in nat-
ural waters. Appl. Geochem. 17, 517–568.
Smedley, P.L., Nicolli, H.B., Macdonald, D.M.J., Barros, A.J. and Tullio, J.O., 2002. Hydrogeochemistry of
arsenic in groundwater from La Pampa, Argentina. Appl. Geochem. 17, 259–284.
Sparks, D.L., 1989. Kinetics of soil chemical processes. Academic Press, San Diego, 210 pp.
Sposito, G., 1983. On the surface complexation model of the oxide-aqueous solution interface. J. Coll. Interf.
Sci. 91, 329–340.
Sposito, G., 1984. The surface chemistry of soils. Oxford Univ. Press, New York, 234 pp.
Stipp, S.L.S., Hochella, M.F., Parks, G.A. and Leckie, J.O., 1992. Cd2+ uptake by calcite, solid-state diffusion,
and the formation of solid-solution: Interface processes observed with near-surface sensitive techniques
(XPS, LEED and AES). Geochim. Cosmochim. Acta 56, 1941–1954.
Stollenwerk, K.G., 1994. Geochemical interactions between constituents in acidic groundwater and alluvium in
an aquifer near Globe, Arizona. Appl. Geochem. 9, 353–369.
Stumm, W., 1992. Chemistry of the solid-water interface. Wiley and Sons, 428 pp.
Stumm, W. and Morgan, J.J., 1996. Aquatic chemistry, 3rd ed. Wiley and Sons, New York, 1022 pp.
Swedlund P.J. and Webster J.G., 2001. Cu and Zn ternary surface complex formation with SO4 on ferrihydrite
and schwertmannite. Appl. Geochem. 16, 503–512.
Thurman, E.M., 1985. Organic geochemistry of natural waters. Nijhoff. Dordrecht, 497 pp.
Tipping, E., 1998. Humic ion-binding model VI: An improved description of the interactions of protons and
metal ions with humic substances. Aq. Geochem. 4, 3–48.
Tipping, E. and Hurley, M.A., 1992. A unifying model of cation binding by humic substances. Geochim.
Cosmochim. Acta 56, 3627–3641.
Tipping, E., Backes, C.A. and Hurley, M.A., 1988. The complexation of protons, aluminium and calcium by
aquatic humic substances: a model incorporating binding-site heterogeneity and macroionic effects. Water
Res. 22, 597–611.
Van Beek, C.G.E.M. and Van der Jagt, H., 1996. Mobilization and speciation of trace elements in groundwater.
Int. Workshop “Natural origin of inorganic micropollutants”, Vienna.
Van Cappellen, P., Charlet, L., Stumm, W. and Wersin, P., 1993. A surface complexation model of carbonate
mineral-aqueous solution interface. Geochim. Cosmochim. Acta 57, 3505–3518.
Van Riemsdijk, W.H., Bolt, G.H., Koopal, L.K. and Blaakmeer, J., 1986. Electrolyte adsorption on heteroge-
neous surfaces: adsorption models. J. Coll. Interf. Sci. 109, 219–228.
Venema, P., Hiemstra, T. and Van Riemsdijk, W.H., 1996a. Comparison of different site binding models for
cation sorption: description of pH dependency, salt concentration and cation-proton exchange. J. Coll.
Interf. Sci. 181, 45–59.
Venema, P., Hiemstra, T. and Van Riemsdijk, W.H., 1996b. Multi site adsorption of cadmium on goethite. J.
Coll. Interf. Sci. 183, 515–527.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

374 Sorption of trace metals

Wang, F.Y., Chen, J.S. and Forsling, W., 1997. Modeling sorption of trace metals on natural sediments by sur-
face complexation model. Env. Sci. Technol. 31, 448–453.
Wehrli, B., Ibric, S. and Stumm, W., 1990. Adsorption kinetics of vanadyl(IV) and chromium(III) to aluminum
oxide: evidence for a two-step mechanism. Coll. Surf. 51, 77–88.
Westall, J.C. and Hohl, H. 1980. A comparison of electrostatic models for the oxide/solution interface. Adv.
Coll. Interf. Sci. 12, 265–294.
Westall, J.C., Jones, J.D., Turner, G.D. and Zachara, J.M., 1995. Models for association of metal ions with het-
erogeneous environmental sorbents. 1. Complexation of Co(II) by Leonardite humic acid as a function of
pH and NaClO4 concentration. Env. Sci. Technol. 29, 951–959.
Zachara, J.M., Girvin, D.C., Schmidt, R.L. and Resch, C.T., 1987. Chromate adsorption on amorphous iron
oxy-hydroxide in the presence of major groundwater ions. Env. Sci. Technol. 21, 589–594.
Zachara, J.M., Cowan, C.E. and Resch, C.T., 1991. Sorption of divalent metals on calcite. Geochim.
Cosmochim. Acta 55, 1549–1562.
Zachara, J.M., Smith, S.C., Resch, C.T. and Cowan, C.E., 1992. Cadmium sorption to soil separates containing
layer silicates and iron and aluminum oxides. Soil Sci. Soc. Am. J. 56, 1074–1084.
Zeltner, W.A. and Anderson, M.A., 1988. Surface-charge development at the goethite aqueous-solution inter-
face-effects of CO2 adsorption. Langmuir 4, 469–474.
Zhang, P.C. and Sparks, D.L., 1990. Kinetics and mechanisms of sulfate adsorption/desorption on goethite
using pressure-jump relaxation. Soil Sci. Soc. Am. J. 54, 1266–1273.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

8

Silicate Weathering

Weathering of silicate minerals is a slow process and the resulting changes in water chemistry will
be gradual and less conspicuous than in carbonate aquifers (Chapter 5). Still, weathering of silicate
minerals is estimated to contribute about 45% of the total dissolved load of the world’s rivers (Stumm
and Wollast, 1990). In the global cycle of CO2, the weathering of silicate minerals acts as an impor-
tant CO2 sink. Furthermore, silicate weathering is the most important pH-buffering mechanism in
sediments without carbonate minerals. Because the rate of silicate dissolution is slow, aquifers in sili-
cate rock are vulnerable towards acidification. Also, forest deterioration due to acid rain has focused
attention on silicate weathering as pH-buffer.

8.1 WEATHERING PROCESSES

Traditionally, weathering of detrital silicate minerals has been studied in soils. Soils may be exposed
to chemical weathering over thousands of years and the variation in mineralogical composition
which develops as a function of depth and time displays even very slow degradation and transfor-
mation processes of minerals.
Figure 8.1 shows the mineralogical variation in a soil formed on a granodiorite. Cosmogenic 10Be in
quartz revealed that weathering has been going on here for at least 90,000 years (Schroeder et al., 2001).

Residual mineral (%)

0 50 100 0 50 100 0 50 100
0 0
Soil

2 2
Saprolite
4 4
Depth (m)

6 6
Bedrock

8 K-feldspar Biotite Fe-oxide Kaolinite 8

10 10
Plagioclase
12 12
0 2000 0 1000 0 200 0 250 500 0 1000 2000
Residual minerals (mol/m3)

Figure 8.1. Mineralogy of a soil developed on granodiorite (White et al., 2001). Weathered rock residue is also
called saprolite.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

376 Silicate weathering

Olivine
Calcic plagioclase

Decreasing weatherability
Augite
Intermediate plagioclase
Hornblende
Sodic plagioclase
Biotite

Potassium feldspar

Muscovite

Quartz

Figure 8.2. The Goldich weathering sequence based on observations of the sequence of their disappearance in
soils (Goldich, 1938).

The parent granodiorite rock consists of plagioclase (32%), quartz (28%), K-feldspar (21%), biotite
(13%), muscovite (7%) and small amounts of hornblende (2%). In Figure 8.1, the pristine bedrock
composition is found at the greatest depth. Already in the uppermost part of the bedrock the plagio-
clase disappears. Biotite and K-feldspar are apparently more resistant to weathering but decrease in
the saprolite (partially weathered rock). The sequential disappearance of different silicate minerals
reflects the overall kinetic control on the distribution of primary silicate minerals during weathering.
This was recognized as early as 1938 by Goldich who used field observations to compile the weather-
ing sequence shown in Figure 8.2. Olivine and Ca-plagioclase are listed as the most easily weather-
able minerals, and quartz as the mineral most resistant to weathering. The Goldich sequence is the
reverse of the Bowen reaction series which predicts the sequence of minerals precipitating during
cooling from basaltic magma. The minerals that formed at the highest temperature in the earth
(olivine) are the most unstable under weathering conditions at the surface of the earth.
The second important observation to be made from Figure 8.1 is the formation of secondary miner-
als like clays, here kaolinite, and Fe-oxides during the weathering process. These are the insoluble rem-
nants which form during incongruent dissolution of primary silicate minerals. Incongruent dissolution
strictly means that the ratio of elements appearing in the solution differs from the ratio in the dissolving
mineral. In silicate weathering studies the term incongruent dissolution is commonly extended to
include the effect of secondary precipitates. Weathering reactions for some common primary minerals
are listed in Table 8.1, where the clay mineral kaolinite is used as an example of a weathering product.

Table 8.1. Weathering reactions for different silicate minerals to the clay mineral kaolinite.

Sodic Plagioclase (Albite) Kaolinite

2Na(AlSi3)O8  2H  9H2O → Al2Si2O5(OH)4  2Na  4H4SiO4
Calcic Plagioclase (Anorthite)
Ca(Al2Si2)O8  2H  H2O → Al2Si2O5(OH)4  Ca2
K-feldspar (Microcline)
2K(AlSi3)O8  2H  9H2O → Al2Si2O5(OH)4  2K  4H4SiO4
Pyroxene (Augite)
(Mg0.7CaAl0.3)(Al0.3Si1.7)O6  3.4H  1.1H2O → 0.3Al2Si2O5(OH)4  Ca2  0.7Mg2
 1.1H4SiO4
Mica (Biotite)
2K(Mg2Fe)(AlSi3)O10(OH)2  10H  0.5O2  7H2O → Al2Si2O5(OH)4  2K  4Mg2
 2Fe(OH)3  4H4SiO4
CO2  H2O → H  HCO 3

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Weathering processes 377

The formation of secondary products is due to the insolubility of Al-compounds. The reactions in
Table 8.1 are therefore written so that aluminum remains conserved in the solid phase. Generally, these
reactions can be derived by following steps 1 to 4.
(1) balance all the elements except H and O,
(2) balance O by adding H2O,
(3) balance H by adding H and
(4) check the charge balance of the resulting equation (for redox reactions consult Chapter 9).
The effect of silicate weathering on the water chemistry is primarily the addition of cations and silica.
Nearly all silicate weathering reactions consume acid and increase the pH. Under unpolluted condi-
tions, carbonic acid and organic acids are the most important sources of protons. As indicated by the
last equation in Table 8.1, bicarbonate will be produced during weathering of silicates. Finally, iron
that is present in silicate minerals, like biotite or hornblende, may form Fe-oxide as an insoluble
weathering product (Figure 8.1).
Weathering reactions will also take place in silicate rocks or sandstones percolated by groundwater,
although the mineralogical transformations here often are more difficult to detect than in soils. Some
examples of groundwater compositions which may result from weathering of silicate minerals are
shown in Figure 8.3. The high silica content indicates active degradation of silicate minerals. The high-
est concentration is found in volcanic rocks (rhyolite and basalt) which contain more reactive material
than rock-types like mica-schists or granite. Sodium contributes in all the 4 groundwaters significantly
to the cations (K which has been added to Na in Figure 8.3 is in most cases a minor constituent) and
its presence is not balanced by chloride as expected when seawater was the source of Na. Sodium is
mainly derived from weathering of Na-feldspar like albite or any member of the plagioclase solid solu-
tion series between albite and anorthite (Ca-feldspar). Ca2 is released both from weathering of pla-
gioclase and of hornblendes (also called amphiboles) and pyroxenes. As suggested by Table 8.1, the
increase in cation concentration is accompanied by an increase in bicarbonate and it appears that in
some cases even carbonate precipitation can be the result of silicate weathering.

Legend

6
Na  K Cl  NO3
Well in basalt
Oregon
Mg SO4
Spring in rhyolite
meq/L or mmol/L

New Mexico
4
Ca ΣCO3
Well in
granitic gneiss
SiO2 Maryland
Well in mica schist
2 North Carolina

Figure 8.3. Examples of groundwater compositions in igneous and metamorphic rock. Dissolved silica
is expressed as mmol/L while charged ions are displayed in meq/L (modified from Hem (1985)).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

378 Silicate weathering

Weathering of primary silicates results in the formation of clay minerals like montmorillonite and
kaolinite, and of gibbsite. Using albite as an example, its transformation to different weathering
products is described by the following equations. For montmorillonite the presence of Mg2, for
example leached from pyroxenes, amphiboles or biotite, is assumed:

3Na(AlSi3 )O8  Mg 2  4H 2O → 2Na 0.5 (Al1.5Mg0.5 )Si 4O10 (OH)2  2Na  H 4SiO 4 (8.1)
albite montmorillonite

2Na(AlSi3 )O8  2H  9H 2O → Al 2Si 2O5 (OH)4  2Na  4H 4SiO 4

(8.2)
albite kaolinite

Na(AlSi3 )O8  H  7H 2O → Al(OH)3  Na  3H 4SiO 4

(8.3)
albitte gibbsite

The alteration of albite to montmorillonite consumes no acid, but with kaolinite and gibbsite as
weathering products, increasing amounts of protons are consumed. Furthermore, when albite alters
to montmorillonite, 89% of the Si is preserved in the weathering product, decreasing to 33% for
weathering to kaolinite and to 0% for gibbsite. The sequence of weathering products, going from
montmorillonite over kaolinite to gibbsite, reflects increasing intensity of leaching, removing silica
and cations from the rock.
The hydrological conditions in combination with the rate of mineral weathering determine the
nature of the weathering product. Montmorillonite is preferentially formed in relatively dry climates,
where the rate of flushing of the soil is low. Its formation is further enhanced when rapidly dissolv-
ing material such as volcanic rock is available. Gibbsite, on the other hand, typically forms in tropi-
cal areas with intense rainfall and well drained conditions. Under such conditions, gibbsite and
other Al-hydroxides may form a thick weathering residue, bauxite, that constitutes the most
important Al-ore.

100
Weight percent in total soil

80

60

Bauxite
40

20
Kaolinite
Montmorillonite

0 20 40 60 80 100 120 140 160

Mean annual rainfall (inches)

Figure 8.4. Weathering products on volcanic rocks on the island of Hawaii as a function of mean annual rain-
fall. Bauxite corresponds to Al-hydroxides and the contents of different minerals are plotted cumulatively as
weight percent of the total soil (Berner, 1971).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Weathering processes 379

A classic example of the effect of precipitation and leaching intensity on the composition of weathering
residues is found in the soils on Hawaii where bauxite, kaolinite and montmorillonite form differently
with altitude on the volcanic slopes (Figure 8.4). Montmorillonite is found in areas with low rainfall and
therefore with a longer residence time of water in the soil, so that the solute concentrations become higher.
At the other end of the scale, bauxite forms in high-rainfall areas where leaching is intense and the resi-
dence time of water in the soil is short, so that dissolved ion concentrations are low. Figure 8.4 also shows
the amounts of weathering product to increase with the amount of rainfall.
Authigenic clay mineral formation due to the alteration of primary silicates is studied extensively
in sedimentary petrology, since the type of clay mineral may seriously affect the permeability of
the rock. Kaolinite forms neat booklets which have only a moderate effect on the permeability,
but illite forms hairy aggregates that clog pore throats and reduce the permeability very
significantly. This feature is well recognized in oil exploration and is illustrated in Figure 8.5.
Sandstones cemented with illite are shown to have a permeability that is at least an order of magni-
tude lower than sandstones with the same porosity, but cemented with kaolinite.
Clay minerals that originate from silicate weathering remain fairly stable during erosion and transport,
and are deposited as coatings on sand grains, often closely intermingled with iron oxides. These coatings
are important for the ion exchange properties of the aquifer. In some cases the clay particles are washed
down from the upper soil layers deeper into the profile (Ryan and Gschwend, 1992). The total amount of
clay that is present in sand deposits is normally less than a few percent and consists in most cases of a mix-
ture of various clay minerals. An example of the relative distribution of clay minerals in a sandy deposit
is shown in Figure 8.6. Illite, montmorillonite and kaolinite are the dominant clay minerals.

2000

1000
800 Areas marked represent
600 the main concentrations
400 of points

200

100
80 Kaolinite
60
cemented
Permeability (md)

40
formation
20

10
8
6
4

1
0.8
0.6
Illite
0.4
cemented
0.2 formation

0.1
0 4 8 12 16 20 24 28
Porosity (percent bulk volume)

Figure 8.5. The effect of type of authigenic clay on the permeability of a sandstone (Blatt et al., 1980).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

380 Silicate weathering

Clay mineral distribution (% wt.)

10 20 30 40 50 60 70 80 90 100
2

6 Kaolinite

10

Illite

Depth (m) 14

18 Montmorillonite

22

26

30

Figure 8.6. The relative distribution of clay minerals in the unsaturated zone of a quaternary sand deposit
(Matthess et al., 1992).

In the upper part of the sequence the montmorillonite content tends to increase with depth and per-
haps montmorillonite is being transformed into illite. An alternative interpretation would be that the
sequence reflects differences in the source areas of the sediments. Sometimes, the source areas of the
clay minerals can be recognized by their
18O composition which is a function of the
18O of the soil
water (Savin and Hsieh, 1998; Salomons and Mook, 1987).

QUESTION:
Write balanced reactions for dissolution or weathering to kaolinite for: Olivine, Mg2SiO4; Olivine
(Hortonolite), (FeMg)SiO4; Intermediate Plagioclase (Oligoclase), (Na0.8Ca0.2)(Al1.2Si2.8)O8; Hornblende
or Amphibole (Glaucophane), Na2(Mg3Al2)Si8O22(OH)2, Mica (Muscovite), KAl2(AlSi3)O10(OH)2. Also
with gibbsite as weathering product.

ANSWER: (for oligoclase only)

(Na 0.8Ca 0.2 )(Al1.2Si 2.8 )O8  3.8H 2 O  1.2H →

0.8Na  0.2Ca 2  0.6Al 2Si 2 O5 (OH)4  1.6H 4SiO 4

and

(Na 0.8Ca 0.2 )(Al1.2Si 2.8 )O8  6.8H 2 O  1.2H →

0.8Na  0.2Ca 2  1.2Al(OH)3  2.8H 4SiO 4

8.2 THE STABILITY OF WEATHERING PRODUCTS

Gibbsite or Al-hydroxide is the most extreme product of silicate weathering and this mineral is used
as a starting point for our discussion of the stability of weathering products. The stability is described
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 The stability of weathering products 381

by the reaction:

Al(OH)3 gibbsite  3H ↔ Al3  3H 2O (8.4)

with the mass action equation

Kgibbsite  [Al3 ] / [H ]3 ≈ 1010 (8.5)

Equation (8.5) indicates that the activity of aluminum in water [Al3], will be strongly dependent on
the pH. Dissolved aluminum has, however, a distinct tendency to form hydroxy complexes which
may increase the solubility of gibbsite significantly. The stability of the most important Al-hydroxy
complexes is listed in Table 8.2.

Table 8.2. The stability of the predominant dissolved Al-

hydroxy complexes at 25°C, taken from PHREEQC.DAT.

Reaction log K

Al3  H2O ↔ Al(OH)2  H 5.09

Al3  2H2O ↔ Al(OH)2  2H 10.1
Al3  4H2O ↔ Al(OH)4  4H

22.7

The total amount of aluminum in solution consists of both complexed and uncomplexed aluminum
and is described by the mass balance equation:
A1  mAl3  mAlOH2  mAl(OH)  mAl(OH) (8.6)
2 4

In some cases also complexes between Al3 and fluoride, sulfate and organic matter have to be
considered (Driscoll, 1980).
The set of equations consisting of the mass action equations for the complexing reactions in
Table 8.2, the equation for gibbsite stability (8.5) and the mass balance equation (8.6) can be solved
by the methods given in the Section 4.2, to describe the solubility of total dissolved Al3 as a func-
tion of pH (Figure 8.7).

0 Solubility

Al(OH)3, (c)
4
log [Al, mol/L]

8

 Al(
H) 4 OH
12 (O
Al Al )2
(O
Al
3

H)
2

16
0 2 4 6 8 10 12 14
pH

Figure 8.7. The solubility of total dissolved aluminum in equilibrium with gibbsite (Al(OH)3) and aqueous
Al-hydroxy complexes.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
382 Silicate weathering

5 ite
bs
gib
−log[Al3]
e
llin
sta
Cr
y ite
bs
gib
e
4 llin
ta
r ys
oc
icr
M

3
4 4.2 4.4 4.6 4.8 5
pH

Figure 8.8. The saturation state of shallow groundwater for gibbsite. Activity calculations have been carried
out with PHREEQC at the in situ temperature of 10°C. The lines indicate saturation for gibbsite according to
Equation (8.5). At 10°C log K is 9.00 for crystalline gibbsite and 10.30 for microcrystalline gibbsite (Hansen
and Postma, 1995).

Below pH 4.5 the aluminum solubility increases and the Al3 concentration may by far exceed the
drinking water limit of 0.2 mg/L. In the near-neutral pH range, the total dissolved Al3 concentration
becomes as low as 1 mol/L. Above pH 7 the aluminate complex dominates dissolved aluminum and
the total solubility increases again.
Figure 8.8 shows the pH and [Al3] activity of some acid groundwaters. Also shown are the stabil-
ity lines for crystalline gibbsite or microcrystalline gibbsite, both corresponding to Equation (8.5).
Clearly the data covariate with the stability lines indicating a solubility control for Al3 in ground-
water by a gibbsite-type mineral. The use of two stability lines in Figure 8.8 illustrates that the gibb-
site phase in a natural sediment is not well defined. Inspection of the PHREEQC databases shows
the log K of Reaction (8.5) to vary from 8.11 for crystalline gibbsite to 10.8 for amorphous Al(OH)3,
that is two orders of magnitude difference in solubility! The microcrystalline gibbsite in Figure 8.8
corresponds to log K  9.0 and is intermediate.
Next after gibbsite, the stability of kaolinite is considered and the dissociation reaction is:

Al 2Si 2O5 (OH)4  6H ↔ 2Al3  2H 4SiO 4  H 2O (8.7)

In logarithmic form the mass action equation is:

log K  2 log[Al3 ]  2 log[H 4SiO 4 ]  6pH  7.4 (8.8)

Equation (8.8) can be rewritten as

log[Al3 ]  3pH  (7.4  2 log[H 4SiO 4 ]) / 2 (8.9)

The log[H4SiO4] of groundwater is mostly in the range from 4 to 3.3 and varies much less than
log[Al3] or pH. If log[H4SiO4] is approximately constant, Equation (8.9) becomes identical to
Equation (8.5). Substitution of a realistic value of 4 for log[H4SiO4] reduces the right hand part of

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Incongruent dissolution of primary silicates 383

Equation (8.9) to 7.7 which is close to the value of 8.11 for crystalline gibbsite. Accordingly, it is
difficult to identify the actual controlling mineral from saturation calculations.
Additional uncertainty concerns the slow dissolution kinetics of clay minerals which makes it
questionable whether true equilibrium is ever attained. This problem is encountered in the interpreta-
tion of field data, as well as in the determination of clay mineral stability by equilibration with water
in the laboratory. May et al. (1986) found it necessary to equilibrate kaolinite with water for 1237 days
in order to approach equilibrium, while smectite (montmorillonite) did not attain equilibrium in that
period. Despite these uncertainties, there is little doubt that the dissolved aluminum concentration in
groundwater is controlled by the solubility of weathering products like gibbsite or kaolinite.

QUESTION:
Find the pH where the solubility of gibbsite surpasses the drinking water limit (0.2 mg Al/L)?
ANSWER: SOLUTION; Al 0.2 mg/kgw; pH 4 Gibbsite; END. Here pH is adjusted to
equilibrium with gibbsite. Note: for pH 8 Gibbsite PHREEQC gives a different
answer, why?

8.3 INCONGRUENT DISSOLUTION OF PRIMARY SILICATES

Primary silicate minerals comprise feldspars, amphiboles, pyroxenes, micas, etc, that are present in
igneous and metamorphic rocks and as detrital minerals in sand and sandstones. The mineralogical
composition of a quaternary sand deposit is illustrated in Figure 8.9 and shows that quartz is by far
the dominant component of the sand. However, more easily weatherable minerals, like plagioclase,
K-feldspar and mica are also present in significant amounts.

GL
Quartz
K-feldspar
Plagioclase
Mica
10 Other
Depth (m)

20

30

0 75 80 85 90 95 100 vol %

Figure 8.9. The mineralogical composition of quaternary glacial outwash sands. The fraction “other” com-
prises heavy minerals, organic matter, etc. (Ohse et al., 1984).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

384 Silicate weathering

Albite
0 Amorphous AI(OH)3
0
Saturation index

Saturation index
2

2
Gibbsite
4
Ridge top pore water
4
Ridge slope pore water
6 Granite ground water

6
5.0 5.5 6.0 6.5 7.0 7.5 5.0 5.5 6.0 6.5 7.0 7.5
pH pH

Figure 8.10. The saturation index (log (IAP/K ) of groundwater for albite and Al(OH)3 in an area with granite
weathering (White et al., 2001).

The stability of these minerals in a groundwater system can be evaluated by calculating the satura-
tion state of the groundwater for a given mineral. For example for albite, we write the dissociation
reaction:

NaAlSi3O8  4H  4H 2O → Na  Al3  3H4SiO4 (8.10)

albite

As for other primary silicate weathering reactions (Table 8.1), the single arrow in the equation is
used to indicate that the reaction is irreversible at temperatures less than 50°C (Stéfansson and
Arnorsson, 2000). The mass action expression for Reaction (8.10) is

log K  log[Na ]  log[Al3 ]  3 log[H 4SiO 4 ]  4pH (8.11)

The saturation index for albite is shown for groundwater in a granitic area in Figure 8.10 and the
groundwater is found to be subsaturated for albite by orders of magnitude. Since albite is present in
the rock, this indicates that the kinetics of albite dissolution are slow. The saturation state for gibb-
site in the same waters is also included and shows approximate equilibrium with a phase ranging
between gibbsite and amorphous Al(OH)3, similar as in Figure 8.8. This pattern demonstrates the
general behavior of primary silicate minerals during weathering. The dissolution of the primary sili-
cate is slow and kinetically controlled while the precipitation of the secondary weathering product is
faster and approaches equilibrium (Helgeson et al., 1969). The dissolution kinetics of the primary
silicate are therefore overall rate limiting in the weathering process.
Although the saturation state approach outlined above is useful, it also has several drawbacks.
First of all, the Al3 concentration in groundwater of near neutral pH is often very low and cumber-
some to analyze accurately. Second, clusters of Al-hydroxy complexes can be present in the water
which may not be removed completely by filtering over 0.45 or even 0.1 m filters, and which are
unaccounted for in the speciation model. Third, even in the simple case of albite (Equation 8.11)
already 4 variables affect the saturation state, and the stability of albite is therefore difficult to dis-
play in a diagram that includes all the relevant parameters. To circumvent these problems, silicate
stability diagrams have been devised which assume that all Al3 is preserved in the weathering product.
Such a diagram for Ca-silicates is shown in Figure 8.11. It contains stability fields for the Ca-feldspar
anorthite and its possible weathering products, gibbsite, kaolinite and Ca-montmorillonite expressed
as a function of log([Ca2] / [H]2) and log[H4SiO4].

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Incongruent dissolution of primary silicates 385

20 [Ca2]
log x Basic rocks
[H]2
o Granitic rocks

Quartz
Anorthite N Norway
V Vosges
17 B Brittany
16 Mc Massif Central
Ca-montmorillonite A Alrance
15
C Corsica
14 Sa Sahara

13 Se Senegal
x Mc
oTc Tc Chad
12 Gibbsite Kaolinite x Ci
o Mc Ci Ivory Coast
C
11 Sa o x K Korhogo
Sa x o x Se
Ma Malagasy

Amorphous silica
10 Co Se
oB
x Ma
9 xV
o
8
A
o oA
7
N x Ko o Ma
oK
6
N o
5 ppm SiO2

0.6 1 6 10 50 100 200 log [H4SiO4]

5 4 3

Figure 8.11. The stability of anorthite and its possible weathering products gibbsite, kaolinite and
Ca-montmorillonite. Also included are the compositions of selected surface water samples from areas with
crystalline rocks (Tardy, 1971).

The peculiar choice of the y-axis parameter can be understood by considering the reaction between
anorthite and gibbsite:

CaAl 2Si 2O8  2H  6H 2O → 2Al(OH)3  Ca 2  2H 4SiO 4 (8.12)

anoorthite gibbsite

In this reaction all the Al3 released from anorthite is preserved in gibbsite. The relative stability
between the two minerals is therefore controlled by dissolved silica, Ca2 and pH. The mass action
equation of Reaction (8.12) is:

log K  log[Ca 2 ]  2 log[H 4SiO 4 ]  2 log[H ]  6.78 (8.13)

Rearranging Equation (8.13) yields:

log K  log([Ca 2 ] / [H ])2  2 log[H 4SiO 4 ]  6.78 (8.14)

Equation (8.14) is plotted as a straight line with slope 2 in the stability diagram (Figure 8.11) and the
four variables have been reduced to two. In general, the expression on the y-axis has the form
log([cationn] / [H]n) and reflects that the charge of released cations must be balanced by consumption
of H. Similarly, the equilibrium between Ca-montmorillonite and kaolinite is described by the reaction:

2H  3Ca.33 [Si7.33Al.67 ][Al 4 ]O 20 (OH)4  23H 2O ↔ 7Al 2Si 2O5 (OH)4  8H 4SiO 4  Ca 2
Ca − montmorillonite kaolinite (8.15)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

386 Silicate weathering

and

log K  log([Ca 2 ] / [H ]2 )  8 log[H4SiO4 ]  15.7 (8.16)

The boundary between kaolinite and gibbsite is described by the reaction:

Al 2Si 2O5 (OH)4  5H 2O ↔ 2Al(OH)3  2H4SiO4 (8.17)

kaolinitee gibbsite

and

log K  2 log[H 4SiO 4 ]  9.8 (8.18)

Accordingly, a H4SiO4 activity of 104.9 indicates equilibrium between kaolinite and gibbsite and
this phase boundary results in a straight line parallel to the y-axis in the stability diagram (Figure
8.11). Finally in the reaction of anorthite forming kaolinite, all silica is preserved in the solid phase
(Table 8.1) and the boundary therefore plots parallel to the log[H4SiO4] axis.

QUESTION:
Write the mass action equation for the reaction of anorthite forming kaolinite. Find the mass action con-
stant by combining Equations (8.18) and (8.13).
ANSWER: log K  6.78  (9.8)  16.58.

The discussed silicate stability diagrams contain the implicit assumption that Al3 is present in the
water, and in equilibrium with the depicted phases. However, a water sample that plots in the kaoli-
nite field in Figure 8.11 may well be subsaturated for kaolinite because of a low [Al3]. A better
statement would be that according to Figure 8.11, kaolinite is more likely to be stable than for exam-
ple gibbsite.
The composition of surface waters from different crystalline massifs in Europe and Africa
plot dominantly in the field of kaolinite (Figure 8.11). Basic rocks (basalts etc.) are more
reactive than granitic rocks and yield higher total dissolved solid concentrations (Figure 8.3).
Accordingly, waters from basic rocks plot in general closer to the stability field of montmorillonite
than those derived from granitic rocks. The leaching intensity also affects the water composition.
Therefore, water from a basic rock area in rainy Norway plots way down in the kaolinite field, while
a sample from a granitic area in arid Chad plots in the montmorillonite field.
Figure 8.11 considers only Ca-feldspar while most rocks contain several feldspars. Stability
diagrams for sodium and potassium feldspars can be constructed in a similar way, but this is still a
simplification, since feldspars in rocks often are solid solutions rather than pure end-member min-
erals. For example, plagioclases consist of a series of solid solutions of albite and anorthite. Also
the variation in the composition of clay minerals such as montmorillonite is large. The stability and
dissolution/precipitation behavior of solid solutions and of pure end-member minerals will affect the
stability boundaries in Figure 8.11 considerably and the conclusions drawn from these diagrams
must be used with caution. The second point of concern is again the slow reaction kinetics of silicate
minerals and it seems questionable whether true equilibrium is ever attained. Tardy (1971) reports
that many waters which plot in the kaolinite stability field were sampled from springs where mont-
morillonite is in the soils of the watershed.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Incongruent dissolution of primary silicates 387

EXAMPLE 8.1. Incongruent dissolution of K-feldspar

Irreversible dissolution of K-feldspar (microcline) as primary mineral, and the precipitation of secondary
weathering products at partial equilibrium is illustrated in the following PHREEQC input file.

SOLUTION 1
pH 4.8 charge; C(4) 1 CO2(g) -1.5
END

USE solution 1
EQUILIBRIUM_PHASES 1
Gibbsite 0 0
Kaolinite -1. 0
PRINT; -reset false
REACTION 1
KAlSi3O8 1; 10 umol in 20

USER_GRAPH
-head Si K-feldspar Gibbsite Kaolinite
-axis_titles Si umol/L
-start
10 graph_x tot("Si") * 1e6
20 graph_y tot("K") * 1e6, equi("Gibbsite") * 1e6, equi ("Kaolinite") * 1e6
-end
END

The initial solution is soil water containing carbonic acid. Equilibrium is specified for the weathering products
gibbsite and kaolinite with initial amounts of zero, implying that precipitation will start once saturation is
reached. The SI for kaolinite is set to 1.0 to comply with the relatively more stable kaolinite used by Tardy
in Figures 8.11 and 8.13. Irreversible dissolution of K-feldspar is simulated with REACTION, the amount
dissolved is given by the K concentration (tot(“K”) in USER_GRAPH).
Figure 8.12 shows the dissolution and precipitation reactions. K-feldspar dissolves stoichiometrically until
1.6 mol have reacted and the solution reaches saturation with gibbsite. Subsequently, gibbsite precipitates while
the concentrations of K and of Si increase. When saturation with kaolinite is reached, gibbsite is converted to
kaolinite by the reverse of Reaction (8.17). The reaction consumes all the Si coming from feldspar dissolution
and [H4SiO4] remains fixed until all gibbsite has been converted to kaolinite. Once gibbsite has disappeared, the
silica concentration increases again. The solution remains, throughout, strongly subsaturated for microcline.

10

6
M

4
Kaolinite

2 K-feldspar
dissolved Gibbsite
0
0 5 10 15 20
Si (µM)

Figure 8.12. Incongruent dissolution of K-feldspar resulting in sequential precipitation of gibbsite and
kaolinite, as modeled by PHREEQC.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

388 Silicate weathering

The water composition pathway is also shown in Figure 8.13, the potassium sister-diagram of Figure 8.11.
Initially the water composition plots at the lower left in the gibbsite stability field. As microcline dissolves
the water moves to the right until the kaolinite boundary. Here gibbsite is recrystallized to kaolinite and then
the water moves towards the microcline field. The pathway in water composition evolution is reasonably
consistent with the plotted water compositions from different localities and rock types.

[K]
log ——
[H]
x Basic rocks
o Granitic rocks

Amorphous silica
10 N Norway
V Vosges
9

Quartz
B Brittany
8 Mc Massif Central
A Alrance
7 Muscovite Microcline
C Corsica
6 Sa Sahara
Se Senegal
5
x Mc Tc Chad
4 oTc Ci Ivory Coast
Mc
o K Korhogo
3 Sa x x oSe x Se
C Sa oB C x Ma Malagasy
oo Ci
2 Gibbsite V
Vo
A x oA
1 o oK x Ma
o xN oK
0 N
Kaolinite
1

2

3 ppm SiO2
0.6 1 6 10 50 100 200 log[H4SiO4]
5 4 3

Figure 8.13. The stability diagram for K-feldspar microcline and its weathering products (modified from
Tardy, 1971). Superimposed is the pathway modeled with PHREEQC for incongruent dissolution of micro-
cline using an initial solution of H2O in contact with [PCO2]  101.5. To model the full pathway, the input
file in Example 8.1 was extended to dissolve a total of 0.5 mM KAlSi3O8.

QUESTIONS:
Plot the concentrations of Figure 8.12 logarithmically with PHREEQC (Note to use log10, cf. RATES in
the PHREEQC manual).
Plot the reaction path with log10([K] / [H]) and log10[H4SiO4] in Figure 8.13.
Find the names of K-feldspars in the PHREEQC databases?
ANSWER: Microcline, K-feldspar, Adularia, Sanidine
Continue dissolution of K-feldspar until saturation in the PHREEQC log10 plot. Discuss the (fixed) end-
point state.
Change the CO2 pressure of initial water to 102.5. Why does the line plot at a higher [K] / [H] than before?

Figures 8.11 and 8.13 include the stability lines for quartz and amorphous silica. For both sub-
stances, the solubility is described by the reaction:
SiO 2(s)  2H 2O ↔ H 4SiO 4 (8.19)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The mass balance approach to weathering 389

H4SiO4 remains undissociated at pH values below 9, and the stability of the SiO2(s) phases is deter-
mined by the solubility product of Reaction (8.19):

K  [H 4SiO 4 ] (8.20)

For quartz, K  103.98 at 25°C is generally used (see PHREEQC.DAT) but a recent study (Rimstidt,
1997) has suggested a higher solubility product of 103.74. Quartz has extremely sluggish reaction kinetics
and solutions grossly supersaturated for quartz are common. The slow precipitation of quartz allows
the formation of less stable forms of SiO2(s) like amorphous SiO2(s) also called opal, found in marine sedi-
ments as diatom tests, chalcedony (cryptocrystalline quartz), cristobalite, tridymite and opal-CT (disor-
dered cristobalite-tridymite) (Williams et al., 1985). The most soluble phase is amorphous SiO2(s) which
has a solubility product of about 102.7 (25°C) and places the upper constraint on the dissolved silica con-
centration.
The water compositions in Figures 8.11 and 8.13 range from subsaturated for quartz to close
to equilibrium with amorphous silica. The silica has probably been released to the water by weath-
ering of silicates like feldspars and because the precipitation of quartz apparently cannot keep pace
with the silica release, the water becomes supersaturated. In contact with slowly weathered granites
(Figure 8.11 and Schulz and White, 1999) and low reactive sandy aquifers (Kjøller, 2001; Ohse et al.,
1983) the silica concentration may approach quartz saturation quite closely. Volcanic rocks on the
other hand often contain volcanic glass that readily dissolves and silica concentrations may become
much higher (Figure 8.11 and Miretzky et al., 2001).

QUESTION:
Will the Al3 concentration change with pH in the kaolinite stability field of Figure 8.11 at fixed
[H4SiO4]  104?
ANSWER: Yes, it follows the pH / Al trend of Figure 8.8.

8.4 THE MASS BALANCE APPROACH TO WEATHERING

The problems encountered when treating silicate-water reactions as equilibrium systems have stim-
ulated the exploration of alternative approaches. One of the best alternatives is the use of mass balance
calculations, which relate changes in water chemistry to the dissolution or precipitation of minerals
and basically have the character of bookkeeping. For reactions between mineral and water we can
write the general reaction:
reactant phase → weathering residue  dissolved ions (8.21)

For the congruent dissolution of calcite, in water containing carbonic acid, the dissolution reaction is:

H 2CO3  CaCO3 → Ca 2  2HCO

3
(8.22)

Water that percolates through a soil with calcite should accordingly be enriched with two moles of
HCO 2
3 for each mole of Ca . For the incongruent dissolution of albite to kaolinite:

2NaAlSi3O8  2H 2CO3  9H 2O → Al 2Si 2O5 (OH)4  2Na  2HCO

3  4H 4SiO 4 (8.23)
albite kaolinite

the release of H4SiO4 is twice the release of HCO 

3 or of Na .

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

390 Silicate weathering

SiO2 Ca2
Calcite
K-feldspar K 
HCO3
HCO
3

SiO2
SiO2
Na
Na-feldspar Na
Biotite Mg2
HCO
3
K

Ca 2 HCO3
Ca-feldspar
HCO
3
SiO2
SiO2 
Na
Ca2 Hornblende Ca2
Pyroxene
Mg2 Mg2
HCO
3

HCO3

Figure 8.14. The composition of waters resulting from alteration of different silicate minerals to kaolinite in
the presence of carbonic acid, according to the reactions listed in Table 8.1 (Garrels and Mackenzie, 1971).

Exactly the same can be done for the other reactions listed in Table 8.1. The results are displayed
graphically in Figure 8.14 and show the expected water quality for each reaction. Note that a given
water composition is not necessarily unique for a specific mineral; the dissolution of calcite
(Reaction 8.22) and the weathering of anorthite to kaolinite (Table 8.1) yield the same relative con-
centrations of Ca2 and HCO 3 (Figure 8.14).
Rocks consist normally of mixtures of minerals and in some cases it is possible to reconstruct the
contributions of different weathering reactions from concentration changes along a flowpath.

EXAMPLE 8.2. Mass balance and the water chemistry of the Sierra Nevada (USA)
In a classic paper, Garrels and Mackenzie (1967) related the groundwater chemistry in a granitic area, Sierra
Nevada (USA), to silicate weathering reactions by the use of mass balance calculations. The approach used by
Garrels and Mackenzie was stepwise substraction of different weathering reactions. The chemistry of snow and
ephemeral spring waters is listed in Table 8.3. The concentrations in snow were subtracted from the spring
water, to obtain the contribution by rock weathering.

Table 8.3. The composition of snow and ephemeral spring water (Garrels and Mackenzie, 1967).
Concentrations in mmol/L.

Snow Spring Rock

weathering

Na 0.024 0.134 0.110

Ca2 0.01 0.078 0.068
Mg2 0.007 0.029 0.022
K 0.008 0.028 0.020
HCO 3 0.018 0.328 0.310
Si 0.003 0.273 0.270
pH 6.2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The mass balance approach to weathering 391

The granites of the Sierra Nevada contain the primary minerals plagioclase, biotite, quartz and K-feldspar
and the weathering product kaolinite. Quartz was not considered because its reactivity is so low. The intuitive
expectation is that weathering of the primary minerals to kaolinite (Table 8.1) may explain the spring water
composition. The average composition of the minerals found in the area are listed in Table 8.4.

Table 8.4. The composition of some minerals

in the granitic rocks in the Sierra Nevada.

Plagioclase Na0.62Ca0.38Al1.38Si2.62O8
Biotite KMg3AlSi3O10(OH)2
K-Feldspar KAlSi3O8
Kaolinite Al2Si2O5(OH)4

The calculation proceeds as follows. First all Na is attributed to weathering of plagioclase. This requires
the alteration of 0.110 / 0.62  0.177 mmol/L plagioclase to kaolinite:

.177Na 0.62 Ca 0.38 Al1.38Si 2.62 O8  .246CO2  .367H 2 O →

.122 Al 2Si 2 O5 (OH)4  .11Na  .068Ca 2  .246HCO

3  .22SiO 2

Subtracting the contribution of plagioclase weathering (Table 8.5) from the water chemistry shows that it also
accounts for all Ca2, as well as most of the HCO3 and SiO2. The next step is to attribute all Mg
2
to biotite
weathering and 0.022 / 3  0.0073 mmol/L biotite is altered to kaolinite:

.0073KMg 3AlSi3O10 (OH)2  .051CO 2  .026H 2 O →

.0037Al 2Si 2 O5 (OH)4  .0073K  .022Mg 2  .051HCO

3  .015SiO 2

This removes most of the HCO  

3 as well as some of the K . The final step is to attribute the remainder of K to
K-feldspar weathering and accordingly 0.013 mmol/L K-feldspar must alter to kaolinite:

.013KAlSi3O3  .013CO 2  .0195H 2 O →

.0065Al 2Si 2 O5 (OH)4  .013K +  .013HCO

3  .026SiO 2

What is left is only a minor amount of SiO2 and the calculation shows that the three weathering reactions may
explain the water chemistry surprisingly well. According to this reaction scheme, the dominant reaction is
the transformation of plagioclase into kaolinite under the consumption of carbonic acid, while minor
amounts of biotite and K-feldspar dissolve.

Table 8.5. The contributions of weathering various silicates to the composition of ephemeral
spring waters in the Sierra Nevada (mmol/L).

Rock weathering Minus plagioclase Minus biotite Minus K-feldspar


Na 0.110 0.000
Ca2 0.068 0.000
Mg2 0.022 0.022 0.000
K 0.020 0.020 0.013 0.000
HCO 3 0.310 0.064 0.013 0.000
Si 0.270 0.050 0.035 0.009

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

392 Silicate weathering

In a mass balance calculation a set of linear equations, that describe the components involved, is
solved and for this purpose dedicated computer programs like NETPATH (Plummer et al., 1991) are
available. The problems that can be solved comprise water composition evolution in terms of min-
eral dissolution and precipitation and simple mixing of end-member waters. The approach is also
termed inverse modeling: from the final result, the chemical reactions are unraveled by backward
calculus. NETPATH handles mass balances including redox reactions, isotopic mole balances, iso-
topic fractionation etc. PHREEQC also contains an option for mass balance calculation or inverse
modeling and it can adjust concentrations in the optimization process to achieve charge and mass
balance. In addition, PHREEQC employs an electron balance equation for redox processes, and an
alkalinity-balance equation. In Example 8.3 the mass balance calculation is illustrated for Sierra
Nevada spring water.

EXAMPLE 8.3. Mass balance calculation of mineral weathering as source for solutes in ephemeral spring
waters in the Sierra Nevada using PHREEQC
The overall mass balance equation is:
initial solution  reactant phases → final solution  product phases (8.24)

and the mass balance equations on elements is:

P
mT ,k  mT , k (final )  mT, k (initial )  ∑  p b p, k (8.25)
p1

For each element k  1 to J.

Here mT,k is the change in total molality of the kth element between final and initial solution, while P is
the number of total reactant and product phases, p is the mass transfer of the pth phase and bp,k denotes the
stoichiometric coefficient of the kth element in the pth phase. For solving the set of linear equations, the
number of unknowns (independent elements) should equal the number of equations. For Equation (8.25),
this means that the number of independent elements equals the number of phases. The mass balance equa-
tion for the various elements becomes:
mT,Na  0.134  0.024  0.62plagioclase
mT,Ca  0.78  0.01  0.38plagioclase (Note that Na and Ca2 are dependent by the stoichio-
metric ratio in plagioclase)
mT,Mg  0.029  0.007  3biotite
mT,K  0.028  0.008  1biotite  1K-feldspar
mT,Si  0.273  0.003  2.62plagioclase  3biotite  3K-feldspar  2kaolinite
mT,C  0.328  0.018  1CO -gas
2

We could also have chosen to balance on mT, Al  0 with the appropriate coefficients instead of on Si,
when doing the calculation by hand. The PHREEQC input file is:
DATABASE wateq4f.dat # used for phlogopite (biotite)
SOLUTION_SPREAD
Number pH Na Ca Mg K Alkalinity Si Cl S(6)
1 5.6 0.024 0.01 0.007 0.008 0.018 0.003 .014 .01
2 6.2 0.134 0.078 0.029 0.028 0.328 0.273 .014 .01
INVERSE_MODELING
-solutions 1 2 # find solution 2 from 1
-uncertainty 0.05 # default is 0.05 or 5%
-phases
CO2(g)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The mass balance approach to weathering 393

Kaolinite force prec

Plagioclase force dis
Phlogopite # iron-free biotite
Adularia
-balances
Cl; S(6) -0.7e-5
PHASES
Plagioclase
Na0.62Ca0.38Al1.38Si2.62O8 + 5.52H+ + 2.48H2O = .62Na+ + .38Ca+2 + 1.38Al+3 + 2.62H4SiO4
log_k 0.0
END

Solutions number 1 and 2 correspond to the snow and ephemeral springs respectively in Table 8.3. The pH
of snow water was not given in Garrels and Mackenzie (1967) but is estimated to be 5.6, corresponding to
equilibrium with the atmospheric PCO2 (Chapter 5). The keyword INVERSE_MODELING controls the
mass balance calculations. -solutions 1 2 indicates that solution 2 is to be balanced from the initial solution 1,
using the constraints:
-uncertainty 0.05 specifies that 5% may be added to or subtracted from all the concentrations to achieve
charge balance in the solutions (E.B.  0) and to obtain mass balance with the phase transfer;
-phases lists the minerals whose mass transfer must be considered to explain the evolution of solution 1 to
2. Following the mineral names, the option “dis” indicates dissolution only, “prec” precipitation only, and
“force” that the mineral should be included in every model considered. -phases reads mineral data from the
database, and since biotite is not listed, we look for another one. Phlogopite is in wateq4f.dat, and is the
name for iron-free biotite. The specific plagioclase (andesine) is not present in any database and is defined
separately under the keyword PHASES;
-balances lists the elements not in -phases, in this case, Cl and SO42. Solution 1 has a large charge
imbalance of 11%, and can be balanced only when the allowed uncertainty for all the elements is increased
to 0.12, or when Cl or SO42 is given a higher uncertainty. Here, an absolute uncertainty of 0.7105
(negative number) is defined for SO42, which is just the amount of SO42 needed for charge-balance in
solution 1.
PHREEQC finds one model and a selection of the output file showing the mineral mass transfer is:
Phase mole transfers:
CO2(g) 6.473e-04 CO2
Kaolinite -1.330e-04 Al2Si2O5(OH)4
Plagioclase 1.783e-04 Na0.62Ca0.38Al1.38Si2.62O8
Phlogopite 7.333e-06 KMg3AlSi3O10(OH)2
Adularia 1.267e-05 KAlSi3O8

Positive values indicates dissolution and a negative values precipitation. Comparison with the manual calcula-
tions in Example 8.2 shows excellent agreement, although more CO2(g) is consumed because the PHREEQC
mass balance includes H2CO3. The most important reaction is still the alteration of plagioclase to kaolinite,
while only small amounts of biotite and K-feldspar dissolve.

The concentration of solutes in rain and snow by evapotranspiration was not considered in the mass
balance calculations of Examples 8.2 and 8.3. It appears unimportant in the example since the Cl
concentrations in the snow and spring water are the same. Actually, average Cl concentrations were
found to be lower in the springwaters than in the precipitation, which suggested sorption of Cl in
the soil (cf. Chapter 2). Evapotranspiration can be included in manual calculations (Example 8.2) by
increasing all the concentrations in the initial water by a factor that includes evapotranspiration and
dry deposition (Chapter 2). In PHREEQC calculations, the phase “Water” can be added under -
phases (it needs the definition of Water; H2O  H2O; log_k 0 with keyword PHASES).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

394 Silicate weathering

Mass balance calculations are a useful approach for identifying possible reactions that may explain
differences in water chemistry along a flow path, not only in the case of silicate weathering, but also
for carbonates, redox reactions, etc. Often they form the first step in elucidating geochemical
processes before attempting the more complex approach of forward modeling. One should always
keep the following limitations of mass balances in mind:

(1) The solution of a mass balance calculation is not necessarily unique. Different choices of phases
may lead to equally consistent reaction schemes.
(2) There are no thermodynamic constraints on mass balance calculations. The mass balance calcu-
lation may predict impossible reactions, like precipitation of gibbsite even though the water is
subsaturated.
(3) Mass balance calculations do not consider what is kinetically consistent. They may predict that
quartz is dissolving, but not plagioclase, although this kinetically would be unreasonable.
(4) Mass balance calculations assume steady state; water samples along a flow path are
usually taken at the same time and the differences in water chemistry are assumed to be due to
reactions with minerals and not to temporal variations in the composition of water entering the
system.
(5) Mass balances assume a homogeneous reaction between the points of analysis. This condition is
hard to test in aquifer systems and certainly questionable in soil systems unless mass balances
are carried out for the different soil horizons.
In other words, use your geochemical common sense in the interpretation of mass balance
calculations. In the Sierra Nevada example, one could plot the water chemistry of the spring in the
silicate stability diagram (Figure 8.11) and check whether kaolinite is stable relative to gibbsite
(do this yourself, neglecting differences between activities and molar concentrations). Also the pref-
erential dissolution of plagioclase compared to K-feldspar as observed in Examples 8.2 and 8.3
appears kinetically reasonable. Note that the rate of mineral dissolution or precipitation can
be derived from a mass balance calculation when the travel distance and the groundwater flow rate
are known.

12
Augite
Si (µmol/g of solid)

Diopside
8

Enstatite
4

Bronzite
0
0 4 8 12 16
Time (days)

Figure 8.15. The release of silica to the solution over time for pyroxenes and amphiboles at pH 6. Experi-
ments with enstatite, augite and diopside were carried out at 50°C while bronzite was dissolved at 20°C and
PO2  0 atm (Schott and Berner, 1985).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of silicate weathering 395

8.5 KINETICS OF SILICATE WEATHERING

The dissolution kinetics of silicate minerals have been assessed both in laboratory experiments and
in the field. In laboratory studies, mineral grains are dissolved in solutions of various composition
(White and Brantley, 1995; Sverdrup, 1990) while monitoring the rate of dissolution from the Si release
(Figure 8.15). This type of results allows to quantity the dissolution kinetics of various minerals and
the change in crystal size during dissolution. Table 8.6 shows the calculated lifetime of 1 mm
crystals at pH 5 for various minerals and indicates a very large range. For example, Ca-feldspar
(anorthite) dissolves about 700 times faster than K-feldspar. More recent data suggest, however, a
smaller difference between the dissolution rates of K-feldspar and albite (Blum and Stillings, 1995).
Qualitatively, the lifetimes calculated from laboratory experiments are in good agreement with the
Goldich weathering sequence (Figure 8.2). The long dissolution times are also in agreement with the
high age of the Panola granodiorite weathering profile shown in Figure 8.1.

Table 8.6. The mean lifetime in years of 1 mm

crystals of various minerals calculated from
laboratory dissolution studies at 25°C and pH 5
(Lasaga, 1984).

Mineral Lifetime

Quartz 34,000,000
Muscovite 2,700,000
Forsterite 600,000
K-feldspar 520,000
Albite 80,000
Enstatite 8,800
Diopside 6,800
Nepheline 211
Anorthite 112

Apart from the mineralogy, other factors like solution composition may influence the dissolution
rate. Equation (8.26) contains the various parameters used in rate equations by Sverdrup and
Warfvinge (1995).

[H ]n 1 [OH ]o [PCO ]0.6 [R ]0.5

r  kH  kH O  kOH  kCO 2
 k org (8.26)
fH 2 fH O f OH 2 f CO f org
2 2

where r is the reaction rate in mol/m2/s, ki are the rate coefficients for the solutes that influence
the rate (mol/m2/s), [i] indicate the solute activities of H, OH, free organic radicals R and the
partial pressure of CO2, n and o are the apparent reaction orders, and fi are inhibition factors.
The first term on the right hand describes the effect of protons on the rate, the second term the rate
contribution due to hydrolysis, the third term entails the contribution of OH, the fourth gives the
influence of CO2 and the last term the effect of organic acids.
The influence of pH on the dissolution kinetics of several silicate minerals is shown in Figure 8.16.
In all cases, the rate shows a distinct minimum at neutral pH and the rate increases both towards
lower and higher pH. However, the absolute rates and the rate dependency on the pH vary signifi-
cantly. For example, in the near neutral pH range the dissolution rate of hornblende is two orders of
magnitude higher than of albite.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

396 Silicate weathering

7
Weathering of silicate minerals
8 Forsterite, pyroxene group
Hornblende, amphibole group
Albite, feldspar group

log dissolution rate (keq/m2/s)

9

10

11

12

13

14

15

16
0 2 4 6 8 10 12 14
Solution pH

Figure 8.16. Experimental dissolution rates for different silicate minerals at 25°C (Sverdrup and Warfvinge,
1988; Chou and Wollast, 1985).

The inhibition factor fH in Equation (8.26) indicates that proton-promoted dissolution is inhibited by
solution elements included in the term fH. The inhibition factors have the form:
x BC xAl
 [BC ]   3 
f H  1   1  [Al ]  (8.26a)
Lim   LimAl,H
 BC, H   
z BC zAl
   [Al3 ] 
[BC ]  
 1   1 
 
fH (8.26b)
2O LimBC, H O LimAl,H O
 2   2 
0.5
 [R ] 
f org 
 1 (8.26c)
 Limorg 

f OH  f CO  1 (8.26d)
2

where Lim is the limiting activity below which the inhibitive effect of the specific solute starts to
become negligible. [BC] indicates the sum of the activities of the base cations Na, K, Mg2 and
Ca2. The exponents xi and zi are empirical.
Equations (8.26a and b) indicate that dissolved aluminum may influence the dissolution rate. The
effect of aluminum on the dissolution rate of albite is shown in Figure 8.17 and apparently, already low
concentrations of dissolved aluminum inhibit the dissolution of albite. Since the Al-concentration under
natural conditions also is strongly pH dependent (Figure 8.7), the two rate-controlling parameters inter-
act in a complex way. The first term in Equation (8.26a) indicates that the presence of base cations in
solution may also slow down silicate dissolution. The fourth and fifth term of Equation (8.26) show that

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of silicate weathering 397

15.0

log rate
15.5

16.0
7 6 5 4 3 2
log (mAI, mol/L)

Figure 8.17. The effect of dissolved Al concentrations on the dissolution rate of albite at pH 3 (Chou and
Wollast, 1985).

both a high CO2 and a high content of free organic acid radicals may accelerate the dissolution rate
(Sverdrup, 1990; Sverdrup and Warfvinge, 1995). The role of CO2 is partly to act as a provider of pro-
tons, but particularly in the near neutral and basic pH range, the carbonate ion may accelerate silicate
weathering directly (Berg and Banwart, 2000). The coefficients for calculating approximate rates with
Equation (8.26) were obtained from Sverdrup and Warfvinge (1995) and Sverdrup (1990) and are listed
in Tables 8.7 and 8.8. An application for K-feldspar dissolution is presented in Example 8.4.

Table 8.7. Coefficients for calculating approximate dissolution rates of silicate minerals in soils for use with
Equation (8.26). Coefficients ki are in mol/m2/s and given as pki  log ki for 8°C, coefficients Limi are 106.
Arrhenius factors for recalculating pki to other temperatures are in Table 8.8. kCO2
1013 mol/m2/s, korg

1012 mol/m2/s, Limorg  5  106.

Mineral pkH nH LimAl xAl LimBC xBC pkH2O zAl zBC pkOH oOH

K-feldspar, K(AlSi3)O8 11.7 0.5 4 0.4 500 0.15 14.5 0.14 0.15 13.1 0.3
Albite, Na(AlSi3)O8 11.5 0.5 4 0.4 500 0.2 13.7 0.14 0.15 11.8 0.3
Anorthite, Ca(Al2Si2)O8 6.9 1.0 4 0.4 500 0.25 13.2 0.14 0.25 12.0 0.25
Pyroxene, Mg0.9Ca0.5Fe0.6Si2O6 9.9 0.7 500 0.2 200 0.3 15.1 0.1 0.3 8.6 0.5
Hornblende, 11.4 0.7 30 0.4 200 0.3 14.0 0.3 0.3 11.2 0.3
NaCa2Mg4(Fe)(Al2Si6)O22(OH)2
Epidote, Ca2(Al2Fe)Si3O12OH 11.4 0.5 500 0.3 200 0.2 15.3 0.2 0.2
Biotite, K(Mg2Fe)(AlSi3)O10(OH)2 12.6 0.6 10 0.3 500 0.2 14.5 0.2 0.2
Muscovite, KAl2(AlSi3)O10(OH)2 12.2 0.5 4 0.4 500 0.1 14.5 0.2 0.1 12.7 0.3
Chlorite, 12.5 0.7 50 0.2 200 0.2 14.9 0.1 0.1 13.0 0.3
(Mg8Fe2Fe3Al2)(Al3Si5)O20(OH)16
Apatite, Ca5(PO4)3OH 10.8 0.7 300 0.4 13.8 0.2

Table 8.8. Temperature factors (Ea / (2.303 R) in the Arrhenius equation (4.63) for calculating silicate
weathering rates ki at another temperature (T K) than 8°C (271 K):

Ea  1 1 
pki,T  log ki,T  pki, 271     (8.27)
2.303 R  T 271 

pkH pkH2O pkOH pkCO2 pkorg

Ea / (2.303 R) 3500 2000 2500 2000 3000

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

398 Silicate weathering

EXAMPLE 8.4. Dissolution kinetics of K-feldspar calculated by PHREEQC

We calculate the dissolution kinetics of K-feldspar in the soil from Example 4.14 and consider the various
influences on the rate given by Equation (8.26). The soil contains 10% K-feldspar as 0.1 mm spheres,
A0  2.26 m2/kg, w  0.3 hence VH O  0.162 L/kg (cf. Example 4.14). The PHREEQC input file is:
2

DATABASE phreeqc.dat # for K-feldspar

SOLUTION 1
temp 10; pH 4; Cl 0.1 charge
EQUILIBRIUM_PHASES 1
Gibbsite 0 0
# CO2(g) -2.5
KINETICS 1
K-feldspar
-m0 2.16 # 10% K-fsp, mol/L pore water
-parms 2.26 0.162 # A0 in m2, V in L
-steps 4.7e7 in 40 # 1.5 years
INCREMENTAL_REACTIONS true

RATES
K-feldspar
-start
# specific rates from Table 8.7 in mol/m2/s
# parm(1) = A in m2, parm(2) = V in L (recalc's sp. rate to mol/kgw)
1 A0 = parm(1); 2 V = parm(2)
# find activities of inhibiting ions. . .
3 a_Al = act("Al+3"); 4 BC = act("Na+") + act("K+") + act("Mg+2") + act("Ca+2")
# temp corrected with the Arrhenius eqn, Table 8.8
# the difference in temperature, TK gives solution temp in Kelvin...
10 dif_T = 1/TK - 1/271
# rate by H+...
20 pk_H = 11.7 + 3500 * dif_T
22 rate_H = 10^-pk_H * act("H+")^0.5 / ((1 + a_Al / 4e-6)^0.4 * (1 + BC / 5e-4)^0.15)
# rate by hydrolysis...
30 pk_w = 14.5 + 2000 * dif_T
32 rate_w = 10^-pk_w / ((1 + a_Al / 4e-6)^0.14 * (1 + BC / 5e-4)^0.15)
# rate by OH-...
40 pk_OH = 13.1 + 2500 * dif_T
42 rate_OH = 10^-pk_OH
# rate by CO2...
50 pk_CO2 = 13.0 + 2000 * dif_T
52 rate_CO2 = 10^-pk_CO2 * (10^SI("CO2(g)"))^0.6
# Sum the rate contributions...
60 rate = rate_H + rate_w + rate_OH + rate_CO2
# normalize to mol/kgw, correct for m/m0 and the approach to equi…
70 rate = rate * A0 / V * (m/m0)^0.67 * (1 - SR("K-feldspar"))
# integrate...
80 moles = rate * time
90 save moles
-end

USER_GRAPH
-head year Si
-axis_titles Years "mmol Si/L"
-start
10 graph_x Total_time / (365 * 24 * 3600)
20 graph_y tot("Si") * 1e3
-end
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetics of silicate weathering 399

The initial solution has a pH of 4 and temperature is 10°C. From the previous discussion, incongruent dis-
solution of K-feldspar is expected and therefore the input file allows gibbsite to precipitate once the solu-
tion becomes supersaturated. The result of this calculation is shown in Figure 8.18 as case a), indicated by
the bold line. Dissolved Si, monitoring K-feldspar dissolution, increases almost linearly until equilibrium
for K-feldspar is reached and then remains constant.

0.5

e
0.4
Si (mmol/L)

0.3 a
c
0.2
d
b
0.1

0
0 0.5 1 1.5
Years

Figure 8.18. PHREEQC calculation of K-feldspar dissolution kinetics using Equation (8.26). The condi-
tions are varied as follows: (a) Equilibrium for gibbsite imposed, initial pH  4 and 10°C. (b) as case (a)
but without equilibrium for gibbsite, (c) as case (a) but using a temperature of 25°C, (d) as case (a) but with
an initial pH of 5, (e) as case (a) but with PCO2  102.5 atm imposed.

Small modifications in the input file allow the exploration of the effects of the different parameters in Equation
(8.26). Only one parameter is changed each time while the others remain the same as in case a). First the equi-
librium constraint for gibbsite was removed, b), and as the result less than half the K-feldspar dissolves to
reach equilibrium. Without gibbsite precipitation, both Al3 and OH remain in solution and saturation with
K-feldspar is reached more rapidly. In c) the temperature is increased from 10 to 25°C and as the result the dis-
solution rate of K-feldspar more than doubles, illustrating the importance of the temperature on weathering
rates. Also, the solubility of K-feldspar increases. In d) the initial pH was increased from 4 to 5 which decreased
the amount dissolved at equilibrium by about 50%. Finally, Equation (8.26) predicts that CO2 stimulates the
dissolution rate and therefore in e) a PCO2 of 102.5 atm was included. CO2 has two different effects. First
the CO2 increases the dissolution rate slightly and second, hydroxyl ions react with H2CO3 and form HCO 3
allowing more K-feldspar to dissolve before saturation is reached.

QUESTIONS:
Include dissolution of quartz (Example 4.15), discuss the effect on the Si concentration with time?
Let quartz precipitate 10 times slower than it dissolves?
The quartz rate is for 25°C, include the temperature effect, the Arrhenius activation energy is Ea  90 kJ/mol?
Compare the dissolution rates of K-feldspar and quartz at 100°C?
Alkalis increase the quartz dissolution rate by (1  1.5 mNa), mNa in mmol/L (Dove and Rimstidt, 1993).
Include the alkali effect assuming that K reacts similar to Na?
ANSWER: add the rate for quartze and modify KINETICS
RATES
Quartz
-start
1 A0 = parm(1); 2 V = parm(2)
10 dif_T = 1/TK - 1/298
20 rate = 10^-(13.7 + 4700*dif_T) * (1 - SR("Quartz"))\
* A0/V * (m/m0)^0.67 * (1 + 1500*tot("K"))
30 if SR("Quartz") > 1 then rate = rate * 0.1
40 moles = rate * time
50 save moles
-end

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

400 Silicate weathering

8.6 FIELD WEATHERING RATES

Silicate weathering rates have been studied extensively in soils (Sverdrup, 1990; Rosén, 1991; White and
Brantley, 1995). However, agreement between field and laboratory dissolution rates is frequently poor
and subject to continuous debate (White and Brantley, 1995; Langan et al., 2001; White et al., 2001).
The kinetic dissolution model PROFILE (Sverdrup and Warfvinge, 1988, 1993; Sverdrup, 1990)
uses laboratory dissolution rate data to estimate the field rate of silicate weathering. The distribution of
minerals in the soil must be determined and the mineral surface areas estimated. Then, using the kinetic
rate laws determined in the laboratory, the model calculates the dissolution of different minerals con-
sidering the effects of the soil solution composition, available surface areas and the temperature. The
total weathering rate is given in terms of amount of acid neutralized by base cation release as:
horizons minerals
ANReq  ∑ zi
i ∑ rj Aw nBCj
ij
(8.28)
i1 j 1

where ANReq is the acid neutralization rate for the soil (eq/m2/yr), zi is thickness of layer i (m),
i
is amount of water in the layer (kg H2O/m3 soil), rj is in mol/m2/yr, Awij the total exposed surface area
(m2/kg H2O) and nBC, j are the equivalents of base cations released per mol (Na  K  2(Ca2 
Mg2) per formula unit).
The model has been applied to a number of watersheds (Sverdrup and Warfvinge, 1993; Hodson
et al., 1997). Results for the Gårdsjön watershed, which is located on a granitic bedrock with shallow
podzolic soils are shown in Figure 8.19. Generally, the weathering rate increases down through the
upper 0.65 m of the profile. Plagioclase, hornblende, epidote and apatite are the main contributors to
weathering. Hornblende and epidote constitute at most 2% of the mineral content of the soil, while the
content of microcline and plagioclase ranges between 12 and 16%. The important role of even minor
amounts of hornblende and epidote is obviously due to their fast dissolution kinetics. This type of
model clearly requires detailed information on soil mineralogy, the dissolution kinetics for the vari-
ous minerals, and particularly, the exposed reactive surface areas of the minerals (Hodson et al., 1997).

K-feldspar
0–5 Plagioclase
Epidote
Soil depth (cm below surface)

Hornblende
Apatite
5–15

15–35

35–65

65–80

0.00 0.05 0.10 0.15

Weathering rate (keq/ha/yr in layer)

Figure. 8.19. Contribution of different minerals to weathering in four soil horizons in the granitic watershed
of Gårdsjön, Sweden (Sverdrup and Warfvinge, 1993).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Field weathering rates 401

An alternative way to estimate field weathering rates is to consider the change in mineralogical com-
position through a soil sequence which has developed over a known period of time (Bain et al., 1991;
Olsson and Melkerud, 1991; Sverdrup, 1990; White et al., 1996). The disappearance of minerals is
calculated relative to an inert tracer such as quartz or zircon. The historical acid neutralization rate
is obtained by summing the relative loss over depth:

 1  x  wS, j
ANReq   
 t  ∑ zi ∑  xS , qu xR, j  xS, j  n
MW j BC, j
(8.29)
i j  R, qu 

where t is the time interval of soil formation, zi is the thickness of the soil layer (m), x indicates the
weight fraction in the parent rock (subscript R) or soil (subscript S) of quartz (qu) or mineral j, wS, j
is the amount of mineral j in the soil layer (g/m3), MWj is the molecular weight of mineral j (g/mol)
and nBC, j gives again the equivalent base cations per formula unit.
Figure 8.20 shows the weathering rates calculated from the mineral distributions in granitic sands
found in alluvial terraces of different age in the Merced chronosequence (White et al., 1996).

17
Hornblende A horizon

Plagioclase A horizon

K-feldspar A horizon

Hornblende BC horizon

Rates incorporating
18 Plagioclase BC horizon
geometric surface areas
K-feldspar BC horizon
log rate constant, kr (mol/cm2/s)

19

Rates incorporating
variable surface roughness
20

21
0 1000 2000 3000
Age (kyr)

Figure 8.20. Calculated dissolution rate constants for specific minerals as a function of age in granitic sands
of the Merced chronosequence. Rates are given both for geometric surface areas and for variable surface rough-
ness based on the BET surface area (White et al., 1996).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

402 Silicate weathering

Rates are expressed both as a function of the geometric surface area and relative to the actual surface
area based on the BET surface area (cf. Chapter 4). Naturally, the BET based rates are much lower
than those based on the geometric surface area. Furthermore, the weathering rates for individual
minerals decrease with soil age, fresh mineral grains dissolve faster than old mineral grains (Bain
et al., 1991). Another uncertainty is that the weathering rate calculated from mineralogical changes
in the soil profile provides an average weathering rate over a long period of time, during which
the conditions controlling the weathering rate, like the pH of the soil solution, may well have
varied.
The third approach towards obtaining field weathering rates is the mass balance as already intro-
duced in Section 8.4 for a groundwater flow path. To obtain the rate of weathering for a whole water-
shed the general mass balance equation for base cations (Ca, Mg, K, Na) becomes:

[flux from weathering]  [flux with outflow]  [flux in exchange pool]

 [flux in biomass]  [flux from atmosphere] (8.30)

In a steady state situation both the changes in the exchange and biomass will must be zero, and the
weathering rate is then obtained as the difference between solute outflow and atmospheric input.
Generally, young systems where the soil develops on freshly exposed rock have a high contribution
of solutes from weathering, while in strongly weathered areas the atmospheric input becomes
increasingly important (Chadwick et al., 1999).
Table 8.9 shows the mass balance for the Hubbard brook catchment area, which we already introduced
in Chapter 2, to highlight some of the problems in obtaining weathering rates from mass balances.

Table 8.9. Yearly element balances in the Hubbard Brook area, New Hampshire, USA (Likens and
Bormann, 1995).

Component Chemical element

Ca Mg Na K N S P Cl

Standing stock (kg/ha)

Aboveground biomass 383 36 1.6 155 351 42 34 *
Belowground biomass 101 13 3.8 63 181 17 53 *
Forest floor 372 38 3.6 66 1256 124 78 *
Annual flux (kg/ha.yr)
Bulk precipitation input 2.2 0.6 1.6 0.9 6.5 12.7 0.04 6.2
Gaseous or aerosol input * * * * 14.2 6.1 * ?
Weathering release 21.1 3.5 5.8 7.1 0 0.8 ? *
Streamwater output
Dissolved substances 13.7 3.1 7.2 1.9 3.9 17.6 0.01 4.6
Particulate matter 0.2 0.2 0.2 0.5 0.1 0.1 0.01 *
Vegetation uptake 62.2 9.3 34.8 64.3 79.6a 24.5a 8.9 *
Litter fall 40.7 5.9 0.1 18.3 54.2 5.8 4.0 *
Root litter 3.2 0.5 0.01 2.1 6.2 0.6 1.7 *
Throughfall and stemflow 6.7 2.0 0.3 30.1 9.3 21.0 0.7 4.4
Root exudates 3.5 0.2 34.2 8.0 0.9 1.9 0.2 1.8
Net mineralization 42.4 6.1 0.1 20.1 69.6 5.7 ? ?
Aboveground biomass accretion 5.4 0.4 0.03 4.3 4.8 0.8 0.9 *
Belowground biomass accretion 2.7 0.3 0.12 1.5 4.2 0.4 1.4 *
Forest floor accretion 1.4 0.2 0.02 0.3 7.7 0.8 0.5 *

*Small, unmeasured, aRoot uptake.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Field weathering rates 403

First of all, the weathering rate is very small compared to the amount of elements fixed in the bio-
mass. Also, the rate of annual element cycling by the biomass is much larger than the release by
weathering. Accordingly, small changes in total biomass, which can be difficult to measure, may
yield significant apparent variations in the weathering rate.
The same problem applies to the pool of exchangeable cations, which, under the stress of acid
input, typically is not in steady state. The change in composition of the exchange pool is also diffi-
cult to analyze and the weathering rates obtained by the mass balance approach may include the
depletion of the pool of exchangeable cations. The buffering effects become clear when weathering
rates are calculated for different seasons in which discharge varies (Drever and Clow, 1995). Usually,
the concentrations in surface water are not simply diluted as discharge increases, but show the hydro-
logical response of the watershed in which water qualities vary with depth in the soil profile, and
most often also laterally, over the area (Richards and Kump, 2003).
For Ca2, some of the problems of estimating the weathering rate from mass balance calculations can
be solved using the strontium isotope ratio 87Sr / 86Sr (Åberg et al., 1989; Bailey et al., 1996; Clow et al.,
1997; Bullen et al., 1997; Stewart et al., 2001). 87Sr is a decay product from 87Rb, a process with a half-
life of 48.8 billion year. Due to similar geochemical properties, Sr2 can be considered as an analogue to
Ca2 for studying Ca2 behavior in catchments. The 87Sr / 86Sr ratio in atmospheric deposition differs
from that in bedrock so that the contribution by weathering can be estimated from the difference between
the 87Sr / 86Sr ratio in deposition and in runoff. This approach assumes isotopic equilibration between
soil solution, exchange complex and runoff. The weathering rate is calculated from the mass balance:
qa
RCa  PCa (8.31)
sq

Where RCa and PCa are the Ca2 weathering and deposition rate, respectively, a is the 87Sr / 86Sr ratio
in the deposition, q the 87Sr / 86Sr ratio in the runoff and s the 87Sr / 86Sr ratio in the mineral matrix.
The combination of weathering rates obtained using the Sr-isotope method, with those estimated from
traditional mass balances also permits to separate of the weathering rate from the depletion of the
cation exchange complex (Åberg et al., 1989; Bailey et al., 1996). The method assumes isotopic equi-
librium between the soil solution and the exchanger which may not always be attained (Bullen et al.,
1997). Furthermore, the 87Sr / 86Sr ratio of different minerals in the bedrock will vary as it depends on
their original Rb / Sr ratio. The weathering rate based on the whole rock 87Sr / 86Sr ratio therefore
assumes the congruent release of Sr. However, since different minerals weather at different rates, and
also may dissolve incongruently, caution is warranted (Bullen et al., 1997; Brantley et al., 1998).
Preferably, the strontium isotopic behavior of the various minerals in the rock should be evaluated.

QUESTIONS:
Estimate the contributions of K-feldspar, plagioclase, epidote and apatite to ANReq in Gårdsjön (Figure 8.19).
ANSWER: K-fsp 0.09, plag 0.28, epi 0.135, ap 0.16 keq/ha/yr. Total: 0.67 keq/ha/yr.
Estimate the background acid input (keq/ha/yr) by 600 mm rain in Sweden (Table 2.2; note to convert NH 4
to NO3 )
ANSWER: 0.087 keq/ha/yr
What causes the weathering of the silicates in Gårdsjön?
ANSWER: CO2 produced in the soil
Estimate the acid deposition with 800 mm rain in Beek (NL) in ’78–’83 and in ’97–’98 (Table 2.2)?
ANSWER: 2.2 and 1.2 keq/ha/yr
Estimate ANReq of a 1 m thick sandy soil with 90% quartz and 10% albite, w  0.1, pH  4.5 (Use Table
8.7, neglect inhibition)?
ANSWER: 0.05 keq/ha/yr

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

404 Silicate weathering

8.7 ACID GROUNDWATER

Acid groundwater is found at an increasing number of places throughout the industrialized world
(e.g. Appelo et al., 1982; Eriksson, 1981; Hultberg and Wenblad, 1980; Böttcher et al., 1985; Moss and
Edmunds, 1992; Hansen and Postma, 1995; de Caritat, 1995; Hinderer and Einsele, 1997; Donovan,
1997). The threat of acidification towards the groundwater resource is illustrated in Figure 8.21,
showing a strong decrease of pH in the youngest groundwater of a carbonate-free sandy aquifer in
the Netherlands.
Several processes may cause the acidification of groundwater. First, there is natural acidification
through CO2 production and root respiration in the soil by the overall reaction:

CH 2O  O 2 → H 2O  CO 2 (8.32)

As was pointed out in Chapter 5, the lower pH limit from CO2 production in soil is around 4.6, in the
absence of buffering processes within the aquifer.
A second potential source of acidification is the excessive use of ammonia and manure as fertil-
izers. Nitrification of ammonia is a major acidifying process in the soil:

NH  
4  2O 2 → NO3  2H  H 2 O (8.33)

If, subsequently, the nitrate is removed by denitrification, the proton production due to (8.33) is
balanced again by the HCO 3 production of denitrification (Chapter 9):

5CH 2O  4NO
3 → 2N 2  4HCO
3  CO 2  3H 2 O
(8.34)

However, the ubiquitous presence of nitrate in aquifers shows that this is not the case and nitrification
of ammonia must be considered as an important acidifying process. The net effect of nitrification on
the pH of recharge water is largely determined by the amount of lime which is applied together with
fertilizers to the soil.

0
Depth (m below surface)

50

100

150

200
4 5 6 7 8 9
pH
Figure 8.21. The pH of groundwater versus depth, collected from different wells in sandy aquifers of the
Veluwe area, the Netherlands (Appelo et al., 1982).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Acid groundwater 405

Table 8.10. The composition of 1938 rain and 1980 acid rain and the effect of evapotranspiration in the
Netherlands (Appelo, 1985). Concentrations are in mmol/L . 3  gives values calculated by PHREEQC
using an evapotranspiration factor of 3 and [PCO2]  103.5, 2) Na  K.

1938 rain 3 1980 rain 3

pH 5.4 5.62 4.52 3.54

 2) 2)
Na 0.07 0.21 0.073 0.219
K 0.004 0.012
Mg2 0.03 0.09 0.009 0.027
Ca2 0.04 0.12 0.014 0.042
NH4 0.03 – 0.13 –

Cl 0.09 0.27 0.085 0.255
Alk 0.06 0.002 – –
SO42 0.045 0.135 0.071 0.213
NO 3 – 0.09 0.061 0.573

A third major acidifying process is the oxidation of pyrite (FeS2). Pyrite is found, at least in small
quantities, in most reduced sediments and lowering the groundwater table may cause the oxidation
of pyrite. This process is discussed in detail in Chapter 9. For our present purpose it is summarized
by the overall reaction:

2FeS2  15⁄2O2  5H2O ↔ 2FeOOH  4SO2
4  8H (8.35)
Pyrite oxidation is one of the most strongly acid-producing reactions found in nature.
Among the possible causes for acid groundwater, acid rain is the one that impacts the largest
tracts of land and its detrimental effect on ecosystems of northern Europe and North America is well
documented (Wright and Henriksen, 1978; Likens and Bormann, 1995; Ulrich et al., 1979; Drabløs
and Tollan, 1980; Kirchner, 1992). Acid rain is, to an increasing extent, the cause of groundwater
acidification (de Caritat, 1995; Hansen and Postma, 1995; Hinderer and Einsele, 1997; Kjøller et al.,
2004). The acid rain problem originates from fossil fuel combustion, which produces nitrous oxides
and SO2, that subsequently are oxidized in the atmosphere and precipitate as dilute sulfuric and nitric
acid solutions (Berner and Berner, 1996, see also Chapter 2).
Table 8.10 compares the composition of rainwater before heavy industrialization in 1938 and
acid rain around 1980 in the Netherlands. A decrease in pH of almost one unit is evident and associ-
ated with significant increases of sulfate and the nitrogen compounds. Evapotranspiration further
concentrates the solutes and also changes the pH of the solution. In the 1938 rainwater the pH
increases from 5.4 to 5.6 upon concentration. However, the pH of 1980 rain changes from an initial
value of 4.52 to 3.54 after concentration. Most of this pH drop is due to oxidation of ammonia, pre-
sent at a high concentration in 1980 rainwater, to nitrate (Reaction 8.33).

8.7.1 Buffering processes in aquifers

The acid water that is introduced in soils and aquifer systems may be neutralized by reactions
with soil and aquifer materials. To predict the extent of neutralization, the first assessment
is to consider the relation between the geology and the sensitivity for acidification. Obviously,
rocks which contain carbonate minerals are unlikely to develop acid groundwater because
of the fast dissolution kinetics of carbonates (Chapter 5). In a regional survey of groundwater
in the UK, Edmunds and Kinniburgh (1986) pointed out that groundwaters low in alkalinity
are most vulnerable to acidification. In fact a decreasing alkalinity over time is a
good warning for groundwater acidification in the future (Hultberg and Wenblad, 1980).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

406 Silicate weathering

pH Al (mM) meq/kg
3 4 5 6 7 0 0.1 0.2 0.3 0.4 0.5 0 2 4 6
43

41 GWT

39
Elevation (m)

AI
37

35 Ca
33 Acidification
front Na K Mg
31

29

Figure 8.22. Groundwater acidification in the saturated zone of a sandy aquifer below a coniferous forest in
Denmark. Displayed are groundwater pH, dissolved aluminum and the exchanger composition (Kjøller et al., 2004).

As carbonate-free rocks, Edmunds and Kinniburgh (1986) listed granites, acid igneous rock and
clean quartz sandstones, which contain mainly slowly dissolving silicate minerals, as the most sus-
ceptible to acidification, while basic and ultrabasic rocks, that are dominated by more quickly dis-
solving silicates, were found at the other end of the scale. According to Sverdrup (1990) the presence
of a few percent of pyroxene, hornblende or biotite should suffice to prevent acidification when the
acid load is less than 0.5 keq/ha/yr.
Figure 8.22 shows groundwater acidification in a low reactive sandy aquifer below a coniferous
forest. The groundwater in the upper part of the profile has a pH of 4.5, and suddenly increases at a
depth to about 5.5–6.0 in what appears as a distinct acidification front. The acidification front is sit-
uated at 3 m below the water table. Since the downward pore water velocity is 1–1.4 m/yr, and acid
rain has been infiltrating the soil for decades, buffering must take place within the aquifer. The upper
groundwater with the low pH, also has a high Al3 concentration and this suggest that buffering is
related to the dissolution of Al3 containing minerals. According to Table 8.10 the pH of the acid rain
that enters the soil is close to 3 while the pH in the uppermost groundwater is slightly above 4 (Figure
8.22). The high Al3 concentration indicates that most mineral dissolution and buffering already
takes place in the unsaturated soil. The minerals being dissolved could be primary silicates, clay
minerals, Al(OH)3, or any combination of these.

EXAMPLE 8.5. Acid groundwater formation and gibbsite buffering

In the Veluwe (the Netherlands), acid groundwater develops from acid rain. Groundwater is potentially even
more acid than rain because acidity is concentrated by evapotranspiration and NH 4 is nitrified. The pH
increases again when gibbsite dissolves or when NO 3 is denitrified. This example shows how to calculate
buffering and dissolution reactions when acid rain becomes concentrated groundwater.
1. Rainwater from 1980 (Table 8.10) is concentrated 3 times by evapotranspiration and becomes ground-
water. All NH 
4 is oxidized to NO3 . Calculate the composition.
2. Al(OH)3 (gibbsite) dissolves in the soil. Calculate pH and Al3-concentration.
3. Nitrate is denitrified in peat layers in the aquifer. Calculate pH and Al3-concentration.

ANSWER:
We use PHREEQC, enter the rain as SOLUTION 0 with the sum of NH  
4 and NO3 as NO3 (N(5)) and
obtain charge balance on pH. From this water H2O is removed (‘evapotranspirated’) with keyword REAC-
TION and the result is saved as SOLUTION 1. In the next simulation equilibrium with gibbsite is imposed.
With USER_PRINT we obtain the pH and the Al3 concentration in mg/L for easy comparison with the

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Acid groundwater 407

drinking water limit of 0.2 mg Al/L. In the third simulation, we add CH2O to the previous solution for deni-
trification ( 5⁄4 times the moles of NO
3 according to Reaction (8.34)).

SOLUTION 0 # Rainwater at Deelen (1980)

temp 10
pH 4.5 charge
-units umol/L
Na 73; K 4; Mg 9; Ca 14
Cl 85; S(6) 71; N(5) 191
REACTION 0
H2O -1; 37.005 # remove 37/55.5 = 2/3 of the water
SAVE solution 1
END

USE SOLUTION 1
EQUILIBRIUM_PHASES 1
Gibbsite
USER_PRINT
-start
10 print 'pH = `, -la("H+"), '. mg Al/L = `, tot("Al") * 26981
-end
SAVE solution 2
END

USE SOLUTION 2; USE equilibrium_phases 1

REACTION 2
CH2O 1.25; 191e-6
END

The result is:

Solution number pH mg Al/L mM NO
3

1, evapotranspired rain 3.54 – 0.573

2, soil water 4.00 7.2 0.573
3, groundwater  peat 4.17 2.3 0.0
Clearly, the Al concentration is above the drinking water limit.

QUESTIONS:
The concentration of SO42 in this rainwater has decreased to 21 M in 2000. Calculate the pH and Al con-
centration in soil- and groundwater?
ANSWER: No.1: pH  3.25; No.2: pH  4.04, 4.5 mg Al/L; No.3: pH  4.7, 0.07 mg
Al/L.
Discuss why the SO42 decrease lowers the pH when NH 
4 is oxidized. (Hint: consider the effect of HSO4 )
In a dry year the acid rain was concentrated 5 times. Calculate the concentrations?
ANSWER: No.2: pH  3.9, 15 mg Al/L.

Example 8.5 suggests that 0.27 mmol Al/L was mobilized in the Dutch rain of the 1980’s. In
Figure 8.22 the Al3-concentration in groundwater is around 0.25 mmol/L showing that the same
conditions apply to Denmark. As found elsewhere in Denmark (Figure 8.8) the activity of dissolved
Al3 and the pH are in near equilibrium with microcrystalline gibbsite. Figure 8.22 shows the sedi-
ment exchanger to be filled nearly exclusively with Al3 in the upper part, while in the lower part
Ca2 is the predominant adsorbed cation with subordinate amounts of Mg2.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

408 Silicate weathering

At exactly the same depth where the Al3 concentration in groundwater is decreasing, also
exchangeable Al3 decreases (Figure 8.22). Apparently, the acid water loses Al3 by ion exchange
(Dahmke et al., 1986; Hansen and Postma, 1995; Kjøller et al., 2004):
1
⁄3Al3  1⁄2Ca-X2 ↔ ⁄3Al-X3  1⁄2Ca2
1
(8.36)
Since gibbsite is present, the adsorption of Al3 will cause dissolution of gibbsite:

Al(OH)3 gibbsite  3H ↔ Al3  3H 2O (8.4)

The result is an increase in pH, controlled by the two simultaneous equilibria.

The progression of the acidification front is restricted by the exchange of Al3. We can apply the
retardation equation on the changes in solid and dissolved Al over the front,
R  1  qAl /mAl (8.37)

Most of the change in solid qAl is located in exchangeable Al-X3. Since the exchanger above the
front is filled with Al, and below the front adsorbed Al is negligible, qAl
mCEC and equal to
3 meq/kg  6.17 mmolAl/L. mAl is about 0.25 mmolAl/L and accordingly, the retardation
R  1  6.17 / 0.25  25.7. Since the downward rate of water transport is 1.0–1.4 m/y, the current
depth of the acidification front at 3 m below the water table, corresponds to 77–55 years of acidifi-
cation, which seems a reasonable estimate considering the history of industrialization.

EXAMPLE 8.6. Modeling acidification with PHREEQC

The file below models the downward migration of acid water into an aquifer in equilibrium with gibbsite
and an exchanger. The SIgibbsite of 0.38 corresponds to field observations and the exchanger capacity of
18.5 meq/L is equivalent to measured values of around 3 meq/kg. The pre-acid water quality is unknown (it
has long been flushed), but is estimated by analogy with the Dutch rainwater trend (Table 8.10) to have half
the SO42 and, 0.2 times the NO 2
3 concentration of the acid groundwater. Furthermore, the Ca / Mg
2
ratio
was probably 2.5, the same as in water below the acidity front, and C(4) was set to equilibrium with
[PCO2]  102. See also Kjøller et al. (2004).

DATABASE phreeqc.dat
SOLUTION 1-20
pH 6 charge; temp 8
Na 1.0; K 0.048; Mg 0.07; Ca 0.17
Cl 1.0; S(6) 0.16; N(5) 0.02; C(4) 0.6 CO2(g) -2
EQUILIBRIUM_PHASES 1-20
Gibbsite 0.38 1.0
SAVE solution 1-20
END

EXCHANGE 1-20
X 18.5e-3; -equil 1
END

SOLUTION 0
pH 4.5 charge; temp 8
Na 1.0; K 0.048; Mg 0.068; Ca 0.02; Al 0.2 Gibbsite 0.38
Cl 1.0; S(6) 0.37; N(5) 0.1; C(4) 0.3
END

TRANSPORT
-cells 20; -lengths 0.25
-shifts 330; -diffusion_coefficient 0

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Acid groundwater 409

-punch_frequency 110
-time_step 7.9e6
PRINT
-reset false
SELECTED_OUTPUT
-file Grind.csv
-totals Al Ca Mg Na K
-molalities AlX3 AlOHX2 NaX CaX2 MgX2 KX
-equilibrium_phases Gibbsite
END

pH AI (mmol/L)
4 4.5 5 5.5 6 0.0 0.1 0.2 0.3 0.4
0

2
Depth (m)

Exchanger (meq/L) Gibbsite dissolved (mmol/L)

0 5 10 15 20 0 0.1 0.2 0.3 0.4 0.5
0

1
Al-X3
2
Depth (m)

4
Ca-X2 Mg-X2

5
NaX  KX
6

Figure 8.23. Groundwater acidification modeled with PHREEQC.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

410 Silicate weathering

The vertical pore water velocity at Grindsted is 1.0–1.4 m/yr and with a cell length of 0.25 m the 330 shifts
in the modeled sequence correspond to 83–59 years. The results are presented in Figure 8.23. Acid water
with 0.2 mM Al3 is transported downward. When it reaches the acidification front, Al3 is sorbed on the
exchanger, mostly as Al-X3 and a small part as AlOH-X2. The decrease of Al3 and the ensuing dissolution
of gibbsite raises the pH. The acidification front therefore reflects the depth where the exchanger is filled
with Al3. The retardation is:

mAl-X  mAlOH-X  mgibbsite (0  5.58)  (0  0.15)  (1  0.6)

R  1 3 2
 1  27.6
mAl3 (0  0.2) (8.38)

Note that the change in gibbsite concentration is positive and reduces the total change qAl (the change
mgibbsite is approximate because some gibbsite dissolves beyond the front due to changes in the solution
composition). Again, for a vertical pore water velocity of 1 m/yr it would take 84 years to move the acidifi-
cation front 3 m downward. This simple model is adequate for explaining the field data in Figure 8.22.
Silicate weathering is not included in the model, and has only a slight effect on reducing the acidity
(cf. Question).

QUESTIONS:
Inspect the calculated compositions before and after the acid front and note the differences?
ANSWER: anions have the same concentration, Ca2 has diminished and is replaced
by Al3.
Include the kinetic dissolution of 10% K-feldspar in the model and note the effect on the Al3 concentration?
ANSWER: The concentration of Al3 decreases with 0.01 mM/m in the acid zone
Estimate the time needed to reduce the Al3 concentration from 0.2 mmol/L to below the drinking water
limit (0.0074 mmol Al/L) by the feldspar buffering reaction?
ANSWER: (0.2  0.0074) / 0.01  19.3 m
20 year

In addition to progression of the front under uniform conditions as in Example 8.6, major varia-
tions in the input of Na, for example derived from sea salt deposition or by concentration by
evapotranspiration in dry years, will be accompanied by 1000-fold larger changes in Al3 by the
salinity jump effect (Section 6.4.2). Displacement of Al3 from the exchanger will then cause
gibbsite precipitation and vice versa (Moss and Edmunds, 1992; Hansen and Postma, 1995; de
Caritat, 1995).
The exchange complex forms a storage medium for acidity when it neutralizes the acid input. If
a remediation scheme is planned for deacidification, for example by applying lime to the soil
(Warfvinge, 1988) the processes occurring during acidification are reversed: the addition of base
induces the precipitation of gibbsite, and thereby the Al3 activity in solution is lowered, causing the
desorption of Al3 from the exchange complex and exchange by Ca2. The amount of base applied
must therefore be sufficient to precipitate all adsorbed Al3 as gibbsite before the pH can increase to
near-neutral values. The high retardation of the acidification front through the aquifer implies a very
long lag period before acid water appears in production wells. Late discovery of aquifer acidifica-
tion entails the storage of large amounts of acidity on the sediment, which make it increasingly more
difficult to remediate the problem.

PROBLEMS
8.1. Which of the following minerals do you expect in a thoroughly weathered sediment: olivine, K-
feldspar, quartz, gibbsite and smectite.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 411

8.2. The pH and aluminum speciation (in molality) of a groundwater sample is shown below.

pH Al3 AlOH2 Al(OH)

2
4.53 6.0105 7.7106 6.0107

What is the pH of the water entering the reservoir when gibbsite dissolution has been the only
buffering process? Neglect differences between activity and molal concentrations.
8.3. Derive the relationship between pH and [SO42] at simultaneous equilibrium between gibbsite
and jurbanite (AlOHSO4  Al3  OH  SO42; K  1017.8).
8.4. Listed below are the compositions of ephemeral and perennial springs in the Sierra Nevada
(Garrels and Mackenzie, 1967). Perennial springs are considered to represent a longer resi-
dence time in the reservoir and the difference between perennial and ephemeral spring is con-
sidered to be due to rock weathering. Concentrations are given as mmol/L.

Perennial spring Ephemeral spring Rock weathering

Na 0.259 0.134 0.125

Ca 0.260 0.078 0.182
Mg 0.071 0.029 0.042
K 0.040 0.028 0.012
HCO3 0.895 0.328 0.567
Si 0.410 0.273 0.137
SO4 0.025 0.010 0.015
Cl 0.030 0.014 0.016
pH 6.8 6.2

The minerals presumed to be present are:

Halite NaCl
Gypsum CaSO4
Plagioclase Na0.62Ca0.38Al1.38Si2.62O8
Biotite KMg3AlSi3O10(OH)2
Kaolinite Al2Si2O5(OH)4
Ca-smectite Ca0.17Al2.33Si3.67O10(OH)2
Calcite CaCO3
(a) Perform a mass balance calculation to explain the change in composition in terms of
dissolution and precipitation of the minerals listed below. Hints: Use the change in Al3
concentration (which is O). Assign all Cl and SO42 to respectively halite and gypsum. Do
not include K in the mass balance.
(b) Plot the water composition of the perennial spring into the stability diagram Figure 8.11
and evaluate whether the chosen weathering products are reasonable.
(c) Are the calculated mineral transfers kinetically consistent?
8.5. Aqua de Penha is a mineral water from a granitic area in Portugal with the composition
(mmol/L, except pH) pH 5.77; Na 0.343; K 0.015; Mg 0.036; Ca 0.06; Cl 0.24; Alkalinity
0.292; SO42 0.007; Si 0.293. Calculate how much Na comes from rain and how much from
weathering of albite? (Assume that the rain has the Na/Cl ratio of seawater). Perform an
inverse model calculation balancing Aqua da Penha water from seawater and distilled water,
with the phases albite; biotite; K-feldspar; anorthite; gibbsite; kaolinite and CO2(g).
Estimate the composition of the plagioclase (NaxCa1x)(Al1xSi3x)O8 in the rock.
8.5. In Example 8.5, include equilibrium with atmospheric oxygen in evapotranspired rain- and soil
water, and calculate the reaction of peat sequentially, first with O2, then with NO3.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

412 Silicate weathering

REFERENCES
Åberg, G., Jacks, G. and Hamilton, P.J., 1989. Weathering rates and 87Sr / 86Sr ratios; an isotopic approach.
J. Hydrol. 109, 65–78.
Appelo, C.A.J., 1985. CAC, computer aided chemistry, or the evaluation of groundwater quality with a geo-
chemical computer model (in Dutch). H2O 26, 557–562.
Appelo, C.A.J., Krajenbrink, G.J.W., Van Ree, C.C.D.F. and Vasak, L., 1982. Controls on groundwater quality
in the NW Veluwe catchment (in Dutch). Soil Protection Series 11, Staatsuitgeverij, Den Haag, 140 pp.
Bailey, S.W., Hornbeck, J.W., Driscoll, C.T. and Gaudette H.E., 1996. Calcium inputs and transport in a base-
poor forest ecosystem as interpreted by Sr isotopes. Water Resour. Res. 32, 707–719.
Bain, D.C., Mellor, A., Robertson, M.S.E. and Buckland, S.T., 1991. Variations in weathering processes and
rates with time in a chronosequence of soils from Glen Feshie, Scotland. In O. Selinus (ed.), Proc. 2nd Int.
Symp. Env. Geochemistry, Uppsala.
Berg, A. and Banwart, S.A., 2000. Carbon dioxide mediated dissolution of Ca-feldspar: implications for silicate
weathering. Chem. Geol. 163, 25–42.
Berner, R.A., 1971. Principles of chemical sedimentology. McGraw-Hill New York, 240 pp.
Berner, E.K. and Berner, R.A., 1996. Global Environment. Prentice-Hall, Englewood Cliffs, 376 pp.
Blatt, H., Middleton, G. and Murray, R., 1980. Origin of sedimentary rocks, 2nd ed. Prentice-Hall, Englewood
Cliffs, 782 pp.
Blum, A.E. and Stillings, L.L., 1995. Feldspar dissolution kinetics. In A.F. White and S.L. Brantley (eds),
Chemical weathering rates of silicate minerals. Rev. Mineral. 31, 291–351.
Böttcher J., Strebel, O. and Duynisveld, H.M., 1985. Vertikale Stoffkonzentrationsprofile im Grundwasser eines
Lockergesteins-Aquifers und deren Interpretation (Beispiel Fuhrberger Feld). Z. dt. geol. Ges. 136,
543–552.
Brantley, S.L., Chesley, J.T. and Stillings, L.L., 1998. Isotopic ratios and release rates of strontium measured
from weathering feldspars. Geochim. Cosmochim. Acta 62, 1493–1500.
Bullen, T.D., White, A.F., Blum, A.E., Harder, J.W. and Schulz, M.S., 1997. Chemical weathering of a soil
chronosequence on granitoid alluvium: II. Mineralogical and isotopic constraints on the behavior of stron-
tium. Geochim. Cosmochim. Acta 61, 291–306.
Chadwick, O.A., Derry, L.A., Vitousek, P.M., Huebert, B.J. and Hedin, L.O., 1999. Changing sources of nutri-
ents during four million years of ecosystem development. Nature 397, 491–496.
Chou, L. and Wollast, R., 1985. Steady-state kinetics and dissolution mechanisms of albite. Am. J. Sci. 285,
963–993.
Clow, D.W., Mast, M.A., Bullen, T.D. and Turk, J.T., 1997. Strontium 87 – strontium 86 as a tracer for mineral
weathering reactions and calcium sources in an alpine/subalpine watershed, Loch Vale, Colorado. Water
Resour. Res. 33, 1335–1351.
de Caritat, P., 1995. Intensifying groundwater acidification at Birkenes, southern Norway. J. Hydrol. 170,
47–62.
Dahmke, A., Matthess, G., Pekdeger, A., Schenk, D. and Schulz, H.D., 1986. Near-surface geochemical
processes in Quaternary sediments. J. Geol. Soc. 143, 667–672.
Donovan, J.J., Frysinger, K.W. and Maher, T.P., 1997. Geochemical response of acid groundwater to neutraliza-
tion by alkaline recharge. Aq. Geochem. 2, 227–253.
Dove, P. and Rimstidt, J.D., 1993. Silica-water interactions. Rev. Mineral. 29, 259–308.
Drabløs, D. and Tollan, A. (eds), 1980. Ecological impact of acid precipitation. Proc. Int. Conf. Sandefjord,
SNSF Project, Oslo, Norway.
Drever, J.I. and Clow, D.W., 1995. Weathering rates in catchments. In A.F. White and S.L. Brantley (eds),
Chemical weathering rates of silicate minerals. Rev. Mineral. 31, 463–483.
Driscoll, C.T., 1980. Aqueous speciation of aluminum in the Adirondack region of New York State, USA. In D.
Drabløs and A. Tollan, op. cit., 214–215.
Edmunds, W.M. and Kinniburgh, D.G., 1986. The susceptibility of UK groundwaters to acidic deposition.
J. Geol. Soc. 143, 707–720.
Eriksson, E., 1981. Aluminum in groundwater possible solution equilibria. Nordic Hydrol. 12, 43–50.
Garrels, R.M. and Mackenzie, F.T., 1967. Origin of the chemical compositions of some springs and lakes. In
W. Stumm (ed.), Equilibrium concepts in natural water systems. Adv. Chem. Series 67, 222–242.
Garrels, R.M. and Mackenzie, F.T., 1971. Evolution of Sedimentary Rocks. Norton, New York, 397 pp.
Goldich, S.S., 1938. A study in rock-weathering. J. Geol. 46, 17–58.
Hansen, B.K. and Postma, D., 1995. Acidification, buffering, salt effects in the unsaturated zone of a sandy
aquifer, Klosterhede, Denmark, Water Resour. Res. 31, 2795–2809.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 413

Helgeson, H.C., Garrels, R.M. and MacKenzie, F.T., 1969. Evaluation of irreversible reactions in geochemical
processes involving minerals and aqueous solutions. II. Applications. Geochim. Cosmochim. Acta 33,
455–481.
Hem, J.D., 1985. Study and Interpretation of the chemical characteristics of natural water, 3rd ed. U.S. Geol.
Surv. Water Supply Paper 2254, 264 pp.
Hinderer, M. and Einsele, G., 1997. Groundwater acidification in Triassic sandstones: prediction with MAGIC
modeling. Geol. Rundsch. 86, 372–388.
Hodson, M.E., Langan, S.J. and Wilson, M.J., 1997. A critical evaluation of the use of the PROFILE model in
calculating mineral weathering rates. Water Air Soil Poll. 98, 79–104.
Hultberg, H. and Wenblad, A., 1980. Acid groundwater in southwestern Sweden. In D. Drabløs and A. Tollan,
op. cit., 220–221.
Kirchner, J.W., 1992. Heterogeneous geochemistry of catchment acidification. Geochim. Cosmochim. Acta 56,
2311–2327.
Kjøller, C., 2001. Nickel mobilization in response to groundwater acidification, Ph.D. thesis, Lyngby, 140 pp.
Kjøller, C., Postma, D. and Larsen, F., 2004. Groundwater acidification and the mobilization of trace metals in
a sandy aquifer. Env. Sci. Technol., 38, 2829–2835.
Langan, S., Hodson, M., Bain, D., Hornung, M., Reynolds, B., Hall, J. and Johnston, L., 2001. The role of min-
eral weathering rate determinations in generating uncertainties in the calculation of critical loads of acidity
and their exceedance. Water Soil Air Poll. Focus 1, 299–312.
Lasaga, A.C., 1984. Chemical kinetics of water–rock interactions. J. Geophys. Res. 89, 4009–4025.
Likens, G.E. and Bormann, F.H., 1995. Biogeochemistry of a forested ecosystem. 2nd ed. Springer, New York,
159 pp.
Matthess, G., Petersen, A., Schenk, D. and Dahmke, A., 1992. Field studies on the kinetics of silicate miner-
als/water interaction. In G. Matthess, F.H. Frimmel, P. Hirsch, H.D. Schulz and E. Usdowski (eds), Progress
in hydrogeochemistry, 298–307, Springer, Berlin.
May, H.M., Kinniburgh, D.G., Helmke, P.A. and Jackson, M.L., 1986. Aqueous dissolution, solubilities and
thermodynamic stabilities of common aluminosilicate clay minerals: kaolinite and smectites. Geochim.
Cosmochim. Acta 50, 1667–1677.
Miretzky, P., Conzonno, V. and Cirelli, A.F., 2001. Geochemical processes controlling silica concentrations in
groundwaters of the Salado River drainage basin, Argentina. J. Geochem. Expl. 73, 155–166.
Moss, P.D. and Edmunds, W.M., 1992. Processes controlling acid attenuation in the unsaturated zone of a
Triassic sandstone aquifer (U.K.) in the absence of carbonate minerals. Appl. Geochem. 7, 573–583.
Ohse, W., Matthess, G. and Pekdeger, A., 1983. Gleichgewichts- und Ungleichgewichtsbeziehungen zwischen
Porenwassern und Sedimentgesteinen im Verwitterungsbereich. Z. dt. geol. Ges. 134, 345–361.
Ohse, W., Matthess, G., Pekdeger, A. and Schulz, H.D., 1984. Interaction water–silicate minerals in the unsatu-
rated zone controlled by thermodynamic disequilibria. IASH Pub. 150, 31–40.
Olsson, M. and Melkerud, P.A., 1991. Weathering rates in mafic soil mineral material. In K. Rosén (ed.),
Chemical weathering under field conditions. Rep. Forest Ec. Forest Soil 63, 63–78, Swedish Univ. Agri. Sci.
Plummer, L.N., Prestemon, E.C. and Parkhurst, D.L., 1991. An interactive code (NETHPATH) for modeling net
geochemical reactions along a flow path. U.S. Geol. Surv. Water Resour. Inv., 91-4078, 100 pp.
Rosén, K. (ed.), 1991. Chemical weathering under field conditions. Rep. Forest Ec. Forest Soil 63, Swedish
Univ. Agri. Sci., 185 pp.
Richards, P.L. and Kump, L.R., 2003. Soil pore-water distributions and the temperature feedback of weathering
in soils. Geochim. Cosmochim. Acta 67, 3803–3815.
Rimstidt, J.D., 1997. Quartz solubility at low temperatures. Geochim. Cosmochim. Acta 61, 2553–2558.
Ryan, J.N. and Gschwend, P.M., 1992. Effect of iron diagenesis on the transport of colloidal clay in an uncon-
fined aquifer. Geochim. Cosmochim. Acta. 56, 1507–1521.
Salomons, W. and Mook, W.G., 1987. Natural tracers for sediment transport studies. Cont. Shelf Res. 7,
1333–1343.
Savin, S.M. and Hsieh, J.C.C., 1998. The hydrogen and oxygen isotope geochemistry of pedogenic clay miner-
als: principles and theoretical background. Geoderma 82, 227–253.
Schott, J. and Berner, R.A., 1985. Dissolution mechanisms of pyroxenes and olivines during weathering. In
J.I. Drever (ed.), The chemistry of weathering, 35–53, D. Reidel, Dordrecht.
Schulz, M.S. and White, A.F., 1999. Chemical weathering in a tropical watershed, Luquillo Mountains, Puerto
Rico III: Quartz dissolution rates. Geochim. Cosmochim. Acta 63, 337–350.
Schroeder, P.A., Melear, N.D., Bierman, P., Kashgarian, M. and Caffee, M.W., 2001. Apparent gibbsite growth
ages for regolith in the Georgia piedmont. Geochim. Cosmochim. Acta 65, 381–386.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

414 Silicate weathering

Stefansson, A. and Arnorsson, S., 2000. Feldspar saturation state in natural waters. Geochim. Cosmochim. Acta
64, 2567–2584.
Stewart, B.W., Capo, R.C. and Chadwick, O.A., 2001. Effects of rainfall on weathering rate, base cation prove-
nance, and Sr isotope composition of Hawaiian soils. Geochim. Cosmochim. Acta 65, 1087–1099.
Stumm, W. and Wollast, R., 1990. Coordination chemistry of weathering. Rev. Geophys. 28, 53–69.
Sverdrup, H.U., 1990. The kinetics of base cation release due to chemical weathering. Lund Univ. Press, 246 pp.
Sverdrup, H.U. and Warfvinge, P., 1988. Weathering of primary silicate minerals in the natural soil environment
in relation to a chemical weathering model. Water Air Soil Poll. 38, 387–408.
Sverdrup, H.U. and Warfvinge, P., 1993. Calculating field weathering rates using a mechanistic geochemical
model PROFILE. Appl. Geochem. 8, 273–283.
Sverdrup, H.U. and Warfvinge, P., 1995. Estimating field weathering rates using laboratory kinetics. In
A.F. White and S.L. Brantley (eds), Chemical weathering rates of silicate minerals. Rev. Mineral. 31,
485–541.
Tardy, Y., 1971. Characterization of the principal weathering types by the geochemistry of waters from some
European and African crystalline massifs. Chem. Geol. 7, 253–271.
Ulrich, B., Mayer, R. and Khanna, P.K., 1979. Deposition von Luftverunreinigungen und ihre Auswirkungen in
Waldökosystemen im Solling. Sauerländer, Frankfurt a.M., 291 pp.
Warfvinge, P., 1988. Modeling acidification mitigation in watersheds. Ph D. Thesis, Lund, 180 pp.
White, A.F. and Brantley, S.L. (eds), 1995. Chemical weathering rates of silicate minerals. Rev. Mineral. 31,
583 pp.
White, A.F., Blum, A.E., Schulz, M.S., Bullen, T.D., Harder, J.W. and Peterson, M.L., 1996. Chemical weather-
ing rates of a soil chronosequence on granitic alluvium: 1. Quantification of mineralogical and surface area
changes and calculation of primary silicate reaction rates. Geochim. Cosmochim. Acta 60, 2533–2550.
White, A.F., Bullen, T.D., Schulz, M.S., Blum, A.E., Huntington, T.G. and Peters, N.E., 2001. Differential rates
of feldspar weathering in granitic regoliths. Geochim. Cosmochim. Acta 65, 847–869.
Williams, L.A., Parks, G.A. and Crerar, D.A., 1985. Silica Diagenesis, I. Solubility Controls. J. Sed. Petrol. 55,
301–311.
Wright, R.F. and Henriksen, A., 1978. Chemistry of small Norwegian lakes, with special reference to acid pre-
cipitation. Limnol. Oceanogr. 23, 487–498.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

9

Redox Processes

Reduction and oxidation processes exert an important control on the natural concentrations of O2,
Fe2, SO24 , H2S, CH4, etc. in groundwater. They also determine the fate of pollutants like nitrate
leaching from agricultural fields, contaminants leaching from landfill sites, industrial spills, or
heavy metals in acid mine drainage. Redox reactions occur through electron transfer from one atom
to another and the order in which they proceed can be predicted from standard equilibrium thermo-
dynamics. However, the electron transfer is often very slow and may only proceed at significant rates
when mediated by bacterial catalysis. An example is the reduction of sulfate by organic matter which
occurs both in aquifers and in marine sediments. The reaction is immeasurably slow abiotically, but
microbes like Desulfovibrio sp. produce enzymes that catalyze the process and the reaction proceeds
rapidly in natural environments.
Redox processes in groundwater typically occur through the addition of an oxidant, like O2 or
NO 3 to an aquifer containing a reductant. However, the addition of a reductant, such as dissolved
organic matter (DOC) that leaches from soils or landfills can also be important. In the following, we
first treat some basic redox theory, and subsequently use the principles to discuss the redox
processes in aquifers.

9.1 BASIC THEORY

As an example of a redox process, consider the reaction between Fe2 and Mn4 as it would take
place in acid solution:

2Fe 2  MnO 2  4H + ↔ 2Fe +  Mn 2  2H 2O (9.1)

In this reaction two electrons are transferred from Fe(2) to reduce Mn(4) in MnO2. Ferrous
iron acts as the reductant and reduces Mn(4), while MnO2 can be called the oxidant
that oxidizes Fe2. An alternative terminology is to call Fe2 the electron donor, and Mn(4) the
electron acceptor. Reaction (9.1) also demonstrates that redox reactions may have a significant
pH effect.
Electrons exchange between atoms and do not exist in a “free” state in solution
(Figure 9.1). Accordingly, electrons do not appear in a balanced redox reaction such as (9.1).
However, when deriving a complete redox reaction it is convenient to consider first the
half-reactions:

Fe 2 ↔ Fe3  e (9.2)
and

Mn 2  2H 2O ↔ MnO 2  4H  2e (9.3)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

416 Redox processes

WAL
TS E
LECT
EXCH RON
ANG
E

NO REASONABLE ELECTRON REFUSED WE

BUY
, SEL
L ON
TRA
NS

OXIDIZE

Figure 9.1. Electrons can only be exchanged (Nordstrom and Munoz, 1994).

Subtracting Reaction (9.2) twice from (9.3) cancels the electrons and produces Reaction (9.1). With
the guidelines listed in Table 9.1, any redox reaction can be balanced.

Table 9.1. Guide to balancing redox equations.

1. For each redox couple, write the oxidized and reduced species in an equation and balance the
amount of element.
2. Balance the number of oxygen atoms by adding H2O.
3. Balance the number of hydrogen atoms by adding H.
4. Balance electroneutrality by adding electrons.
5. Subtract the two half-reactions, canceling electrons to obtain the complete redox reaction.

For example, in Reaction (9.1) we could replace pyrolusite (MnO2) by birnessite (MnO1.9).
Birnessite contains small and variable amounts of Mn2 or Mn3 instead of Mn4 (the framework of
oxygens remains intact with all the oxygens having valence O(2)):
Step from Table 9.1
1. MnO1.9  Mn 2
2. MnO1.9  Mn 2  1.9H 2O
3. MnO1.9  3.8H  Mn 2  1.9H 2O
4. MnO1.9  3.8H  1.8e  Mn 2  1.9H 2O
5. MnO1.9  3.8H  1.8Fe 2 ↔ Mn 2  1.9H 2O  1.8Fe3 (9.1b)

QUESTION:
Balance the following redox reactions:
NO3  Fe
2
↔ N2  FeOOH
SO2
4  C ↔ CO2  HS
C2HCl3  Fe ↔ C2H3Cl  Cl  Fe2 (trichloroethene reacts with iron to vinylchloride)
ANSWER: NO 3  5Fe
2
 7H2O ↔ 1⁄2N2  5FeOOH  9H; SO24  2C  H

↔
  
2CO2  HS ; C2HCl3  2Fe  2H ↔ C2H3Cl  2Cl  2Fe 2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Basic theory 417

Reactions (9.1) to (9.3) can be written in their general form as

bBred  cCox → dDox  gGred (9.4)

and half-reactions

bBred → dDox  ne (9.5)

gGred → cCox  ne (9.6)

In terms of Gibbs free energy (Section 4.3.1) we may write for Reaction (9.4)

[Dox ]d [Gred ]g
Gr  Gr0  RT ln
[ Bred ]b [Cox ]c

The Gibbs free energy of a reaction can be related to the voltage developed by a redox reaction in an
electrochemical cell by the relation

G  nFE (9.7)
E is the potential (emf) in Volts, F the Faraday’s constant (96.42 kJ/Volt gram equivalent) and n the num-
ber of electrons transferred in the reaction. Substitution of (9.7) into (4.24) produces the Nernst equation:

RT [ D ]d [G ]g
E  Eo  ln ox b red c (9.8)
nF [ Bred ] [Cox ]

Here E0 is the standard potential (Volt) where all substances are present at unit activity at 25°C and
1 atm, similar to Gr0. As before, R is the gas constant (8.314103 kJ/deg/mol), and T the absolute
temperature. For oxidation of H2 the half-reaction is:

H 2 ↔ 2H  2e (9.9)

By definition Reaction (9.9) has Gr0  0, at 25°C and 1 atm, and according to (9.7) E0  0 Volt.
Substituting (9.9) for half-reaction (9.6) in (9.8) gives:

RT [ Dox ]d PH
E  E0  ln 2
(9.10)
nF [ Bred ]b [H ] 2

The setup for such a redox cell is illustrated in Figure 9.2. The left hand part is the standard hydro-
gen electrode, consisting of a Pt-electrode over which H2 gas is bubbled in a solution of pH  0, so
that standard state conditions are fulfilled. In the right hand part of the cell, an inert
Pt electrode is placed in a solution containing Fe2 and Fe3, corresponding to half-reaction (9.2).
The two electrodes are connected to a voltmeter and the electrical circuit is closed by a salt bridge.
Under these conditions, both PH2 and [H] are equal to one in (9.10). When Fe2 and Fe3 are pres-
ent in solution at unit activity, the voltmeter will register the E0 of (9.2) but at other activity ratios of
Fe3 and Fe2, different E values are measured.
Since both PH2 and [H] are always equal to one in this setup, they are usually omitted from
Equation (9.10) and instead indicated by adding the postscript h from hydrogen to E:
RT [ Dox ]d
Eh  E 0  ln (9.11)
nF [ Bred ]b

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

418 Redox processes

H2-electrode Volt Pt-electrode

H2
Salt bridge

Fe2

[H]1 Fe3

Figure 9.2. A schematic drawing of a redox cell.

The Eh and E0 of a half-reaction are thus defined as potentials relative to the standard state H2 / H
reaction. The value of the standard potential E0 of a half-reaction indicates the tendency to release or
to accept electrons. Standard potentials for a few reactions are listed in Table 9.2 and such a list is
useful to identify possible reactions. There is some confusion concerning sign conventions of redox
reactions. In many texts half-reactions are written as reduction reactions instead of oxidation reac-
tions which are used here. The advantage of writing oxidation reactions is that the oxidized species
appears in the numerator of the mass action equation, similar to the Nernst equation. When reduc-
tion reactions are written instead, a minus sign is added in Equation (9.7). With both conventions
reducing agents have a more negative E0 and oxidizing agents a more positive E0.
If we apply the Nernst equation to Reaction (9.2), using E0 from Table 9.2, we obtain:

RT [Fe3 ]
Eh  E 0  ln
1F [Fe 2 ]
(8.314103 )( 298.15)( 2.303) [Fe3 ] (9.12)
 0.77  log
96.42 [Fe 2 ]
[Fe ]3
 0.77  0.059 log
[Fe 2 ]

Table 9.2. Standard potentials for a few reactions at 25°C, 1 atm.

Reaction E0, Volt

Fe(s) ↔ Fe2  2e 0.44

Cr2 ↔ Cr3  e 0.41
H2 ↔ 2H  2e 0.00
Cu ↔ Cu2  e 0.16
 
S2  4H2O ↔ SO2 4  8H  8e 0.16
As(s)  3H2O ↔ H3AsO3(aq)  3H  3e 0.25
Cu(s) ↔ Cu  e 0.52
H3AsO3(aq)  H2O ↔ H3AsO4(aq)  2H  2e 0.56
Fe2 ↔ Fe3  e 0.77
Fe2  3H2O ↔ Fe(OH)3  3H  e 0.98
2H2O ↔ O2(g)  4H  4e 1.23
Mn2  2H2O ↔ MnO2  4H  2e 1.23

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Basic theory 419

where the factor 2.303 converts natural to base ten logarithms. Similarly for Reaction (9.3):
RT [H ]4
Eh  E 0  ln
2 F [Mn 2 ]
(8.314103 )( 298.15)( 2.303) [H ]4 (9.13)
 1.23  log
2  96.42 [Mn 2 ]

[H ] 4
 1.23  0.03 log
[Mn 2 ]
Equation (9.13) contains neither MnO2 nor H2O, since they both have unit activity. At equilibrium
between the two half-reactions, the Eh for both reactions should be the same. In other words, for a
given Eh the distribution of all redox equilibria is fixed (Example 9.1).

EXAMPLE 9.1. Calculation of redox speciation with the Nernst equation

A water sample contains [Fe2]  104.95 and [Fe3]  102.29 with a pH  3.5 at 25°C. What would be
[Mn2] if this water sample were in equilibrium with sediment containing MnO2?

ANSWER:
First rewrite Equation (9.12) as
Eh  0.77  0.059(log[Fe3 ]  log[Fe 2 ])
Substitute iron activities

Eh  0.77  0.059(2.29  (4.95))  0.927 Volt.

Next rewrite Equation (9.13) as

Eh  1.23  0.03(4 pH  log[Mn 2 ])
log [Mn 2 ]  (1 / 0.03)(1.23  Eh  0.12 pH)

substitute the given pH and the Eh calculated from the Fe3 / Fe2 couple
log [Mn 2 ]  33.33(1.23  0.927  0.42)  3.90
→ [Mn 2 ]  103.90

Thus, for any lower [Fe3] / [Fe2] ratio, or lower pH value, the [Mn2] will increase and vice versa. The
example illustrates that once the Eh is evaluated from one redox couple, the distribution of all other redox
couples is fixed.

In order to obtain the E 0 for Reaction (9.1), we simply subtract the E0 of (9.13) from the E0 of (9.12)
(without multiplying the E0 with the number of electrons transferred since E0 already is expressed
as potential per electron according to Equation (9.8)).
E 0  0.77  1.23  0.46 Volt.
The negative voltage indicates that the reaction should proceed spontaneously to the right when all
activities are equal to one. For redox reactions like (9.1), we return to the general form of the Nernst
Equation (9.8):
RT [Fe3 ]2 [Mn 2 ]
E  E0  ln
2F [Fe 2 ]2 [H ]4 (9.14)
[Fe3 ]2 [Mn 2 ]
 0.46  0.03 log
[Fe 2 ]2 [H ]4

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

420 Redox processes

In this fashion, the Nernst equation can be used to express the distribution of species in any redox
reaction at equilibrium.
As indicated by Equations (9.7) and (9.8), redox potentials can be related to Gibbs free energies.
The practical significance of this relation is that it allows us to calculate the standard potential of any
redox reaction directly from thermodynamic tables (see Section 4.3.1) since we may rewrite
Equation (9.7) for standard conditions as:

Gr0  nFE 0 (9.15)

EXAMPLE 9.2. Calculate E 0 from Gr0

For Reaction (9.2)

Fe 2 ↔ Fe3  e
G 0f Fe2  78.9 kJ/mol;
G 0f Fe3  4.7 kJ/mol;

G 0f e  0.0 kJ/mol (by definition).

Gr0  G 0f Fee3  G 0f e  G 0f Fe2  4.7  0  (78.9)  74.2 kJ/mol

Using (9.15) E 0  Gr0 / nF  74.2 / 96.45  0.77 Volt

QUESTIONS:
Calculate the concentration of Fe2 in equilibrium with solid iron at Eh  0 and 0.3 V?
ANSWER: Using Table 9.2, 81014 and 5.6104 mol Fe2/L for Eh  0 and 0.3 V
(huge concentration!)
And of Cu2 in contact with solid copper at Eh  0?
ANSWER: E 0  0.52  0.16 V; Cu2  8.91024 mol/L (very low concentration)
Which metal is more stable in water, Cu(s) or Fe(s)?
ANSWER: Cu
Calculate Gf0 of Cu from Table 9.2?
ANSWER: 50.15 kJ/mol

9.1.1 The significance of redox measurements

Theoretically, the Eh determines the distribution of all redox equilibria in a similar way as the pH
expresses the distribution of all acid-base equilibria. In contrast to pH, unfortunately Eh cannot be
measured unambiguously in most natural waters.
Eh-measurements are performed using an inert Pt-electrode against a standard electrode of a
known potential (see Grenthe et al., 1992 and Christensen et al., 2000 for detailed procedures). What
we really want to know is the potential relative to the standard hydrogen electrode. However, since
the latter is rather impractical to carry around in the field, a reference electrode of known potential
is used and measured potentials are corrected accordingly

Eh  Emeas.  Eref (9.16)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Basic theory 421

0.780 Symbol Redox couple

Fe3/Fe2
0.630
O2 (aq) / H2O
HS/SO2
4
0.480
HS/Srhombic
Field-measured Eh (V)

NO 
2 /NO3
0.330
NH 
4 /NO3
NH4 /NO

2
0.180 CH4 (aq)/HCO
3
NH
4 /N2 (aq)
0.030 Fe2/Fe(OH)3(s)

0.120

0.270

0.420
0.500 0.325 0.150 0.025 0.200 0.375 0.550 0.725 0.900 1.075
Eh (V) computed from redox couples

Figure 9.3. Comparison of groundwater field Eh measurements with potentials calculated for individual
redox species (Lindberg and Runnels, 1984, Science, 225, 925–927 Copyright 1985 by the AAAS).

For example, the potential of a calomel reference electrode (KCl(sat), Hg2Cl2(s)|Hg(l)) is Eref  244.4 mV
at 25°C.
Although waters from oxidized environments generally yield higher Eh values than those from
reduced environments, it has proven very difficult to obtain a meaningful quantitative interpretation
in the sense of the Nernst equation. This is illustrated in Figure 9.3 were Eh values measured in the
field are compared with those calculated with the Nernst equation from the analytic data for several
half-reactions. The results show disturbingly large variations between the two sets of data. For exam-
ple, for the important reaction which relates the O2 content to the Eh,

2H 2O ↔ O 2(g)  4H  4e (9.17)

the Pt-electrode is apparently unaffected by the O2 concentration. The results for the other half-
reactions are not much better.
There are two reasons for the large discrepancies: lack of equilibrium between different redox
couples in the same water sample (Lindberg and Runnels, 1984) and analytical difficulties in mea-
suring with the Pt-electrode (Stumm and Morgan, 1996). The latter include lack of electroactivity at
the Pt surface, like for O2, mixing potentials, and poisoning of the electrode. An example of poison-
ing is the precipitation of FeOOH on the Pt-electrode which occurs when the electrode is immersed
in an anoxic, Fe2-rich, sample because of O2 adsorbed on the electrode surface (Doyle, 1968).
Others have used graphite electrodes instead but also these have their limitations (Walton-Day et al.,
1990; Grenthe et al., 1992).
Probably, Eh measurements are applicable in acid mine waters, where high concentrations of
Fe2 and Fe3 appear to control the electrode response (Nordstrom et al., 1979). Encouraging results
of Eh measurements in the iron system at higher pH have also been obtained (Macalady et al., 1990;

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

422 Redox processes

Grenthe et al., 1992). Still, the overall conclusion remains that Eh-measurements only should be inter-
preted quantitatively when there is a rigorous control on what is really being measured.

9.1.2 Redox reactions and the pe concept

An alternative theoretical treatment of redox reactions simplifies the algebra considerably. In this
approach the law of mass action is used in redox half-reactions. For example, for Reaction (9.2):

[Fe3 ][e ]
K   1013.05 (9.18)
[Fe 2 ]

where K is the equilibrium constant. In contrast to the Nernst equation, the electron activity appears
explicitly in the activity product. The electron activity should not be interpreted in terms of a con-
centration of electrons, since electrons are only exchanged, but rather as the tendency to release or
accept electrons. In analogy to pH, one may define the parameter pe:
pe  log[e ] (9.19)

Just as for Eh, high positive values of pe indicate oxidizing conditions and low negative values
reducing conditions. Rewriting (9.18) in logarithmic form yields the equation:
log K  log[Fe3 ]  pe  log[Fe 2 ]  13.05 (9.20)
In a similar fashion we may write for Reaction (9.3)

log K  4pH  2pe  log[Mn 2 ]  41.52 (9.21)

Equations (9.20) and (9.21) can be compared with the corresponding Nernst Equations (9.12) and
(9.13). The calculation of redox speciation, using the pe concept is demonstrated in Example 9.3. It
is a direct analogue to Example 9.1 where the Nernst approach was used.

EXAMPLE 9.3. Calculation of redox speciation using the pe concept

A water sample contains [Fe2]  104.95 and [Fe3]  102.29 with a pH  3.5 at 25°C. What would be
[Mn2] if this water sample were in equilibrium with sediment containing MnO2?

ANSWER:
First pe is calculated from Equation (9.20) which is then substituted into Equation (9.21). Rewriting (9.20)
yields:

pe   log K  log[Fe3 ]  log[Fe 2 ]

Substituting values gives:

pe  (13.05)  (2.29)  (4.95)  15.71

Rewriting (9.21) yields:

log[Mn 2 ]   log K  4 pH  2pe

Substituting log K, pH and the pe calculated above yields

log[Mn 2 ]  (41.52)  4  3.5  2  15.71  3.90

Which is the same result as obtained in Example 9.1.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 423

The values of K can be calculated using thermodynamic tables just as any other mass action constant
(Example 9.4).

EXAMPLE 9.4. Calculation of K from thermodynamic data

This example is an analogue to Example 9.2 where the Nernst equation was used. The mass action constant
is related to Gr0 by Equation (4.26)

Gr0  RT ln K

At 25°C this is equal to

Gr0  5.701 log K

Following Example 9.2 the Gr0 of the Fe2 / Fe3 reaction is:

Gr0  G 0f Fe3  G 0f e  G 0f Fe2  4.7  0  (78.9)  74.2 kJ/mol

Note again that Gf0e  0.0 kJ/mol by definition. Substitution results in

log K  74.2 / 5.701  13.02

There must of course be a simple relationship between Eh and pe, which can be found by combining
Equations (4.26), (9.7) and (9.8) written for a half-reaction. It yields:
2.303 RT
Eh  pe (9.22)
F
At 25°C this is equal to
Eh  0.059 pe (Volt) (9.23)

Both the Nernst equation and the pe-concept are commonly used in the literature. The advantage of
the pe concept is that the algebra of redox reactions becomes identical to other mass action expres-
sions, allowing the same algorithm to be used in computer programs. The disadvantage of the use of
pe is, that it is a non-measurable quantity. For the sake of keeping calculations as simple as possible,
we will in this book mainly use the pe concept.

QUESTION:
What are the pe values corresponding to Eh  0.59 and 0.118 Volt, and what is the activity of the
electrons?
ANSWER: pe  10 and 2, [e]  1010 and 102

9.2 REDOX DIAGRAMS

The number of dissolved species and mineral phases for which the redox conditions affect the
stability can be overwhelming. Furthermore many redox reactions, like for Example (9.3), are
strongly influenced by pH. In order to retain an overview in such complicated systems, redox
diagrams are useful to display the stability of both dissolved species and minerals as a function

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

424 Redox processes

of pe (or Eh) and pH. The prime force of the diagrams is that possible stable phases and
species can be identified at a glance. Garrels and Christ (1965) and Brookins (1988) have
presented redox diagrams for many systems serving that purpose, although more detailed
calculations are often needed to confirm critical aspects. The construction of redox diagrams,
and some of their limitations are discussed below. Detailed guidelines for construction of
redox diagrams can be found in Garrels and Christ (1965), Stumm and Morgan (1996) and Drever
(1997).

9.2.1 Stability of water

Although H2O is not intuitively considered a redox sensitive substance, it may take part in the
following redox reactions
H2O ↔ 2e  2H  1⁄2O2(g) (9.24)
2
where O in water is oxidized, and
1
⁄2H2 ↔ H  e (9.25)
where H is being reduced. Very strong oxidants, that drive Reaction (9.24) to the right, cannot per-
sist in natural environments because they react with H2O. Likewise, very strong reductants will
reduce H2O. Accordingly, the stability of water sets limits to the possible redox conditions in natural
environments. These can be quantified with the mass action equations for Reactions (9.24) and
(9.25). For Reaction (9.24) it is:
log K  1⁄2 log[PO2]  2pe  2pH  41.55 (9.26)
Substituting the atmospheric concentration of oxygen, [PO2]  0.2, leads to
pe  20.60  pH (9.27)
Similarly for Equation (9.25):
log K  pH  pe  1⁄2 log[PH2] (9.28)

Here log K is zero by definition. To assign a value to [PH2] is more arbitrary, but the upper limit at
the earth’s surface must be a value of one, which reduces (9.28) to

pe  pH (9.29)

Relationships (9.27) and (9.29) can be plotted in a pe–pH diagram (Figure 9.4) to delineate the
range of redox conditions to be expected in natural environments. Together with the upper and
lower stability limits of water, ranges of pe/pH conditions encountered in natural environments are
shown. Some caution is warranted since these ranges are based on Eh measurements in nature
(Section 9.1.1). Groundwater environments are seen to cover a broad range from oxidizing to reduc-
ing environments.

QUESTION:
If a more realistic value for PH2 of 105 is used in Equation (9.28), how would this displace the H2O / H2
line in Figure 9.4?
ANSWER: pe  pH  1⁄2 log[H], pe  pH  2.5

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 425

O
20 2 (0
H O .2 atm
2 )
EN
VIR 1.0
O
15 Min NMEN
ew T
ate S IN
r C
Ra ONTA
in C
10 Riv T WIT
er H
Gro Oc ATMO 0.5
und ean SP

Eh (V)
TR
AN
wa
ter Lak HERE
pe

5 SIT e
EN ION
VIR Gro ZO
ON Pea und NE
ME t wa
HO NT t e r
0 2 S IS 0
H OL Hyp
2 (1 A o
atm Pad TED lymn
) dy FR ion
5 Gro ield OM
f
und s AT
wa MO
ter SP
HE
RE 0.5
10

0 2 4 6 8 10 12 14
pH

Figure 9.4. The stability of water the ranges of pe- and pH-conditions in natural environments (Modified from
Garrels and Christ, 1965).

9.2.2 The stability of dissolved species and gases: Arsenic

Arsenic will be used to illustrate the construction of redox diagrams for dissolved species. As
already discussed in Chapter 1, a high arsenic concentration in groundwater is a threat towards the
health of millions of people in West Bengal and Bangladesh.
Arsenic is found in groundwaters as arsenate, As(5) (Smedley et al., 1998) and arsenite, As(3)
(Mukherjee et al., 2000). We discuss here how to plot the stability fields of the dissolved As species
in a redox diagram. Both As(5) and As(3) form protolytes, which may release protons stepwise in
the same way as carbonic acid. The mass action constants for the reactions among the selected
species can be found in the WATEQ4F.DAT database distributed with PHREEQC, and are given in
Table 9.3.

Table 9.3. Log K’s for reactions among (selected) As species.

Reaction log K

As(5) H2AsO
4 ↔ HAsO2
4 H

6.76 (9.30)
 
As(3) H3AsO3 ↔ H2AsO3  H 9.23 (9.31)
As(5) /As(3) H3AsO3  H2O ↔ H2AsO 
4  3H  2e

21.14 (9.32)

For the second dissociation of H3AsO4 (Reaction (9.30) in Table 9.3) the mass action equation is:

log[HAsO24 ]  pH  log[H2 AsO

4 ]  6 . 76
(9.33)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

426 Redox processes

20

PO 
ⴚ 0.2
10 H2AsO4 2

HO
2
pe

HAsO42ⴚ

0
HO
2 H3AsO3
H
2

ⴚ
H2AsO3
10
2 4 6 8 10 12
pH
Figure 9.5. A partial pe–pH stability diagram for dissolved As species. Boundaries indicate equal activities of
both species.

In case of equal activity of H2AsO 4 and HAsO4 , Equation (9.33) reduces to pH  6.76; at lower
2

pH the species H2AsO4 is the dominant form and at higher pH, HAsO2 4 predominates. The stabil-
ity fields of the two species are separated by a vertical line at pH 6.76 in a pe–pH diagram (Figure
9.5), since the boundary is independent of the pe. It is important to realize that in the stability field
of H2AsO 2
4 , HAsO4 is also present albeit, at a lower concentration.
In the same way the first dissociation reaction of H3AsO3 gives:
log [H 2 AsO
3 ]  pH  log[H3 AsO3 ]  9 . 23 (9.34)

Again a vertical line indicating equal activities of both species at pH 9.23 is shown in our pe–pH
diagram (Figure 9.5).
The boundary between the stability fields of the As(5)-species H2AsO 4 and the As(3)-species
H3AsO3 is determined from Reaction (9.32) in Table 9.3 and results in:
log[H 2 AsO
4 ]  3pH  2pe  log[H3 AsO3 ]  21 . 14 (9.35)
which for equal activities of both species reduces to
2pe  3pH  21.14
This boundary depends on both pe and pH and is also plotted in Figure 9.5.

The boundaries between HAsO2 2
4 and H3AsO3, and between HAsO4 and H2AsO3 are found by
2
combining reactions from Table 9.3. For the HAsO4 / H3AsO3 boundary, we add (9.30) to (9.32) and
obtain:

H3AsO3  H 2O ↔ HAsO 24  4H  2e (9.36)

and
log[HAsO 24 ]  4 pH  2pe  log[H3AsO3 ]  27.9
(9.37)
or 2 pe  4pH  27.9

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 427

And similar for the HAsO2

4 / H2AsO3 boundary (Figure 9.5):

2pe  3pH  18.67

QUESTIONS:
At what pe and pH have the 3 species H2AsO 2
4 , HAsO4 and H3AsO3 equal activity?
ANSWER: pH  6.76, pe  0.43
Find the pH where HAsO2 3
4 and AsO4 have equal activity from the WATEQ4F database?
ANSWER: pH  11.6
Find log K for Reaction (9.32) from MINTEQ.DAT?
ANSWER: 21.68 (note the difference with WATEQ4F.DAT!)

Calculating redox speciation with PHREEQC

We check the answer to the first question above with PHREEQC, using the input file:

DATABASE wateq4f.dat
SOLUTION 1
pH 6.76; pe 0.43
As 1e-3 # 75 ug As/L
END

The speciation of As(3) and As(5) is here calculated from the input pe. Since there is only a very
small amount of As in distilled water, the activity coefficients will be close to unity and the output
shows indeed almost equal concentrations of the three species:

Species Molality Activity

As(3) 3.329e-07 (← this is total As(3) in mol/L)

H3AsO3 3.318e-07 3.318e-07
As(5) 6.671e-07
HAsO4-2 3.349e-07 3.333e-07
H2AsO4- 3.322e-07 3.318e-07

Alternatively, we can input the concentrations of As(3) and As(5), together with a default pe (4):

SOLUTION 1
pH 6.76; pe 4
As(3) 0.333e-3; As(5) 0.667e-3
REACTION
END

In this case the input As speciation is maintained and PHREEQC calculates the pe corresponding to
the As(3) / As(5) redox couple:

Redox couple pe Eh(volts)

As(3)/As(5) 0.4299 0.0254

The pe of 4 is used in the initial equilibrium calculation for all other redox couples, e.g. of
H2O / O2. Next, the keyword REACTION signals to do an equilibrium calculation in which the pe is
adapted to achieve overall equilibrium. In this case, only the As species are of importance, and the

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

428 Redox processes

pe becomes 0.43 “adjusted to redox equilibrium” in agreement with the initial concentrations of
As(5) and As(3).

QUESTION:

At what pe and pH have the 3 species HAsO2
4 , H3AsO3 and H2AsO3 equal activity?
ANSWER: pH  9.23, pe  4.51
Check the answer with PHREEQC?

Nitrogen
Nitrogen is an important component in the biogeochemical cycle, it is found in organic matter, in
dissolved species and as various gases. In contrast, minerals containing N are generally very soluble
and therefore rare in nature (but adsorption of NH 4 can be highly significant). Increasing fertilizer
usage and spreading of manure with subsequent leaching of nitrogen from the soil has raised an
interest in the fate of nitrogen in aquifers. Section 9.5 will deal with this subject.
Nitrogen is found in nature in valences ranging from 5 in NO 
3 , to 3 in NH4 , and a reduction
series can be written as:
NO  
3 → NO 2 → N 2(g) → NH 4

Intermediates between NO 2 and N2(g), such as NO(g) and N2O(g) are known to occur in aquifers,
although seldom in significant amounts. Bacteria play an essential role in catalyzing the reactions of
nitrogen in nature, but the stability relationships among the nitrogen species will guide us in what can
be expected in various environments. From PHREEQC.DAT we obtain the constants listed in Table 9.4.

Table 9.4. Log K’s for redox reactions among N species.

Reaction log K

N(5)/N(3) NO2  H2O ↔ NO3  2H  2e 28.57 (9.38)

N(5)/N(0) 1
⁄2 N2  3H2O ↔ NO 
3  6H  5e

103.54 (9.39)
N(0)/N(3) NH 4 ↔ 1⁄2 N2  4H  3e 15.54 (9.40)

First, we calculate the boundaries of N2 with NO 

3 as the most oxidized species and with NH4 as the
most reduced species. For NO 3 and N 2 the mass action equation of Reaction (9.39) is:
log K  log[NO
3 ]  6pH  5pe  ⁄2 log[PN2]  103.54
1
(9.41)
And for N2 and NH
4 the mass action equation of Reaction (9.40):

log K = 1⁄2 log[PN2]  4pH  3pe  log[NH

4 ]  15.54 (9.42)

In order to plot these relations in a redox diagram, we have to substitute values for the activities
of NO 
3 , N2, NH4 . The pressure of N2 in the atmosphere is 0.77 atm and is a reasonable value to chose.
The choice for [NO 
3 ] and [NH4 ] is more arbitrary, but an activity of 10
3
for NO
3 is commonly found
in polluted groundwater and is used here. For consistency, we may then choose the same
activity for NH4 even though it is usually much lower in groundwater. Substitution of these values sim-
plifies Equations (9.41) and (9.42) to:

6pH  5pe  101.0 for the NO

3 / N 2 equilibrium (9.43)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 429

20

1.0
15
NOⴚ
NO 
3 /N 3
O
2
1 03
10
O 0.5
2
HO
2

5 NO 
3 /N

Eh (V)
O
2
pe

N2 1
0 0
ⴙ
NH4

5

0.5
10
HO
NH3 H
2

2
15
0 2 4 6 8 10 12 14
pH

Figure 9.6. pe–pH diagram for the nitrogen system at 25°C. The diagram is drawn for [PN ]  0.77 and 103
2
as activity of dissolved species unless otherwise specified. Metastable boundaries are indicated by dashed lines.

and
4pH  3pe  18.48 for the N 2 / NH
4 equilibrium (9.44)
These equations are displayed in Figure 9.6. The results show a large stability field for N2. NO 3 is
only stable near the upper stability limit of water and NH 4 first becomes stable near the lower sta-
bility limit of water. Above pH 9.25, NH 4 deprotonates to NH3(aq) and the pe-pH slope for the
N2 / NH3 boundary changes to 1.
In order to evaluate the stability of NO2 , which is intermediate in the redox sequence between
NO 3 and N2, we consider Reaction (9.38) from Table 9.4:

log K  log[NO 
3 ]  2 pH  2pe  log[NO 2 ]  28 . 57 (9.45)
Equation (9.45) can be drawn by assuming equal activities for NO 
3 and NO2 . The resulting line
plots in Figure 9.6 in the N2 stability field. Since N2 is more reduced than NO 2 , nitrite is unstable
relative to N2 with the chosen activities. One might argue that the assumption of equal activities of
NO  
3 and NO2 is unreasonable, since the concentration of NO2 in groundwater normally is much
lower than of NO3 . Therefore, we replot Equation (9.45) with the activity of NO

3 a thousand times
larger than of NO  
2 . The NO3 / NO2 boundary now moves upward to a higher pe but NO2 remains


unstable relative to N2. Only when the NO3 / NO2 ratio is increased to above 10 , would NO
  7
2 attain
its own stability field. But in that case NO2 becomes unmeasurable low, which leads us to conclude
that the presence of NO 2 in groundwater is kinetically controlled as an intermediate product in the
reduction of NO3 to N2. In the same way it can be shown that intermediates like NO(g) and N2O(g)
are unstable. In general, the example illustrates how unstable boundaries are identified during the
construction of a redox diagram.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

430 Redox processes

Figure 9.6 shows that at the PO2 of the earth’s atmosphere, N2-gas is thermodynamically unstable
relative to NO3 . The atmosphere consists of 77% N2 and 21% O2, and if N2 should react with O2
to form nitrate, the atmosphere would become depleted of oxygen. The associated proton
production would acidify ocean waters to around 1.7 when mineral buffering reactions are disre-
garded. Fortunately, equilibrium thermodynamics do not predict a feasible reaction in this
particular case.

EXAMPLE 9.5. Oxidation of the atmosphere’s N2 content to nitrate

Thermodynamically, N2 is unstable relative to nitrate at the PO of the atmosphere. What would be
2
the pH of ocean water if all the O2 in the atmosphere was consumed by oxidation of N2 to NO
3 ? The
O2 content of the atmosphere is 3.710 mol. The oceans contain 13.71023 g H2O with an alkalinity
19

of 2.3 meq/L.

ANSWER:
For oxidation of N2 to NO
3 we may write

N 2  2.5O 2  H 2O → 2NO
3  2H


Accordingly, the 3.71019 moles of O2 in the atmosphere could produce 3.01019 mol H, or
21.6 mmol H per kg seawater. Subtracting the alkalinity content of seawater would leave us with
19.3 mmol H/L, corresponding to a pH of 1.7. Dissolution of carbonate and silicate minerals would
in the long run probably buffer most of this acidity, but it would still leave us with an atmosphere deprived
of oxygen.

QUESTIONS:
How would the NO 
3 / N2 line move when [NO3 ]  10
2
instead of 103?
ANSWER: upward by 1⁄5 pe unit
Evaluate the fate of NH 4 , commonly present in rainwater (Table 2.2)?
ANSWER: it should oxidize to NO 3

Non-equilibrium redox in PHREEQC

The lack of reaction among nitrogen species may conflict with the equilibrium calculations of
PHREEQC. For example, you can dissolve the chemical Fe(NO3)2 and prepare a 1 mM solution in
an anoxic glove box. Under anoxic conditions the concentrations of Fe2 and NO 3 remain unchanged.
However, the simultaneous presence of Fe2 and NO 3 is not a stable combination (compare Figure 9.6
with 9.8) and if we calculate the equilibrium composition with PHREEQC, the result is that almost
all Fe2 oxidizes to Fe3, while 0.2 mM NO 3 reduces to N2. The reaction can be blocked in
PHREEQC by decreasing the stability of NO 2 and N2 relative to NO 3:

SOLUTION_SPECIES
NO3-  2H  2e-  NO2-  H2O; log_k 28.570 # 28.57 in database
2NO3-  12H  10e-  N2  6H2O; log_k 207.08 # 207.08
SOLUTION 1
Fe(2) 1
N(5) 2
REACTION
END

Of course this also disables all redox reactions involving nitrate in further calculations.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 431

When nitrate becomes reduced in aquifers the predominant reaction product is N2. Further reduction of
N2 to NH3 does not occur. In the PHREEQC database it is therefore useful to decouple NH3 from the
rest of the nitrogen system. In the database under SOLUTION_MASTER_SPECIES, N(-3) has been
deleted and a substitute "Amm" species is defined:

SOLUTION_MASTER_SPECIES
#N(-3) NH4+ 0.0 14.0067
Amm AmmH+ 0.0 AmmH 17.0

The molecular weight of Amm is the same as for NH3, and the default input species to recalculate
input from mg/L to molar units is AmmH ( NH4 ). In the database, all the SOLUTION_SPECIES,
EXCHANGE_SPECIES, etc. which contain NH3 have been redefined as Amm.

Sulfur and carbon

Also redox reactions involving sulfur and carbon play crucial roles in biogeochemical cycling of
organic matter. The redox diagrams for these elements are shown in Figure 9.7. The stable valences of
dissolved sulfur are sulfate with S(6) and hydrogen sulfide with S(2). At near neutral pH both H2S
and HS can be found, but never S2 because of the very small second dissociation constant of H2S,
log K  13.9. Elemental sulfur S(0) is shown as an intermediate wedge at the low pH side. This is
a solid and the methods to delineate solids in redox diagrams will be presented in the next section.

1.2 1.2

1.0 1.0
P P
O  O 
2 1 2 1
b ar bar
0.8 0.8
HSOⴚ
4
0.6 0.6
H2CO3
0.4 0.4
SO2ⴚ HCOⴚ
Eh (V)

3
Eh (V)

S 4
0.2 0.2
na CO32ⴚ
tiv
eC
0.0 0.0
H2S CH
CH 4(
0.2 P 0.2 4 aq
)
H  P
2 1 H 
bar 2 1
ba
0.4 0.4 r
HSⴚ
0.6 0.6

0.8 0.8
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
pH pH

Figure 9.7. Redox diagrams for sulfur and carbon at 25°C. Total activity of dissolved species is 103. Native
C indicates dissolved CH2O. S indicates elemental sulfur (Brookins, 1988).

The redox diagram of carbon displays C(4) at high Eh in the form of the pH dependent carbonate
species. Under highly reducing conditions methane, C(4), becomes stable. The intermediate
wedge of “native C” is given for aqueous CH2O, but is indicative for the stability of carbon com-
pounds with C(0) in general.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

432 Redox processes

Comparison of redox diagrams immediately delineates the thermodynamically stable species of dif-
ferent elements. For example, at a high pe (or Eh), O2, NO 2 2
3 , SO4 and CO3 are stable according to
Figures 9.4, 9.6, and 9.7.

QUESTION:
Find the stable species of C, N and S in highly reducing conditions in water?
ANSWER: CH4, NH 4 and H2S

9.2.3 The stability of minerals in redox diagrams: iron

The construction of redox diagrams containing minerals is analogous to those for dissolved species,
although the number of assumptions tends to increase. The construction of stability fields for min-
erals will be illustrated with a diagram for the iron system (Figure 9.8), considering the aqueous
species and solids listed in Table 9.5.

Table 9.5. Log K’s for reactions among Fe species (Figure 9.8).

Reaction log K

FeOH2 Fe3  H2O ↔ FeOH2  H 2.4 (9.46)

Fe(3)/ Fe(2) Fe2 ↔ Fe3  e 13.05 (9.20)
Ferrihydrite Fe(OH)3(s)  3H ↔ Fe3  3H2O 3.7 (9.47)
Siderite FeCO3 ↔ Fe2   CO32 10.45 (9.48)
Fe(OH)2(s) Fe(OH)2(s)  2H ↔ Fe2  2H2O 13.9 (9.49)

Stability lines corresponding to Reactions (9.46) and (9.20) can be plotted in Figure 9.8 by again
assuming equal activities for the dissolved species Fe2, Fe3 and Fe(OH)2.

1.0
15 Fe3ⴙ
Fe(OH)2ⴙ
O
10 2
HO
2
0.5

5 Fe2ⴙ Fe(OH)3
Eh (V)

HO
pe

0 2 0
H
2 FeCO3
103
104

105.5

5

0.5
10

Fe(OH)2
15
1.0
0 2 4 6 8 10 12
pH

Figure 9.8. Stability relations in the system, Fe-H2O-CO2 at 25°C. TIC  102.5. Solid/solution boundaries are
specified for different [Fe2]. Heavy lines indicate “realistic” boundaries that correspond to usual field conditions.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 433

Hematite

Goethite

Maghemite

Lepidocrocite

Ferrihydrite

37 38 39 40 41 42 43 44
log([Fe3] [OH]3)

Figure 9.9. The stability ranges of common Fe-oxyhydroxides, hematite (-Fe2O3), goethite (-FeOOH),
maghemite (-Fe2O3), lepidocrocite (-FeOOH) and amorphous Fe(OH)3 or ferrihydrite (5Fe2O3 9H2O). The
activity products are given for 25°C and 1 atm (modified after Langmuir, 1969, using data tabulated by Cornell
and Schwertmann, 2003).

To depict the stability field of iron oxide we need to select one among several minerals. Some
common iron oxide minerals comprise hematite (-Fe2O3), goethite (-FeOOH), maghemite
(-Fe2O3), lepidocrocite (-FeOOH) and freshly precipitated hydrous ferric oxide or ferrihydrite
(5Fe2O3 9 H2O) (Cornell and Schwertmann, 2003). For the dissociation of lepidocrocite and
goethite we may write:

FeOOH  H 2O → Fe3  3OH (9.50)

for hematite and maghemite:

⁄2 Fe2O3  3⁄2 H2O → Fe3  3OH
1
(9.51)
and for ferrihydrite:

Fe(OH)3 ↔ Fe3  3OH (9.52)

In all three cases the mass action equation is

K  [Fe3 ][OH ]3 (9.53)

Figure 9.9 compares the solubility of the various iron oxides and demonstrates a range in stability of
about six orders of magnitude.

QUESTIONS:
Find log K for ferrihydrite in the PHREEQC databases?
ANSWER: PHREEQC.DAT, MINTEQ.DAT, WATEQ4F.DAT: 37.11; LLNL.DAT: 36.35
Find log K for goethite in the PHREEQC databases?
ANSWER: PHREEQC.DAT, WATEQ4F.DAT: 43; MINTEQ.DAT: 41.5;
LLNL.DAT: 41.47

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

434 Redox processes

The most unstable form is freshly precipitated hydrous ferric oxide (Fe(OH)3) while the most stable
forms are hematite and goethite. Even the more crystalline minerals goethite and hematite show a
range in solubility which partly is due to experimental uncertainty, but also covers a real variation
due to different crystallinity, crystal size, solid solution, etc. In aquifers a mixture of different iron
oxides is commonly found. Since only the least soluble iron oxide is stable, the presence of other Fe-
oxides reflects the slow formation kinetics of the more stable phases. For Figure 9.8 we have used
amorphous iron oxyhydroxide, Fe(OH)3, with K  [Fe3][OH]3  1038.3.
Substituting the dissociation reaction of H2O yields Reaction (9.47) in Table 9.5, and the mass
action equation in logarithmic form:

log[Fe3 ]  3pH  3.7 (9.54)

To display Equation (9.54) in the redox diagram, we select a small value for [Fe3] (for example
106) to indicate a low concentration under most field conditions. Actually, the situation at hand is
more complex, since the aqueous complex Fe(OH)2 is also present in significant amounts. For total
ferric iron in solution the mass balance is:
Fe(3)  mFe3  mFeOH2 (9.55)

The total concentration of dissolved ferric iron in equilibrium with Fe(OH)3 is found by substituting
the mass action equations for Reactions (9.46) and (9.47) in Equation (9.55). We disregard the dif-
ference between activity and molar concentration and obtain:

Fe(3)  103.7 [H ]2 ([H ]  102.4 ) (9.56)

This equation was solved for Fe(3) equal to 103, 104 and 105.5 M, and the resulting boundaries are
displayed in Figure 9.8. Note that the aqueous complex Fe(OH)2 doubles the solubility of Fe(OH)3 at
pH  2.4.
The boundary between the Fe2 and Fe(OH)3 is obtained by subtracting Reaction (9.47) from (9.20)
and writing the mass action equation:

log K  3pH  pe  log[Fe 2 ]  16.8 (9.57)

Usually a low Fe2 concentration is chosen to draw a stability line in contact with Fe-oxide and a
boundary for [Fe2]  105.5 is marked by a full line in Figure 9.8. Note the strong pH dependency of
this boundary. The size of the stability field of Fe-oxide will vary with the stability of the selected Fe-
oxide. In Figure 9.8 we used the most soluble form, ferrihydrite, but for Figure 9.10 (discussed later)
the least soluble Fe-oxide (hematite) was taken and the Fe-oxide stability field becomes much larger.
For siderite (FeCO3, Reaction (9.48) in Table 9.5) we need to define the carbonate speciation. Assume
TIC to be constant, and obtain CO2 3 as a function of pH with the speciation factor   mCO3/TIC,
2

where (cf. Equation 5.13):

[H ]2 [H ]
1   1 (9.58)
K1 K 2 K2

Substituting in Equation (9.48) and filling in constants yields:

1010.45  [H ]2 [H ] 

TIC   16.6  10.3  1 (9.59)

[Fe ]  10
2 10 
This equation has been solved for [H], with TIC  102.5 M and [Fe2] activities of 103, 104
and 105.5. The size of the siderite stability field changes with [Fe2] and for a boundary concentration

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 435

of 106 the siderite stability field would disappear altogether. In this case the choice of a low equilib-
rium [Fe2] is not reasonable since siderite in nature is found in environments high in [Fe2] (Postma,
1981, 1982). The boundaries at [Fe2]  103 and 104 are therefore the most realistic ones.
The boundary between FeCO3 and Fe(OH)3 is found graphically as the intersection of the
Fe2 / FeCO3 and Fe2 / Fe(OH)3 boundaries, or more formally by combining Reactions (9.20),
(9.47) and (9.48):
FeCO3  3H 2O ↔ Fe(OH)3  CO32  3H  e (9.60)

where the CO2

3 activity should again be related to TIC and pH. The resulting boundary between FeCO3
and Fe(OH)3 is also shown in Figure 9.8.

QUESTIONS:
Find log K for Reaction (9.60)?
ANSWER: 27.2
Find [H] and [CO2 3 ] for TIC  10
2.5
M and [Fe2]  103 in equilibrium with siderite?
6 8
ANSWER: 1.2710 (pH  5.90), [CO2 3 ]  3.5510
In addition to the previous question, find the pe for equilibrium with ferrihydrite?
ANSWER: 2.05
Calculate the pe/pH for these conditions with PHREEQC? The solubility products of siderite and ferrihy-
drite used in the manual calculation differ from those in PHREEQC.DAT. Saturation indices following the
mineral names are used to apply the manual calculation constants in PHREEQC.
ANSWER: SOLUTION 1; C(4) 3.16; Fe 1; pH 7 Siderite 0.44;
pe 4 Fe(OH)3(a) -1.19;END
gives pH  6.02, pe  1.78 (Why are the results still different?)
Find the pH for the boundary between siderite and Fe(OH)2(s) in Figure 9.8?
ANSWER: 10.87

In summary, boundary lines in redox diagrams have different meanings: between dissolved species
they usually indicate equal activities, between solids and dissolved species they indicate equilibrium
for a specified concentration, and if other components, such as dissolved carbonate or sulfur species,
are involved their concentrations need to be stipulated. Therefore the specifications in small print
below redox diagrams should be read carefully.

Iron and manganese

Compare, for example, Figure 9.8 with the often reproduced diagram for the iron system in Figure
9.10. The more stable Fe-oxide hematite (Figure 9.10) has a much larger stability field than Fe(OH)3
(Figure 9.8), while magnetite (Fe3O4), which was not considered in Figure 9.8, replaces Fe(OH)2.
Figure 9.10 includes sulfur in the system and a stability field for pyrite appears within that of
siderite. Inspection of the figure caption shows the diagram to be drawn for an unrealistically high TIC
content of 1 M, while S is very low. These values were chosen in order to display both pyrite and
siderite stability fields together in the same diagram. With a more realistic (smaller) TIC, the siderite
field would simply disappear. Clearly the diagram is not suitable to evaluate the relative stability of
pyrite and siderite.
The great value of redox diagrams is to obtain a quick overview of complex geochemical systems, as
illustrated in a comparison of the diagrams for Fe and Mn (Figure 9.10). The presence of MnO2 in sedi-
ments indicates strongly oxidizing conditions. In contrast hematite is stable over a much broader pe range.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

436 Redox processes

1.4 1.4

20 20
1.2 1.2

Fe3ⴙ 1.0 1.0

15 15

0.8 0.8

10 10
0.6 0.6
O MnO2 O
2 2
H H

Eh (V)
2O
0.4 Mn2ⴙ 2O
0.4
Fe2ⴙ Fe2O3
pe

5 5

0.2 0.2
Mn2O3
0 H
2O
FeCO3 0 0 H
2O MnCO3 0
H H
2 2
0.2 0.2
FeS2 Mn3O4
5 5
0.4 0.4

10 FeCO3 Fe3O4 0.6 10

MnO 0.6

0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
pH pH

Figure 9.10. Stability relations for iron and manganese at 25°, both assuming that S  106 and TIC  100 M.
Solid-solution boundaries are drawn for [Fe2]  106 (Modified from Krauskopf, 1979).

Dissolved Mn2 is stable over a wide range in contact with hematite, while reversely Fe2 is unstable
in contact with MnO2. Furthermore, rhodochrosite (MnCO3) is stable over a wide pe range, while the
presence of siderite (FeCO3) indicates strongly reducing conditions. Finally, Mn-sulfide does not
appear in the diagram, because it is much more soluble than Fe-sulfide, and therefore Mn-sulfide is
extremely rare in recent environments. These important conclusions can be derived immediately from
pe / pH diagrams, and alternatively would require many hours of calculations.

0.00

0.10 ic Hematite
lfid
e su ts Fe2O3
ri n e n
Ma edim
0.20 s
Magnetite
Eh (V)

0.30 Pyrite FeS2 Fe3O4

Pyrrhotite
Siderite
FeS
0.40 FeCO3

0.50
H2

0.60
4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
pS2

Figure 9.11. Eh–pS2 diagram where pS2  log[S2], pH  7.37, PCO2  102.4, 25°C (Berner, 1971).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Redox diagrams 437

(a) 10  (b) 50 

(c) 10  (d) 30 

Figure 9.12. SEM micrographs of some common authigenic Fe(2) minerals. (a) Framboidal pyrite which are
aggregates of pyrite crystals, (b) polyframboidal pyrite, (c) siderite and (d) a crystal fragment of vivianite. From
Postma (1982).

In some cases the stability relations among redox species are better visualized in diagrams using
variables other than pe and pH. For example, the relative stability of pyrite and siderite is relatively
unaffected by pe and pH. The decisive factor is the insolubility of iron sulfide. It is only once dis-
solved sulfide becomes exhausted, that the Fe2 concentration can increase sufficiently to stabilize
siderite. This point will become clear when in the redox diagram of iron, the pH is replaced by
pS2 as controlling variable (Figure 9.11). Siderite first becomes stable at extremely low dissolved
sulfide concentrations. In practice this implies an anoxic environment where sulfate is absent or has
become exhausted. Therefore, siderite may be found in freshwater environments low in sulfate, like
swamps (Postma, 1977, 1981, 1982) and lakes (Anthony, 1977) and is here often accompanied by the
ferrous phosphate mineral vivianite (Fe3(PO4)2 8 H2O). Ferrous-iron rich groundwater is often
supersaturated for siderite (Margaritz and Luzier, 1985; Morin and Cherry, 1986; Ptacek and
Blowes, 1994; Jakobsen and Postma, 1999) indicating slow precipitation kinetics. Siderite is a com-
mon constituent of sedimentary rocks either in finely dispersed form or as concretions. Furthermore,
siderite may form solid solutions with calcite and dolomite. Some SEM micrographs of siderite and
other authigenic iron minerals are shown in Figure 9.12.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

438 Redox processes

9.3 SEQUENCES OF REDOX REACTIONS AND REDOX ZONING

Figure 9.13 shows the evolution in groundwater composition along the flow path in the Middendorf
aquifer. The groundwater is already anoxic near the zone of recharge but nitrate and sulfate are still
present as electron acceptors. Downstream, first nitrate is reduced, and next the reduction of Fe-oxides
leads to an increase of the Fe2 concentration. Subsequently sulfate is reduced and the precipitation of
iron sulfide causes a decrease in the Fe2 concentration. Finally, methane appears in the groundwater.

10 100

Dissolved sulfate or nitrate (µM)

methane (nM), or iron (µM)
Dissolved hydrogen (nM),

8 H2 80
NO
3
6 60
CH4
2
4 SO4 40

2 Fe2 20

0 0
0 50 100 150
Down flowpath (km)

Figure 9.13. Redox zoning along the flow path in the Cretaceous Middendorf aquifer, South Carolina, USA
(Lovley and Goodwin, 1988).

The Middendorf aquifer is a good example of a regional aquifer, with long groundwater res-
idence times, where the slow degradation of sedimentary organic material generates sequential
changes in the water chemistry. On a much smaller scale Figure 9.14 displays the water chemistry
in a Pleistocene phreatic aquifer near Bocholt, Germany. In the profile, the residence time
of the groundwater increases with depth as discussed in Chapters 3 and 11. Going downwards
in the borehole, first O2 disappears, then NO 2
3 , and then the concentration of SO4 decreases,
similar to what is observed with increasing flow distance in the Middendorf aquifer.

mg/L

0 40 80 120
0 0.02 0.04
15 20 25 30
0 4 8 12 0 80 160 240
0

O2
H2S
Depth (m)

10
NO
3
SO42
N2
20

Figure 9.14. Redox zoning in a sandy Pleistocene aquifer, Bocholt, Germany. The borehole (DFG6) is located
in the upstream part of the aquifer with coherent flowlines (Leuchs, 1988).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sequences of redox reactions and redox zoning 439

Apparently, as water contacts reductants in the subsoil, it loses its oxidants in a sequence that follows
the pe from high to low and the changes in water chemistry can be predicted by the redox diagrams.
In the upper part of the Bocholt aquifer the groundwater age increases with 1 yr/m depth and it takes
about 5 years before O2 is fully depleted. In the Middendorf aquifer hundreds of years are involved.
Taken together, the two aquifers show that the sequence of redox reactions can be predicted from
equilibrium thermodynamics, but the reaction rates are rather variable.
The sequence of predominant redox half reactions that can be predicted from the redox diagrams
is summarized in Figure 9.15 for pH 7. The upper part lists the reduction reactions going from O2
reduction, NO3 reduction and reduction of Mn-oxides that occur at a high pe, to the reduction of Fe-
oxides, sulfate reduction and methanogenesis taking place at a lower pe. The lower half of Figure
9.15 lists in the same way the oxidation reactions, with the oxidation of organic matter having the
lowest pe. A reduction reaction will proceed with any oxidation reaction that is located (origin of the
arrow) at a lower pe. For example, the reduction of sulfate can be combined with the oxidation of
organic matter, but not with the oxidation of Fe(2).

Reductions O2 Reduction

Denitrification

Mn(4) oxide → Mn(2)

Fe(3) oxide → Fe(2)

2
SO4 Reduction

CH4 Fermentation

Oxidat. org. mat.

2
Sulfide → SO4

Oxidat. of Fe(2)

 
NH4 → NO3

Oxidations Oxidat. of Mn(2)

20 10 0 10 20
pe

Figure 9.15. Sequences of important redox processes at pH  7 in natural systems (modified and corrected
after Stumm and Morgan, 1996).

Taking organic matter as the driving reductant, the water chemistry may change as illustrated schemat-
ically in Figure 9.16. For some electron acceptors (O2, NO 2
3 , SO4 ) it is the disappear-
2 2
ance of a reactant while in other cases (Mn , Fe , H2S and CH4) it is the appearance of a reaction
product that is notable in the groundwater composition. In the region dominated by sulfate reduction
the Fe2 concentration may decrease as the result of precipitation of iron sulfides. Redox environments
are often characterized by the dominant ongoing redox process as indicated by the water chemistry
(Champ et al., 1979; Berner, 1981). Berner (1981) (Figure 9.16) distinguishes between oxic and anoxic
environments, i.e. whether they contain measurable amounts of dissolved O2 ((106 M).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

440 Redox processes

Concentration Berner (1981)

O2 Oxic Environment Characteristic phases

NO
3
6
I. Oxic (mO2 1 10 ) Hematite, goethite, MnO2-type
minerals: no organic matter
6
II. Anoxic (mO2  10 )
Mn2 A. Sulfidic (mH S 1 106) Pyrite, marcasite, rhodochrosite,
Post-oxic 2 alabandite: organic matter
6
B. Nonsulfidic (mH S  10 )
2
1. Post-oxic Glauconite and other Fe2–Fe3
Fe2 silicates (also siderite, vivianite,
rhodochrosite): no sulfide
minerals: minor organic matter

Anoxic
2
Depth

SO4 2. Methanic Siderite, vivianite, rhodochrosite:

earlier formed sulfide minerals;
organic matter
Sulfidic
H2S

Fe2 Methanic
CH4

Figure 9.16. The sequence of reduction processes as displayed in groundwater chemistry. At right is
Berner’s (1981) classification of redox environments together with solids expected to form in each zone.

The increase in nitrate in the oxic zone is due to oxidation of ammonia released from oxidizing organic
matter. The anoxic environments are subdivided in post-oxic, dominated by the reduction of nitrate,
Mn-oxide and Fe-oxide, sulfidic, where sulfate reduction occurs and finally the methanic zone.
A further subdivision of the post-oxic zone into a nitric (mNO3 ( 106 M, mFe 2 0 106 M) and a
ferrous zone (mFe 2 ( 106 M, mNO3 0 106 M) can be useful in groundwater environments.
Other classifications of groundwater redox zoning are given by Bjerg et al. (1995) and Chapelle
et al. (1995). Not all zones are necessarily visible in a reduction sequence and in many cases the
groundwater never passes the post-oxic state. In other cases, the sulfidic zone may seem to follow the
oxic zone directly, but the post-oxic zone is probably always present on a microscopic scale. Figure
9.16 also shows some minerals which could be stable in the different redox zones. Their occurrence can
be derived from the redox diagrams on the preceding pages. To characterize the aquifer redox state
from the water composition is certainly a useful approach but some caution is warranted. For example,
the enrichment of groundwater with Fe2 may be caused by partial oxidation of pyrite (see Section
9.4.2) a process totally different from the reduction of iron oxides by organic matter. Reaction products
like methane can be transported over long distances through the aquifer and the mere presence of
methane in groundwater does not implicate ongoing methanogenesis at the site of sampling.

EXAMPLE 9.6. Calculation of redox zonation with PHREEQC

The sequence of redox reactions can be modeled by adding organic carbon stepwise through an irre-
versible reaction to a mixture of water and sediment that contains oxidants. A PHREEQC input file might
look like:

SOLUTION 1
pH 6.0
Na 1.236; K 0.041; Mg 0.115; Ca 0.067
Cl 1.467; N(5) 0.058; S(6) 0.085; Alkalinity 0.26
O(0) 0.124

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sequences of redox reactions and redox zoning 441

EQUILIBRIUM_PHASES 1
Goethite 0 2.5e-3
FeS(ppt) 0 0
Pyrolusite 0 4e-5

REACTION 1
C; 0.572E-3 in 26 steps
INCREMENTAL_REACTIONS true

USER_GRAPH
headings C O2 NO3 Mn Fe SO4 S(-2) C(-4)
axis_titles "Carbon added, mmol/L" "Concentration, mol/L"
-start
10 graph_x step_no * 0.572/26
20 graph_y tot("O(0)")/2, tot("N(5)"), tot("Mn(2)"), tot("Fe(2)"),
\tot("S(6)"), tot("S(-2)"), tot("C(-4)")
-end
END

The results are plotted in Figure 9.17 and show the characteristic stepwise development in redox zones.
Hydrogen sulfide appears only at low concentrations in the water because of excess iron and the precipita-
tion of FeS.

0.14

0.12
Concentration (mmol/L)

0.1
2
SO4
0.08 Fe2

0.06 NO CH4

3

0.04
O2
Mn2
0.02

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Carbon added (mmol/L)

Figure 9.17. Development of redox zones modeled by PHREEQC using stepwise addition of carbon to an
oxidized sediment.

QUESTIONS:
Plot pe and pH in Figure 9.17 on the secondary y-axis?
What is the effect of defining pe 16 in SOLUTION 1?

Redox zoning may develop in an opposite way downstream of landfills (see Christensen et al., 2000
for a review). Figure 9.18 displays the extent of different redox zones in the plume of the Vejen land-
fill, Denmark. The different redox zones are delineated based on the water composition, simi-
lar to the classification given in Figure 9.16. Directly below the landfill the conditions are

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

442 Redox processes

methanogenic, but they change downstream to sulfate reducing, iron and manganese reducing, then
to nitrate reducing and about 400 m downstream of the landfill the groundwater is oxic (this is the
pristine groundwater). In contrast to the natural situation of infiltrating rainwater, the landfill plume
becomes less reduced downstream and electron acceptors must be provided to oxidize the organics
in the plume. The most important electron acceptors are the Fe-oxides present in the sediment
(Heron and Christensen, 1995) through which the leachate passes, and the iron reducing zone also
has the largest extent in Figure 9.18. Dissolved electron acceptors, like sulfate, nitrate and oxygen
can only be introduced into the organic-rich leachate plume through mixing, dispersion and diffu-
sion. Therefore, a sulfate reducing zone is not logically generated downstream of a methanogenic
zone. In this case, local sulfate rich sources within the landfill, like gypsum waste, are emitting sul-
fate enriched groundwater which is mixed with the organic leachate. Likewise, a nitrate reducing
zone can only result from mixing of organic leachate with nitrate-bearing groundwater. The oxida-
tion of a large amount of organics in the leachate plume may generate abundant CO2 which can
cause the dissolution of carbonates. Similar plumes of organic contaminants are often found in
aquifers below cities in developing countries due to infiltration of unsewered urban wastewaters
(Lawrence et al., 2000).

Vejen Distance from landfill (m)

0 100 200 300 400

Landfill
B1 B2 B3 B4 B5 B6 B7 B8 B9
38
Water table
m above sea level

34

30 Methanogenic
Clay/silt
Sulfate-reducing
26 Iron/manganese-
reducing
22 Clay Nitrate-reducing
Aerobic

Figure 9.18. Redox zones in a sandy aquifer downstream the Vejen Landfill, Denmark (Lyngkilde and
Christensen, 1992).

9.3.1 Decomposition of organic matter

Organic matter in aquifers is mostly of sedimentary origin and consists of the remains of plants
in terrestrial sediments, supplemented by plankton and animal residues in the case of marine
and lacustrine sediments. The organic carbon content is commonly lowest in sand deposits
and chalk but higher in clays and silts. Generally, the degradation of sedimentary organic matter
is very slow. Table 9.6 provides the rate of organic carbon degradation of some unpolluted
aquifers and shows rates to vary from a few micromoles to more than millimoles per liter of ground-
water per year. The older aquifer systems generally contain organic matter that reacts more
slowly although the reactivity of organic matter will depend on the diagenetic history.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sequences of redox reactions and redox zoning 443

Table 9.6 also lists the inverse of the first order reaction rate constant, derived by arbitrarily assum-
ing that all the aquifers contain 12‰ organic C (
6 mol C/L groundwater), i.e. (k)1 in the equa-
tion dC / dt  kC. With this organic C content, k1 is approximately equal to the age of
the sediment, illustrating in a general sense how organic matter becomes less reactive as it grows
older (Middelburg, 1989; Boudreau, 1997). Middelburg (1989) found a relation that can be
reworked to:

k  0.16 / t 0.95 (9.61)

where t is the age of the organic matter. The organic matter reactivity is ultimately the engine
that drives most redox reactions and the difference in organic matter reactivity of more than
five orders of magnitude is the key towards understanding differences in redox behavior among
aquifers.

Table 9.6. Rates of organic matter degradation in aquifers. Modified from Jakobsen and Postma (1994).

Aquifer Rate (mmol C/L/yr) Sediment age k1 (yr) Reference

Rømø, Denmark 1.0–90101 Holocene 1.3103 Jakobsen and Postma (1994)

Tuse Næs, Denmark 3.0–7.4101 Pleistocene 1.2104 Jakobsen and Postma (1994)
Bocholt, Germany 4.6102 Pleistocene 1.3105 Leuchs (1988)
Führberg, Germany 2.8102 Pleistocene 2.1105 Böttcher et al. (1989)
Sturgeon Falls, Canada 2.0–3.2102 Pleistocene 2.4105 Robertson et al. (1989)
Fox Hills, USA 4104 Cretaceous 1.5107 Thorstenson et al. (1979)
Florida, USA 2104 Tertiary 3107 Plummer (1977)
Black Creek, USA 1.4–30105 Cretaceous 108 Chapelle and McMahon (1991)

The most abundant source of sedimentary organic matter is plant material, consisting mainly
of lignin and polysaccharides. Since lignin is largely recalcitrant, the degradation of carbohy-
drates becomes the most important process and it is mediated by different groups of micro-
organisms (Conrad, 1999). In an overall sense, this is what happens. As the first step, fermenting
bacteria excrete enzymes that hydrolyze the polysaccharides and break them down to alcohols,
fatty acids and H2. During step two, other bacteria degrade the alcohols and long chained fatty
acids further to acetic acid (CH3COOH), formic acid (HCOOH), H2 and CO2. In the final step, the
fermentative intermediates acetate, formate and H2 are oxidized by oxidants or so-called
terminal electron accepting processes (TEAP’s). For the microbial reduction of Fe-oxide,
Mn-oxide and sulfate, and for methanogenesis, the intermediate products of the fermentation
process are necessary. For the reduction of oxygen and nitrate the pathway is slightly different
since the microorganisms mediating these processes may metabolize the products of step one
directly.
Figure 9.19 illustrates the overall pathway of organic matter decomposition, combining step
one and two in the production of H2, acetate and formate, and indicating step three with the com-
petition of different oxidants for the same pool of substrate. In a kinetic process consisting of sev-
eral successive reaction steps, the slowest step will become overall rate limiting. The rate limiting
step can be identified from the behavior of the pool of reaction intermediates: if the first step is
rate limiting then the concentration of intermediates remains low since everything produced is
quickly consumed by step two. Conversely when step two is rate limiting, the intermediates are
accumulating.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

444 Redox processes

Fe-oxide reducing
bacteria

FeOOH Fe2

Organic
material
Acetate
Formate H2S
SO42

H2 Sulfate reducing
bacteria
CO2
Fermenting
bacteria
CH4  CO2

CH4  H2O
Methanogenic bacteria

Figure 9.19. Schematic pathway of organic matter decomposition under anaerobic conditions (Courtesy
Rasmus Jakobsen).

Figure 9.20 shows the redox chemistry and the distribution of acetate, formate and H2 in the anoxic
aquifer at Rømø, Denmark (Hansen et al., 2001; Jakobsen and Postma, 1999). H2 is present at a con-
centration level of a few nanomoles per liter, and acetate and formate are at a micromolar level.
Since the oxidants are turned over on a scale of mmol/L, as reflected by the increases in Fe2 and
CH4 and the decrease in SO2 4 , the first (fermenting) step must be rate limiting. The residence time
for H2, acetate and formate can be calculated from rates and pool sizes and ranges from minutes to
hours, indicating a highly dynamic behavior of the intermediates. If the fermentive step is rate lim-
iting and the overall process therefore kinetically controlled why can the sequence of the different
redox reaction be predicted from equilibrium thermodynamics? As a first approximation Postma and
Jakobsen (1996) proposed a partial equilibrium model where organic matter fermentation provides
the overall kinetic control, while the subsequent oxidation step approaches equilibrium. The order of
the redox processes is in this model derived from equilibrium calculation using, for example, H2 as
electron donor (Table 9.7).
On a microbial level, the different microbial communities mediating the oxidation step are com-
peting for the energy available from the oxidation of H2, acetate or formate. For each microbial com-
munity there is an energy threshold in the range 3–25 kJ/mol H2, corresponding to the energy needed
to sustain microbial activity. This energy threshold is the reason why H2 concentrations measured in
the field (Figures 9.13 and 9.20) are about an order of magnitude higher than calculated for the reac-
tion at equilibrium (cf. Question). The microbial communities utilize the available energy, bringing it
down to the minimum threshold where they can operate (Hoehler, 1998; Conrad, 1999; Jakobsen and
Postma, 1999). Reactions with a high energy gain (Table 9.7) are therefore able to bring down the H2
concentration to a lower level than reactions with a lower energy gain. As the result, the reactions requir-
ing a higher H2 concentration become inhibited because they cannot pass their energy threshold.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sequences of redox reactions and redox zoning 445

Fe2 (mM) SO42 (mM) CH4 (mM)

0 0.2 0.4 0.6 0.8 1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
0

3
Depth (m)

H2 (nM) Acetate (µM) Formate (µM)

0 1 2 3 4 5 6 0 2 4 6 8 10 12 0 2 4 6 8 10 12
0

3
Depth (m)

Figure 9.20. The distribution of Fe2, SO2 4 and CH4 and of fermentative intermediates H2, acetate and
formate in the Rømø aquifer (Hansen et al., 2001).

For example, microbes reducing Fe(OH)3 may, due to the higher energy yield of the process (Table
9.7), lower the H2 concentration, and thereby the energy gain, to below the level where a sulfate
reducer can be active. The result is that different electron accepting processes proceed in the same order
as predicted for equilibrium even though in reality they never attain equilibrium because of the
energy thresholds.

Table 9.7. Energy gains of reactions between H2 and electron acceptors under
standard conditions at pH 7 (Lovley and Goodwin, 1988).

Reaction Released energy (kJ/mol H2)

2H2  O2 → 2H2O 237

5H2  2NO3  2H

→ N2  6H2O 224
H2  2Fe(OH)3  4H → 2 Fe2  6H2O 50

4H2  SO2
4 H → HS  4H2O 38
4H2  H  HCO 3 → CH4  3H2O 34

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

446 Redox processes

As hinted above, the energy gain of the various redox reactions is related to the H2 concentration. In
earlier studies (Lovley and Goodwin, 1988; Chapelle and Lovley, 1992; Vroblesky and Chapelle,
1994; Chapelle et al., 1995) it was therefore proposed to use the H2 level as an indicator for an ongo-
ing electron accepting process. It has been observed indeed, that H2 concentrations increase as con-
ditions become more reducing (Figures 9.13 and 9.20). However, the concentrations of H2 will be
influenced by the activities of the other reactants in the redox reaction, notably the concentration of
SO24 and the type of iron-oxyhydroxide available. Furthermore, the redox reaction in which H2 is
consumed is endothermal and the equilibrium constant decreases with temperature (Hoehler et al.,
1998; Jakobsen et al., 1998; Problem 9.9).

QUESTIONS:
Use the reaction 1⁄2H2(aq) ↔ H  e; log K  1.57, to estimate the concentration of H2 in nM, at pH 7,
 
in equilibrium with 0.3 mM O2 / H2O; 0.3 mM SO2 4 / 1 M HS ; 5 mM HCO3 / 0.3 mM CH4 (Hint: Find
log K values from PHREEQC.DAT)
ANSWER: H2  3.81036, 0.04, 0.4nM for O2 / H2O, SO2  
4 / HS , HCO3 / CH4,
respectively.
Compare the estimated concentrations of H2 with observations in Figure 9.20.
ANSWER: equilibrium concentrations are about 10 times smaller than observed.
Plot the concentration of H2 during the reaction in the graph of Example 9.6?

9.4 OXYGEN CONSUMPTION

The atmosphere has a PO 2 of 0.21 atm, and using Henry’s law the calculated dissolved oxygen con-
tent is 0.27 mmol/L (8.6 mg/L) at 25°C, increasing to 0.40 mmol/L (13 mg/L) at 5°C. Water saturated
with oxygen by contact with the atmosphere is continuously infiltrating through soils into aquifers
and O2 is therefore an important oxidant in aquifer systems.

EXAMPLE 9.7. Applying Henry’s law to oxygen dissolution

K  [O 2(aq) ]/[ PO ]  102.898
2

and log K  7.5001  7.8981103  T  2.0027105 / T2

The PHREEQC input file to calculate the O2 concentration at different temperature(s) (beware, some data-
bases may contain an incorrect K; compare with the values above):
SOLUTION 1
EQUILIBRIUM_PHASES
O2(g) -0.678 #log 0.21
REACTION_TEMPERATURE
25 10 in 2
END

Figure 9.21 shows the O2 distribution in a sandy aquifer. The unsaturated zone is 15 m thick
and the upper 12 m of the saturated zone displays a nearly constant O2 content. Since the aver-
age annual temperature is near 8°C, the water is in equilibrium with the O2 content of the
atmosphere (Example 9.7). Any O2 consumption that may have occurred in the soil zone is
apparently resupplied by gaseous O2 transport through the permeable sandy soil and there
is no significant reduction of O2 in the upper 12 m of the saturated zone. Below 27 m depth
O2 is rapidly consumed, because the water enters a layer containing reduced substances.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Oxygen consumption 447

O2 (mg/L)
0 2 4 6 8 10 12
10

20

Depth (m)

30

40

Figure 9.21. The O2 distribution in a sandy aquifer (Postma et al., 1991).

In the absence of such reducing substances, O2 saturated groundwater may travel a long way through
aquifers. Winograd and Robertson (1982) reported groundwaters more than 10,000 years old that
traveled up to 80 kilometers from their point of recharge and were still rich in oxygen.
Figure 9.22 illustrates how the thickness of the unsaturated zone influences the O2 content
of the groundwater in the saturated zone. At Alliston the water table is located 4 m below the
surface and the travel time of water through the unsaturated zone is almost a year. During this
time, DOC may react with O2 which is resupplied by gas diffusion from the atmosphere.

Rodney Alliston
0 0

NO3-N
Cl DOC DO
NO3-N
1 5
Cl DO
Depth (m)

Depth (m)

2 10
DOC

3 15
0 20 40 0 5 10 0 20 40 0 6 12
mg/L mg/L mg/L mg/L

Figure 9.22. Dissolved organic carbon (DOC), dissolved oxygen (DO) and nitrate in a sandy aquifer at two
locations in a sandy aquifer; Rodney (left) with a groundwater table at 1 m below the surface and at Alliston
(right) with a water table at 4 m depth (Modified from Starr and Gillham, 1993).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

448 Redox processes

As the result, most DOC is degraded and the groundwater remains oxic. At the Rodney site,
the water table is at 1 m below the surface and, due to the short residence time, the DOC leaching
from the soil survives transport through the unsaturated zone. Therefore the DOC content in the
top of the saturated zone is high and the groundwater anoxic. Clearly, the oxygen concentration
of groundwater is influenced by the residence time of DOC in the unsaturated zone in the
recharge area.
When the water table is high and the soils are less permeable, the O2 concentration of
groundwater may also become depleted due to reduced diffusion of O2 through the soil, in combi-
nation with the O2 consumption by organic matter oxidation. Gas diffusion in the soil pores
depends on the porosity of the soil and the fraction of the porosity that is water filled. The effective
diffusion coefficient can be estimated with one of the Millington-Quirk relations (Jin and Jury,
1996):

De, a  Da eg2 / e0.67 (9.62)

where Da is the diffusion coefficient in “free” air (Da
105 m2/s), and g and  are the fractions of
gas-filled and total porosity, respectively. In a completely dry soil with g    0.4, the effective
diffusion coefficient becomes:

De, a  Da e1.33  0.3105 m 2/ s

If the same soil is partly saturated (for example w  0.32 and g   – w  0.08), De,a diminishes to:

De, a  Daeg2 / e0.67  0.01105 m 2/ s

5 5E6
log (suction head/cm water)

4
4E6
3
clay 3E6
De (m2/s)

2 clay
sand 2E6
1

1E6 sand
0

1 0E0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Water content, εw() Water content, εw()

Figure 9.23. (Left) pF-curves which relate soil suction (log h, h in cm water) and water content w for a sandy
soil and a clay. At equilibrium, h equals height above the water table. (Right) Estimated effective diffusion coef-
ficient (De,a m2/s) as function of the water filled porosity (w). (Da
105 m2/s)

The relation between De,a and w as predicted by Equation (9.62) is shown in Figure 9.23 (right) for
a clay soil with a high porosity and a sand with a lower porosity. With an increasing water content,
more air-filled pores become isolated, and gas transport is hindered as it must diffuse via water with
a diffusion coefficient that is approximately 104 times lower than in air (cf. Chapter 3). We can esti-
mate the yearly flux of oxygen in a soil and compare it with the carbon productivity (Example 9.8).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Oxygen consumption 449

EXAMPLE 9.8. Compare oxygen flux and carbon productivity in a soil

A loamy soil, with the water table at 1 mbs, has   0.4, and average g  0.15 in the upper 0.9 m. Below
that depth, g is smaller than 0.05 and O2 diffusion is negligible in the gas phase. Estimate the yearly flux of
oxygen into the soil.

ANSWER:
Assume a zero O2 concentration at 0.9 m depth, and a linear gradient. Following Fick’s first law the O2
flux is:
cair  0
F  eg De, a
0  0.9

with g  0.15,   0.4

De,a  0.152 / 0.40.67105 (m2/s)  3.15107 (s/yr)  13 m2/yr
cair  P (atm) / RT (atm L/mol)  0.21 (atm) / 24 (atm L/mol) 103 (L/m3)  8.75 mol/m3
we find FO 2  19 mol/m2/yr.
This flux can oxidize 19 mol CH2O/m2/yr, or 5.7 ton CH2O/ha/yr, which amounts to the productivity of a
rich soil (Russell, 1973).

QUESTION:
Show, using the steady state diffusion equation (Equation 5.59) that a linear concentration gradient for
oxygen is valid when organic carbon degrades with a first order rate coefficient of less than 1/yr (Hint:
Find k1 and k2 in Equation (5.61) using cO2 at 0 and 0.9 m)

Example 9.8 suggests that soils with a deep water table will not accumulate organic matter because
O2 transport from the atmosphere is sufficient to oxidize all the carbon produced. The O2 gradient in
a wet soil will be steep (e.g. if all O2 is consumed at 0.09 m depth, the gradient is 10 times higher)
but since the effective gas-diffusion coefficient is much lower, the flux of O2 is smaller than in a dry
soil and organic matter may accumulate to generate peaty soils.
As Equation (9.62) shows, the diffusion coefficient of a soil depends strongly on its water con-
tent, which in turn depends on the soil type and weather conditions. The distribution of soil moisture
over depth can be predicted from the so-called pF-curve which relates the water content of the soil
to soil suction, and may be defined as:
e
ew  (9.63)
1  (a h )n

where h is soil suction in cm water, and  and n are Van Genuchten fitting parameters. In a soil pro-
file at equilibrium, the soil suction h equals the height above the water table. The parameters  and
n depend on the characteristics of the soil. For a sandy soil   0.05 and n  1.2 may be found,
while in a clay soil   0.01 and n  0.5 can apply. Figure 9.23 (left) shows how with increasing
depth, the water content increases which in turn decreases the diffusion coefficient of O2 and thereby
the downward flux of O2.

QUESTION:
Estimate the diffusional oxygen flux when the soil in Example 9.8 is fully water saturated?
ANSWER: F  w De grad(mO2) (cf. Chapter 3). Take a gradient mO2 / z  0.3 /
0.9 mol/m2, and find F  1.7 mmol O2/m2/yr

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

450 Redox processes

9.4.1 Pyrite oxidation

The oxidation of pyrite and other metal-sulfide minerals by oxygen has a large environmental impact
and plays a key role in acid mine drainage and the formation of acid sulfate soils resulting from
drainage of lowlands. It acts as a source of sulfate and iron in groundwater, and of heavy metals in gen-
eral in the environment. The oxidation of pyrite has been studied for almost a century but some aspects
remain unclarified (Lowson, 1982; Nordstrom, 1982; Alpers and Blowes, 1994; Evangelou and Zhang,
1995; Herbert, 1999; Rimstidt and Vaughan, 2003). The overall process is described by the reaction:

FeS2  15⁄4O2  7⁄2H2O → Fe(OH)3  2SO2
4  4H (9.64)

It illustrates the strong generation of acid by pyrite oxidation. In extreme cases this may result in
negative pH values, corresponding to a concentrated sulfuric acid solution (Nordstrom et al., 2000).
The full oxidation process involves both the oxidation of the disulfide S2 2
2 and of the Fe .
The initial step is the oxidation of the disulfide to sulfate by O2:

FeS2  7⁄2O2  H2O → Fe2  2SO2
4  2H (9.65)

Subsequently Fe2 is oxidized by oxygen to Fe3:

Fe2  1⁄4O2  H → Fe3  1⁄2H2O (9.66)

The oxidation of disulfide proceeds at a lower redox potential (Figures 9.7, 9.8) than Fe2 oxidation.
Incomplete pyrite oxidation, due to an insufficient supply of electron acceptor, therefore results in a
solution enriched in Fe2 and SO42 following Reaction (9.65). Unless pH is extremely low, Fe3
will precipitate according to:

Fe3  3H 2O → Fe(OH)3  3H (9.67)

Equation (9.67) is highly pH dependent and generates three quarters of the acidity in the overall
Reaction (9.64).

9.4.2 Kinetics of pyrite oxidation

Laboratory experiments show the oxidation of pyrite by O2 to be a slow process, a conclusion con-
trary to the dramatic results of pyrite oxidation in the field. This discrepancy has induced extensive
research in the kinetics of pyrite oxidation (Williamson and Rimstidt, 1994; Rimstidt and Vaughan,
2003). In order to oxidize an S22 group of pyrite to sulfate, 14 electrons have to be transferred and
therefore complicated reaction kinetics are to be expected. The different reactions involved in the
oxidation of pyrite are summarized in Figure 9.24.
Initially FeS2 reacts with O2 following Reaction (9.65), either through a direct reaction ((a) in
Figure 9.24)) or through dissolution followed by oxidation (a), but in both cases the rates remain
low. The second pathway of pyrite oxidation is by reaction with Fe3:

FeS2  14Fe3  8H 2O → 15Fe 2  2SO24  16H (9.68)

The reaction between pyrite and Fe3 (Figure 9.24(c)) is fast and yields a low pH. The produced Fe2
may become oxidized by O2 to Fe3 (Reaction 9.66, Figure 9.24(b)). The kinetics of Fe2 oxidation
have already been discussed in Chapter 4 and are slow at low pH and increase steeply with increas-
ing pH. On the other hand, only at low pH does Fe3 remain in solution since it otherwise precipi-
tates as Fe(OH)3 (Equation 9.67). In a purely inorganic system the rate of Fe2 oxidation (b)
therefore rapidly becomes rate limiting as the pH decreases. However, in a natural setting iron oxi-
dizing bacteria may accelerate process (b) by orders of magnitude (Kirby and Elder Brady, 1998).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Oxygen consumption 451

2
Fe(II)  S2

a  O2

(a) 2
FeS2(s)  O2 SO4  Fe(II)
Fast

 O2 (b) (c)  FeS2(s)

Slow
Fe(III) Fe(OH)3
(d)

Figure 9.24. Reaction pathways in the oxidation of pyrite (Stumm and Morgan, 1996).

Bacterial catalysis therefore enables rapid pyrite oxidation at low pH (pH  4) through reactions (b)
and (c). Towards higher pH, pyrite oxidation by O2 becomes dominant because the Fe3 pathway is
inhibited by the low solubility of Fe(OH)3 (reaction (d)) and Equation (9.67)) which keeps the Fe3
concentration very low.
Figure 9.25 shows experimental results for both pathways. At high pH, oxidation by oxygen is the
dominant pathway but the process is slow. Thiosulfate appears as an intermediate oxidation product
between sulfide and sulfate. The specific rate of pyrite oxidation by O2 is described by (Williamson
and Rimstidt, 1994):

r  108.19 mO0.5 m0.11 (mol/m 2/ s)

H
(9.69)
2

The rate has a square root dependency on the oxygen concentration, indicating a large effect at low O2
concentrations while at higher O2 concentration the effect is small. This corresponds to the pyrite sur-
face becoming saturated with O2 (Nicholson et al., 1988). The effect of pH on the rate is very small.

12 S 150
O2
120
 Sulfur (M/g pyrite)

9 Initial pH  2
Sulfur (M/g pyrite)

90
6
Thiosulfate
60

3
30
Sulfate Fe3ⴙ
0 0
0 40 80 120 0 40 80 120
Time (min)

Figure 9.25. Pyrite oxidation by an oxygen saturated solution at pH 9 (left), and by ferric iron at pH 2 (right).
Note the differences in scale on the y-axis implying that oxidation by Fe3 is much faster (Moses et al., 1987).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

452 Redox processes

The rate of pyrite oxidation by Fe3 is many times faster (Figure 9.25) and in the presence of O2 it is
described by (Williamson and Rimstidt, 1994):
0.93 m0.40
r  106.07 mFe (9.70)
3 Fe 2

and when O2 is absent by:

0.3 m0.47 m0.32
r  108.58 mFe (9.71)
3 Fe 2 H

Pyrite is a semiconductor where sulfide atoms are oxidized at an anodic site, releasing electrons that
are transported through the crystal to a cathodic Fe(II) site, where they are acquired by the aqueous oxi-
dant. The overall rate determining step is the reaction at the cathodic site (Rimstidt and Vaughan, 2003).
Fe2, and H compete with Fe3 for sorption on the cathodic sites on the pyrite, and therefore have a
negative effect on the rate, as is expressed in the rate equations. Other cations can be expected to com-
pete with Fe3 as well, but the effects have not been quantified yet (Evangelou and Zhang, 1994).
The rates as calculated from Equations (9.70) and (9.71) will increase infinitely as the Fe2 and

H concentrations decrease, whereas the competitive effect should become negligible towards low
concentrations. To avoid this artefact, numerical models contain an inhibition factor which approaches
1 when the concentration of the inhibiting ion becomes smaller than a limiting concentration:
(1  mFe2 / mLim )n (9.72)

where mLim is a small limiting concentration and n an empirical coefficient. The data compilation by
Williamson and Rimstidt (1994) was used to obtain coefficients of modified rate equations that
include an inhibition factor, using a non-linear estimation algorithm (cf. Example 7.1), which gave
an optimized value of mLim  1.3 M. Using instead a round number of mLim  1 M, the rate equa-
tion for the case with oxygen present becomes:
r  6.3104 m0.92
3
(1  mFe2 / 106 )0.43 (9.70a)
Fe

and without oxygen:

r  1.9106 m0.28
3
(1  mFe2 / 106 )0.52 m0.3 (9.71a)
Fe H

with r in mol/m2/s. The maximum pH in the data is 2.5 which limits the applicability of the rates
to the pH range 0.5–3.0. However, the rate will go to zero anyway for pH ! 3.0 because the Fe3
concentration becomes limited by the solubility of iron oxyhydroxide.
All the rates given by Equations (9.69, 9.70a and 9.71a) must be multiplied with the surface area
of pyrite in m2/L to obtain the reaction in mol/L/s. Experimentally, the reactive surface area is found
to be smaller than geometric surface area, and furthermore, to decrease in the course of the reaction
as result of armoring by precipitates of Fe(OH)3 and iron-sulfates (Wiese et al., 1987; Nicholson
et al., 1990). Observed rates in tailings and waste dumps may therefore be smaller than estimated
from the laboratory rates (cf. Questions). Higher rates are also found because microbes enhance the
reactions in the field (Herbert, 1999).

QUESTIONS:
Estimate the maximal yearly rate of oxidation of framboidal pyrite in the presence of oxygen and
calcite. Assume 3 m framboids, pyr  5 g/cm3.
ANSWER: pH  7, A  0.4 m2/g, mO 2  3  10 4 M yield 0.007 mol FeS2/(g pyrite)/yr.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Oxygen consumption 453

What is the rate at pH  2, oxygen present?

ANSWER: From solubility of Fe(OH)3: mFe3  1040 / (1012)3  10 4 M (hence Fe2 
1010 M) gives 1.7 mol FeS2/g/yr.
And with initial conditions as previous, but no resupply of oxygen?
ANSWER: From electron balance, Fe(OH)3 : FeS2  14 : 1, hence mFe2 / mSO42  15/2.
From electrical balance, 2mFe2  2mSO42  0.01. Together, mSO42  0.77mM,
FeS2  0.38mM. (you may want to check with PHREEQC?).

9.4.3 Oxygen transport and pyrite oxidation

Apart from the kinetics of pyrite oxidation, the transport of oxygen towards pyrite may become
the limiting factor for pyrite oxidation in the field. Similar to the different oxidation rates of organic
carbon in wet and dry soils, pyrite oxidation also depends strongly on moisture conditions. When
pyrite is situated below the water table advective transport of dissolved O2 is the only mode of O2
transport. Since air saturated groundwater contains about 0.33 mM O2 the maximum increases are
(4 / 7)  0.33  0.19 mM SO2 4 and (2 / 7)  0.33  0.09 mM Fe
2
for incomplete pyrite oxidation
(Reaction 9.65) or (8 / 15) 0.33  0.18 mM SO4 in the case of complete pyrite oxidation (Reaction
2

9.64). This situation is illustrated at the left in Figure 9.26 where the upper part of the saturated zone in
a sandy aquifer has a constant O2 content that matches air saturation, while the increase in SO2 4 and
Fe2, at the depth where O2 disappears corresponds to the ones predicted by Reaction (9.65).

0 2 4 6 8 pH
mM mM mM 0 10 20 30 40
Fe (mM)
0 0.2 0.4 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0 20 40 60 80 SO (mM)
4

pH Fe SO4
46
1
42
Elevation (m)

38 2
Depth (m)

34
3

30

26 4
2ⴚ
O2 SO4 Fe2ⴙ
22
5

Figure 9.26. Pyrite oxidation by oxygen supplied by purely advective flow (left) and (right) by diffusive influx
under stagnant water conditions. In the first case (Postma et al., 1991) groundwater, air-saturated with O2, is
transported through a pyritic layer. In the diagram at right, drained pyritic swamp sediments are oxidized by
gaseous O2 diffusion through the unsaturated zone into the pyritic layer (Postma, 1983). Note the huge differ-
ence in the amounts of pyrite oxidation and the associated effects.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

454 Redox processes

Saturation (%) Sulfide (% by wt)

0 20 40 60 80 100 0 2 4 6 8
0
Depth from the tailings surface (cm)

20

40

60

0 4 8 12 16 20
Oxygen concentration (%)

Figure 9.27. Sulfide oxidation in mine tailings: (left) O2 concentration in the gas phase and the water saturation
percentage. (right) sulfide distribution in the solid phase (Elberling and Nicholson, 1996).

A completely different water chemistry results when the pyrite is located above the groundwater
table. Gaseous diffusion may now convey a much larger flux of oxygen and produce dramatic
changes in water chemistry. This situation causes the severe effects of acid mine drainage and critically
deteriorates soils which are drained for agricultural ue and contain pyrite. The resulting sulfate and
iron concentrations (Figure 9.26, right) are huge and the pH decreases to values close to 2.
Figure 9.27 illustrates the O2 transport through a mine tailing deposit. The upper 60 cm of the tail-
ing are unsaturated and the O2 content in the soil gas is already close to zero at 40 cm below the surface.
Oxygen is transported downward by diffusion. The right hand graph shows the distribution of the
sulfide in the solid phase which mirrors the O2 gas distribution. If downward diffusion of O2 was the
sole process to limit the rate of sulfide oxidation, then a sharp front is expected in the distribution of
sulfide in the sediment. The more gradual increase in sulfide distribution over depth indicates that
O2 transfer over the air-water interface and reaction kinetics are both of importance. Therefore the
O2 profile is different from what is expected for simple diffusion with linear retardation
(cf. Question with Example 9.8). The shape suggests that O2 is mostly consumed at 30 cm while
above that depth pyrite has become unreactive or shielded by reaction products.

QUESTIONS:
Estimate the yearly flux of O2 in the mine tailings of Figure 9.27, cf. Example 9.8?
ANSWER: take   0.3, then g  0.3  (100 – 20) / 100  0.24. cO2 / z  ((0.21 – 0.17) /
24  1000) / 0.2  8.3 mol/m4. The flux is 81 mol O2/m2/yr.
Estimate the amount of sulfide oxidation in the tailings?
ANSWER: 81  4 / 15  22 mol FeS2/m2/yr
Estimate sulfide oxidation if the tailings are wetted by sprinklers to 90% saturation?
ANSWER: 0.04 mol FeS2/m2/yr

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Oxygen consumption 455

Transport of gaseous O2 occurs not only by simple diffusion. When O2 is used up from soil air, the
total gas pressure decreases by 20% resulting in downward advective transport of air into the soil.
This advective gas transport brings in nitrogen, which builds up pressure in the soil and tends to dif-
fuse back into the air. A multicomponent diffusion process takes place which is further complicated
by temperature and atmospheric pressure variations (Thorstenson and Pollock, 1989; Massmann and
Farrier, 1992; Elberling et al., 1998).
The drop in gas pressure also depends on accompanying buffering reactions. If the proton
production by pyrite oxidation (Reaction 9.64) causes calcite dissolution according to:
CaCO3  2H → Ca 2  CO 2  H 2O (9.73)
then the net volume loss of the two reactions would be close to 10%. Wisotzky (1994) and Van Berk
and Wisotzky (1995) estimated that the contributions of O2 transport by diffusion and advection are
about the same. In the field barometric pressure changes and the action of wind may also contribute
to advective transport of O2 in the unsaturated zone.

EXAMPLE 9.9. Modeling gas loss during pyrite oxidation with oxygen
Unsaturated aquifer sediment containing pyrite was incubated in gas-impermeable polymer bags containing
a gas mixture with 8% O2 and 92% N2 (Andersen et al., 2001). The sediment contains CaCO3 that functions
as pH buffer. The gas composition in contact with the incubated sediment is followed over time as shown in
Figure 9.28.

9 100
8
7 98
P (O2 or CO2) (%)

6 O2 N2 P (N2) (%)
96
5
4
94
3 CO2
2 92
1
0 90
0 200 400 600 800 1000 1200 1400 1600
Time (hours)

Figure 9.28. Development in gas phase composition during the oxidation of pyrite in unsaturated aquifer
sediment incubated in gas-impermeable polymer bags. The initial gas volume was 21.3 ml, the gas/water
ratio 3:1, the sediment weight 82.79 g and the FeS2 content 1.59 mmol/kg. Symbols indicate experimental
data and lines are modelled with PHREEQC.

Since N2 is a conservative gas, and is quite insoluble in water, its increase indicates a loss of total volume
given by

(
Vloss  V0 1  PN
2(0)
/ PN
2(t )
) (mL) (9.74)

Here (0) and (t) refer to times zero and t. The change in N2 gas pressure in Figure 9.28 indicates a total loss
of 1.1 mL, corresponding to 5% of the initial gas volume.The evolution of the gas phase composition
during pyrite oxidation was modeled using PHREEQC and the input file is shown below.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

456 Redox processes

RATES
Pyrite # rates from Equations (9.69), (9.70a)
# and (9.71a)
-start
1 A = 15e3 * m0 # initial surface area in m2 for 0.01 um crystals
10 if SI("Pyrite") >0 then goto 100 # step out when supersaturated...
20 fH = mol("H+")
30 fFe2 = (1 + tot("Fe(2)")/1e-6)
40 if mol("O2") < 1e-6 then goto 80
50 rO2 = 10^-8.19 * mol("O2") * fH^-0.11 # ...rate with oxygen
60 rO2_Fe3 = 6.3e-4 * tot("Fe(3)")^0.92 * fFe2^-0.43 # rate O2 + Fe3+
70 goto 90
80 rem # rate with Fe3+ without oxygen,
# and for pH < 3...
81 rFe3 = 1.9e-6 * tot("Fe(3)”)^0.28 * fFe2^-0.52 * fH^-0.3
90 rate = A * (m/m0)^0.67 * (rO2 + rO2_Fe3 + rFe3) * (1 - SR("Pyrite"))
100 save rate * time
-end

SOLUTION_SPECIES # make N2 the only N species...

2NO3- + 12H+ + 10e- = N2 + 6H2O; log_k 500 # 207.080 in the database

SOLUTION 1
-water 0.0069239
-temp 20
pH 7 charge; pe 14 O2(g) -1.0878
Ca 1 Calcite; C 1 CO2(g) -2.6021
Fe 1e-3 Goethite 2; N 1.3 N2(g) -0.0382

EQUILIBRIUM_PHASES 1
Goethite 2; Calcite 0; Gypsum 0 0

KINETICS 1
Pyrite; -m0 1.32e-4; -step 0 5e5 5e5 5e5 5e5 5e5 5e5 5e5 5e5 5e5 5e5
INCREMENTAL_REACTIONS true

GAS_PHASE 1
-fixed_pressure # default 1 atm
-volume 0.02127; -temp 20.0
CO2(g) 0.0025; O2(g) 0.0817; N2(g) 0.9157

USER_GRAPH
-head time CO2 O2 N2
-axis_titles "Time/hour" "P_O2 and P_CO2/atm" "P_N2/atm"
-start
10 graph_x total_time/3600
20 graph_y 10^si("CO2(g)"), 10^si("O2(g)"); 30 graph_sy 10^si("N2(g)")
-end
END

The rate of pyrite oxidation is defined according to Equations (9.69)–(9.71a). In the experiment only the rate
with O2 gives a reaction, but the other terms were added for illustration. The overall rate is multiplied with
(1
SR(“Pyrite”)) to let it zero out towards equilibrium (line 90). N2 must behave like a conservative gas dur-
ing the simulation and is made the only N-species by increasing its stability relative to NO
3 . (Note that
N(
3) should also be minimized if defined in the database). The SOLUTION specifies the actual amount of
water in the incubation bag, and EQUILIBRIUM_PHASES and KINETICS define the solids. GAS_PHASE
states the volume and composition of the gas, and lastly, USER_GRAPH plots the data.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Oxygen consumption 457

Pyrite oxidation is simulated as kinetic reaction, fitting the rate to the experimental data by adapting the spe-
cific surface (which appears to be rather high, line 1 in the Pyrite rate). The reaction consumes O2 (Reaction
9.64). The modeled gas concentrations are given as solid lines in Figure 9.28 and show good correspondence
with the measured compositions. Note that the amount of CO2 that appears in the gas phase is smaller than
predicted by the sum of Reactions (9.64) and (9.73) because CO2 is more soluble in water than N2 and O2;
its dissolution explains part of the loss of volume.

QUESTION:
Find the specific surface of 4 wt% pyrite (Figure 9.27) necessary to consume 22 mol O2/m2/yr
from 25 to 35 cm depth? Hint: use the rate of PHREEQC and a gas phase that contains 22 mol O2
ANSWER: 0.005 m2/mol pyrite (valid for 3 cm crystals). Note the increase in rate for
pH  3, when the rate with Fe3 and O2 takes the lead. Also note that allow-
ing for precipitation of JarositeH gives sulfate concentrations close to field
observations (Figures 9.26, 9.29; 7.1)

C pH C
North South

Dam

345 Road
Elevation (m)

4.5
340 4.5
5.0
5.5
6.0
335
6.5
330
Bedrock
Till

Dam
Fe (102 ppm)
345 Road
Elevation (m)

30
40
340
50
40
335 30
20

330 10
Bedrock
Till

Dam
SO42 (102 ppm)
345 Road
Elevation (m)

100 60 40
340 120
140
120 40
100
335 80
60 40
330
Bedrock
Till 0 5 10 m

Figure 9.29. A plume of contaminants, including iron sulfate and low pH emanating from sulfidic
mine tailings into a sandy aquifer. The tailings are located just north of the dam. (Modified from
Dubrovsky et al., 1984).

Figure 9.29 shows drainage water from sulfide mine tailings entering an aquifer. A plume of
contaminants spreads through a sandy aquifer from sulfide mine tailings located at the left side of
the dam. In its center the plume contains more than 11,000 ppm sulfate, more than 5000 ppm

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

458 Redox processes

dissolved Fe, and the pH is less than 4.5. Sulfide ores contain, apart from pyrite, a range of
sulfide minerals like sphalerite (ZnS), chalcopyrite (CuFeS2) and arsenopyrite (FeAsS). These min-
erals provide an ample source of heavy metals for groundwater contamination. The low pH values
causes large increases in dissolved Al (Chapter 8) and a variety of secondary minerals may precipi-
tate under such extreme pyrite oxidation conditions. These include gypsum, ferric hydroxysulfates
like jarosite (KFe3(SO4)3 9H2O and several others (van Breemen, 1976; Nordstrom, 1982; Postma,
1983). Generally these precipitates are soluble and will be leached out by infiltration, except for
FeOOH. The plume displayed in Figure 9.29 was found to migrate at a much slower rate than the
groundwater flow velocity, and neutralization and precipitation of sulfate and ferrous iron must
occur at the margins of the plume. Morin and Cherry (1986) found equilibrium with gypsum
throughout the plume and saturation to supersaturation for siderite. They proposed that calcite,
which is present in small amounts in the sediments, dissolves. The resulting high Ca2, Fe2 and dis-
solved carbonate concentrations apparently generate the simultaneous precipitation of gypsum and
siderite, the latter possibly as solid solutions with calcite. Carbonate reactions were also observed
and evaluated in the Pinal Creek aquifer (Glynn and Brown, 1996).

9.5 NITRATE REDUCTION

Nitrate pollution of groundwater is an increasing problem in all European and North American countries
(e.g. Strebel et al., 1989; Korom 1992; Spalding and Exner, 1993; Feast et al., 1998; Tesoriero et al.,
2000) and poses a major threat to drinking water supplies based on groundwater. The admissible nitrate
concentration in drinking water (see Chapter 1) is 50 mg NO 
3 /L (corresponding to 11 mg NO3 -N/L or
0.8 mmol/L) and the recommended level is less than 25 mg/L NO3. A high nitrate concentration in drink-
ing water is believed to be a health hazard because it may cause methaemoglobinaemia in human infants,
a potentially fatal syndrome in which oxygen transport in the bloodstream is impaired.
The groundwater nitrate content is derived from various point and non-point sources, including
cattle feed lots, septic tanks, sewage discharge and the oxidation of organically bound nitrogen in soils.
However, the main cause for the increasing nitrate concentration in shallow groundwater is, without doubt,
the excessive application of fertilizers and manure in agriculture since the early sixties (cf. Example 2.3).
The relation between land-use and nitrate pollution of aquifers is illustrated in Figure 9.30.
Little nitrate is leached from the forest and heath areas, but downstream from the arable land,
plumes of nitrate containing waters spread through the aquifer. The question is to what extent nitrate
is transported as a conservative substance through the aquifer, and how geochemical processes
within the aquifer may attenuate the nitrate concentration.

Arable Forest and heath Arable

T1 T2 T3 T4 T5 T6
60 T7
Elevation (m)

T8 Water table
40
? ?

20

NO
0.5 km
3 ! 0.1 mM O2  0.05 mM
0

Figure 9.30. Nitrate pollution plumes emanating from agricultural fields into an unconfined sandy aquifer
(Rabis Creek, Denmark). The groundwater flows from right to left. Numbers T1 through T8 refer to
locations of multilevel samplers on which the plume distribution is based (Postma et al., 1991).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Nitrate reduction 459

9.5.1 Nitrate reduction by organic matter oxidation

Nitrate forms neither insoluble minerals that could precipitate, nor is it adsorbed significantly.
Therefore the only way for in situ nitrate removal from groundwater is by reduction:

2NO 
3  12H  10e
 → N  6H O
2 2 (9.75)

The redox diagram for nitrogen (Figure 9.6) indicates that nitrate is only stable under highly oxidiz-
ing conditions while N2 is the stable form at intermediate pe. Ammonia is stable only under highly
reduced conditions.
In microbiology (Krumbein, 1983; Zehnder, 1988) the two important overall reactions for
N-cycling are denitrification and nitrification (Figure 9.31). Denitrification is the term used for the
microbial reduction of nitrate to N2 by organic carbon. The overall reaction comprises a transfer of
five electrons per N-atom and proceeds through a complicated pathway with several metastable
intermediates (see also Section 9.2.2 on nitrogen):

NO 
3(aq) → NO 2(aq) → NO(enzyme complex) → N 2 O(gas) → N 2(gas)
(9.76)

Intermediates like NO 2 and N2O are often found at trace levels in natural environments and are then
used as evidence for ongoing denitrification. Nevertheless, N2 is always the predominant reaction
product. Denitrification is not a reversible reaction; fortunately there are no bacteria which are able
to live on the energy available from oxidizing N2 to NO 3 (Example 9.5). Dissimilatory nitrate reduc-
tion to NH 4 is possible in groundwater systems (Smith et al., 1991a) but normally plays a subordi-
nate role. During nitrification, bacteria oxidize amines from organic matter to nitrite and nitrate and
this process is of most importance in the soil zone.

N2

Denitrification (N2O) N2 – Fixation

(microbiological
NO
2 and industrial)
Organic Diss.
.
Ass N-compounds
Industrial Ass.
NO
3 NH
4

Ass. nitrate red.

NO
2

Nitrification

Figure 9.31. Pathways in the redox reactions of nitrogen. Ass. is assimilation into, diss. is dissimilation from
organic matter.

The reduction of nitrate by organic matter is well documented for soils and marine sediments
and is also important in aquifers (Smith and Duff, 1988; Bradley et al., 1992; Korom, 1992;
Starr and Gillham, 1993; Smith et al., 1996; Bragan et al., 1997). The overall reaction stoi-
chiometry is:

5CH 2O  4NO 
3 → 2 N 2  4HCO3  H 2 CO3  2H 2 O (9.77)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

460 Redox processes

O2 (mg/L) NO
3 (mg/L) NO
2 (mg/L) pH HCO
3 (mg/L)
2 4 6 50 100 150 0.1 0.2 5 6 7 8 100 200 300
Depth (m below GWT)

2
4
6
8
10
12

Figure 9.32. Nitrate reduction by organic matter oxidation. Average concentrations in the saturated zone at the
Mussum waterworks, Germany. Modified from Obermann (1982).

Figure 9.32 illustrates the reduction of nitrate by oxidation of organic matter in a sandy aquifer. At
the groundwater table the O2 concentration is already depleted to less than half of the value at air sat-
uration. As predicted thermodynamically, O2 is preferentially reduced compared to nitrate but the
presence of small amounts of nitrite confirms that denitrification takes place as well. The oxidation
of organic matter is reflected by increases in HCO 3 and pH, resulting from the increase in
HCO 3 / H2 CO 3 ratio, predicted by Equation (9.77). As demonstrated in Example 9.10, 60% of the
decrease in nitrate (and O2) between the water table and 8 m depth can be explained by the increase
in total dissolved inorganic carbon, which strongly suggests organic matter to be the predominant
electron donor.

EXAMPLE 9.10. Construct a redox balance for nitrate reduction by organic matter oxidation
Based on the changes in concentrations between the water chemistry at the top of the saturated zone (Figure
9.32) and those at 8 m depth, the following redox balance can be constructed (modified from Obermann,
1982). All concentrations are in mmol/L.

Water table 8 m Depth Difference

NO3 2.18 0.81 1.37
Ca2 3.37 3.24 0.13
HCO 3 3.11 4.42 1.31
TIC 3.75 4.74 0.99
O2 0.11 0.03 0.08

If the decrease in nitrate is caused by oxidation of organic matter according to Reaction (9.77), it should
be balanced by an increase in total dissolved inorganic carbon (TIC) of 5 / 4 1.37  1.71 mmol/L.
The observed TIC amounts to 0.99. Additional processes which affect TIC may be CaCO3 precipita-
tion induced by increasing pH and HCO 3 during nitrate reduction, and also oxidation of organic
matter by O2 that produces an equivalent amount of TIC. Correcting for these reactions yields
TIC  0.99  0.13  0.08  1.04 mmol/L. In other words, 1.04 / 1.71100%  61% of the decrease in
nitrate can be explained by organic matter oxidation while the remainder probably is due to variations in
nitrate input. Sulfate concentrations in the profile are constant so that pyrite oxidation (see next section) is
of no importance.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Nitrate reduction 461

QUESTIONS:
What is the pH effect of Reaction (9.77) when the pH of the NO 3 water is 6.3, 6.9, or 7.3?
ANSWER: At pH  6.3, [HCO 3 ] equals [H 2 CO 3], thus pH increases. At pH  6.9,
1
⁄4[HCO 3 ]  [H2CO3] and Reaction (9.77) is pH neutral. At pH  7.3,
[HCO 3 ]  10 [H2CO3], thus pH decreases.
Is it likely that the pH increase with depth in Figure 9.32 is caused by denitrification?
ANSWER: Yes, down to 6 m depth, the pH tends to the pH-neutral value of 6.9.
Use PHREEQC to estimate the pH at the water table and 8 m depth from HCO 3 and TIC in the table
(temp  10°C)?
ANSWER: pH  7.13 at w.t., pH  7.57 at 8 m (Probably, TIC was higher in the
groundwater, CO2 escaped before analysis).

The gradual decreases of oxygen and nitrate with depth (Figure 9.32) indicates that O2 and
NO 3 reduction rates are slow compared to the downward rate of water transport. The reactivity of
organic matter is probably the overall rate controlling factor.
More precise evidence for the rate of denitrification in aquifer sediment can be obtained with
the acetylene block technique. In the presence of acetylene the bacterial reduction of N2O to
N2 is inhibited. An increase in N2O is much easier to measure than an increase in N2 because the
background level of N2 is so high. In the technique, acetylene is added to sediment incuba-
tions and the denitrifying activity is quantified as the rate of accumulation of N2O. Rates of
denitrification obtained by the acetylene block technique are in the range of 0.08 to 85 mmol/L/yr
(Starr and Gillham, 1993; Smith et al., 1996; Bragan et al., 1997). This broad range again
reflects the huge variability in the reactivity of organic carbon in aquifers. However, the highest
values suggest that the conditions during sediment incubation differ from those in situ in the
aquifer and therefore, yield higher rates (cf. Table 9.6; Question below Figure 9.33).
Figure 9.33 shows the rate of denitrification versus the organic content of the sediment for
incubations in the presence of acetylene. The good correlation of the N2O production rate with
the organic content suggest organic carbon to be the controlling parameter limiting the rate of
denitrification.

2
N2O production rate
(nmol/g/h)

0
0 1 2
Organic content (%)

Figure 9.33. The rate of denitrification versus sediment organic carbon content. Results from sediment
incubations using the acetylene block technique where N2O accumulation gives the rate of denitrification
(Bradley et al., 1992).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

462 Redox processes

QUESTIONS:
Estimate from Figure 9.33 the denitrification rate in mmol C/L/yr for 1% carbon.
ANSWER: 0.7109 (mol/g/hr)  8760 (hr/yr)  1000 (g/kg)  6 (L/kg)  37 mmol/L/yr.
Estimate the age of the organic carbon used by Bradley et al. (1992)?
ANSWER: k  (0.037 mol C/L/yr)/(5 mol C/L)  0.00736/yr. Using Equation (9.61),
t  26 yr.

Another method to identify nitrate reduction is the N2 / Ar method (Wilson et al., 1994; Feast et al.,
1998; Blicher-Mathiesen et al., 1998). The reduction of nitrate produces N2 gas and since N2 is virtu-
ally inert, denitrification should result in excess dissolved N2 in the groundwater. The background con-
centration of N2 in groundwater depends on the temperature at the time of infiltration and the
entrapment of air bubbles during recharge. In order to correct for variations in background N2 concen-
tration, the amount of excess N2 is usually evaluated from the N2/Ar ratio, where argon is used as tracer.
Also nitrogen stable isotopes, 14N / 15N, have been used both to identify the process of denitrification
and the source of nitrate in groundwater (Feast et al., 1998; Wilson et al., 1994; Böttcher et al., 1990).

9.5.2 Nitrate reduction by pyrite and ferrous iron

Organic carbon is not the only electron donor available for the reduction of nitrate in aquifers. Also
the reduction of NO 2
3 by reduced groundwater components, such as Fe , H2S and CH4, is thermo-
dynamically favored. However, the nitrate concentration introduced into the aquifer by fertilizers
usually by far exceeds the reducing capacity of these dissolved species and the required electron
donor must be found in the solid phase. Besides organic matter, pyrite (FeS2) is the other important
solid phase electron donor for nitrate reduction. The energy yield of NO 3 reduction with organic
matter is larger than with pyrite (Figure 9.15) and, thermodynamically, nitrate reduction by organic
matter should occur before the reduction by pyrite. However, the relative sequence of these two reac-
tions is also strongly affected by the reaction kinetics.
Nitrate reduction coupled with pyrite oxidation in aquifers has been reported widely (Kölle et al.,
1983; Strebel et al., 1985; Van Beek et al., 1988, 1989, Postma et al., 1991; Robertson et al., 1996;
Tesoriero et al., 2000). The process involves the oxidation of both sulfur and Fe(2):

5FeS2  14NO
3  4H
 → 7 N  5Fe 2  10SO 2  2H O
2 4 2 (9.78)

and
5 Fe2  NO
3  7 H2O → 5 FeOOH  ⁄2 N2  9 H
1 
(9.79)
The energy gain for sulfide oxidation is larger than for Fe(2) oxidation (Figure 9.15). In the presence of
excess pyrite, Fe2 will therefore remain unoxidized. Inorganic oxidation of pyrite by nitrate does not
seem possible (Postma, unpublished results) and bacterial catalysis is therefore required. Thiobacillus
denitrificans is able to oxidize sulfur in pyrite (Kölle et al., 1987) while Fe2 may be oxidized with
nitrate by Gallionella ferruginea (Gouy et al., 1984) and Escherichia coli (Brons et al., 1991).
Nitrate reduction by pyrite oxidation is illustrated in Figure 9.34 for the sandy Rabis Creek
aquifer. The groundwater in the upper part is derived from forest and heath (Figure 9.30) and
is free of nitrate, while the water in the lower part is derived from arable land and contains nitrate.
The O2 content is near constant and close to air saturation in the oxidized zone. At the redoxcline
oxygen and nitrate disappear simultaneously, and therefore the reduction process must be fast
compared to the downward water transport rate of about 0.75 m/yr. The oxidation of pyrite is
reflected by increases in sulfate and Fe2, in agreement with the pyrite distribution in the sediment.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Nitrate reduction 463

mM mM mM mmol/kg
0.1 0.3 0.2 0.5 0.1 0.3 5 10

45 2
O2 SO4 Fe2 FeS2

40
Elevation (m)

35

30

25

mM mM meq/L mmol/kg
0.5 1.0 0.5 1 2 3 100 200 300

45 
NO3 TIC ∑ anions TOC

40
Elevation (m)

35

30

25

Figure 9.34. Pyrite oxidation by nitrate and oxygen in multilevel sampler T2 (for location see Figure 9.30) in
the saturated zone of the Rabis Creek aquifer, Denmark. The dashed line indicates the depth where O2 disap-
pears and the shaded parts indicate nitrate contaminated water derived from arable land which is overlain by
nitrate free water from a forested area (Modified from Postma et al., 1991).

However, organic carbon (TOC) is also present in the sediment, and in fact at a higher concen-
tration than pyrite. The slight increase in TIC over the redoxcline could indicate the oxidation of
organic carbon. The relative contributions of organic matter and pyrite oxidation can be esti-
mated from an electron balance across the redoxcline (Postma et al., 1991; Tesoriero et al., 2000).
The processes expected are listed in Table 9.8 and the contribution of each half reaction in
the system is estimated by multiplying concentrations with the number of electrons transferred.

Table 9.8. Electron equivalents for dissolved redox species.

Reaction Electron equivalents

NO
3 → 1⁄2N2 5e
5 mNO3
O2 → 2O2 4e 4 mO2
CH2O → CO2 4e 4 TIC
SFeS2 → SO42 7e 7 mSO23
FeFeS2 → FeOOH 1e 0.5 (mSO23  2mFe2)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

464 Redox processes

The electron equivalents obtained in this way are only valid for the specified half reactions and car-
bonates should not dissolve or precipitate, a condition which is fulfilled in the Rabis aquifer. As
shown in Example 9.10, carbonate dissolution/precipitation can to some extent be corrected for by
taking Ca2 concentrations into account.
Electron equivalents are plotted cumulatively for two boreholes in Figure 9.35. T5 displays nitrate
containing groundwater derived from agricultural fields. The disappearance of nitrate over depth is
balanced by the oxidation of pyrite-S(1) as the main electron donor. The contributions of both the
oxidation of organic carbon and of pyrite-Fe(2) are minor. The organic matter present in these sedi-
ments consists of reworked Miocene lignite fragments, and apparently has a very low reactivity.
Borehole T2 shows water from the forested area, free of nitrate, on top of water from arable land
containing nitrate. Compared to the input of only O2, the contamination with nitrate increases the
load of electron acceptor on the aquifer by a factor of five. But even then, the nitrate/pyrite interface
moves downward at a rate of less than 2 cm/yr in the Rabis aquifer (Postma et al., 1991) and even
slower in other systems (Robertson et al., 1996). Apparently the pyrite content of aquifer sediment
may attenuate groundwater nitrate over long periods of time but not indefinitely.
Fe(2) in sediment is another possible electron donor for nitrate reduction, according to the over-
all reaction:

10Fe 2  2 NO
3  14H 2 O → 10FeOOH  N 2  18H
 (9.80)
In sediments, Fe(2) is present in clay minerals, detrital silicates like amphiboles, pyroxenes and
biotite and in magnetite and its precursor green rust. Ernsten et al. (1998) reported Fe(2) in clay min-
erals to be able to reduce nitrate. Postma (1990) found that amphiboles and pyroxenes could reduce
nitrate at low rates, but only in the presence of secondary reaction products, including FeOOH.

Electron equivalents
2 4 6 8 10 12 2 4 6 8 10 12

O2

40 O2 40 NOⴚ
3
Elevation (m)

TIC

NOⴚ
3
2ⴚ
SO4
2 Fe (III) prod. Fe (III) prod.
30 SO4 30

TIC
T2 T5

Figure 9.35. Cumulative distribution of electron equivalents in Rabis Creek multisampler T2 and T5. Electron
equivalents are defined in Table 9.8 (Postma et al., 1991).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Iron reduction and sources of iron in groundwater 465

Hansen et al. (1996) found that green rusts (mixed Fe(2,3)oxides) readily reduced nitrate but
the reaction product was ammonia! Probably the reduction of nitrate by the oxidation of structural
Fe(2) in minerals is most important in fine grained sediments with slow groundwater transport
rates.

QUESTIONS:
Estimate the reactivity of Miocene (107 yr) organic carbon in mmol C/L/yr using Equation (9.61).
ANSWER: k C  0.16 / (107)0.95 (/yr) 100 (mmol C/kg) 6 (kg/L)  2  105 mmol
C/L/yr.
Estimate the time for reducing 1 mM NO 3 by reaction with Miocene OC?
ANSWER: 58,000 yr

9.6 IRON REDUCTION AND SOURCES OF IRON IN GROUNDWATER

Ferrous iron is a common constituent of anoxic groundwater. Its origin may be the partial oxidation
of pyrite, the dissolution of Fe(2) containing minerals, or reductive dissolution of iron oxides. Redox
diagrams (Figures 9.8, 9.10) suggest Fe2 to be the dominant form of dissolved iron in the pH range
of most groundwaters (5–8), since under these conditions Fe3 is insoluble. During drinking water
production, anoxic groundwater containing Fe2 is aerated and Fe-oxyhydroxides precipitate.
Although iron in drinking water is not poisonous and perhaps even beneficial, Fe-oxyhydroxides
may clog distribution systems, and stain clothing and sanitary installations, and iron is therefore
removed during water treatment. In hand-pumped wells, slimy deposits from Fe2-oxidizing bacte-
ria can also be a problem. Most of the following discussion applies for manganese as well. However,
Mn-oxides become reduced at a higher pe than Fe-oxides (Figure 9.10) and manganese is much less
abundant than iron in aquifers.

9.6.1 Iron in aquifer sediments

The distribution of iron in a sandy aquifer sediment is shown in Figure 9.36. The light fraction
(a) contains quartz and feldspars and here iron is present as coatings of iron oxides covering
the grains. Fraction (b) contains primary Fe(2)-bearing silicates such as amphiboles, pyroxenes
or biotite. Finally fraction (c) contains iron containing oxide minerals comprising magnetite and
ilmenite.
Under anoxic conditions, the dissolution of Fe(2)-bearing silicates, including amphiboles and
pyroxenes, and magnetite may release Fe2 to the groundwater. Because the dissolution rates of such
minerals generally are very low, the resulting Fe2 concentration will also be low. Postma and
Brockenhuus Schack (1987) found amphiboles and pyroxenes in a sandy aquifer with distinct disso-
lution features and a Fe2 concentration of 10–13 M. Under oxic conditions the Fe2 released dur-
ing dissolution precipitates as an iron oxyhydroxide coating that inhibits further dissolution (Schott
and Berner, 1983; Ryan and Gschwend, 1992; White, 1990). Common iron oxides and oxyhydrox-
ides found in sediments are ferrihydrite, goethite, lepidocrocite and hematite (Figure 9.9, Cornell
and Schwertmann, 2003). For example, the color of red sandstones is due to thin hematite coatings
covering all sediment grains. In areas of groundwater discharge, Fe2-rich groundwater comes into
contact with atmospheric oxygen and the precipitation of iron oxyhydroxides forms deposits in
streams beds and bog iron ores.
Iron speciation in aquifer sediments can be determined by solid state methods such as Mössbauer
spectroscopy and X-ray diffraction and/or physical separation as shown in Figure 9.36. However, in
most cases iron speciation in sediments is determined using wet chemical extraction methods.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

466 Redox processes

% Fe
0 0.2 0.4 0.6 0.8

10

Depth (m)

20

c
a b
30

Figure 9.36. Cumulative iron distribution (wt%) in an oxidized sandy aquifer based on physical separation
methods: (a) light fraction with quartz and feldspar, (b) weak magnetic fraction containing pyroxenes and
amphiboles, (c) strong magnetic fraction with magnetite and ilmenite (Postma and Brockenhuus-Schack, 1987).

These methods are empirically defined and basically non-mineral specific. Popular methods for
extracting poorly crystalline material, mainly ferrihydrite, comprise extraction with ammonium
oxalate (Schwertmann, 1964) or 0.5 M HCl (Heron and Christensen, 1995). Less reactive iron
oxides, including hematite and goethite, can be extracted with dithionite-citrate-bicarbonate (Mehra
and Jackson, 1960), and Ti-citrate-EDTA-bicarbonate (Ryan and Gschwend, 1991). A method to
quantify the reactivity of iron oxides in sediments is given by Postma (1993).

9.6.2 Reductive dissolution of iron oxides

The reduction of iron oxides by organic matter is of major importance for the evolution of water
quality in pristine aquifers (Jakobsen and Postma, 1999). Also, iron oxides often form an important
redox buffer that may limit the spreading of organic pollutants (Tucillo et al., 1999; Heron and
Christensen, 1995). The overall reaction for the reduction of iron oxide by organic carbon is:
CH 2O  4FeOOH  7H → 4Fe 2  HCO
3  6H 2 O (9.81)
2
The process entails a strong increase in the Fe concentration and a large proton consumption.
Microbial catalysis is important for the reduction of iron oxides by organic matter and the microbes
mediating the reaction must be in direct contact with the iron oxide surface, or excrete complex formers
and reductants that enhance the dissolution (Thamdrup, 2000). Humics have been proposed to function
as electron carriers between iron reducing bacteria and the iron oxide (Nevin and Lovley, 2000).
For oxidants like O2 and nitrate, the overall kinetic control by organic matter fermentation has
been emphasized in Section 9.3.1. When iron is the oxidant, the reactivity of the iron oxide present
in the sediment may exert additional kinetic control. In a sediment containing highly reactive organic
carbon the reactivity of iron oxide is likely to become rate limiting, while in the presence of poorly
reactive organic matter the fermentation step becomes rate limiting. Accordingly, Albrechtsen et al.
(1995) found iron oxide reactivity to limit the rate of iron reduction in the proximal part of a landfill
plume while organic matter limited the rate in the more distal part.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Iron reduction and sources of iron in groundwater 467

The reaction kinetics of inorganic reductive dissolution of iron oxides may be described by the gen-
eral rate equation for crystal dissolution (Postma, 1993):

A  m n
Rk 0   g (c) (4.66)
V  m0 

where R is the overall rate (mol/L/s), A0 the initial surface area (m2), V the solution volume (L), k the spe-
cific rate constant (mol/m2/s), m0 the initial mass of crystals (mol), m the mass of undissolved crystals at
time t (mol), n a constant, and g(c) contains terms representing the influence of the solution composition.
In initial rate experiments a negligible amount of mineral matter is dissolved and (m/m0)n
remains constant (1). Such experiments are used to find the specific rate r  k g(c) and to study
the effect of solution composition (Zinder et al., 1986; Banwart et al., 1989; Hering and Stumm,
1990 and others). Dissolution of iron oxides by the attack of protons proceeds slowly and the rate
depends on the pH of the solution. For goethite, the experiments of Zinder et al. (1986) yield:

r  1011 [H ]0.45 ( mol/m 2/ s) (9.82)

The fractional exponent for [H] may be related to formation of the protonated surface complex, e.g.
Hfo_wOH 2 , but the relation between the rate and this complex is not simply linear (Stumm, 1992;
Problem 9.8).
Organic compounds that form a ligand with iron at the surface of the oxide enhance the dissolution
rate considerably (Figure 9.37). Small amounts of Fe2 in solution accelerate the process even more and
apparently the oxalate-Fe2 complex acts as a catalytic electron shuttle in which Fe2 is regenerated
after dissociation of an iron atom from the crystal lattice (Sulzberger et al., 1989). Also a reductant like
ascorbate facilitates dissolution by electron exchange with surface Fe(3) through an inner sphere com-
plex. The reduction rate is found to increase with the ascorbate concentration until surface saturation
with ascorbate is reached (Banwart et al., 1989). The combination of a ligand and a reductant gave the
highest rate of hematite dissolution (Figure 9.37). In these laboratory experiments relatively simple syn-
thetic organic compounds were used but natural organic compounds, such as phenols, tannic acid and
cysteine are also known to reduce Fe-oxides (LaKind and Stone, 1989; Lovley et al., 1991). Finally,
hydrogen sulfide is probably the most powerful reductant of iron oxides (Yao and Millero, 1996).

50
100 M ascorbate

40 50 M oxalate
[Fe(2)], [Fe(3)] (106 mol/L)

Reductive dissolution
e

30 te te
at
te

ba rba
rb
rba

or co
co

c as
as
co

as

M
as

M 10
M
M

50
0
0

20
10
50

Ligand-promoted dissolution
10 orbate
, 0.0 asc
oxalate
50 M
Proton-assisted dissolution
0
0 10 20 30 40 50
Time (hours)

Figure 9.37. Dissolution of hematite at pH 3, by proton assisted dissolution, ligand (oxalate) promoted disso-
lution, reductant (ascorbate) promoted dissolution and combined ligand and reductant promoted dissolution
(Banwart et al., 1989).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

468 Redox processes

The effect of the iron oxide reactivity on the reduction rate can be evaluated by using a high ascor-
bate concentration. When the surface of the Fe-oxide is saturated with ascorbate, the rate of reduc-
tive dissolution is no longer affected by small variations in the ascorbic acid concentration (Banwart
et al., 1989; Postma, 1993). Figure 9.38 shows the results of iron oxide dissolution in 10 mM ascor-
bate. While ferrihydrite is completely dissolved within two hours, the dissolution of a poorly crys-
talline goethite takes about five days. The solid lines correspond to data fits of the integrated form
of Equation (4.66) (Larsen and Postma, 2001). The effect of iron oxide reactivity is contained in the
specific rate r  k g(c), with g(c) being constant, and the exponent n. The specific rate r varies
from 7.6104/s for ferrihydrite to 5.4106/s for the poorly crystalline goethite (Larsen and
Postma, 2001), or more than two orders of magnitude. The exponent n (Section 4.6.4) was found to
be close to 1 for goethite and lepidocrocite and to vary between 1 and 2.3 for ferrihydrite and indi-
cates a strong decrease in the rate of dissolution as the iron oxide dissolves (Figure 9.38). The sum
these two factors suggest a variation in the rate of iron oxide dissolution of at least three orders of
magnitude, depending on the initial mineralogy and the extent of dissolution.

Goethite
2-line ferrihydrite
Lepidocrocite
m/m0

0.5

0
0 100000 200000 300000 400000 500000
Time (sec)

Figure 9.38. The rate of dissolution of iron oxides in 10 mM ascorbic acid at pH 3. m / m0 is the remaining frac-
tion of mineral mass (Modified from Larsen and Postma, 2001).

The reduction of iron oxide by organic pollutants has been studied extensively at an oil spill
due to a pipeline rupture near Bemidji, Minnesota (Lovley et al., 1989). The spill released soluble
BTEX (Benzene, Toluene, Ethylbenzene, Xylene) compounds to the groundwater, resulting in a
reducing contaminant plume (Baedecker et al., 1993; Tucillo et al., 1999; Cozarelli et al., 2001).
The BTEX reduced iron oxides in the aquifer and created also methanogenic conditions according
to the reactions:

C6H6  30FeOOH  54H → 6HCO

3  30Fe
2  42H O
2 (9.83)

and
8
⁄3 C6H6  18 H2O → 10 CH4  6 HCO
3  6H

(9.84)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Iron reduction and sources of iron in groundwater 469

A Oil body A
425
Elevation above
sea level (m)

40
0.4
420 4.0

Average direction of
groundwater flow mg/L Fe in 1992

415
50 0 50 100 150
Distance from centre of oil body (m)

Figure 9.39. Dissolved ferrous iron (Fe2) contours in groundwater below the Bemidji oil spill. The screens
of observation wells are shown as bars (Cozzarelli et al., 2001).

In the previously aerobic aquifer, the Fe2 concentration reached more than 40 mg/L in 1992,
13 years after the accident (Figure 9.39). The increase in concentrations over time, directly below the
floating oil, is shown in Figure 9.40. The concentrations of Mn2 and Fe2 went up first, but three
years later the CH4 concentration had increased markedly while Fe2 stabilized at about 1 mM.
After 13 years about 30% of HCl extractable iron-oxide in the sediment below the oil had reacted
(8 mmol Fe(3)/kg), and methanogenic conditions appeared to extend further in the aquifer than iron-
reduction. This sequence indicates a gradual depletion of the pool of easily reducible iron-oxides.

10
Site A
Concentration (mM)

1
Fe2ⴙ

0.1

Mn2ⴙ
0.01 CH4

0.001
1984 1985 1986 1987 1988

Figure 9.40. The changing concentrations of Mn2, Fe2 and CH4 with time in groundwater directly below
the Bemidji oil spill (Baedecker et al., 1993).

Reaction (9.83) will result in the precipitation of siderite with the net effect of increasing the pH
while lowering the CO2 pressure:
C6H6  30FeOOH  24H  24HCO
3 → 30FeCO3  42H 2 O (9.85)
To maintain the observed, neutral pH, Reaction (9.85) must be balanced by the acidifying Reaction
(9.84). As the pool of reactive iron oxides becomes depleted, Reaction (9.85) becomes less impor-
tant resulting in an increase of the CO2 pressure and a decrease of pH. In good agreement, the field
data show that, as the increase in Fe2 levels off, there is a slight decrease of pH and an increase of
the CO2 pressure.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

470 Redox processes

EXAMPLE 9.11. Reaction of benzene with iron-oxide and via methanogenesis

Benzene polluting an aquifer may be degraded by iron-reducing and methane generating bacteria. When the
most reactive iron oxides have been removed by reduction, the importance of Fe(3) reduction decreases and that
of methanogenesis increases. In the following PHREEQC input file, benzene degradation occurs at a constant
rate and is controlled by REACTION. The concentration of CH4 is limited to 1 mM, by using SI CH4(g)  0.3,
i.e. excess CH4 is degassing. The dissolution of goethite over a period of 13 years is described by a kinetic rate
law (RATES) similar to Equation (4.66) and its effect is monitored in the lower part of Figure 9.41. Most of the
released Fe2 is reprecipitated as siderite. The reduction of iron oxide is proton consuming (Reaction 9.48) and
causes the initial increase in pH in Figure 9.41. As the reduction of iron oxides diminishes, methanogenesis
takes over, Reaction (9.85) is impeded compared to Reaction (9.84), and the pH must decrease.

0.7 8.5

0.6
Fe2 8
0.5
Fe2 (mM)

7.5
0.4

pH
0.3 7
0.2
6.5
0.1 pH

0 6

10
FeOOH-reduced (mM)

0
1 2 3 4 5 6 7 8 9 10 11 12 13
Years

Figure 9.41. The degradation of benzene by initially iron reduction (lower graph) followed by methano-
genesis. As the latter process becomes dominant the pH decreases. The Fe2 concentration is controlled by
equilibrium with siderite and will in consequence increase.

# Bemidji oil spill... react C6H6 with goethite and/or ferment

# kinetically controlled dissolution of goethite
# Limit CH4 to 1e-3 by degassing, Fe+2 precipitates in siderite
RATES
Goethite
-start
10 moles = 1e-10 * (m/m0) # ...empirical rate
20 save moles * time
-end

KINETICS 1
Goethite; -m0 0.012
-steps 4.1e8 in 13 steps
INCREMENTAL_REACTIONS true

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Iron reduction and sources of iron in groundwater 471

SOLUTION 1
pH 7 charge; -temp 10
Ca 1 Calcite; C 1 CO2(g) -2.

REACTION 1; C6H6 1; 10.0e-3 in 13

EQUILIBRIUM_PHASES 1
Calcite; CH4(g) -0.3 0; Siderite 1.2 0 # Limit CH4 to 1 mM
# SI_Siderite = 1.2

USER_GRAPH
-headings yr Fe+2 go pH
-start
10 graph_x total_time / (3600 * 24 * 365)
20 graph_y tot (*Fe(2)*)*le4, kin("Goethite")*1e3
30 graph_sy -la("H+")
-end
END

EXAMPLE 9.12. pH buffering by pyrite and kinetically dissolving iron oxides

In pH–pe diagrams, the pyrite/iron-oxyhydroxide equilibrium is located at near-neutral pH (Figure 9.10).
This suggests that acid mine drainage leaking into an aquifer may be buffered to neutral pH by the reaction
with these minerals. Pyrite oxidation consumes ferric iron (Fe3), causing the dissolution of iron oxides
which again raises the pH. The following PHREEQC file calculates the reaction when both goethite and fer-
rihydrite are present. We use the rate constant from Equation (9.82) for goethite and a 100 times larger rate
for ferrihydrite (cf. Figure 9.38)
RATES
Ferrihydrite
-start
1 A0 = 54000 * m0 # initial surface area, 600 m2/g
2 SS = (1 - SR("Goethite")/1e3) # 1000 * more soluble than goethite
10 if SS < 0 then goto 30 # dissolve only
20 moles = A0 * (m/m0)^2.3 * 10^-9 * act("H+")^0.45 * SS
30 save moles * time
-end

Goethite
-start
1 A0 = 5400 * m0 # initial surface area, 60 m2/g
2 SS = (1 - SR("Goethite")) # SI = 0 for goethite
10 if SS < 0 then goto 30
20 moles = A0 * (m/m0) * 10^-11 * act("H+")^0.45 * SS
30 save moles * time
-end

SOLUTION 1
pH 1.5; pe 10 O2(g) -0.01
Fe 1. Goethite 1
S 1. charge

EQUILIBRIUM_PHASES 1; Pyrite

KINETICS 1
Ferrihydrite; -m0 0.025; -form FeOOH
Goethite; -m0 0.1
-step 0 2e6 2e6 5e6 5e6 5e6 5e6 5e6 5e6 5e6 5e6 5e6 5e6 5e6 5e6 3e7
INCREMENTAL_REACTIONS true

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

472 Redox processes

USER_GRAPH
-head time Fh Go pH SI_go
-start
10 graph_x total_time/3.15e7
20 graph_y kin("Ferrihydrite"), kin("Goethite")
30 graph_sy -la("H+"), SI("Goethite")
-end
END

0.1 7
Goethite 6
0.08
pH 5

pH or SI
0.06 4
mol/L

0.04 3
Ferrihydrite SIgoethite 2
0.02
1
0 0
0 1 2 3
Time (years)

Figure 9.42. pH buffering by dissolution of iron-oxides in acid mine water. The dissolution of goethite
stops when SIgoethite ! 0 and, because of the higher solubility (Figure 9.9) for ferrihydrite when SIgoethite ! 3.

QUESTION:
Explain what happens when m0 of ferrihydrite is increased to 0.05?

9.7 SULFATE REDUCTION AND IRON SULFIDE FORMATION

Sulfate reduction by organic matter is catalyzed by various strains of bacteria according to the over-
all reaction:

2CH 2O  SO 24 → 2HCO

3  H 2S (9.86)
Here CH2O is used as a simplified representation of organic matter. H2S gives the foul smell of rot-
ting eggs and is also highly toxic. For both reasons H2S is undesirable in drinking water. Sulfate
reduction is a common process in anoxic aquifers (Böttcher et al., 1989; Robertson et al., 1989;
Chapelle and McMahon, 1991; Stoessell et al., 1993; Jakobsen and Postma, 1999). The importance
of sulfate reduction in the aquifer depends on both the availability of reactive organic matter and the
supply of sulfate. Freshwaters are generally low in sulfate. However, important local sources of sul-
fate can be the dissolution of gypsum, the oxidation of pyrite, and mixing of fresh water with sea-
water. Also acid rain, sea spray, fertilizers and waste deposits may be sources of sulfate. Enhanced
circulation around production wells may stimulate sulfate reducing bacteria to grow, clogging the
well screens with iron sulfides (Van Beek and Van der Kooij, 1982). Sulfate reduction is also an
important electron accepting process in landfill plumes (Christensen et al., 2000) and can accom-
pany BTEX pollution (Chapelle et al., 1996; Wisotzky and Eckert, 1997).
Ongoing sulfate reduction is often manifested by the presence of H2S in groundwater (Figure 9.14).
However, the produced H2S may react with Fe-oxides present in the sediment and form iron sulfide min-
erals (see Section 9.7.1). If excess Fe-oxide is available, this may consume even all H2S (Jakobsen and
Postma, 1999). Particularly in aquifers where mixing of freshwater and seawater occurs, the Cl / SO2
4

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sulfate reduction and iron sulfide formation 473

ratio may indicate sulfate reduction. Seawater has a fixed Cl / SO2 4 molar ratio of 19 so that signifi-
cantly higher values in combination with an increased Cl concentration, point to sulfate reduction.
Stable sulfur isotopes are also a suitable indicator since during sulfate reduction, 32S is preferentially
consumed compared to 34S. The
34S value in rain is less than 10‰ and a higher
34S in combination
with decreasing sulfate concentrations (Figure 9.43) must indicate sulfate reduction. Rates of sulfate
reduction in sediments can be measured directly by using a radiotracer method (Jørgensen, 1978).

2
Tritium (TU) SO4 (mg/L) Cl (mg/L)
34SSulfate D.O. (mg/L)
0.1 1 10 100 Age 0 5 10 15 20 25 30 0 0.5 1.0 1.5 0 10 20 30 40 50 0 1 2 3
0
1 1980
2 1970
3
4 1960
5
6 7 1950
Depth (m)

8 9
Sand

10 11

12 12 0.1

14
15
16 1940

18
19 0.3
20

Figure 9.43. Sulfate reduction as indicated by
34S values in a sandy aquifer (Robertson et al., 1989).

Ludvigsen et al. (1998) spiked sediment incubations with 35SO4 to identify sulfate reduction, using
aquifer sediment downstream a landfill. Jakobsen and Postma (1994, 1999) and Cozzarelli et al. (2000)
determined in situ rates of sulfate reduction by injecting a trace amount of 35SO 42 in a sediment
core which was incubated for 24 hours in a borehole, followed by determination of the fraction of
tracer that was reduced. Some results are displayed in Figure 9.44 and show considerable variation.
However, a good correlation exists between the depth range where sulfate is depleted and the occurrence
of high rates of sulfate reduction. At some depths apparently highly reactive organic matter is present and
rates as high as several mM/yr are measured while at other depth ranges there is little or no activity.

Fe2, CH4 (mM) SO42 (mM) SO42 red. rate (mM/year)

0.0 0.4 0.8 0.0 0.1 0.2 0.3 0.0 0.2 0.4 0.6
1

2
Depth (m)

4 Fe2

5
CH4
6

Figure 9.44. Depth distribution of ferrous iron, methane and sulfate and rates of sulfate reduction measured
by the radiotracer method in the sandy aquifer at Rømø, Denmark. The H2S concentration is below the detec-
tion limit (Jakobsen and Postma, 1994).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

474 Redox processes

Sulfate reducing bacteria have a size range of 0.5–1 m and it has been suggested that they are too big
to pass the pore necks of fine grained sediments (Fredrickson et al., 1997). Fermenting organisms, in
contrast, are smaller and a spatial separation of the two processes can result. This is illustrated in Figure
9.45, where fermentation occurs in low permeable clayey aquitard layers, producing formate and acetate
that diffuse into sandy aquifer layers where the dissolved organics are consumed by sulfate reduction.

30
Acetate Formate

40

Aquitard

50 Aquifer
Depth (m)

60

70

80

90
0 10 20 30 40 50 60 70
Concentration (µM)

Figure 9.45. Concentrations of dissolved formate and acetate in pore waters of aquitard and aquifer sedi-
ments, indicating diffusion of dissolved organics into aquifer layers and consumption by sulfate reduction
(McMahon and Chapelle, 1991).

Although pH–pe diagrams (Figures 9.7 and 9.8) suggest that iron reduction takes place at a higher
pe than sulfate reduction, the two reactions may proceed concurrently when the available iron-oxide
has low solubility (Postma and Jakobsen, 1996). An equilibrium reaction for the two redox couples
can be written as:

8Fe 2  SO 24  20H 2O ↔ 8Fe(OH)3  HS  15H (9.87)

The reaction will be displaced to the right when sulfate reduction is favorable, and to the left when
Fe(OH)3 reduction gives more energy. However, the reaction depends also on the activities of the
other reactants and we wish to eliminate some variables to simplify. Sulfate reducing environments
are commonly found to be near equilibrium with FeS (Wersin et al., 1991; Perry and Pedersen, 1993)
and we can write:
H  FeS ↔ Fe 2  HS (9.88)

Reactions (9.87) and (9.88) can be combined in two ways, eliminating Fe2 or HS. Both options may
be valid for natural conditions, since due to the insolubility of FeS, Fe2 and dissolved sulfide normally
are mutually exclusive in anoxic environments. Eliminating HS, the reactions combine to:

9Fe 2  SO 24  20H 2O ↔ 8Fe(OH)3  FeS  16H (9.89)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sulfate reduction and iron sulfide formation 475

with

[H ]16 (9.90)

K( 9.89 ) 
[Fe 2 ]9 [SO 24 ]

The logarithmic form of Equation (9.90) is displayed for different Fe-oxides in Figure 9.46, using pH
and log[Fe2] as variables and at a fixed log[SO2 4 ]  3. For each iron oxyhydroxide, the solid line
indicates equilibrium. On the right side of the equilibrium line sulfate reduction is favored ([Fe2] or
pH is higher than at the assumed equilibrium condition and the reaction tends to proceed to the
right). Similarly, on the left side of the line Fe(3) reduction will take place. In the presence of amor-
phous Fe(OH)3, iron reduction is favored under most aquatic conditions ([Fe2]  103 and
pH  8.5). However, if goethite or hematite are present, sulfate reduction is often more favorable.
Figure 9.46 demonstrates the major influence of pH on the relative stability of iron and sulfate reduc-
tion. In contrast, the effect of [SO2
4 ] variations is small as indicated by the dotted lines for Fe(OH)3.
Figure 9.46 also illustrates that Fe(3) reduction and sulfate reduction thermodynamically
may proceed simultaneously when the most unstable Fe-oxides are absent or have dissolved com-
pletely. Data from natural environments are included in Figure 9.46 where iron and sulfate reduction
were observed to proceed concomitantly. All environments are able to reduce the least stable Fe-
oxides, Fe(OH)3 and lepidocrocite, but compared to goethite and hematite, sulfate reduction is more
favorable except in case (5).

2
log
log

[SO 2 ]
[SO

2 ]
4

3
4

3
5
1
[Fe2]

Fe

4
(O

4 2
H) 3
log

1
Le
pid
oc

7
roc

5 6
5
Go

ite
He

eth
ma

ite
tite

6
5 6 7 8 9
pH

Figure 9.46. Stability relations for simultaneous equilibrium between Fe(3) and sulfate reduction at equilib-
rium with FeS in an Fe2-rich environment. The solid lines correspond to Equation (9.89) at log[SO2 4 ]  3
for different Fe-oxides. Shaded areas reflect field data with the simultaneous occurrence of Fe(3) and sulfate
reduction taken from: 1. Canfield et al. (1993, and pers. comm.), marine sediment, core S4 ! 1 cm depth and
S6 ! 3 cm depth, 2. Leuchs (1988), aquifer, core DGF7 ! 2.5 m depth, 3. Jakobsen and Postma (1999) core 6,
4.45–6 m depth, 4. White et al. (1989), lake sediment, 17 m site, 5. Marnette et al. (1993), lake Kliplo sediment,
2–10 cm, 6. Wersin et al. (1991), lake sediment, 0–10 cm, 7. Simpkins and Parkin (1993), aquifer, Nest 1/4,
6–18 m depth. (Postma and Jakobsen, 1996).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

476 Redox processes

QUESTIONS:
For an enrichment 34SH2S/SO42  27‰, estimate
34S when the SO2 4 concentration decreases from
13 mg/L (
34S  5‰) to 5 mg/L by bacterial reduction (Hint: use the Rayleigh formula)
ANSWER:
34S  30.8‰, cf. Figure 9.43, 15 m depth
From Figure 9.46, estimate log K(9.89) with hematite and ferrihydrite?
ANSWER: 51 (hematite) and 104 (ferrihydrite)
What is difference in solubility of hematite and ferrihydrite in Figure 9.46?
ANSWER: log Kferrihydrite  log Khematite  6.7 (cf. Figure 9.9)

9.7.1 The formation of iron sulfides

Pyrite (cubic FeS2) and its polymorph marcasite (orthorhombic FeS2) are the two most abundant
iron sulfide minerals in ancient sedimentary rocks. Pyrite (Gf0  160.2 kJ/mol) is more
stable than marcasite (Gf0  158.4 kJ/mol) and pyrite is also the form that is normally encoun-
tered in recent environments. In ancient deposits pyrite is particularly abundant in fine
grained sediments, but it may also be present in sand-sized deposits. It occurs as dissemi-
nated single crystals, globular aggregates of crystals with a size of up to several hundreds
microns (Love and Amschutz, 1966) called framboids (Figure 9.12), and as concretions.
Marcasite is less stable, its formation is poorly understood, and it is often found as concretions
in carbonate rocks. The overall pathway of pyrite formation is in most cases initiated by the
reaction:

2FeOOH  3HS → 2FeS  S°  H 2O  3OH (9.91)

This reaction comprises the reduction of iron oxide by H2S which generally is a fast
process (Canfield et al., 1992). While part of the sulfide reduces Fe(3) and produces S°, the
remainder of dissolved sulfide precipitates as FeS. FeS is also called acid volatile sulfide (AVS),
since in contrast to pyrite it readily dissolves in HCl, and this property forms the base of the
analytical assay. FeS stains the sediment black and consist of amorphous FeS and extremely fine
grained minerals like mackinawite (Fe1xS) and greigite (Fe3S4). FeS is less stable than pyrite
but precipitates rapidly and is formed therefore due to the more sluggish precipitation kinetics
of pyrite. The transformation of FeS to FeS2 is an oxidation process and the overall reaction can be
written as:

FeS  S° → FeS2 (9.92)

The process probably consists of a continuous sulfurization of FeS (Schoonen and Barnes, 1991) and
proceeds presumably through a dissolution-precipitation pathway (Wang and Morse, 1996). The oxi-
dation of FeS may also proceed through the reduction of protons according to the reaction (Drobner
et al., 1990; Rickard, 1997).

FeS  H 2S → FeS2  H 2 (9.93)

This reaction enables continuous pyrite formation in a completely anoxic environment. In contrast,
the consumption of S° by Reaction (9.92) is twice its production, relative to FeS, by Reaction (9.91)
and an external oxidant is therefore required to sustain continuous pyrite formation. In some cases,
where dissolved sulfide concentrations remain very low, fast direct formation of pyrite seems also
possible (Howarth, 1979; Howarth and Jørgensen, 1984).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 The formation of methane 477

QUESTION:
Find the H2 concentration where Reaction (9.93) proceeds if H2S  106M? (Use K values from
PHREEQC.DAT).
ANSWER: log K(9.93)  4.43. H2(aq)  0.027 M.

9.8 THE FORMATION OF METHANE

The final stage in the reductive sequence is methane formation. Methane is a common constituent of
anoxic groundwater (Barker and Fritz, 1981; Leuchs, 1988; Grossman et al., 1989; Simkins and Parkin,
1993; Aravena et al., 1995; Zhang et al., 1998). Its presence can be a problem for water supplies due to
its potentially explosive nature (Aravena et al., 1995). Methane is also a common constituent of
groundwater pollution plumes downstream of landfills containing organic waste (Adrian et al., 1994;
Albrechtsen et al., 1999). Finally, methane is the second most important atmospheric greenhouse gas.
The release of methane from wetlands (Avery and Martens, 1999) and landfills (Chanton and Liptay,
2000) contributes significantly to the increase of atmospheric methane (Dlugokencky et al., 1998).
The origin of methane may either be bacterial, produced by microbial activity, or thermogenic. The
latter takes place non-biologically at several kilometers depth and is often related to oil formation.
The produced methane may, however, migrate upwards into aquifers through fracture zones etc.
Distinction between the two types of methane is usually made by comparison of stable carbon and
hydrogen isotopes (Clark and Fritz, 1997). Thermogenic methane has
13C values in the range 50
to 20‰, while bacterial methane has
13C values of less than 50‰ (Whiticar, 1999). Biogenic
methane from landfill sites is reported to have intermediate
13C values. Biogenic methane forms
through a series of complex biogenic reactions (Conrad, 1999; Whiticar, 1999; Vogels et al., 1988). In
general, the predominant processes of methane formation are the reduction of CO2 by free hydrogen:

CO 2  4H 2 → 2H 2O  CH 4 (9.94)

and the fermentation of acetate

CH3COOH → CH 4  CO 2 (9.95)

In the latter reaction the acetate methyl group is transferred directly to CH4. Both Reactions (9.94)
and (9.95) are based on competitive substrate reactions, i.e. methanogenic bacteria have to compete
with sulfate reducing and iron reducing bacteria for the same pools of hydrogen and acetate
(Figure 9.19). There are also non-competitive substrates, such as methylated amines which are fer-
mented directly to CH4 by methanogens (Whiticar, 1999). The relative importance of non-competitive
substrates in natural environments has not yet been clarified.
The pathway of methane formation in sediments has been determined by radiotracer methods.
The sediment is incubated with trace amounts of either 14CO2 or 14CH3COOH and the production
rate of 14CH4 is determined. Both pathways of CH4 production are found to proceed simultaneously
in sediments and their relative importance may vary from close to 100% acetate fermentation
(Phelps and Zeikus, 1984) to close to 100% CO2 reduction (Lansdown et al., 1992). Generally,
acetate fermentation is the primary pathway in lake sediments (Phelps and Zeikus, 1984; Kuivila
et al., 1989) while CO2 reduction tends to dominate in marine sediments (Crill and Martens, 1986;
Hoehler et al., 1994). Sulfate is abundant in seawater, and sulfate reduction is therefore an important
precursor for methanogenesis in marine sediments and at the same time accounts for most acetate
metabolism. Below the zone of sulfate reduction, CO2 reduction will become the dominant pathway
of methanogenesis. In a typical freshwater lake sediment there is little sulfate and therefore more

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

478 Redox processes

reactive organic carbon is available for methanogenesis. Acetate fermentation may then be the dom-
inant pathway (Whiticar, 1999).
The deuterium content of methane is another, albeit rather complicated key for deriving the dom-
inant pathway of methane formation in sediments (Whiticar et al., 1986; Clark and Fritz, 1997;
Whiticar, 1999). In the interpretation of Whiticar, CO2 reduction produces
DCH4 ranges from 250
to 150‰ while fermentation of acetate produces generally lower values. However, Sugimoto and
Wada (1995) found much lower
DCH4 values for CO2 reduction in freshwater sediments. Recently
Waldron et al. (1999) showed that the
DCH4 in low sulfate freshwater environments may be inde-
pendent of the methanogenic pathway but strongly influenced by the
DH2O.
Figure 9.47 shows the depth distribution of methane in the Rømø aquifer. Below 2 m depth,
methane is present in the groundwater and radiotracers were used to determine the source of CH4.
CO2 reduction was found to be the dominant pathway of methane formation. In the depth range
4–5.5 m the rate of CO2 reduction is particularly high with values up to 3.5 mM/yr. However, acetate
fermentation also occurs throughout the profile and in depth intervals where the rate of CO2 reduc-
tion is low, acetate fermentation may contribute with up to 50% of the total methane production rate.

CH4 (mM) CO2 CH4 (mM/Year) Acetate CH4 (mM/Year)

0 0.1 0.2 0.3 0.4 0.5 0.6 0 1 2 3 4 0 0.2 0.4 0.6 0.8
0

3
Depth (m)

Figure 9.47. The distribution of methane and rates and pathways of methane formation in the Rømø aquifer.
Rates of methanogenesis were measured by injecting 14HCO 14
3 or CH3COOH in sediment cores and measur-
ing the production of 14CH4 after incubation (Hansen et al., 2001).

The rate of methane formation is rather variable over depth but this variation is not reflected in the
methane distribution over depth. Once formed, methane is quite stable in anoxic aquifers and it may
be transported further downstream without showing any reaction.
However, when methane enters an aerobic environment, by degassing into soil or a landfill cover,
or by mixing in an aquifer, bacteria will oxidize methane in the presence of O2. Chanton and Liptay
(2000) found that 10–60% of methane emitted from landfills is reoxidized in the overlying cover.
Smith et al. (1991b) studied methane oxidation by O2 in groundwater using a natural gradient
approach and described the results using Michaelis-Menten kinetics. Anaerobic oxidation of methane
by sulfate has in marine sediments been studied by Iversen and Jørgensen (1985) and Hoehler et al.
(1994). However, in aquifer sediments the process was found to be insignificant (Hansen, 1998).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 479

PROBLEMS
9.1 Groundwater samples from wells in the Nadia district of West Bengal show the following composition
(Mukherjee et al., 2000). Concentrations are in micrograms per liter.

Sample pH Eh/mV As(3) As(5)

S1 6.90 152 84.7 88.3

a. How do these samples compare with drinking water limits.

b. Calculate the Eh from dissolved arsenic.
c. Compare with measured Eh values and discuss differences.

9.2 a. List, using Figure 9.10, those Fe and Mn minerals which are stable together. Do the same for combina-
tions of dissolved species and minerals.
b. Calculate and plot in Figure 9.8 the redox boundary between Fe2 and hematite.
c. Drainwater from the Haunstrup browncoal mine, in Denmark, showed the following variations over
time:

8
pH
6
pH

Mn2
2
1
mg/L

2
Fe2
1

71 72 73 74 75
Years

Find two explanations, in terms of pe/pH variations why Mn2 is released earlier than Fe2.
d. An analysis of groundwater gave the following concentrations:

pH NO3 NO2
7.16 110 2.1 (mg/L)

Calculate the amount of Mn2 in water, when the aquifer contains MnO2.

9.3 Construct a pe–pH diagram for Pb2 with the following data:
1) PbO2  4H  2e ↔ 2H2O  Pb2 log K1  49.2
2) Pb2  2e ↔ Pb log K2  4.26
3) PbO  2H ↔ Pb2  H2O log K3  12.7
4) Pb2  3H2O ↔ Pb(OH)
3  3H

log K4  28.1
Use activity levels of Pb2 and Pb(OH)
3 10
6
at the solid/solution boundaries.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

480 Redox processes

9.4 a. Calculate the amount of MnCO3 (rhodochrosite) that can dissolve in water, if PCO2  101.5

MnCO3 ↔ Mn 2  CO32 ; K  109.3

b. Explain the terms “congruent/incongruent” dissolution of a mineral.

c. Which type of rhodochrosite dissolution is expected in a well-aerated soil: congruent dissolution or
incongruent dissolution? Which mineral(s) will form? What is the effect on the PCO 2?

9.5 Write a balanced equation for oxidation of organic material (CH2O) by KMnO4 (i.e. determination of COD)

9.6 Draw a pe/pH-diagram for sulfur-species. Consider S(s), H2S(aq), HS, SO2
4 . Activity of a species in
solution is 10 mmol/L.

1) H2S(aq) ↔ HS  H K  107

 
2) H2S(aq)  4H2O ↔ SO2
4  10H  8e K  1040.6
 
3) S(s)  4H2O ↔ SO2
4  8H  6e K  1036.2

If S precipitates from spring-water, what can be concluded about pH and H2S-content?

9.7 Zinder et al. (1986) measured the rate of dissolution of goethite as function of pH:

pH r, mol/m2/s

1.5 8.72109
2 4.46109
3 1.02109
3.5 8.91109
3.5 1.56109
3.7 2.34109

a. Determine a rate law that includes the pH dependency (cf. Example 7.1, Equation (9.82))
b. Find a rate law that includes the effect of the protonated surface complex, Hfo_wOH
2 . (Hint: regress
on the calculated fraction of the surface complex).
9.8 A threshold concentration of H2 is needed to trigger acetogenesis, methanogenesis and sulfate reduction,
in energy terms amounting to 2 to 4 kJ per transferred electron. By how much must H2 exceed the
equilibrium concentration to correspond to a energy threshold of 3 kJ/e?
9.9 The enthalpy for the reaction 4H2  H CO
3 H

↔ CH4  3H2O is Hr0  237.8 kJ/mol. Estimate
the change in H2 concentration when temperature decreases from 25 to 10°C in methanogenic aquifers?
9.10. The free energy of the reaction CH3COO  4H2O ↔ 4H2  2HCO 
3  H is Gr  144.29 kJ/mol.
0

Plot the concentration of acetate in Example 9.6?

REFERENCES
Adrian, N.R., Robinson, J.A. and Suflita, J.M., 1994. Spatial variability in biodegradation rates as evidenced by
methane production from an aquifer. Appl. Env. Microbiol. 60, 3632–3639.
Albrechtsen, H.-J., Heron, G. and Christensen, T.H., 1995. Limiting factors for microbial Fe(III)-reduction in a
landfill leachate polluted aquifer (Vejen, Denmark). FEMS Microbiol. Ecol. 16, 233–248.
Albrechtsen, H.-J., Bjerg, P.L., Ludvigsen, L., Rügge, K. and Christensen, T.H., 1999. An anaerobic field injec-
tion experiment in a landfill leachate plume, Grindsted, Denmark. 2. Deduction of anaerobic methanogenic,
sulfate- and Fe(III)-reducing redox conditions. Water Resour. Res. 35, 1247–1256.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 481

Alpers, C.N. and Blowes, D.W. (eds), 1994. Environmental geochemistry of sulfide oxidation. ACS Symp. Ser.
550, 681 pp.
Andersen, M.S., Larsen, F. and Postma, D., 2001. Pyrite oxidation in unsaturated aquifer sediments; reaction
stoichiometry and rate of oxidation. Env. Sci. Technol. 35, 4074–4079.
Anthony, R.S., 1977. Iron-rich rhythmically laminated sediments in Lake of the Clouds, northeastern
Minnesota. Limnol. Oceanogr. 22, 45–54.
Aravena, R., Wassenaar, L.I. and Barker, J.F., 1995. Distribution and isotopic characterization of methane in a
confined aquifer in southern Ontario. Can. J. Hydrol. 173, 51–70.
Avery Jr, G.B. and Martens, C.S. 1999. Controls on the stable isotopic composition of biogenic methane pro-
duced in a tidal freshwater estuarine sediment. Geochim. Cosmochim. Acta 63, 1075–1082.
Baedecker, M.J., Cozzarelli, I.M., Siegel, D.I., Bennett, P.C. and Eganhouse, R.P., 1993. Crude oil in a shallow
sand and gravel aquifer. Appl. Geochem. 8, 569–586.
Banwart, S., Davies, S. and Stumm, W., 1989. The role of oxalate in accelerating the reductive dissolution of
hematite (-Fe2O3) by ascorbate. Coll. Surf. 39, 303–309.
Barker, J.F. and Fritz, P., 1981. The occurrence and origin of methane in some groundwater flow systems. Can.
J. Earth Sci. 18, 1802–1816.
Berner, R.A., 1971. Principles of chemical sedimentology. McGraw-Hill, New York, 240 pp.
Berner, R.A., 1981. A new geochemical classification of sedimentary environments. J. Sed. Petrol. 51, 359–365.
Bjerg, P.L., Rügge, K., Pedersen, J.K. and Christensen, T.H., 1995. Distribution of redox-sensitive groundwater
quality parameters downgradient of a landfill. Env. Sci. Technol. 29, 1387–1394.
Blicher-Mathiesen, G., McCarty, G.W. and Nielsen, L.P., 1998. Denitrification and degassing in groundwater
estimated from dissolved dinitrogen and argon. J. Hydrol. 208, 16–24.
Böttcher, J., Strebel, O. and Duynisveld, W.H.M., 1989. Kinetik und Modellierung gekoppelter
Stoffumsetzungen im Grundwasser eines Lockergesteins-Aquifers. Geol. Jb. C 51, 3–40.
Böttcher, J., Strebel, O., Voerkelius, S. and Schmidt, H.L., 1990. Using isotope fractionation of nitrate-
nitrogen and nitrate-oxygen for evaluation of microbial denitrification in a sandy aquifer. J. Hydrol. 114,
413–424.
Boudreau, B.P., 1997. Diagenetic models and their implementation. Springer, Berlin, 414 pp.
Bradley, P.M., Fernandez, M. and Chapelle, F.H., 1992. Carbon limitation of denitrification rates in an anaero-
bic groundwater system. Env. Sci. Technol. 26, 2377–2381.
Bragan, R.J., Starr, J.L. and Parkin, T.B., 1997. Shallow groundwater denitrification rate measurement by acety-
lene block. J. Env. Qual. 26, 1531–1538.
Brons, H.J., Hagen, W.R. and Zehnder, A.J.B., 1991. Ferrous iron dependent nitric oxide production in nitrate
reducing cultures of Escheridia coli. Arch. Microbiol. 155, 341–347.
Brookins, D.G., 1988. Eh-pH diagrams for geochemistry. Springer, Berlin, 176 pp.
Canfield, D.E., Raiswell, R. and Bottrell, S., 1992. The reactivity of sedimentary iron minerals toward sulfide.
Am. J. Sci. 292, 659–683.
Canfield, D.E., Thamdrup, B. and Hansen, J.W., 1993. The anaerobic degradation of organic matter in Danish
coastal sediments: Iron reduction, manganese reduction, and sulfate reduction. Geochim. Cosmochim. Acta
57, 3867–3883.
Champ, D.R., Gulens, J. and Jackson, R.E., 1979. Oxidation-reduction sequences in ground water flow systems.
Can. J. Earth Sci. 16, 12–23.
Chanton, J. and Liptay, K., 2000. Seasonal variation in methane oxidation in a landfill cover soil as determined
by an in situ stable isotope technique. Global Biogeochem. Cycles 14, 51–60.
Chapelle, F.H. and Lovley, D.R., 1992. Competitive exclusion of sulfate reduction by Fe(III)-reducing bacteria:
A mechanism for producing discrete zones of high-iron ground water. Ground Water 30, 29–36.
Chapelle, F.H. and McMahon, P.B., 1991. Geochemistry of dissolved inorganic carbon in a coastal plain aquifer.
1. Sulfate from confining beds as an oxidant in microbial CO2 production. J. Hydrol. 127, 85–108.
Chapelle, F.H., McMahon, P.B., Dubrovsky, N.M., Fujii, R.F., Oaksford, E.T. and Vroblesky, D.A., 1995.
Deducing the distribution of terminal electron-accepting processes in hydrologically diverse groundwater
systems. Water Resour. Res. 31, 359–371.
Chapelle, F.H., Bradley, P.M., Lovley, D.R. and Vroblesky, D.A., 1996. Measuring rates of biodegradation in a
contaminated aquifer using field and laboratory methods. Ground Water 34, 691–698.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

482 Redox processes

Christensen, T.H., Bjerg, P., Banwart, S.A. Jakobsen, R., Heron, G. and Albrechtsen, H.J., 2000.
Characterization of redox conditions in groundwater contaminant plumes. J. Contam. Hydrol. 45, 165–241.
Clark, I.D. and Fritz, P., 1997. Environmental isotopes in hydrogeology. CRC Press, Boca Raton, 328 pp.
Conrad, R., 1999. Contribution of hydrogen to methane production and control of hydrogen concentrations in
methanogenic soils and sediments. FEMS Microbiol. Ecol. 28, 193–202.
Cornell, R.M. and Schwertmann, U., 2003. The Iron Oxides, 2nd ed. Wiley-VCH, Weinheim, 664 pp.
Cozarelli, I.M., Bekins, B.A., Baedecker, M.J., Aiken, G.R., Eganhouse, R.P. and Tucillo, M.E., 2001.
Progression of natural attenuation processes at a crude-oil spill site: 1. Geochemical evolution of the plume.
J. Contam. Hydrol. 53, 369–385.
Cozzarelli, I.M., Suflita, J.M., Ulrich, G.A., Harris, S.H., Scholl, M.A., Schlottmann, J.L. and Christenson, S.,
2000. Geochemical and microbiological methods for evaluating anaerobic processes in an aquifer con-
taminated by landfill leachate. Env. Sci. Technol. 34, 4025–4033.
Crill, P.M. and Martens, C.S., 1986. Methane production from bicarbonate and acetate in an anoxic marine sedi-
ment. Geochim. Cosmochim. Acta 50, 2089–2097.
Dlugokencky, E.J., Masarie, K.A., Lang, P.M. and Tans, P.P., 1998. Continuing decline in the growth rate of the
atmospheric methane burden. Nature 393, 447–450.
Doyle, R.W., 1968. The origin of the ferrous ion-ferric oxide Nernst potential in environments containing fer-
rous iron. Am. J. Sci. 266, 840–859.
Drever, J.I., 1997. The geochemistry of natural waters. 3rd ed. Prentice Hall, Englewood-Cliffs, 436 pp.
Drobner, E., Huber, H., Wächtershauser, G., Rose, D. and Stetter, K.O., 1990. Pyrite formation linked with
hydrogen evolution under anaerobic conditions. Nature 346, 742–744.
Dubrovsky, N.M., Morin, K.A., Cherry, J.A. and Smyth, D.J.A., 1984. Uranium tailings acidification and sub-
surface contaminant migration in a sand aquifer. Water Poll. Res. J. Can. 19, 55–89.
Elberling, B. and Nicholson, R.V., 1996. Field determination of sulphide oxidation rates in mine tailings. Water
Resour. Res. 32, 1773–1784.
Elberling, B., Larsen, F., Christensen, S. and Postma, D., 1998. Gas transport in a confined unsaturated zone
during atmospheric pressure cycles. Water Resour. Res. 34, 2855–2862.
Ernsten, V., Gates, W.P. and Stucki, J.W., 1998. Microbial reduction of structural iron in clay – A renewable
source of reduction capacity. J. Env. Qual. 27, 761–766.
Evangelou, V.P. and Zhang, Y.L., 1995. A review: Pyrite oxidation mechanisms and acid mine drainage preven-
tion. Crit. Rev. Env. Sci. Technol. 25, 141–199.
Fredrickson, J.K., McKinley, J.P., Bjornstad, B.N., Long, P.E., Ringelberg, D.B., White, D.C., Krumholz, L.R.,
Suflita, J.M., Colwell, F.S., Lehman, R.M. and Phelps, T.J., 1997. Pore-size constraints on the activity and
survival of subsurface bacteria in a late Cretaceous shale-sandstone sequence, northwestern New Mexico.
Geomicrobiol. J. 14, 183–202.
Feast, N.A., Hiscock, K.M., Dennis, P.F. and Andrews, J.N., 1998. Nitrogen isotope hydrochemistry and deni-
trification within the chalk aquifer system of north Norfolk, UK. J. Hydrol. 211, 233–252.
Garrels, R.M. and Christ, C.L., 1965. Solutions, Minerals, and Equilibria. Harper and Row, New York, 450 pp.
Glynn, P. and Brown, J., 1996. Reactive transport modeling of acidic metal-contaminated ground water at a site
with sparse spatial information. Rev. Mineral. 34, 377–438.
Gouy, J., Bergé, P. and Labroue, L., 1984. Gallionella ferruginea, facteur de dénitrification dans les eaux pau-
vre en matière organique. C. R. Acad. Sc. Paris 298, 153–156.
Grenthe, I., Stumm, W., Laaksuharju, M., Nilson, A.-C. and Wikberg, P., 1992. Redox potentials and redox reac-
tions in deep groundwater systems. Chem. Geol. 98, 131–150.
Grossman, E.L., Coffman, B.K., Fritz, S.J. and Wada, H., 1989. Bacterial production of methane and its influ-
ence on ground-water chemistry in east-central Texas aquifer. Geology 17, 495–499.
Hansen, H.C., Koch, C.B., Krogh, H.N., Borggaard, O.K. and Sørensen, J., 1996. Abiotic nitrate reduction to
ammonium: key role of green rust. Env. Sci. Technol. 30, 2053–2056.
Hansen, L.K., 1998. Biogeochemistry of methane in a shallow sandy aquifer. Ph.D. thesis, Lyngby, 111 pp.
Hansen, L.K., Jakobsen, R. and Postma, D., 2001. Methanogenesis in a shallow sandy aquifer, Rømø, Denmark,
Geochim. Cosmochim. Acta 65, 2925–2935.
Herbert, R.B., 1999. Sulfide oxidation in mine waste deposits – A review with emphasis on dysoxic weathering.
MIMI 1999:1, 45 pp. (www.mimi.kiruna.se).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 483

Hering, J.G. and Stumm, W., 1990. Oxidation and reductive dissolution of minerals. In M.F. Hochella and A.F.
White (eds), Mineral water interface geochemistry. Rev. Mineral. 23, 427–465.
Heron, G. and Christensen, T.H., 1995. Impact of sediment-bound iron on redox buffering in a landfill leachate
polluted aquifer (Vejen, Denmark). Env. Sci. Technol. 29, 187–192.
Hoehler, T.M., Alperin, M.J., Albert, D.B. and Martens, C.S., 1994. Field and laboratory studies of methane oxi-
dation in an anoxic marine sediment. Global Biogeochem. Cycles 8, 451–463.
Hoehler, T.M., Alperin, M.J., Albert, D.B. and Martens, C.S., 1998. Thermodynamic control on H2 concentra-
tions in anoxic sediments. Geochim. Cosmochim. Acta 62, 1745–1756.
Howarth, R.W., 1979. Pyrite: Its rapid formation in a salt marsh and its importance in ecosystem metabolism.
Science 203, 49–51.
Howarth, R.W. and Jørgensen, B.B., 1984. Formation of 35S-labelled elemental sulfur and pyrite in coastal
marine sediments (Limfjorden and Kysing Fjord, Denmark) during short-term 35SO2 4 reduction measure-
ments. Geochim. Cosmochim. Acta 48, 1807–1818.
Iversen, N. and Jørgensen, B.B., 1985. Anaerobic methane oxidation rates at the sulfate-methane transition in
marine sediments from Kattegat and Skagerrak (Denmark). Limnol. Oceanogr. 30, 944–955.
Jakobsen, R. and Postma, D., 1994. In situ rates of sulfate reduction in an aquifer (Rømø, Denmark) and impli-
cations for the reactivity of organic matter. Geology 22, 1103–1106.
Jakobsen, R. and Postma, D., 1999. Redox zoning, rates of sulfate reduction and interactions with Fe-reduction
and methanogenesis in a shallow sandy aquifer, Rømø, Denmark. Geochim. Cosmochim. Acta 63, 137–151.
Jakobsen, R., Albrechtsen, H.J., Rasmussen, M., Bay, H., Bjerg, P.L. and Christensen, T.H., 1998. H2 concen-
trations in a landfill leachate plume (Grindsted, Denmark): In situ energetics of terminal electron acceptor
processes. Env. Sci. Technol. 32, 2142–2148.
Jin, Y. en Jury, W.A., 1996. Characterizing the dependence of gas diffusion coefficient on soil properties. Soil
Sci. Soc. Am. J. 60, 66–71.
Jørgensen, B.B., 1978. A comparison of methods for the quantification of bacterial sulfate reduction in coastal
marine sediments. I. Measurement with radiotracer techniques. Geomicrobiol. J. 1, 11–27.
Kirby, C.S. and Elder Brady, J.A., 1998. Field determination of Fe2+ oxidation in acid mine drainage using a
continuously-stirred tank reactor. Appl. Geochem. 13, 509–520.
Korom, S.F., 1992. Natural denitrification in the saturated zone: a review. Water Resour. Res. 28, 1657–1668.
Kölle, W., Strebel, O. and Böttcher, J., 1987. Reduced sulfur compounds in sandy aquifers and their interactions
with groundwater. Int. Symp. Groundwater Monitoring, Dresden, 12 pp.
Kölle, W., Werner, P., Strebel, O. and Böttcher, J., 1983. Denitrifikation in einem reduzierenden
Grundwasserleiter. Vom Wasser 61, 125–147.
Krauskopf, K.B., 1979. Introduction to Geochemistry. McGraw-Hill, New York, 617 pp.
Krumbein, W.E. (ed.), 1983. Microbial Geochemistry. Blackwell, Oxford, 330 pp.
Kuivila, K.M., Murray, J.W., Devol, A.H. and Novelli, P.C., 1989. Methane production, sulfate reduction and
competition for substrates in the sediments of Lake Washington. Geochim. Cosmochim. Acta 53, 409–416.
LaKind, J.S. and Stone, A.T., 1989. Reductive dissolution of goethite by phenolic reductants. Geochim.
Cosmochim. Acta 53, 961–971.
Langmuir, D., 1969. The Gibbs free energies of substances in the system Fe-O2-H2O-CO2 at 25 °C. U.S. Geol.
Surv. Prof. Paper, 650-B, B180–B184.
Lansdown, J.M., Quay, P.D. and King, S.L., 1992. CH4 production via CO2 reduction in a temperate bog: A
source of 13C-depleted CH4. Geochim. Cosmochim. Acta 56, 3493–3503.
Larsen, O. and Postma, D., 2001. Kinetics of reductive bulk dissolution of lepidocrocite, ferrihydrite and
goethite. Geochim. Cosmochim. Acta 65, 1367–1379.
Lawrence, A.R., Gooddy, D.C., Kanatharana, P., Meeslip, W. and Ramnarong, P., 2000. Groundwater evolution
beneath Hat Yai, a rapidly developing city in Thailand. Hydrogeol. J. 8, 564–575
Leuchs, W., 1988. Vorkommen, Abfolge und Auswirkungen anoxischer redoxreaktionen in einem pleistozanen
Porengrundwasserleiter. Bes. Mitt. dt. Gewässerk. Jb. 52, 106 pp.
Lindberg, R.D. and Runnels, D.D., 1984. Ground water redox reactions: An analysis of equilibrium state applied
to Eh measurements and geochemical modeling. Science 225, 925–927.
Love, L.G. and Amstutz, G.C., 1966. Review of microscopic pyrite from the Devonian Chattanooga Shale and
Rammelsberg Banderz. Fortschr. Miner. 43, 273–309.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

484 Redox processes

Lovley, D.R. and Goodwin, S., 1988. Hydrogen concentrations as an indicator of the predominant terminal electron-
accepting reactions in aquatic sediments. Geochim. Cosmochim. Acta 52, 2993–3003.
Lovley, D.R., Baedecker, M.J., Lonergan, D.J., Cozzarelli, I.M., Phillips, E.J.P. and Siegel, D.I., 1989. Oxidation
of aromatic compounds coupled to microbial iron reduction. Nature 339, 297–299.
Lovley, D.R., Phillips, E.J.P. and Lonergan, D.J., 1991. Enzymatic versus nonenzymatic mechanisms for Fe(III)
reduction in aquatic sediment. Env. Sci. Technol. 25, 1062–1067.
Lowson, R.T., 1982. Aqueous oxidation of pyrite by molecular oxygen. Chem. Rev. 82, 461–497.
Ludvigsen, L., Albrechtsen, H.-J., Heron, G., Bjerg, P.L. and Christensen, T.H., 1998. Anaerobic microbial redox
processes in a landfill leachate contaminated aquifer (Grindsted, Denmark). J. Contam. Hydrol. 4, 1–26.
Lyngkilde, J. and Christensen, T.H., 1992. Redox zones of a landfill leachate pollution plume (Vejen,
Denmark). J. Contam. Hydrol. 10, 273–289.
Macalady, D.L., Langmuir, D., Grundl, T. and Elzerman, A., 1990. Use of model-generated Fe3+ ion activities
to compute Eh and ferric oxyhydroxide solubilities in anaerobic systems. In D.C. Melchior and R.L. Basset
(eds), Chemical modelling of aqueous systems II. ACS Symp. Ser. 416, 350–367.
Magaritz, M. and Luzier, J.E., 1985. Water-rock interactions and seawater-freshwater mixing effects in the
coastal dunes aquifer, Coos Bay, Oregon. Geochim. Cosmochim. Acta 49, 2515–2525.
Marnette, E.C.L., Van Breemen, N., Hordijk, K.A. and Cappenberg, T.E., 1993. Pyrite formation in two fresh-
water systems in the Netherlands. Geochim. Cosmochim. Acta 57, 4165–4177.
Massmann, J. and Farrier, D.F., 1992. Effects of atmospheric pressure on gas transport in the vadose zone. Water
Resour. Res. 28, 777–791.
McMahon, P.B. and Chapelle, F.H., 1991. Microbial production of organic acids in aquitard sediments and its
role in aquifer geochemistry. Nature 349, 233–235.
Mehra, O.P. and Jackson, M.L., 1960. Iron oxide removal from soils and clays by a dithionite-citrate sytem
buffered with sodium bicarbonate. Clays Clay Min. 5, 317–327.
Middelburg, J.J., 1989. A simple rate model for organic matter decomposition in marine sediments. Geochim.
Cosmochim. Acta 53, 1577–1581.
Morin, K.A. and Cherry, J.A., 1986. Trace amounts of siderite near a uranium-tailings impoundment, Elliot
Lake, Ontario, Canada, and its implication in controlling contaminant migration in a sand aquifer. Chem.
Geol. 56, 117–134.
Moses, C.O., Nordstrom, D.K., Herman, J.S. and Mills, A.L., 1987. Aqueous pyrite oxidation by dissolved oxy-
gen and ferric iron. Geochim. Cosmochim. Acta 51, 1561–1572.
Mukherjee, M., Sahu, S.J., Roy, D., Jana., J., Bhattacharya, R., Chatterjee, D., De Dalal, S.S., Bhattacharya, P.
and Jacks, G., 2000. The governing geochemical processes responsible for mobilisation of arsenic in sedi-
mentary aquifer of Bengal Delta Plain. In P.L. Bjerg., P. Engesgaard and Th.D. Krom (eds), Groundwater
2000, 201–202, Balkema, Rotterdam.
Nevin, K.P. and Lovley, D.R., 2000. Potential for nonenzymatic reduction of Fe(III) via electron shuttling in
subsurface sediments. Env. Sci. Technol. 34, 2472–2478.
Nicholson, R.V., Gillham, R.W. and Reardon, E.J., 1988. Pyrite oxidation in carbonate-buffered solution: 1.
Experimental kinetics. Geochim. Cosmochim. Acta 52, 1077–1085.
Nicholson, R.V., Gillham, R.W. and Reardon, E.J., 1990. Pyrite oxidation in carbonate-buffered solution: 2.
Rate control by oxide coatings. Geochim. Cosmochim. Acta 54, 395–402.
Nordstrom, D.K., 1982. Aqueous pyrite oxidation and the consequent formation of secondary iron minerals. In
D.K. Nordstrom (ed.), Acid Sulphate Weathering. Soil Sci. Soc. Am. Spec. Pub. 10, 38–56.
Nordstrom, D.K. and Munoz, J.L., 1994. Geochemical thermodynamics, 2nd ed. Blackwell, Oxford, 493 pp.
Nordstrom, D.K., Jenne, E.A. and Ball, J.W., 1979. Redox equilibria of iron in acid mine waters. In E.A. Jenne
(ed.), Chemical Modelling in Aqueous Systems. ACS Symp. Ser. 93, 49–79.
Nordstrom, D.K., Alpers, C.N., Ptacek, C.J. and Blowes, D.W., 2000. Negative pH and extremely acidic mine
waters from Iron Mountain, California. Env. Sci. Technol. 34, 254–258.
Obermann, P., 1982. Hydrochemische/hydromechanische Untersuchungen zum Stoffgehalt von Grundwasser
bei landwirtschaftlicher Nutzung. Bes. Mitt. dt. Gewässerk. Jb. 42.
Perry, K.A. and Pedersen, T.F., 1993. Sulphur speciation and pyrite formation in meromictic ex-fjords.
Geochim. Cosmochim. Acta 57, 4405–4418.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 485

Phelps, T.J. and Zeikus, J.G., 1984. Influence of pH on terminal carbon metabolism in anoxic sediments from a
mildly acidic lake. Appl. Env. Microbiol. 48, 1088–1095.
Plummer, L.N., 1977. Defining reactions and mass transfer in part of the Floridan aquifer. Water Resour. Res.
13, 801–812.
Postma, D., 1977. The occurrence and chemical composition of recent Fe-rich mixed carbonates in a river bog.
J. Sed. Petrol. 47, 1089–1098.
Postma, D., 1981. Formation of siderite and vivianite and the pore-water composition of a recent bog sediment
in Denmark. Chem. Geol. 31, 255–244.
Postma, D., 1982. Pyrite and siderite formation in brackish and freshwater swamp sediments. Am. J. Sci. 282,
1151–1183.
Postma, D., 1983. Pyrite and siderite oxidation in swamp sediments. J. Soil Sci. 34, 163–182.
Postma, D., 1990. Kinetics of nitrate reduction by detrital Fe(II)-silicates. Geochim. Cosmochim. Acta 54, 903–908.
Postma, D., 1993. The reactivity of iron oxides in sediments: A kinetic approach. Geochim. Cosmochim. Acta 57,
5027–5034.
Postma, D. and Brockenhuus-Schack, B.S., 1987. Diagenesis of iron in proglacial sand deposits of late- and
post-Weichselian age. J. Sed. Petrol. 57, 1040–1053.
Postma, D. and Jakobsen, R., 1996. Redox zonation: Equilibrium constraints on the Fe(III)/SO4-reduction inter-
face. Geochim. Cosmochim. Acta. 60, 3169–3175.
Postma, D., Boesen, C., Kristiansen, H. and Larsen, F., 1991. Nitrate reduction in an unconfined sandy aquifer:
Water chemistry, reduction processes, and geochemical modeling. Water Resour. Res. 27, 2027–2045.
Ptacek, C.J. and Blowes, D.W., 1994. Influence of siderite on the pore-water chemistry of inactive mine-tailings
impoundments. In C.N. Alpers and D.W. Blowes (eds), Environmental Geochemistry of sulfide oxidation.
ACS Symp. Ser. 550, 172–189.
Rickard, D., 1997. Kinetics of pyrite formation by the H2S oxidation of iron(II)monosulfide in aqueous solu-
tions between 25 and 125°C: The rate equation. Geochim. Cosmochim. Acta 61, 115–134.
Rimstidt, J.D. and Vaughan, D.J., 2003. Pyrite oxidation: A state-of-the-art assessment of the reaction mecha-
nism. Geochim. Cosmochim. Acta 67, 873–880.
Robertson, W.D., Cherry, J.A. and Schiff, S.L., 1989. Atmospheric sulfur deposition 1950–1985 inferred from
sulfate in groundwater. Water Resour. Res. 25, 1111–1123.
Robertson, W.D., Russel, B.M. and Cherry, J.A., 1996. Attenuation of nitrate in aquitard sediments of southern
Ontario, J. Hydrol. 180, 267–281.
Russell, E.W., 1973. Soil conditions and plant growth. Longman, London, 849 pp.
Ryan, J.N. and Gschwend, P.M., 1991. Extraction of iron oxides from sediments using reductive dissolution by
titanium(III). Clays Clay Min. 39, 509–518.
Ryan, J.N. and Gschwend, P.M., 1992. Effect of iron diagenesis on the transport of collodial clay in an uncon-
fined sand aquifer. Geochim. Cosmochim. Acta 56, 1507–1521.
Schoonen, M.A.A. and Barnes, H.L., 1991. Reactions forming pyrite and marcasite from solution: II Via FeS
precursor below 100°C. Geochim. Cosmochim. Acta 55, 1505–1514.
Schott, J. and Berner, R.A., 1983. X-ray photoelectron studies of the mechanism of iron silicate dissolution dur-
ing weathering. Geochim. Cosmochim. Acta 47, 2233–2240.
Schwertmann, U., 1964. Differenzierung der Eisenoxide des Bodens durch Extraktion mit Ammoniumoxalat-
lösung. Z. Pflanzenern. Bodenk. 105, 194–202.
Simpkins, W.W. and Parkin, T.B., 1993. Hydrogeology and redox geochemistry of CH4 in a late Wisconsinan till
and loess sequence in central Iowa. Water Resour. Res. 29, 3643–3657.
Smedley, P.L., Nicolli, H.B., Barros, A.J. and Tullio, J.O., 1998. Origin and mobility of arsenic in groundwater
from the Pampean Plain. In G.B. Arehart and J.R. Hulston (eds), Proc. 9th Water Rock Interaction Symp.,
275–278, Balkema, Rotterdam.
Smith, R.L. and Duff, J.H., 1988. Denitrification in a sand and gravel aquifer. Appl. Env. Microbiol. 54,
1071–1078.
Smith, R.L., Howes, B.L. and Duff, J.H., 1991a. Denitrification in nitrate-contaminated groundwater: occur-
rence in steep vertical geochemical gradients. Geochim. Cosmochim. Acta 55, 1815–1825.
Smith, R.L., Howes, B.L. and Garabedian, S.P., 1991b. In situ measurement of methane oxidation in ground-
water by using natural-gradient tracer tests. Appl. Env. Microbiol. 57, 1997–2004.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

486 Redox processes

Smith, R.L., Garabedian, S.P. and Brooks, M. H., 1996. Comparison of denitrification activity measurements in
groundwater using cores and natural-gradient tracer tests. Env. Sci. Technol. 30, 3448–3456.
Spalding, R.F. and Exner, M.E., 1993. Occurrence of nitrate in groundwater – a review. J. Env. Qual. 22,
392–402.
Starr, R.C. and Gillham, R.W., 1993. Denitrification and organic carbon availability in two aquifers. Ground
Water 31, 934–947.
Stoessell, R.K., Moore, Y.H. and Coke, J.G., 1993. The occurrence and effect of sulfate reduction and sulfide
oxidation on coastal limestone dissolution in Yucatan cenotes. Ground Water 31, 566–575.
Strebel, O., Böttcher, J. and Kölle, W., 1985. Stoffbilanzen im Grundwasser eines Einzugsgebietes als
Hilfsmittel bei Klärung und Prognose von Grundwasserqualitätsproblemen (Beispiel Fuhrberger Feld). Z.
dt. geol. Ges. 136, 533–541.
Strebel, O., Duynisveld, W.H.M. and Böttcher, J., 1989. Nitrate pollution of groundwater in western Europe.
Agr. Ecosys. Env. 26, 189–214.
Stumm, W., 1992. Chemistry of the solid-water interface. Wiley and Sons, New York, 428 pp.
Stumm, W. and Morgan, J.J., 1996. Aquatic chemistry, 3rd ed. Wiley and Sons, New York, 1022 pp.
Sugimoto, A. and Wada, E., 1995. Hydrogen isotopic composition of bacterial methane: CO2/H2 reduction and
acetate fermentation. Geochim. Cosmochim. Acta 59, 1329–1337.
Sulzberger, B., Suter, D., Siffert, C., Banwart, S. and Stumm, W., 1989. Dissolution of Fe(III)-(hydro)oxides in
natural waters: laboratory assessment on the kinetics controlled by surface coordination. Mar. Chem. 28,
127–144.
Tesoriero, A.J., Liebscher, H. and Cox, S.E., 2000. Mechanism and rate of denitrification in an agricultural
watershed: Electron and mass balance along groundwater flow paths. Water Resour. Res. 36, 1545–1559.
Thamdrup, B., 2000. Bacterial manganese and iron reduction in aquatic sediments. In B. Schink (ed.), Adv.
Microbial Ecol. 16, 41–84.
Thorstenson, D.C. and Pollock, D.W., 1989. Gas transport in unsaturated zones: Multicomponent systems and
the adequacy of Fick’s laws. Water Resour. Res. 25, 477–507.
Thorstenson, D.C., Fisher, D.W. and Croft, M.G., 1979. The geochemistry of the Fox Hill-basal Hell Creek
aquifer in the southwestern North Dakota and northwestern South Dakota. Water Resour. Res. 15,
1479–1498.
Tuccillo, M.E., Cozzarelli, I.M. and Herman, J.S., 1999. Iron reduction in the sediments of a hydrocarbon-
contaminated aquifer. Appl. Geochem. 14, 655–667.
Van Beek, C.G.E.M. and Van der Kooij, D., 1982. Sulfate-reducing bacteria in ground water from clogging and
nonclogging shallow wells in the Netherlands river region. Ground Water, 20, 298–302.
Van Beek, C.G.E.M., Boukes, H., Van Rijsbergen, D. and Straatman, R., 1988. The threat of the Netherlands
waterworks by nitrate in the abstracted groundwater, as demonstrated on the well field Vierlingsbeek. Water
Supply 6, 313–318.
Van Beek, C.G.E.M., Hettinga, F.A.M. and Straatman, R., 1989. The effects of manure spreading and acid dep-
osition upon groundwater quality in Vierlingsbeek, the Netherlands. IAHS Pub. 185, 155–162.
Van Berk, W. and Wisotzky, F., 1995. Sulfide oxidation in brown coal overburden and chemical modeling of
reactions in aquifers influenced by sulfide oxidation. Env. Geol. 26, 192–196.
Van Breemen, N., 1976. Genesis and solution chemistry of acid sulfate soils in Thailand. Agric. Res. Rep.,
Wageningen, 263 pp.
Vogels, G.D., Keltjens, J.T. and Van der Drift, C., 1988. Biochemistry of methane production. In A.J.B. Zehnder
(ed.), op. cit., 707–770.
Vroblesky, D.A. and Chapelle, F.H., 1994. Temporal and spatial changes of terminal electron-accepting
processes in a petroleum hydrocarbon-contaminated aquifer and the significance of contaminant biodegra-
dation. Water Resour. Res. 30, 1561–1570.
Waldron, S., Landsdown, J.M., Scott, E.M., Fallick, A.E. and Hall, A.J., 1999. The global influence of the
hydrogen isotope composition of water on that of bacteriogenic methane from shallow freshwater environ-
ments. Geochim. Cosmochim. Acta 63, 2237–2245.
Walton-Day, K., Macalady, D.L., Brooks, M.H. and Tate, V.T., 1990. Field methods for measurement of ground
water redox chemical parameters. Ground Water Mon. Rev. 10, 81–89.
Wang, Q. and Morse, J.W., 1996. Pyrite formation under conditions approximating those in anoxic sediments:
I. Pathway and morphology. Mar. Chem. 52, 99–121.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 487

Wersin, P., Höhener, P., Giovanoli, R. and Stumm, W., 1991. Early diagenetic influences on iron transformations
in a freshwater lake sediment. Chem. Geol. 90, 233–252.
White, A.F., 1990. Heterogeneous electrochemical reactions associated with oxidation of ferrous oxide and sil-
icate surfaces. Rev. Mineral. 23, 467–509.
White, J.R., Gubala, C.P., Fry, B., Owen, J. and Mitchell, M.J., 1989. Sediment biogeochemistry of iron and sul-
fur in an acidic lake. Geochim. Cosmochim. Acta 53, 2547–2559.
Whiticar, M.J., 1999. Carbon and hydrogen isotope systematics of bacterial formation and oxidation of
methane. Chem. Geol. 161, 291–314.
Whiticar, M.J., Faber, E. and Schoell, M., 1986. Biogenic methane formation in marine and freshwater envi-
ronments: CO2 reduction vs. acetate fermentation-isotope evidence. Geochim. Cosmochim. Acta 50,
693–709.
Wiese, R.G., Powell, M.A. and Fyfe, W.S., 1987. Spontaneous formation of hydrated iron sulfates on laboratory
samples of pyrite- and marcasite-bearing coals. Chem. Geol. 63, 29–38.
Williamson, M.A. and Rimstidt, J.D., 1994. The kinetics and electrochemical rate-determining step of aqueous
pyrite oxidation. Geochim. Cosmochim. Acta 58, 5443–5454.
Wilson, G.B., Andrews, J.N. and Bath, A.H., 1994. The nitrogen isotope composition of groundwater nitrates
from the East Midlands Triassic Sandstone aquifer, England. J. Hydrol. 157, 35–46.
Winograd, I.J. and Robertson, F.N., 1982. Deep oxygenated ground water: Anomaly or common occurrence.
Science 216, 1227–1230.
Wisotzky, F., 1994. Untersuchungen zur Pyritoxidation in Sedimenten des Rheinischen Braunkohlenreviers und
deren Auswirkungen af die Chemie des Grundwassers. Bes. Mitt. dt. Gewässerk. Jb. 58, 153 pp.
Wisotzky, F. and Eckert, P., 1997. Sulfat-dominierter BTEX-Abbau im Grundwasser eines ehemaligen
Gaswerksstandortes. Grundwasser 2, 11–20.
Yao, W. and Millero, F.J., 1996. Oxidation of hydrogen sulfide by hydrous Fe(III) oxides in seawater. Mar.
Chem. 52, 1–16.
Zehnder, A.J.B. (ed.), 1988. Biology of anaerobic microorganisms. Wiley and Sons, New York, 872 pp.
Zhang, C., Grossman, E.L. and Ammerman, J.W., 1998. Factors influencing methane distribution in Texas
ground water. Ground Water 36, 58–66.
Zinder, B., Furrer, G. and Stumm, W., 1986. The coordination chemistry of weathering: II Dissolution of Fe(III)
oxides. Geochim. Cosmochim. Acta 50, 1861–1869.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

10

Pollution by Organic Chemicals

We cannot think of a modern society without manufactured organic chemicals. Their production and
application is now regulated and spilling in the environment happens, at least in principle, only by acci-
dent. However, a few decades ago dumping of industrial organic chemicals at a waste site was common
practice, storage tanks with gasoline and petrol were usually leaking into the underlying soil, and agri-
cultural chemicals were spread over the land without considering that groundwater might be polluted
by their application and rendered undrinkable. Figure 10.1 shows the number of wells in Denmark that
were closed because the concentrations of nitrates, pesticides or other chemicals exceeded drinking
water limits. The large increase since 1993 of wells abandoned due to pesticides is highly disturbing.
The pollutants of former practice and of recent accidents may still reside in the subsoil and will
not have traveled far, given that the flow velocity of groundwater is usually limited to a few meters
per year. To describe and model the hazards of transport of organic chemicals into and with ground-
water, we consider in this chapter (Figure 10.2; Tiktak et al., 2000) volatilization as a mechanism that
reduces the mass that remains in the soil (but it may pollute the atmosphere and cause a larger
spreading), leaching into groundwater, sorption that retards the movement and lastly, transformation
and decay by chemical and microbiological reactions.

10.1 GAS–WATER EXCHANGE

The volatilization of organic chemicals is important because these pollutants can form individual,
neutral gas molecules, which contrasts with most elements that dissolve as ions in water.

80
Nitrate
70 Other contaminants
No. of wells closed

60 Pesticides
50
40
30
20
10
0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
Year

Figure 10.1. Abandoned groundwater wells in Denmark due to excess of pesticides, nitrate, or other chemicals
(Miljøprojekt, 1998).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

490 Pollution by organic chemicals

FATE OF ORGANIC CHEMICALS (OC) Detoxication

O OC Crop
C removal
C

O
Photo-
decomposition
OC
Ru
no
Volatilization ff
Absorption & OC
exudation

OC
OC OC OC
Adsorption
O
Chemical OC OC OC OC Biological
OC
decomposition Clay degradation
O OC OC OC C
C OC OC O

OC OC
OC OC
OC OC Organic matter OC Oxides
OC
OC
OC
OC

Leaching OC
Capillary
OC flow

Groundwater

Figure 10.2. Processes affecting the spread and attenuation of organic pollutants (OC) in the environment
(Weber and Miller, 1989).

Volatilization is a function of the difference between the actual gas pressure of a substance and
the gas pressure in equilibrium with its concentration in water. The equilibrium concentrations
are related by the law of mass action (Henry’s law for the link between gas pressure and solute
concentration).
For example, for benzene (C6H6):

C6H6(aq) ↔ C6H6(g) (10.1)

with

PC H
KH  6 6
 5.5 (10.2)
mC H
6 6 ( aq )

where KH has the units atm L/mol (K in the law of mass action would be non-dimensional). Often,
the gas concentration is expressed in ppmv (parts per million by volume, equal to 106 atm for a total
pressure of 1 atm). For example, in urban air the benzene concentration may be 0.01 ppmv, which
yields according to Equation (10.2) an equilibrium concentration of 1.8 nM or 0.14 g benzene/L in
the ponds of the city.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Gas–water exchange 491

ca

za Boundary layer, air

ca
Air/water interface
zw Boundary layer, water
cw

z cw
c

Figure 10.3. Diffusion of a chemical through two boundary layers at the gas–water interface. At the interface
itself, the concentrations in air, ca, and in water, cw, are in equilibrium: KH  ca/ cw.

The gas–water exchange rate can be calculated with diffusion models (Danckwerts, 1970;
Schwarzenbach et al., 1993). The major difficulty of modeling the process is to characterize the
interface where the mass transfer takes place. In the two-film model of Liss and Slater (1974), the
diffusive flux is calculated for two boundary layers at the air / water interface, cf. Figure 10.3.
In this model, the concentration in air (ca) is expressed in mol/L air, i.e. the pressure of the chem-
ical (in atm) is divided by the gas constant (R  0.08206 atm L/mol/K) and the temperature (K)
(RT  24 atm L/mol at 20°C). At the interface, the gas and water concentration are considered to be
in equilibrium,

KH c′
K H′   air (10.3)
RT cw′

where KH has the dimension (L water) / (L air), i.e. indicates how many liters of water contain the
same mass of chemical as 1 liter air at equilibrium. For the model, we eliminate the concentrations
at the interface as follows.
The flux through the water boundary-layer is given by the concentration gradient multiplied with
the diffusion coefficient:

cw′  cw
Fw  D f (mol/m 2 /s) (10.4)
zw

where zw is the thickness of the (stagnant) boundary layer of water. Likewise, the flux through the air
boundary-layer is:
ca  ca′
Fa  Da (10.5)
za

At steady state, the two fluxes are equal (Fw  Fa  F). By introducing the interface equilibrium
(Equation 10.3), the concentrations at the interface can be expressed in the (known) bulk concentra-
tions and eliminated from the combination of Equations (10.4) and (10.5). Thus, cw in the water film is:

(Da / za )ca  (D f / zw )cw

cw′  (10.6)
K H′ Da / za  D f / z f

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

492 Pollution by organic chemicals

This equation is inserted in the flux through the water layer (Equation 10.4) to obtain:

1  ca 
Fw   cw   (10.7)
zw / D f  za /( Da K H′ )  K H′ 

The term (cw  ca / KH )  (cw  cw (in equilibrium with air)) is the concentration difference that drives the
mass transport, while the denominator in (10.7) represents the resistance to transport.
Similarly, by eliminating the concentration ca in the flux through the air boundary-layer (Equation
10.5) we find:

Fa 
1
( c K ′  ca
zw K H′ / D f  za / Da w H
) (10.8)

In Equation (10.8) it is the concentration difference in the air boundary-layer that drives the mass
transport, but the same flux is obtained as with Equation (10.7).
The two terms in the denominator of Equations (10.7) and (10.8) represent the resistances against
diffusive transport in the water and air films (resistances may be added, as in an electrical circuit).
The inverse of this resistance has the dimension m/s, and indicates a velocity. Thus, Equation (10.7)
has two velocity terms, for the water layer vw  Df / zw, and for the air boundary layer va 
Da / (za KH ), in which the diffusion coefficients are pretty well fixed (Df
109 m2/s and
Da
105 m2/s), but where the thickness of the diffusion layers needs to be defined. To do so, let us
calculate the evaporation of water with these formulas. For water molecules in the water film, the
resistance for transport will be near-zero since the layer itself consists of water. In other words, the
transfer velocity is infinite, vw  Df / zw  %, or zw  0. Thus, evaporation is determined solely by
the transfer velocity va through the air film. The evaporation of water can be measured with an evap-
oration pan, and, depending on the wind speed, the incoming radiation, the temperature and the
moisture content of the air, it ranges from 0.6–3 m/yr at various locations on earth. We can now esti-
mate the thickness of the air boundary-layer za to range from 1–5 mm, with an average of about 3 mm
for small open waterbodies in the temperate climate.

EXAMPLES
Find the transfer velocity in the air film, za  3 mm, Da  105 m2/s.
ANSWER: va  3.3103 m/s.
Estimate for this va the yearly water evaporation rate E in air with a relative humidity of 50% at 25°C,
KH  2.3105 L water/L air.
ANSWER: E  Fa / cw,H2O  va(KH  KH / 2)  3.3103  (2.3105  1.65105) 
3.8108 m/s  1.2 m/yr.
Estimate the theoretical evaporation rate when limited by resistance in the water layer, zw  20 m for the
oceans.
ANSWER: Etheor  vw  Df / zw  1734 m/yr.
Calculate the thermodynamic K for the reaction H2O(l) ↔ H2O(g).
ANSWER: K  (2.3105 L water/L air)  (24 atm L air/mol)  (55.6 mol/L water) 
0.03(). Compare with PHREEQC.DAT where the reaction is written as
H2O(g) ↔ H2O(l).

Calculating a water evaporation rate from the resistance in the water film does not make sense, as can
be seen from the example calculation above. However, the calculation does show that the flux is not
only determined by the concentration gradient, but also by the concentration in the medium. In the case
of water, the air concentration is small, as expressed by the small Henry constant of 2.3105 L

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Gas–water exchange 493

water/L air. On the other hand, for sparingly soluble gases the concentration in the water film
is much less than in the air film, for example for O2, KH  33 L water/L air. Transport is now
limited by the water film, and again, the thickness of the boundary layer can be determined
experimentally. It is found to depend strongly on the wind speed in the form (Schwarzenbach et al.,
1993):

vwO  4107 ( v10 )2  4106 ( m/s) (10.9)

2

where v10 is the wind speed in m/s at 10 m above the water surface. The quadratic dependence on
windspeed is related to the fact that the stress of the wind on the water surface, and hence the force
tending to reduce the liquid phase resistance, is proportional to the square of the friction velocity
(Liss, 1973; but it may be cubic, Wanninkhof and McGillis, 1999). With Df,O2  1.1109 m2/s, zw
is in the range of 20–200 m (note that mass transfer over the ocean/air interface (zw
20 m) is
relatively large because of wind and waves). It is assumed that the transfer velocity for other chem-
icals varies in proportion to the ratio of the diffusion coefficients (i.e. the film thickness is the same
for all). Multiplying the velocity with the surface area permits to derive the total mass exchange,
e.g. of freon that enters the oceans (Example 10.1).

EXAMPLE 10.1. Estimate the flux of freon-11 (CCl3F) into the sea (Liss and Slater, 1974)
The concentration of freon in the marine atmosphere was 5105 ppmv in 1973, and in
surface seawater 7.6109 cm3 freon/L was found. KH  5 L water/L air. The surface area of the oceans is
3.61014 m2. Find the flux in or out of the oceans relative to oxygen with vwO2  5.5105 (m/s), and
compare with the yearly production of about 3 1011 g/yr (in 1973).

ANSWER:
First express ‘ppmv’ in concentration per L air, then calculate with Henry’s constant the equilibrium con-
centration per L water and compare with the analyzed concentration.
The atmospheric concentration of 5105 ppmv  5105  (106 L/ppmv)  (103 cm3/L) 
5108 cm3 CCl3F/L air. Water in equilibrium would contain 1108 cm3 CCl3F/L water which is higher
than analyzed. Thus, the flux is into the ocean.
For the given KH, the transfer velocity is limited by the water film and can be calculated
relative to oxygen. The ratio of the diffusion coefficients, Df,CCl3F / Df,O2  (32  (138  18) / (138 
(32  18)) ⁄  0.85 (cf. Equation 3.60).
1
2

The flux is:

F  0 . 8 5 v w O 2  c  0 . 8 5  5 . 5  1 0  5 m / s  ( 1 0  8  0 . 7 6 1 0  8 d m 3 / m 3 )
 1.121013 dm3/m 2/s.
Multiply with the surface area of the oceans and s/yr, and find a total transfer of 1.29109 dm3 freon/yr.
1 dm3 gas is 1 / RT  0.0409 mol, meaning that we had a flow of 7.2109 g freon or 2.4% of the world
production into the oceans in 1973.

QUESTION:
The freon concentration in surface sea water was 3.5 pM in 1993 (Yeats and Measures, 1998); find the
concentration in air from Figure 3.9 and estimate the flux.
ANSWER: (3.51012 mol/L)  (24,400 cm3 air/mol)  8.5108 cm3 freon/L. From Fig-
ure 3.9, freon in air was 3104 ppmv in 1993, and the equilibrium concen-
tration in seawater should be 6108 cm3 freon/L. The flux would be out
of the oceans, but the depth profiles in the sea indicate a downward flux:
apparently the data do not agree, or are insufficiently accurate.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

494 Pollution by organic chemicals

QUESTION:
For which value of KH are the transfer velocities in the water and the air film about equal? Assume
za  3 mm, zw  0.25 mm.
ANSWER: KH  (za / Da) / (zw / Df)  2  103 L water/L air.

It is not very likely that the film thickness is the same for all substances that pass the air-water inter-
face. Another model considers the distance covered by chemicals by diffusion,    , and
(2Dt)
assumes that the mass transfer is related by the square root of the diffusion coefficients of the chemi-
cals (Danckwerts, 1970). For the case of freon in Example 10.1, it would mean that the transfer
velocity of freon is  0.85
  0.92 less than of oxygen, which is not really much different. In experi-
mental studies, the relationship lies mostly in between a square root and a linear one (Schwarzenbach
et al., 1993).
For chemicals that form complexes in water, the water transfer rate may be increased by
transport of the complexes. For example for CO2 exchange, the concentration difference in the water
film is:

c  ( mCO  mCO
′ )  ( mHCO  mHCO
′  )  ( mCO2  mC′ O2 ) (10.10)
2 2 3 3 3 3

At equilibrium, mHCO 3
 mCO2  K1 / [H] and mCO32  mCO2  K1K2 / [H]2 for the bulk solution, and
similar for the interface. Hence, Equation (10.10) becomes:

c  (1  K1 / [H ]  K1 K 2 / [H ]2 )( mCO  mCO

′ )  (1 )(mCO  mCO
′ ) (10.11)
2 2 2 2

where  is the speciation factor (Equation 5.13).

Thus, at pH  6.3 and 8.3, the transfer rate would increase by factors of 2 and 102, respectively.
However, the observed pH dependency of the CO2 transfer is much smaller, which indicates that the
conversion among the species and to CO2 (which only can pass the interface) is relatively slow.
Notably, the dehydration of HCO 3 is rate limiting.
By comparing characteristic times, the rates of mass transfer over the interface and of hydration
of CO2 can be compared. For transport through the water layer with zw  100 m, the residence time
is   zw / vw  104 / 5.5105  1.8 s. The dehydration reaction of HCO 3 has a half-life of
approximately 100 s at pH  8.3 (Problem 10.4), much longer than the residence time. Thus, HCO 3
gives only a very small contribution to CO2 transfer.

10.1.1 Evaporation of a pure organic liquid

The evaporation of a pure organic liquid can be estimated by analogy with the evaporation of water
assuming that transport is controlled by the transfer velocity through the air film, va  Da / za 
3.3103 m/s. We neglect differences in diffusion coefficients (cf. Example 10.1), which permits to
estimate the evaporation from the concentration difference over the air film:

E  3.3103 ( ca  ca′ ) (10.12)

where ca is the concentration at the vapor pressure of the substance, and ca may be
taken 0 (all the chemical is rapidly dispersed). For example, the vapor pressure for benzene,
P 0 0.126 atm. Then, ca  0.126 / (24.4 atm L/mol)  (78.1 g/mol)  (1000 L/m3)  409 g/m3.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Gas–water exchange 495

TCE concentration in soil gas (ppmv)

Cover
0

Source

3,

1,
00

30
00
zone
Section 1

0
0
line

Depth (m)
10

10

30
30,000

0
,00
2

0
3
Top of capillary fringe

0 2 4 6 8 10
(A) Radial distance (m)

100
80
TCE (ppt)

60
40
20
0
0 2 4 6 8
(B) Radial distance (m)

Figure 10.4. TCE concentrations in soil gas, 18 days after 60 L TCE were placed in the source zone. Figure
10.4B compares the observed concentration profile and the diffusion model, Equation (10.14). Data from
Conant et al., 1996.

The estimated evaporation is E  3.3103  409  (86,400 s/day)  117 kg/m2/day, or, with the
density of benzene   0.88 kg/L, 0.13 m/day. Thus, we may express evaporation of substance X
relative to water at a given location as:

PX0 MWX 1
E X  EH (10.13)
2O PH0 O (1  rh ) 18 
2 X

where rh is the relative humidity of air at the site.

By taking ca  0, we assume that all the chemical is removed rapidly by transport through the air,
but what happens in a soil, where the air is stagnant? Figure 10.4 shows concentrations of TCE vapor
in the unsaturated zone in an experiment where a volume of pure liquid TCE was mixed with sand
and placed in the soil. The observed vapor concentration profile after 18 days shows the typical
exponential decay of concentrations that indicates diffusion.
An approximate solution of the diffusion equation in radial direction is (cf. Equation 3.58):

c  ci  ( c0  ci )erfc
( (r  r0 )

1
4

r
2
De,a
t
R ( (10.14)

where r0 is the radius of the cylinder with TCE, R is the retardation through sorption on the solid, and
the factor (1  2 / r) corrects for radial instead of linear diffusion.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

496 Pollution by organic chemicals

Gas diffusion in the soil pores will depend on the pore volume of the soil and the water content, and
the effective diffusion coefficient can be estimated with the Millington-Quirk relations (Jin and Jury,
1996; Equation 9.62):

De,a  Daeg2 / e0.6 (10.15)

where g and  are the gas-filled and the total pore volume, respectively. For the sandy soil in the
experiment of Figure 10.4, g  0.3 and   0.4, which yields:

De,a  8.1106 (0.32 / 0.40.67 )  1.34106 m 2 /s.

After 18 days, with retardation R  4.8 and r0  0.6 m (Conant et al., 1996), and this diffusion coef-
ficient, the calculated concentrations are compared with the experimental data in Figure 10.4B.
In applying Equation (10.14) to the experiment of Figure 10.4 we neglected evaporation from the
soil surface into the air and advective vapor transport of the dense TCE molecules down to the cap-
illary fringe (note that a cover layer was placed above the TCE in the experiment to reduce evaporation).
Furthermore, TCE dissolves into the (saturated) groundwater (Jellali et al., 2003). Nevertheless, it is
clear that our simple equation provides a good approximation for describing the TCE vapor concen-
tration in the soil.
Gas diffusion in a soil (Equation 10.14) is much reduced when gas content decreases to
less than about 0.1 (Troeh et al., 1982). Apparently, gas then resides in isolated, unconnected
pockets, and only is transported by diffusion via water. This phenomenon of a residual occupation
of the soil pores expresses itself in the persistence of organic chemicals in the soil upon water
flooding.

QUESTIONS:
The vapor pressure of TCE is 0.08 atm at 20°C. Estimate the concentration of TCE in ppmv?
ANSWER: 80,000 ppmv
Henry’s constant for TCE at 20°C is KH  7.2 atm L/mol. Calculate the retardation for TCE vapor, when
it dissolves into soil water, g  0.3, w  0.1?
ANSWER: KH  0.3 Lw ater / Lair. Hence, R  1  KH g / w  1.9. (Note that the retarda-
tion given by Conant et al. includes sorption of TCE by organic carbon; its
contribution will be calculated in Section 10.3).

10.2 TRANSPORT OF PURE ORGANIC LIQUIDS THROUGH SOIL

Organic chemicals spilled in excess of their aqueous solubility may form a separate non-aqueous
phase liquid, a NAPL. The liquid may be denser than water, a DNAPL, and it will sink through
the groundwater to larger depths where it may be halted by impermeable layers to accumulate in
depressions in the aquifer (Figure 10.5). The liquid can be lighter than water, a LNAPL, and it
will then float on the groundwater and be confined to the capillary zone from where it evaporates
and also dissolves in percolating soil water. An intermediate form of transport is exhibited by
highly viscous tar products which have fingered from above and form tar blobs in the aquifer.
Examples of DNAPLs are degreasing solvents such as trichloroethylene (C2HCl3, “tri” or “TCE”)
and tetrachloroethylene (C2Cl4, “per” or “PCE”) with densities of 1.46 and 1.63 g/cm3 (Pankow and
Cherry, 1996). The most common LNAPL is gasoline, and tar products are often found below coke-
gas factories.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Transport of pure organic liquids through soil 497

Leaky tank
Oil
LNAPL pool

BTEX plume

Leaky tank
Solvent

TCE plume

DNAPL pool

Figure 10.5. Examples of spreading in groundwater of a dense non-aqueous phase liquid (DNAPL, TCE), and
a light one (LNAPL, gasoline) (Pankow and Cherry, 1996).

If we would take samples from a borehole through the zone affected by the DNAPL spill
of Figure 10.5 and analyze it for water and TCE content, a profile like in Figure 10.6 may be
the result. In the upper, unsaturated part of the profile the water content follows the pF-curve
that relates soil water suction and water content (cf. Figure 9.22). TCE evaporates quickly from
the unsaturated zone when the leakage has been stopped and is already gone in Figure 10.6.
Below the water table, the proportions of TCE and water follow an erratic pattern that depends on
whether a TCE tongue has passed the location and whether the TCE has been displaced again by the
groundwater flow. Down in the profile a pool of TCE has accumulated above the aquiclude.

Air

Water
Soil ∆P  Air entry pressure

Water
Depth

Residual
Aquifer DNAPL

Residual
water
DNAPL

Aquiclude
0 100% Water saturation
100 0% Air/DNAPL saturation

Figure 10.6. Relative saturations of water, air and TCE in a profile through a TCE polluted, phreatic aquifer.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

498 Pollution by organic chemicals

Quite conspicuous in the concentration profile of Figure 10.6 is, that the TCE never attains full sat-
uration, there is always some water left in the pores at the location of the DNAPL pool. Likewise for
TCE, when it has entered the pores for once, a residual part remains entrapped that is very difficult
to remove with water.
When water and air, or water and TCE reside both in the pores of the soil, the interface between
the fluids is curved as result of the pressure difference (Figure 10.7):

P  2 / r (10.16)

where P is pressure (N/m2),  is the interfacial tension (N/m) and r is the pore radius (m). In Figure
10.7, the convex side of the interface points downward, indicating that the pressure in air is higher
than in water. The capillary pressure results from differences in molecular properties, cohesion
within the fluid and strong adhesion of water to the solid capillary.
For the water/air interface,   0.073 N/m, and we can calculate for a soil that pores with r 
10 m will be filled with water above the water table up to 2  0.073 / (105 m  1000 kg/m3 
9.8 m/s2)  1.5 m (in fine sandy soils this may correspond to 10% water saturation). For
water / DNAPL’s, 
0.035 N/m, and similarly, we can calculate that 0.48 m of a DNAPL, with den-
sity 1.5 g/cm3, can displace water from 10 m pores.
The important thing is to realize that transport of water and DNAPL is coupled to the pore
space occupied by each fluid. Initially, when DNAPL reaches the water table, a minimum pressure
is necessary to enable further flow through the water saturated soil. This is similar to the air entry
pressure (Figure 10.6), which is the required minimal pressure before air can start to flow through a
water-saturated soil. It corresponds to the pressure needed to create a network of interconnected
pores from which water has been removed, corresponding to a reduction of the water content
to about 0.8–0.95 of full water saturation. At this stage the air or NAPL enters the largest pores,
with the largest relative permeability. Further displacement of water requires a higher pressure,
which can be achieved by the DNAPL when its downward flow is halted by an aquitard with a finer
pore space.
The simplest approximation for calculating transport is the Green and Ampt model (Weaver
et al., 1994; Hussein et al., 2002):

 NAPL  K NAPL ( h  d  H f ) / d (10.17)

where v is the specific discharge (m/day), K the hydraulic conductivity (m/day), h the pressure head
of the NAPL where it enters the aquifer (m water), d is the depth of the front (m) and Hf is the NAPL
entry pressure (m water). The equation states that as the infiltration depth (d) increases, the pressure
head difference ((h  d  Hf) / d ) decreases, and we recognize the Darcy equation.

2
P  r

Figure 10.7. Water adheres to the walls of a glass capillary and rises in it because it is at lower pressure than
in the free fluid.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 499

In the Green and Ampt approach we assume that the NAPL filled porosity NAPL is constant, and we
can integrate Equation (10.17) with v  NAPLdd / dt from d  0 at time t  0:

e NAPL   h  d  Hf  
t   d  ( h  Hf )ln    (10.18)
K NAPL   h  Hf  

The permeability is a function of the pores occupied, and of the viscosity of the liquid. Assuming that
the viscosities of the NAPL and of water are approximately equal and KNAPL  2 m/day for
NAPL  0.3, and that Hf  0.1 m and h  1 m, we can calculate that it takes only 2.6 days to reach
20 m depth. Introductions to more complete calculations of DNAPL transport have been given by
Bedient et al. (1994) and McWhorter and Kueper (1996).

QUESTIONS:
Discuss the effects of h on NAPL and KNAPL, and on t.
ANSWER: Smaller h will reduce NAPL but we expect that KNAPL reduces more. Anyhow,
t increases.
Compare rates of TCE evaporation and its infiltration into a sandy soil?
ANSWER: Equation (10.13) gives E
0.08 m/day into the air, Equation (10.17) gives v

210 m/day into the soil. Clearly, leakage into the soil is much larger than
evaporation.

10.3 SORPTION OF ORGANIC CHEMICALS

Many organic pollutants are hydrophobic, which indicates that these substances have a low affinity for
solution in water (a polar liquid), and prefer solution in a-polar liquids. These pollutants are readily
taken up in organic matter of sediments. The tendency to become absorbed (i.e. the distribution coef-
ficient for these organic chemicals) is related to the distribution coefficient of the chemical between
water and an a-polar liquid like octanol. The latter is termed a partition constant or extraction coef-
ficient by analytical chemists; the soil sorption process, by analogy, is also envisaged as a partition
process, where the hydrophobic pollutant partitions itself between water and the soil organic matter
(Chiou et al., 1979, 1987; Karickhoff, 1984). Such solution of the hydrophobic compound into
organic matter is appropriately termed absorption, rather than adsorption.
The distribution coefficient between water and octanol is obtained in a separatory funnel as
shown in Figure 10.8. The organic chemical is introduced in a funnel containing water and octanol.

Octanol, co co
Kow 
Water, cw cw

Figure 10.8. Separatory funnel used to obtain octanol/water distribution coefficients.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

500 Pollution by organic chemicals

The funnel is shaken and the two phases are separately collected. Analysis of the concentrations in
the water- and octanol-phases gives cw and co respectively, from which the distribution coefficient

K ow  co / cw (10.19)

is readily obtained. The distribution coefficient is highly correlated with the distribution coefficient
between organic carbon and water, Koc (Karickhoff, 1981; Schwarzenbach and Westall, 1985).
Karickhoff (1981) suggests

log K oc  log K ow  0.35 (10.20)

for the chemicals listed in Table 10.1. For groups of these chemicals, Schwarzenbach and Westall
(1985) propose a linear regression of the form:

log K oc  a log K ow  b (10.21)

Values of a and b are given in Table 10.2. More values of octanol/water distribution coefficients and other
related data can be found in Verschueren (2001), and especially Lyman et al. (1990) and Schwarzenbach
et al. (1993) are recommended for guidelines for estimating properties of organic pollutants.

Table 10.1. Partition coefficients for octanol–water (Kow) and organic carbon–water (Koc) (Karickhoff, 1981).
Compound log Kow log Koc Compound log Kow log Koc

Hydrocarbons and chlorinated hydrocarbons Carbamates

3-methyl cholanthrene 6.42 6.09 Carbaryl 2.81 2.36
Dibenz[a,h]anthracene 6.50 6.22 Carboturan 2.07 1.46
7,12-dimethylbenz[a]anthracene 5.98 5.35 Chlorpropham 3.06 2.77
Tetracene 5.90 5.81
Organophosphates
9-methylanthracene 5.07 4.71
Malathion 2.89 3.25
Pyrene 5.18 4.83
Parathion 3.81 3.68
Phenanthrene 4.57 4.08
Methylparathion 3.32 3.71
Anthracene 4.54 4.20
Chlorpyrifos 3.31 4.13
Naphthalene 3.36 2.94
Benzene 2.11 1.78 Phenyl ureas
1,2-dichloroethane 1.45 1.51 Diuron 1.97 2.60
1,1,2,2-tetrachloroethane 2.39 1.90 Fenuron 1.00 1.43
1,1,1-trichloroethane 2.47 2.25 Linuron 2.19 2.91
Tetrachloroethylene 2.53 2.56 Monolinuron 1.60 2.30
 HCH (lindane) 3.72 3.30 Monuron 1.46 2.00
 HCH 3.81 3.30 Fluometuron 1.34 2.24
 HCH 3.80 3.30
1,2-dichlorobenzene 3.39 2.54 Miscellaneous compounds
pp’DDT 6.19 5.38 13Hdibenzo[a,i]carbazole 6.40 6.02
Methoxychlor 5.08 4.90 2,2’biquinoline 4.31 4.02
22’,44’,66’PCB 6.34 6.08 Dibenzothiophene 4.38 4.05
22’,44’,55’PCB 6.72 5.62 Acetophenone 1.59 1.54
Bromacil 2.02 1.86
Chloro-s-triazines Terbacil 1.89 1.71
Atrazine 2.33 2.33
Propazine 2.94 2.56
Simazine 2.16 2.13
Trietazine 3.35 2.74
Ipazine 3.94 3.22
Cyanazine 2.24 2.26

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 501

Table 10.2. Estimation of Koc from Kow by the expression log Koc  a log Kow  b (Schwarzenbach and
Westall, 1985).

Regression
coefficient
Correlation Number of
a b coefficient compounds Type of chemical

0.544 1.337 0.74 45 Agricultural chemicals

1.00 0.21 1.00 10 Polycyclic aromatic hydrocarbons
0.937 0.006 0.95 19 Triazines, nitroanilines
1.029 0.18 0.91 13 Herbicides, insecticides
1.00 0.317 0.98 13 Heterocyclic aromatic compounds
0.72 0.49 0.95 13 Chlorinated hydrocarbons alkylbenzenes
0.52 0.64 0.84 30 Substituted phenyl ureas and alkyl-N-phenyl carbamates

The log Koc refers to partitioning between water and a 100% organic carbon phase; the actual
distribution coefficient for the soil or sediment is then obtained as

K d′  K oc f oc (10.22)

where foc is the fraction of organic carbon. This relationship holds when foc ! 0.001, otherwise sorp-
tion on non-organic solids can become relatively important (although giving only low Kd’s,
Karickhoff, 1984; Curtis et al., 1986; Madsen et al., 2000). The solubility of the (solid) organic com-
pound in water is another parameter which has been suggested in order to estimate the organic car-
bon partitioning coefficient. It is clearly related to the octanol/water distribution coefficients as
shown in Figure 10.9.

107
 
2,4,5,2,4,5-PCB
n–octanol / water partition coefficient, log scale

Leptophos
DDT 
2,4,5,2,5-PCB
106
DDE
4,4-PCB
Dichlofenthion
105 Chlorpyrifos
Ronnel
Dialifor
Phosalone Methyl chlorpyrifos
Diphenyl ether
104 Parathion
Dicapthon Naphthalene
Fenitrothion p-dichlorobenzene
Iodobenzene
103 Malathion Bromobenzene
Chlorobenzene
Phosmet 2,4-D Toluene
Carbon tetrachloride
Tetrachlorethylene Salicylic acid
102 Benzene
Fluorobenzene Chloroform
Nitrobenzene Benzoic acid
Phenylacetic acid
Phenoxyacetic acid
10
103 102 101 1 10 102 103 104 105 106
Solubility in water (µmoles/L, log scale)

Figure 10.9. Relationship between solubility of a number of compounds and their octanol/water distribution
(or partition) coefficient (Chiou et al., 1979).
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
502 Pollution by organic chemicals

EXAMPLE 10.2. Retardation of Lindane and PCB

Calculate the retardation of Lindane and of PCB with respect to groundwater flow in a sediment with 0.3%
organic carbon.

ANSWER:
From Table 10.1, find the distribution coefficient Koc for lindane, which is 103.3, and obtain with Equation
(10.22):
log Kd  3.3  log 0.003  0.8
Kd  6.3 mL/g, and Kd  6.3 b / w
38.
Similarly for PCB:
log Kd 5.8  log 0.003  3.3
Kd
2000 mL/g, and Kd  2000 b / w
12,000.
Lindane (-HCH) would have a velocity which is about 40 times less than the water velocity, whereas PCB
would move imperceptibly slowly at 12,000 times less than the water velocity.

QUESTION:
Estimate the retardation (R) for TCE in the volatilization experiment (Figure 10.4). Koc  125 L/kg. Organic
carbon is 0.02% on average in the sediment, and 2% in the organic-rich layer.
ANSWER: We noted already that the ratio gaseous TCE / dissolved TCE  1/ 0.9. Express
sorbed TCE with respect to dissolved TCE, b  1.8 g/cm3, w  0.1:
qTCE  (125 L/kg OC)  (2 104 kg OC/kg)  (1.8 kg/L)/ (0.1 ()) 
(0.9 mmol TCE/L)  0.40 mmol TCE/L for the sediment, and 40 mmol TCE/L
for the organic-rich layer. It gives R  1  (0.9  0.4)/ 1  2.3 and
1  (0.9  40) / 1  42 for gas transport in the two layers, respectively. The fit-
ted R  4.8 points to the importance of the organic layer in retarding transport
(although a smaller diffusion coefficient would have a similar effect).

When sediment contains more than 0.1% organic carbon, the adsorption of nonionic organic chem-
icals is wholly attributed to organic carbon. One might expect that the deposition environment has
an influence on organic matter content, e.g. organic matter will be low in sediments deposited in gla-
cial times, and higher during interglacial periods. However, the scarce data available on sandy sedi-
ments in the Netherlands show no definite trend in this respect (Figure 10.10). On the other hand,
the figure does illustrate that organic matter content is so low in the common aquifers that our Kd
approximation method may not be valid. Since sorption to mineral surfaces is also small, Kd’s
smaller than 1 L/kg are expected.
Use of a single Koc assumes that all natural organic matter can be lumped together. However, dif-
ferences of an order of magnitude have been observed among different humic and fulvic acids
(Garbarini and Lion, 1986, Chiou et al., 1987). Grathwohl (1990) investigated organic matter from
sediments with different ages, and a different degree of maturing or ripening, ranging from recent
organic matter to precursors of oil. He found that a higher H/O-ratio in the organic matter gives an
increase of the distribution coefficient for sorption of hydrophobic organic chemicals (Figure
10.11). This effect is associated with lower polarity and higher hydrophobicity of organic matter as
the H/O ratio increases. Also, diffusion in coaled organic matter is slower than in soft humic acid,
which may lead to curved isotherms in short-term experiments and a decreasing curvature of the
isotherm when time progresses (Weber et al., 2001).
The H / O ratio of young organic matter is much more uniform than was used for the experiments
in Figure 10.11. In recent soils, Koc tends to be constant to within 0.3 log Koc units for a given sub-
stance (Schwarzenbach et al., 1993). Nevertheless, trends may exist even within this small range.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 503

Wt - % in light fraction % CaCO3 % org. C

0 4 8 12 16 0 .4 .8 0 .06 .12
0

Boring 20 B - 24

Interglacial
40
Form.v. Urk
Form.v. Sterksel
Depth below surface (m)

Muscovite
80
K - feldspar

Albite
Plagioclase
Anorthite
Form.v. Enschede
120
0.3% in all samples

Glacial
160

Form.v. Hardewijk (young)

200

Form.v. Hardewijk (old)

Figure 10.10. Contents of CaCO3, weatherable silicates, and organic matter of Pleistocene sediments in a
borehole in The Netherlands.

Kile et al. (1999) and Ahmad et al. (2001) investigated a multitude of soils and sediments and found that
Koc increased with the aromaticity of organic carbon. However, the overall variation was small; it can be
calculated from their data that log Koc  1.88  0.12 for tetrachloromethane (CCl4), log Koc  2.58 
0.13 for carbaryl, and Koc  3.82  0.15 for phosalone (both are pesticides, cf. Table 10.1 for carbaryl).

4
Anthracite Shales, coals
3.5 Bituminous Garbarini & Lion, 1986
coal
Shale
3 (Jurassic)
Lignite Shale (Tertiary)
2.5 Peat
Lignin Zein
log Koc

2 Humic acid (Aldrich)

Humic acid
1.5
Tannic acid

1
Fulvic acid
0.5
Cellulose
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
log H/O (atomic ratio)

Figure 10.11. The distribution coefficient Koc of trichloroethylene (TCE) increases with increasing H/O ratio
(Grathwohl, 1990).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

504 Pollution by organic chemicals

Also, with increasing pH the surface charge of organic matter will increase, rendering it more polar and
perhaps less penetrable for a-polar organic chemicals (Kan and Tomson, 1990). Again, the observed
reduction of log Koc was minor, in this case for naphthalene decreasing from 2.9 to 2.7 when pH
increased from 5 to 9.

10.3.1 Sorption of charged organic molecules

If the organic chemical develops a negative charge, sorption to organic carbon and sediments
in general is much reduced (Haderlein and Schwarzenbach, 1993; Broholm et al., 2001). The
herbicide 2-methyl-4,6-dinitrophenol (C6H2CH3(NO2)2OH, abbreviated DNOC-OH) dissociates
according to

DNOC-OH  DNOC-O  H; log K a  4.31 (10.23)

The fraction DNOC-OH of total DNOC is (cf. Equation 5.13):

DNOC-OH  DNOC/(1  K a /[H ]) (10.24)

In case only the neutral species is sorbed, the distribution coefficient becomes:

K d  K oc f oc /(1  K a /[H ]) (10.25)

For pH  pKa, the denominator of Equation (10.25) is 1, and the distribution coefficient is equal
to the organic carbon sorption coefficient. When pH  pKa, the distribution coefficient is just half
of this value, and it will further decrease when pH increases and the negative species becomes dom-
inant. The effect is illustrated in Figure 10.12 for DNOC (Broholm et al., 2001).
Actually, the sediment in Figure 10.12 had a low organic carbon content of 0.02%. With
log Kow  2.12 for DNOC, the estimated Kd  0.026, much smaller than observed (Figure 10.12).
The difference was attributed to sorption on mineral surfaces, but iron-oxyhydroxides which are
positively charged in the pH-range of this study would sorb the anionic form rather than the neutral
form (Clausen and Fabricius, 2001).
If the chemical develops a positive charge, sorption to the dominantly negatively charged soil
particles is enhanced below the pKa. We can use PHREEQC to model the sorption of the neutral
fraction as surface complexation reaction, and of the charged form as ion exchange as shown in
Example 10.3.

16 16
Observed y  14x 14 Observed
14
12 Model R 2  0.96 12 Model
Kd (L /kg)
Kd (L/kg)

10 10
8 8
6 6
4 4
2 2 pKa  4.31
0 0
0 0.2 0.4 0.6 0.8 1 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
(A) Neutral fraction of DNOC (B) pH

Figure 10.12. pH dependent sorption of DNOC-OH on sediment (Broholm et al., 2001).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 505

EXAMPLE 10.3. Calculate sorption and ion exchange of Quinoline with PHREEQC
Quinoline becomes protonated for pH’s smaller than pKa  4.94. The neutral form has Koc  100 L/kg.
Zachara et al. (1986) measured Kd of quinoline for a soil horizon with foc  0.0024, CEC  84 meq/kg, in
10 mM CaCl2 solution at pH 4.2 and 7.5. Estimate the Kd for these pH’s with PHREEQC.

ANSWER:
We make an input file where quinoline is defined as aqueous species that can associate with H. Sorption
of the neutral form to organic matter is modeled with SURFACE, for the positive species EXCHANGE
is used. The log K of the surface species is defined in a special way to simulate the simple distribution
reaction.

SOLUTION_MASTER_SPECIES # define quinoline...

Quin Quin 0 129 129
SOLUTION_SPECIES
Quin = Quin; -log_k 0
Quin + H+ = QuinH+; -log_k 4.94

SURFACE_MASTER_SPECIES # define absorption into OC..

Org_c Org_c
SURFACE_SPECIES
Org_c = Org_c; -log_k 0
Org_c + Quin = Org_cQuin; -log_k -98.0 # divide K_oc = 100 by 10100

EXCHANGE_SPECIES
X- + QuinH+ = XQuinH; -log_k 0 # equal to K_NaX
END

SOLUTION 1; pH 4.2; Ca 10; Cl 20; Quin 1e-3 # the experimental solutions

SOLUTION 2; pH 7.5; Ca 10; Cl 20; Quin 1e-3
END

USER_PRINT
-start
10 K_d_OC = mol("Org_cQuin") / tot("Quin")
20 K_d_X = mol("XQuinH") / tot("Quin")
30 print "pH = ",-la("H+"), ". K_d(OC) =",K_d_OC, ". K_d(X) =",K_d_X,\
". K_d_tot (L/kg) =", K_d_OC + K_d_X
-end

USE solution 1
SURFACE 1; Org_c 0.0024e100 1 1; -equil 1 # sites of Org_c times 10100
EXCHANGE 1; X 0.084; -equil 1
END
USE solution 2
SURFACE 2; Org_c 0.0024e100 1 1; -equil 2
EXCHANGE 2; X 0.084; -equil 2
END

The names of the chemical must be defined starting with a capital letter followed by lower case letters in
keyword SOLUTION_MASTER_SPECIES. The species in the second column (here “Quin”) must also be
defined in an identity reaction with log_k  0 under SOLUTION_SPECIES. The protonated form of quino-
line is defined as well.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

506 Pollution by organic chemicals

In the Kd approach, organic carbon is a phase which does not enter in the reaction equation and it is also not
affected by the reaction. To translate the Kd formula into a formal chemical reaction that is valid in
PHREEQC terms, we rewrite the distribution a bit. For the surface complexation reaction

Org_c  Quin  Org_cQuin,

the mass action equation is

[Org_cQuin ] [Org_cQuin]
 K PHREEQC or  ([Org_c]K PHREEQC )  K oc
[Quin ][Org_c] [Quin ]

A constant distribution coefficient requires that the product ([Org_c] KPHREEQC)  Koc is constant. By
imposing a huge concentration of surface sites mOrg_c, [Org_c] will not change when a relatively small frac-
tion associates to form [Org_cQuin]. The value for the association constant of the SURFACE_SPECIES then
becomes KPHREEQC  Koc / mOrg_c, or log KPHREEQC  log Koc  log mOrg_c. In the example, mOrg_c is multi-
plied with 10100, and K is accordingly divided by 10100 (log KPHREEQC  log(102 / 10100)  98). Since the
surface species is neutral, charge effects are absent and we do not care about the surface area of the organic
carbon and simply make it 1.
The exchange reaction is defined in the usual manner (cf. Chapter 6), we estimate that K for a singly
charged species is identical to the one for Na.
The result is:

--------------------------------------User print------------------------------------
pH = 4.2000e+00 . K_d(OC) = 3.2175e-02 . K_d(X) = 2.4637e-01. K_d_tot (L/kg) = 2.7855e-01
pH = 7.5000e+00 . K_d(OC) = 2.4088e-01 . K_d(X) = 9.2441e-04. K_d_tot (L/kg) = 2.4180e-01

Zachara et al. (1986) found Kd  3.5 and 0.75 L/kg for pH  4.2 and 7.5. Clearly, at low pH where the ion
exchange reaction underestimates observed sorption, we are fairly far off with our model. Schwarzenbach
et al. (1993) modeled the same data, but they used K  20 for the exchange reaction.

QUESTION:
Find Kd when K  20 for the exchange reaction?
ANSWER: Kd  5 L/kg.

10.3.2 Sorption in stagnant zones

In chapter 6 we discussed the effects of stagnant layers on transport of solutes retarded by ion
exchange. Peat layers, buried soils, and in general clayey sediments with a high organic carbon con-
tent are particularly effective absorbers which diminish the spreading of organic pollutants. These
layers have a relatively low permeability and water flow through them is restricted. Typically for sit-
uations where permeable and stagnant zones alternate, the shape of the breakthrough curve displays
an early advent of the chemical and a slow bleeding upon flushing (Section 6.5.1). A column exper-
iment with naphthalene shows the effects (Figure 10.13).
The experiment is to be modeled with PHREEQC (Problem 10.3), but it is instructive to obtain the
general picture first. The soil contains natural organic carbon which sorbs naphthalene (Koc  302 L/kg,
foc  0.014) and retards its flow velocity by 8.5. However, the high flow velocity in the experiment
incited disequilibrium for 1/3 of the sorption sites. In the breakthrough curve, this disequilibrium is
expressed by an early arrival of c0.5 (appears at PV  5.3 instead of 8.5) and tailing towards c1.0.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 507

Virgin soil

1.0 CSAC soil with coal tar

0.8
C/C0

0.6

0.4

0.2

0.0
0 50 100 150 200
Pore volumes

Figure 10.13. Breakthrough curves of naphthalene in column experiments with soil with natural organic
carbon (“virgin soil”) and with the same soil amended with 4 g coaltar/kg soil (Bayard et al., 2000).

When coaltar was added to the soil (4 g/kg with foc  0.38, Koc  2026 L/kg, note the much higher Koc),
the breakthrough of c0.5 of napthalene was retarded only slightly more (to PV  5.8). However, the over-
all breakthrough occurred much later because sorption to the coaltar was mostly kinetic. The same hap-
pened in the elution stage, but note carefully that the model, though excellent for sorption, is not as good
for desorption where the release rate is overestimated (the complete elution of naphthalene occurs too
quickly in the model). This is because the first order exchange model assumes the same rate for uptake
and release, whereas the physical process involves a continuing uptake in the inner realms of the immo-
bile zone, even when the outer layers of the stagnant part release the chemical already to the mobile zone.
The first order exchange process between two boxes is calculated with the formula (Section 6.5.2):

dM im dc
 Vim Rim im  ( cm  cim ) (6.40)
dt dt

where subscripts m and im indicate mobile and immobile, and  is the exchange factor (1/s). The
exchange factor can be related to the physical properties of the stagnant zone:

De eim
 (6.45)
( af s →1 )2

where De,a is the effective diffusion coefficient in the stagnant region, a is the size (m), and fs→1 is a
conversion factor which depends on the shape of the stagnant body.
Let us calculate transport of CCl4 (“tetra”) in a heterogeneous aquifer in which clay
lenses retard transport. The aquifer is a regular sequence of sandy and clayey layers, of 0.66 m
and 0.33 m thickness, respectively, and with equal porosity of 0.36, sketched in Figure 10.14.
Flow velocity in the sandy layer is 50 m/year and zero in the clay layer. A 5 year pulse of
1 mg CCl4/L is followed by 10 years flushing of the chemical. The retardation is zero in the
sand, and estimated to be 4 in the clay, given that the clay contains 0.5% OC, Koc  100 L/kg
and b /   6 kg/L. The effective diffusion coefficient is De,a  0.331010 m2/s in the clay.
We compare transport of Cl as a conservative tracer and of CCl4, and consider the effect of the clay
lenses. The shape factor fs→1 is 0.53 for plane sheets.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

508 Pollution by organic chemicals

0.33 m

Model unit
0.66 m

750 m

Figure 10.14. Small section of an aquifer with uniform sand and clay layer alternations.

The regular alternation of clay and sand means that the concentrations are symmetric
around planes midway of the layers. Therefore, the configuration can be modeled as a
unit of 0.33 m sand in contact with 0.165 m clay. The exchange factor   0.331010 
0.36 / (0.165  0.53)2  1.55109/s. Table 10.3 lists the input file for PHREEQC.

Table 10.3. Input file for sorption of CCl4 in a dual porosity aquifer.

SOLUTION_MASTER_SPECIES; Tetra Tetra 0.0 Tetra 1.0

SOLUTION_SPECIES; Tetra = Tetra; log_k 0.0
SURFACE_MASTER_SPECIES; Sor Sor
SURFACE_SPECIES
Sor = Sor; log_k 0.0
Sor + Tetra = SorTetra; log_k -99.52 # log(Kf = 3) - log(m_Sor)

SOLUTION 1-201 # The sand and clay aquifer

SURFACE 102-201 # Sorption in stagnant cells (clay layer)

Sor 1e100 1 1
END

SOLUTION 0; Cl 1; Tetra 1 # The polluted water

PRINT; -reset false
TRANSPORT; -cells 100; -shifts 25 1; -timest 0.73e7 # 5 years pollution
-length 10; -disp 4.999
-stagnant 1 1.55e-9 0.24 0.12 # 1 stagnant layer, alpha, mobile por, immobile por
END

SOLUTION 0 # Clean water enters

TRANSPORT; -shifts 50 1 # 10 years cleaning
-punch_fr 50 # punch/graph only 50th shift
-punch 1-100 # punchout mobile flow
# -punch 102-201 # punchout stagnant area
USER_GRAPH
-heading Dist CCl4 Cl
-start
10 graph_x Dist
20 graph_y tot("Tetra") * 1e3, tot("Cl") * 1e3
-end
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 509

Note that under keyword TRANSPORT, the option stagnant is invoked with the porosities of the
mobile and immobile parts defined as fractions of the total volume. However, in Equation (6.45) the
porosity corrects for the surface area accessible to diffusion, and it is a fraction of the stagnant volume
there.

0.6
A: Mobile part
0.5

0.4 Cl/2
(cons)
mmol/L

0.3 Cl

0.2
CCl4
0.1

0
0 200 400 600 800 1000
Distance (m)
0.6
B: Immobile part
0.5

0.4
mmol/L

0.3
Cl
0.2
CCl4
0.1

0
0 200 400 600 800 1000
Distance (m)

Figure 10.15. Concentration profiles of Cl and CCl4 in the mobile part (A) and the stagnant part (B) of a dual
porosity aquifer after 5 years pollution and 10 years flushing.

The concentration profiles are plotted in Figure 10.15. The thin dotted line in Figure 10.15A indi-
cates the expected “conservative” concentration of Cl (divided by 2 for scaling to the y-axis) in the
absence of dispersion and without clay layers. In this case, the forward front of the pulse is located
at 750 m and the backward end is at 500 m. In the model situation, Cl shows rounding due to dis-
persion and retardation as part of the concentration moves into the immobile zone. Furthermore, the
pulse has become slightly asymmetric, with a broader backward than forward limb. The concentra-
tion of CCl4 has decreased more than of Cl because the chemical is sorbed in the clay. During flush-
ing, the clay releases CCl4 and the concentration remains higher than of Cl for an extended time
period. Finally, note that the mass of chemical in the system is given by the sum of the concentra-
tions (mol/L) times the volume fraction of water in the system, to which the sorbed amounts must be
added as well.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

510 Pollution by organic chemicals

QUESTIONS:
Run the PHREEQC file (download from www.xs4all.nl/appt) to obtain a chart of transport
without dispersion and without stagnant zones. The concentration peak is located between .......... and
.......... m.
Next, include dispersion and run the file (do not forget to fix the chart to allow for an easy
intercomparison).
Stagnant cells are numbered as:
cell_no  1  n  end_cell_no
where cell_no is the cell with mobile water to which the stagnant cell is connected, n is the
number of the stagnant layer, and end_cell_no is the last cell of the flowtube.
Include the stagnant zones, but display only the mobile zone.
Run the file again to display the concentrations in the stagnant zone. Why is the concentration of CCl4
smaller than of Cl?
Run the file to display the sorbed concentrations of CCl4 (plot mol (“SorTetra”)*1e3).
Investigate the effect of increasing the effective diffusion coefficient in the stagnant region
to 1010 m2/s.

10.3.3 Release from stagnant zones and blobs

An obnoxious aspect is the continuous bleeding of chemicals from stagnant zones or pools. In the
case of NAPL’s this may go up to their aqueous solubility with subsequent dilution by dispersal with
groundwater flow. Usually, the release of BTEX compounds (benzene, toluene, ethylbenzene and
xylene) from pools and tar blobs is modeled with a linear driving force expression that is similar to
the first order exchange relation (Miller et al., 1990, Powers et al., 1994):

M x  k rr A (cx, s  cx ) (10.26)

where Mx is the mass-flow of chemical x (mol/L/s), krr is the specific release rate (m2 s1),
A is the interfacial area among the blob and water (m2), cx,s is the concentration of a BTEX
compound in equilibrium with the blob (mol/L), and cx is the actual concentration of the
compound in the groundwater. To define the changing interface, it may be assumed that the
blobs consist of spheres, with a radius a (m) related to the mass of BTEX in the pool, mBTEX
in mol/L:
a  (mBTEX) ⁄
1
3 (10.27)

The surface area of the spheres is:

A  a2  (mBTEX) ⁄ 2
3 (10.28)

which gives in Equation (10.26):

(10.29)
Mx  korr (mBTEX / mBTEX,0) ⁄ (cx,s  cx)
2
3

where mBTEX,0 is the pool at the start (mol/L), and korr is the overall release rate from the pool
( krr A). The value of the exponent, 2⁄3, is the same as used in defining the kinetic dissolution rate
for ideal spheres or cubes (Chapter 4), but we have noted already that for minerals usually a higher
value is found because the surface characteristics change during dissolution. It is also interesting to
derive the value for BTEX pools from some observations.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 511

100

124_TMB

g/L
10

Benzene
1
0 20 40 60 80
Time (days)

Figure 10.16. Concentrations of BTEX compounds during the remediation of the aquifer below a former ben-
zene factory. TMB is trimethylbenzene (Eckert and Appelo, 2002).

The flux of chemical x will become independent of the concentration difference when the actual con-
centration cx is much smaller than the saturated concentration, cx,s. We expect then, that the rate
becomes:

M x ~ dmBTEX / dt  k orr ( mBTEX / mBTEX, 0 ) f cx, s

where f is the constant to be derived. When f  1, the pool decreases logarithmically with time, and con-
sequently, the BTEX concentrations in water also decrease logarithmically in time. When f  2⁄3, the
pool will decrease with the cubed power of time, and the concentrations decrease more quickly than
logarithmic.
Both concentration patterns were observed in a BTEX remediation scheme where groundwater
with KNO3 was injected to enhance the oxidation capacity (Figure 10.16). In an observation well the
concentration of 1,2,4-trimethylbenzene decreased logarithmically, while the concentration of
benzene decreased more rapidly as a result of biodegradation. The exponent of 2⁄3 was found applica-
ble in column experiments with toluene and benzene (Geller and Hunt, 1993).

Pools with mixtures of chemicals

The value of cx,s in Equation (10.26) is the aqueous solubility of the pure compound, corrected for
the activity in the pool of BTEX. These pools behave thermodynamically nearly ideal, so that the
activity is given by the mole fraction in the mixture (Feenstra and Guiguer, 1996; Eberhardt and
Gratwohl, 2002). For example, for trichloroethane, TCA, in a mixture of j chemicals:
mTCA
[TCA]  TCA  (10.31)
j
∑ mi
i1

and the saturated concentration is less than the solubility of the pure compound:
cCTA  STCA TCA (10.32)

where STCA is the solubility of TCA (in water STCA
1300 mg/L).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

512 Pollution by organic chemicals

We can estimate the fractions in the pool from concentrations in water as shown in Example 10.4.

EXAMPLE 10.4. Estimate the composition of a DNAPL pool in an aquifer

The solubilities of 1,1,1-trichloroethane (TCA) and tetrachloroethylene (PCE) in water are 1300 and
200 mg/L, respectively. Observed concentrations in groundwater are 2.9 and 1.7 mg/L. Estimate the mole
fractions of TCA and PCE in the pool assuming that only TCA and PCE are present and that the low
concentrations are the result of dilution.

ANSWER:
For TCA we have: 2.9 mg/L  STCA TCA / f
and for PCE: 1.7 mg/L  SPCE PCE / f  SPCE (1  TCA) / f

where f is the dilution factor.

We have two equations with two unknowns, which can be solved to:
f  93.2, TCA  0.21 and PCE  0.79.
The dilution factor of 93.2 indicates that polluted water, saturated with the chlorinated compounds in contact
with the pool, is mixed with uncontaminated groundwater in the ratio 1:92.2. When more compounds are
present, another equation is added which permits to estimate the unknown mole fraction.

QUESTION:
Estimate the mole fractions in the pool when also 4 mg 1,1,2-trichloroethane/L is present? This more polar
compound has a higher solubility of 4400 mg/L. Hint: start finding f  1 / (ci / Si).
ANSWER: f  85.9, PCE  0.73, TCA  0.19, 1,1,2-TCA  0.08.

As a result of different solubilities, the composition of a pool of NAPL’s will change in time because
the most soluble components disappear most rapidly. Likewise, the ratios of the aqueous concentra-
tions of the various compounds will change with time. For example, we can imagine an experiment
where TCA and PCE are mixed in equal molar amounts in a beaker, whereafter a number of times a
volume of water is added, and removed again after equilibration. Initially, TCA will have the highest
concentration of the two chemicals in water. However, since more TCA dissolves from the pool than
PCE, the mole fraction of TCA decreases. The lower activity of TCA gives lower aqueous concen-
trations in successive extractions until the pool is emptied of TCA. Since the mixture of organic liq-
uids is thermodynamically ideal (mole fraction  activity), we can model the experiment with
PHREEQC, using keyword SOLID_SOLUTIONS (Example 10.5).

EXAMPLE 10.5. Model the extraction of a DNAPL pool with PHREEQC

Mix 0.8 mmol TCA (106.7 mg) and 0.8 mmol PCE (132.6 mg) in a separatory funnel and add 100 mL water.
Shake the funnel and separate the aqueous solution. Repeat the extraction with water 9 times. Find the
aqueous concentrations with PHREEQC.

ANSWER:
The PHREEQC input file to model the experiment is:

SOLUTION_MASTER_SPECIES # Define the substances...

Tca Tca 0 Tca 133.4 # 1,1,1-trichloroethane C2H3Cl3, TCA
Pce Pce 0 Pce 165.8 # Tetrachloroethene C2Cl4, PCE

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 513

SOLUTION_SPECIES
Tca = Tca; log_k 0
Pce = Pce; log_k 0
# Define the solubility...

PHASES
Tca_lq; Tca = Tca; log_k -2.01
Pce_lq; Pce = Pce; log_k -2.92
# weigh the chemicals in the funnel...

SOLID_SOLUTIONS 1
Pool
-comp Tca_lq 0.0008
-comp Pce_lq 0.0008
# take 0.1 L pure water...

SOLUTION 1
-water 0.1
END
# mix water and chemicals...

USE solution 1
USE solid_solutions 1
SAVE solid_solutions 1 # the pool of chemicals after dissolution

USER_GRAPH # plot the composition...

-connect_simulations
-headings Extr_no TCA PCE
-start
10 graph_x sim_no - 1 # Extraction No. on x-axis
20 graph_y tot(“Tca”) * 133.4e3, tot(“Pce”) * 165.8e3
-end
END
# 2nd extraction..

USE solid_solutions 1; USE solution 1; SAVE solid_solutions 1

END
# 3rd extraction..
USE solid_solutions 1; USE solution 1; SAVE solid_solutions 1
END
# etc... 6 times more

The solubility of the pure chemical is defined with keyword PHASES. The name of the phase is given,
followed by the reaction and the equilibrium constant, here the solubility of the liquid chemical in
mol/L. The name is used again when the mixture of organic liquids is defined with keyword SOLID_
SOLUTIONS.
After the preparations (keyword END, which ends the first simulation), we start the experiment with USE
solution 1, USE solid_solutions 1. The composition of the mixture of organic liquids after extraction is saved
with SAVE solid_solutions 1. The extraction is repeated 9 times, each time the original solution 1 is used (it
keeps the initial composition if not saved) and the varying solid_solution (changing with each SAVE). Run the
file, and inspect the chart (Figure 10.17).
The concentrations in the first extraction are close to half of the solubility of the pure substance, in line
with the initially equal molar concentrations in the liquid mixture. In the subsequent extractions, the con-
centration of TCA decreases, while the concentration of PCE increases. When the concentration of TCA is
very small the concentration of PCE approaches the solubility of pure PCE (200 mg/L). In the 8th extrac-
tion the organic liquid dissolves completely, the available amount of PCE is insufficient to reach the
solubility limit.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

514 Pollution by organic chemicals

500

400

300

mg/L
TCA
200
PCE
100

0
1 2 3 4 5 6 7 8 9 10
Extraction no.

Figure 10.17. The varying aqueous concentrations of TCA and PCE in the extraction experiment of a
liquid mixture.

QUESTIONS:
Which chemical is more soluble, TCA or PCE?
ANSWER: TCA
From the input file, find the solubilities of TCA and PCE in mmol/L.
ANSWER: 9.7 mM TCA, 1.2 mM PCE
What would be the initial aqueous concentrations if the size of the pool were very large?
ANSWER: 4.85 mM TCA (650 mg/L), 0.6 mM PCE (100 mg/L)
Plot the fractions of TCA and PCE in the organic liquid on the secondary Y-axis. Hint: use the special
BASIC function s_s(“Tca_lq”) which gives the moles of TCA, and similar for PCE (note that the name of
the component is to be used and not the chemical formula).
Add 1.2 mmol of 1,1,2-TCA to the initial organic liquid and redo the extractions. The solubility in water is
4.4 g/L. What do you expect?
ANSWER: 1,1,2-TCA is extracted first, followed by a peak of 1,1,1-TCA, and
lastly of PCE.

Figure 10.18 shows an example where PCE and TCA concentrations are declining during the
remediation of a plume (Feenstra and Guiguer, 1996). The ratio PCE/TCA in an observation
well increases with time, in agreement with the lower solubility of PCE compared to TCA which
is the first to be dissolved completely. We expect that the dilution factor (f ) will increase with
time as the pool dissolves and is removed, as appears to be the trend in Figure 10.18, but there are
irregularities.
Actually, many problems arise when dilution factors and pool sizes are estimated. In the field, the
pools may be large and diffusion within the blob may limit the release rate of the compounds.
Moreover, sorption in aquifers will be different for the chemicals, which leads to chromatographic
separation. Both effects will change the concentration ratios of the components in water from the
calculated ones in equilibrium with an ideal pool.
Figure 10.19 shows results from an experiment where an equimolar mixture of benzene and toluene
was injected in the center of a column and subsequently eluted with water (Geller and Hunt, 1993). As
expected, benzene, with a solubility of 1780 mg/L, was eluted quicker than toluene which has a solubil-
ity of 515 mg/L. The concentrations in the effluent did not reach the theoretical solubilities because of the
high water flow velocity, and it can be calculated (cf. Question) that the dilution factor was 9 initially.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Sorption of organic chemicals 515

3 10 500

8 400

PCE/TCA ratio

Dilution factor
2
6 300
Dilution factor
mg/L

TCA
4 200
1
PCE
2 100
Ratio
0 0 0
1978 1979 1980 1981 1978 1979 1980 1981
(A) Year (B) Year

Figure 10.18. Concentrations of PCE and TCA in an observation well during remediation (A), and PCE/TCA
concentration ratio and water/(NAPL saturated water) dilution factor (B). (Concentrations from Feenstra and
Guiguer, 1996).

Thus, the NAPL formed a stagnant zone that exchanged relatively slowly with water that flowed
around it. In the course of the experiment, the dissolution rate increased because the contact surface
of the blobs and water increased. Hence, the dilution factor decreased. Furthermore, the dissolution
took place in heterogeneous fashion along the column. The NAPL in the first part of the column,
already emptied of more soluble benzene, released toluene at a higher concentration than was dis-
solved from the mixture of NAPL’s downstream. Besides these details (and of course, dispersion in
the column), the overall elution pattern is very similar to the batch extraction experiment that we
calculated before in Example 10.5.

125
 Toluene
 Benzene
100

75
mg/L

50

25

0
0 50 100 150 200
Pore volumes

Figure 10.19. Concentrations of toluene and benzene in the effluent of a column with an initially equimolar
amount of both chemicals in the column (Geller and Hunt, 1993).

QUESTIONS:
Give the theoretical (equilibrium) concentrations in the initial phase of Geller and Hunt’s experiment?
ANSWER: ctol  Stol  0.5  207 mg/L, cben  Sben  0.5  890 mg/L.
Find the dilution factor and toluene in Geller and Hunt’s experiment at PV  1 and 50?
ANSWER: PV  1: ctol  30 mg/L, cben  80 mg/L. f  9.7, tol  0.56.
PV  50: ctol  57 mg/L, cben  68 mg/L. f  6.1, tol  0.80.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

516 Pollution by organic chemicals

10.4 TRANSFORMATION REACTIONS OF ORGANIC CHEMICALS

Organic chemicals are transforming via manifold pathways in the environment. The reactions may
involve the substitution or elimination of atoms or groups at the carbon framework of the organic chem-
ical and are usually accompanied by oxidation or reduction. Take methyl iodide, a chemical that you may
have used while preparing microscopic thin slides. It is also produced by marine algae, and is probably
one source of the distinct odor of the coastal environment. With OH, it transforms to methanol:

OH  CH3I → CH3OH  I (10.33)

The reaction follows the kinetic rate:

dmCH I
3
 kOH mOH mCH I (10.34)
dt 3

where kOH  104.2/s/(mol/L) (Schwarzenbach et al., 1993, Figure 12.5). Two species figure in this
bimolecular rate. We can calculate the half-life of methyl iodide with this reaction in seawater with
pH  8.3, mOH  2.6106 mol/L:

ln(0.5)  104.2 mOH t ⁄ , or t ⁄ 4.2109 s  134 years.

1
2
1
2
(10.35)

However, the reaction with water is quicker:

H 2O  CH3I → CH3OH  H  I (10.36)

which has the same bimolecular rate mechanism, with kH2O  108.9/s/(mol/L). The half-life for this
reaction is:

ln(0.5)  108.9 mH t ⁄ , or t ⁄ 0.3 years.

1 1 (10.37)
2O 2 2

(since mH2O  55.6 mol/L).

An even faster reaction transforms methyl iodide into methyl chloride:

Cl  CH3I → CH3Cl  I (10.38)

with kCl  105.5/s/(mol/L). With this reaction the half-life of methyl iodide in seawater is only
0.012 years, and the toxic methyl chloride is produced. Much of this substance escapes to the strato-
sphere, where it is an ozone depleting gas.

QUESTION:
What is the half-life of methyl iodide in fresh water?
ANSWER: t ⁄  0.3 yr, mCl is too small for Reaction (10.38) to be effective.
1
2

All the methyl halogenides are thermodynamically unstable with respect to CO2 and the halogenide
ion, but they persist in nature because the reactions are kinetically controlled. The reaction rates for
anion substitutions have been determined in the laboratory or were deduced from field observations,
and were linked in the nucleophilicity index for anions. The nucleophilicity is the logarithm of
the reaction rate for an anion, relative to H2O. In the above examples of methyl iodide reactions,
chloride has a nucleophilicity index of (5.5  (8.9))  3.4. Since the concentration of water is

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Transformation reactions of organic chemicals 517

constant at mH2O  55.6 mol/L, a concentration of chloride of 55.6 /103.4  22 mmol/L suffices for
an equal reaction rate for chloride and water. The nucleophilicity index can be used to estimate the
kinetic reaction rates of similar chemicals (Example 10.6).

EXAMPLE 10.6. Estimate the hazard of groundwater pollution by methyl bromide

Methyl bromide is used as fumigant (disinfectant gas) in greenhouse gardening. The bimolecular rate con-
stant of the hydrolysis reaction with water is 108.4/s/(mol/L) (three times faster than for methyl iodide). The
most important anion in the soil water, HCO 3 , has a concentration of 3 mM, and its nucleophilicity index is
3.8. Estimate the hazard of groundwater pollution.

ANSWER:
The rate constant for hydrolysis by water is 108.4 mH2O  108.4/s/(mol/L)  55.6 mol/L  106.65/s. The rate
for substitution by HCO 3 is 10
8.43.8
 107.12/s, slightly smaller than for water. Hence, the half-life is
approximately 100 days. In five half-lives, or 11⁄2 year, the concentration will have diminished to 3%, and with
the usual travel time of groundwater of a few m/yr, the hazards of methyl bromide pollution are estimated to be
small. The Br concentration in groundwater may increase, however.

Nucleophilicity expresses the ease with which the carbon nucleus is approached and the substitution
reaction accomplished. If access is facilitated, the reaction will speed up. One way is to alter the shape
and orientation of the organic molecule by sorbing it onto a solid surface, thus to divulge the nucleus
where the reaction takes place. Enzymes in the cells of microbes play a similar role in adapting the ori-
entation of the organic molecule, rendering it more susceptible to the reactions. The half-life of methyl
bromide in experiments with unamended soil samples was about 100 hrs (Gan et al., 1998), which is
20 times faster than calculated for the pure hydrolysis reaction in Example 10.6, and low concentrations
of methyl bromide could be degraded in the order of minutes by soil microbes (Hines et al., 1998).
Another important type of reaction involves the elimination of a molecule or group, removing it
altogether, and installing a double bond instead. For example, 1,1,2,2-tetrachloroethane may convert
to trichloroethylene by reaction with OH:
Cl 2HC-CHCl 2  OH → Cl 2CCHCl  H 2O  Cl (10.39)

which is again a bimolecular reaction with kOH  2/s/(mol/L). At pH  7, it gives a half-life of

40 days for 1,1,2,2-TCA.
Trichloroethylene can be reduced with H2 or with Fe(0) (Gillham and O’Hannesin, 1994; Arnold
and Roberts, 2000) to dichloroethylene:
Cl 2CCHCl  H 2 → ClHCCHCl  H  Cl (10.40)

and quickly on to vinyl chloride:

ClHCCHCl  H 2 → ClHCCH 2  H  Cl (10.41)

which again may be reduced to the more innocent molecules ethylene (H2CCH2) and ethane
(H3C-CH3). The reaction with Fe(0) involves the reduction of H2O to produce H2 and Fe2:

Fe(0)  2H 2O → Fe 2  H 2  2OH (10.42)

where the resulting hydrogen enhances the reduction of the chlorinated compounds, perhaps assisted
by the pH-increase of the reaction. When Fe(0) is introduced in aquifers in reactive iron barriers,
the pH increases to 10–11, and various precipitates and sorption complexes remove Ca2 and Mg2
(Mayer et al., 2001; Yabusaki et al., 2001).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

518 Pollution by organic chemicals

The highly toxic and mobile vinyl chloride accumulates in aquifers when reducing agents such as
pyrite or an iron barrier wall are absent. However, as a fairly reduced substance, it is quite susceptible
to oxidation:

ClHCCH 2  2.5 O 2 → 2 CO 2  H 2O  H  Cl (10.43)

Generally, highly reduced BTEX compounds (benzene, toluene, ethylbenzene and xylene) are more
easily oxidized than reduced. They may be utilized as substrate by various microbes, including
denitrifiers, Fe(3) and SO2
4 reducers and fermentation bacteria (Baedecker et al., 1993; Chapelle,
2001). The oxidation/reduction reactions are catalyzed in sorbed form and accelerated by microbes
as discussed extensively in the next section.

10.4.1 Monod biotransformation kinetics

The concentrations of phenol and biomass during methanogenic degradation in a laboratory experi-
ment are shown in Figure 10.20 (Bekins et al., 1998). The phenol concentration decreases linearly
with time (characteristic for a zeroth order degradation rate), but the degradation seems to decline at
very small concentrations.

40
Phenol
Biomass
Phenol concentration (mg/L)

Monod no-growth
30 model

20

10

0
0 10 20 30 40
Time (days)

Figure 10.20. Phenol and biomass concentrations (both in mg/L) during methanogenic degradation of phenol
in a laboratory experiment (Bekins et al., 1998).

QUESTION:
Write a balanced reaction for the methanogenic degradation of phenol (C6H5OH) (cf. Table 9.1)?
ANSWER: C6H5OH  4H2O → 3.5CH4  2.5CO2.

The breakdown was modeled with the Monod kinetic rate equation:

dS S
 RMonod  k max (10.44)
dt k⁄ S
1
2

where S is the concentration of phenol (mg/L), t is time (s), kmax is the maximal rate
(mg/L/s), and k ⁄ is the half-saturation constant (mg/L). The degradation rate is just half of
1
2

the maximal rate when the concentration of phenol (S ) equals the half-saturation constant.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Transformation reactions of organic chemicals 519

1.4
1.2
kmax

Monod rate (mg/L/d)

1
0.8
0.6
0.4
k½
0.2
0
0 5 10 15
Concentration (mg/L)

Figure 10.21. The rate of a Monod reaction depends on the concentration of the reactant. The values k ⁄  1
2

1.7 mg/L and kmax  1.39 mg/L/d were found in the phenol experiment (Figure 10.20).

The Monod rate depends on the concentration of S and on the value of k ⁄ as illustrated in 1
2

Figure 10.21. When the concentration of S is small, S  k ⁄ , the rate increases with S. At higher con- 1
2

centrations, when S ! k ⁄ , the rate increases less and becomes more and more independent of S. The
1
2

Monod equation has been developed for modeling bacterial metabolism and S stands for the sub-
strate or the nutrient utilized by the microbes. At low substrate concentrations the bacteria consume
substrate in proportion with the concentration, but at high concentrations the consumption is
restricted to what can be utilized by the biomass, for example because an enzyme or another nutri-
ent is limiting the reaction.
Thus, when S can be neglected compared to k ⁄ in the denominator of Equation (10.44), it 1
2

becomes:

dS k (10.44a)
  max S
dt k⁄ 1
2

The rate is now first order with respect to S.

On the other hand, when k ⁄ is small compared to S, Equation (10.44) simplifies to:
1
2

dS
 k max (10.44b)
dt
The rate is now fixed to the maximal rate. It is zero order with respect to S, and therefore, the concen-
tration of S drops linearly with time as in the initial stage of the phenol experiment. The linear decrease
of the SO42 concentration in a marine sediment shown in Figure 4.14 is also typical for this domain.

EXAMPLE 10.7. PHREEQC model of phenol degradation

In their experiment of biodegradation of phenol, Bekins et al., found kmax  1.39 mg phenol/L/day and
k ⁄  1.7 mg phenol/L. Model the data with PHREEQC.
1
2

We make an input file in which phenol is defined as species and the Monod rate is programmed.
SOLUTION_MASTER_SPECIES; Phenol Phenol 0 94 94
SOLUTION_SPECIES; Phenol = Phenol; -log_k 0

RATES
S_degradation
# dS/dt = -k_max * S / (k_half + S). k_max is maximal growth rate, mol_phenol/L/s.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

520 Pollution by organic chemicals

# S is nutrient conc, mol phenol/L. k_half is half saturation conc., mol_phenol/L.

-start
1 k_max = parm(1); 2 k_half = parm(2) # parms defined in KINETICS
10 S = tot("Phenol")
20 if < 1e-9 then goto 50 # quit at low conc.
30 rate = - k_max * S /(k_half + S)
40 dS = rate * time
50 save dS
-end

SOLUTION 1; -units mg/L; Phenol 40.0

KINETICS 1
S_degradation
-formula Phenol 1
-m0 0
-parms 1.71e-10 1.81e-5 # k_max in mol phenol/L/s, k_half in mol/L
-steps 3.3e6 in 20 # 40 days
INCREMENTAL_REACTIONS true

USER_GRAPH
-headings time phenol
-axis_titles "Time / days" "mg phenol/ L"
-start
10 graph_x total_time / 86400
20 graph_y tot("Phenol") * 94e3
-end
END

Running this file will give the model line of Figure 10.20. The small program that defines the Monod rate
starts with assigning the values of parm(1) and parm(2) to kmax and k ⁄ in lines 1 and 2. These parameters are
1
2

defined under keyword KINETICS. Next, in line 10, the concentration of phenol is obtained in mol/L
(g/L) with the special function tot(“…”). If the concentration is below a limit of 109 mol/L, we are done
and leave the degradation calculation with dS undefined and thus equal to zero (goto line 50). Otherwise,
line 30 is processed, where the rate is written just like Equation (10.44). The rate is multiplied with “time”
to give the reaction amount dS in line 40. In the last line, dS is passed on to PHREEQC with “save”.

The Michaelis-Menten approach

The functioning of the Monod equation can be understood in terms of enzymes in the microbial cell
which bind the chemical and facilitate the reaction. Take the enzyme E, which associates with S to
form ES, and assume that the reaction rate depends completely on the concentration of ES:
dS / dt  k ES (10.45)
We rewrite ES in terms of the total enzyme concentration and the substrate concentration S. Thus, the
association reaction
[ ES ]
E  S ↔ ES ; K ES  (10.46)
[ E ][S ]
and a total enzyme concentration Et  E  ES give:

ES  1 
Et   ES  ES   1 (10.47)
K ES S  K ES S 

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Transformation reactions of organic chemicals 521

or

Et K ES S Et S
ES   (10.48)
1  K ES S ( K ES )1  S

This is the Michaelis-Menten equation. (Note that it is similar to the Langmuir isotherm, cf.
Equation (7.9), also note that we have left out the activity signs).
On combining Equations (10.45) and (10.48), we obtain:

dS k Et S
 (10.49)
dt ( K ES )1  S

Returning now to the Monod equation, we can see that the half-saturation constant, k ⁄ , equals the 1
2

inverse of the association constant, KES, in the Michaelis-Menten equation. Since the association has
the specific function to promote a reaction, KES will be on the high side in the range of association
constants, say 105 (cf. PHREEQC.DAT). The value of k ⁄ is therefore expected to be in the mol/L
1
2

range. (Note that we assumed that the complex ES forms instantaneously at equilibrium. If not, k ⁄ 1
2

may be larger than (KES)1, cf. Berg. et al., 2002).

Biomass effects
The Michaelis-Menten approach shows that the rate depends on the total enzyme concentration,
Et. Since the enzymes are located in bacterial cells or produced by them, the rate depends on the
number of microbes, and the latter may grow as substrate S is consumed. Thus, in the Monod equa-
tion, the rate of substrate consumption also depends on the number of bacteria present. Actually, in
Equation (10.44) the bacterial mass concentration was included in kmax:

k max   max B / Y (10.50)

where max is the specific bacterial growth rate (s1), B is the biomass (mol CH1.4ON0.2/L), and Y is
the yield factor. The biomass formula is for dry bacterial cells according to Rittmann and McCarty
(2001).
The yield factor indicates the fraction of substrate that is converted to biomass. Usually, Y is
given in (g biomass) / (g substrate) irrespective of the chemical composition of S (Madigan et al.,
2003). For a more direct link between reaction energetics and cell growth, Rittmann and McCarty
(2001) suggest to express Y in (electron-equivalents needed for cell growth)/(electron-equivalents
from S-degradation). For the same reason, and often resulting in simpler arithmetics, VanBriesen
(2002) uses (mol CH1.4O0.4N0.2/mol C in S).
For organic matter and methanogenic conditions, Y is small, about 0.05 (mol CH1.4O0.4N0.2/mol
C in S ) as discussed later. Now, one bacterial cell contains about 1014 mol carbon (Balkwill et al.,
1988), and thus, a yield factor of 0.05 means that degradation of 40 mg phenol/L will have increased
the bacterial population with:
(0.04 g phenol)  (0.05 mol C in biomass / mol C in phenol)  (6 mol C / mol phenol) /
(94 g/mol phenol) / (1014 mol C/bacterium)  1.31010 bacteria.
The bacterial cells consist of CH1.4O0.4N0.2 and are equivalent to
1.31010  1014  (22.6 g/mol CH1.4O0.4N0.2)  2.9 mg.
which is about the increase shown in Figure 10.20 (squares). In the phenol degradation experiment,
kmax was assumed constant and fitted together with k ⁄ on the experimental data.
1
2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

522 Pollution by organic chemicals

4
Experimental data
Ln phenol concentration (mg/L) First-order fit A
2 First-order fit A First-order fit B
kA  0.06/day
r 2  0.96
0
First-order fit B
kB  0.43/day
2
r 2  0.90

4

6
0 10 20 30 40
Time (days)

Figure 10.22. A logarithmic plot of phenol concentration vs time, showing two apparent first order rates
(Bekins et al., 1998). Note: the logarithmic plot appears in Example 10.7 when line 20 in USER_GRAPH is
changed to: 20 graph_y log(tot(“Phenol”) * 1  103).

Bekins et al. also plotted their data assuming first order degradation of phenol (Figure 10.22). For a
first order rate, the concentration of phenol should decrease logarithmically with time, but the plot
showed a non-linear trend of ln(phenol) vs time. The data were divided in two linear portions, each
with a different rate constant. Interestingly, the plot shows that the apparent first order rate constant
is greater for larger time than the initial rate constant (initially 0.06/day, finally 0.43/day). This
behavior in logarithmic plots is typical for Monod type kinetics and will be noted when the degra-
dation rate is monitored for both large and small concentrations of S compared to k ⁄ . 1
2

If the bacterial mass does vary, the change of B can be coupled to substrate decrease and to
biomass decay according to:
dB dS
 Y  k Bd B (10.51)
dt dt
where kBd is the first order death rate of the bacteria (s1). Neglecting the death rate (kBd  0), the
growth can be included in Equation (10.44):

dS ( B  Y ( S0  S )) S
 RMonod   max 0 (10.52)
dt Y k ⁄ S 1
2

where B0 is the initial bacterial mass (mol CH1.4O0.4N0.2/L), and S0 is the initial substrate concentra-
tion (mol C/L). Y(S0  S ) gives the cells which grew out of degradation of S0  S. This equation can
be integrated analytically (Alexander and Scow, 1989), but it may be easier to look at a solution with
PHREEQC as in Example 10.8.

EXAMPLE 10.8. Xylene degradation with biomass growth

Schirmer et al. (1999) measured breakdown of xylene (C8H10) in a batch experiment with aquifer
sediment. The vial with pristine sediment contained 1.47 mg biomass/L, of which 0.2% was active in
degrading aromatic hydrocarbons. When 16 mg xylene/L were added to the reaction vessel, the concentra-
tion dropped immediately to 8.6 mg/L which indicated that 46% of the chemical was sorbed to the sediment and

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Transformation reactions of organic chemicals 523

evaporated into the headspace of the vial. In other words, the distribution coefficient Kd  (16  8.6) /
8.6  0.86, and breakdown of the solute concentrations was retarded by R  1.86. The data can be modeled
with the Monod rate equation, with B0  7.5108 mol CH1.4O0.4N0.2/L, max  2.49105/s,
k ⁄  7.45 mol xylene/L and Y  0.305 (mol CH1.4O0.4N0.2/mol xylene-C) (experimental conditions at 10°C,
1
2

oxygen at 0.3 mM in excess). We extend the input file from Example 10.7 with a rate for biomass increase, and
the experimental details from Schirmer et al. Note that Schirmer’s parameters were recalculated from mg/L to
mol/L (cf. Questions).

SOLUTION_MASTER_SPECIES; Xylene Xylene 0 106 106 # C8H10

SOLUTION_SPECIES; Xylene = Xylene; -log_k 0

RATES
S_degradation
# dS/dt = -mu_max * (B/(Y * 8)) * (S / (k_half + S)) / R (mol xylene/L/s)
# mu_max is maximal growth rate, 1/s. B is biomass, mol C/L.
# Y * 8 is yield factor, mol biomass-C/mol xylene. S is xylene conc, mol/L.
# k_half is half saturation concentration, mol xylene/L. R is retardation, 1 + K_d
# K_d = q_xylene / c_xylene.
-start
1 mu_max = parm(1); 2 k_half = parm(2); 3 Y = parm(3); 4 R = 1 + parm(4)
10 S = tot("Xylene")
20 if S < 1e-9 then goto 60
30 B = kin("Biomass") # kin(".i.") gives moles of "Biomass"
40 rate = -mu_max * (B / (Y * 8)) * (S /(k_half + S)) / R
# 40 rate 5 -mu_max * (B / (Y * 8)) * (S /(k_half + S + S^2/8.65e-4)) / R
50 dS = rate * time
60 save dS # d mol(C8H10)
70 put(rate, 1) # Store dS/dt for use in Biomass rate
-end

Biomass
# dB/dt = dBg/dt - dBd/dt
# dBg/dt = - Y dS/dt is biomass growth rate. dBd / dt = k_Bd * B is biomass death rate.
# k_Bd is death rate coefficient, 1/s.
-start
1 Y = parm(1); 2 R = 1 + parm(2); 3 k_Bd = parm(3)
10 rate_S = get(1) * R # Get degradation rate, multiply by Retardation
20 B = m
30 rate = -(Y * 8) * rate_S - k_Bd * B
40 dB = rate * time
50 save -dB # dB is positive, counts negative to solution
-end

SOLUTION 1; -units mg/L; Xylene 8.6

KINETICS 1
# xylene, Schirmer et al., 1999. JCH 37, 69-86, expt. Figure 6b...
S_degradation; -formula Xylene 1; -m0 0
-parms 2.49e-5 7.45e-6 0.305 0.86 # mu_max, k_half, Y, K_d
Biomass; -formula C 0;-m0 0.75e-7 # 1.33e-7 with Haldane inhibition
-parms 0.305 0.86 0 # Y, K_d, k_Bd
-steps 0.6e6 in 50
INCREMENTAL_REACTIONS

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

524 Pollution by organic chemicals

USER_GRAPH
-headings time c_xylene Biomass
-axis_scale y_axis 0 10 2 1; -axis_scale x-axis 0 6 1 0.5
-axis_titles "Time / days" "mg / L"
-start
10 graph_x total_time / 86400
20 graph_y tot("Xylene") * 106e3, kin("Biomass") * 22.6e3
-end
END

Results of the calculation are shown in Figure 10.23, together with the experimental data. Quite conspi-
cuous is the delay of xylene breakdown because the microbes which degrade the substance first must grow
and increase in number. When the biomass is sufficient, xylene is quickly gone. The rate for S_degradation
takes this into account by multiplying with B / (Y  8) (mol xylene/L). Initially, B is small
(B0  “m0”  75 nmol/L in “Biomass”, keyword KINETICS), but as xylene is used for energy and growth,
B increases. Note how the degradation rate (dS / dt) is passed on to the biomass growth rate with the special
BASIC functions “put” and “get”. Because the degradation rate for solute xylene is retarded by sorbed and
headspace xylene, it is multiplied with the retardation factor R  (1  Kd) to obtain the rate for total xylene.
Line 50 of the Biomass rate definition “saves” negative moles, and because dB is positive, the kinetic reac-
tant (mass m of Biomass) increases. Furthermore, note that the death rate of the microbes is set to zero in
this example (parm(3)  0 in the call for Biomass in KINETICS).

10
Xylene Biomass
8

6
mg/L

0
0 1 2 3 4 5 6
Days

Figure 10.23. Microbial degradation of m-xylene in batch experiments (Schirmer et al., 1999). Datapoints
and model line for solute xylene, dotted line for biomass.

The kinetic reactant “Biomass” is not a solute species and its reaction does not change the solution compo-
sition, which is indicated by the zero coefficient in “-formula C 0”. In fact, these kinetic calculations could
be done without any change in solution composition, only referring to the moles of kinetic reactants.
When Schirmer et al. added initially more xylene to the reaction vessel, the lag time increased. This was
modeled with an inhibition factor of the form S2 / kHD in the denominator of the Monod rate (with
kHD  0.865 mmol S/L, the denominator becomes (k ⁄  S  S2 / 8.65104)). Also, they used the initial
1
2

concentration of B0  133 nmol/L. If the initial biomass is made 100 mol/L, and inhibition is absent,
degradation starts immediately and the xylene concentration decreases linearly with time, similar to phenol
in Figure 10.20.
When doing the following questions, it is helpful to compare the results of the simulations directly, fix-
ing the chart of PHREEQC as can be modified in Edit, Preferences, Miscellaneous.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Transformation reactions of organic chemicals 525

QUESTIONS:
Explain the increase of biomass quantitatively?
ANSWER: (8.6 mg xylene/L)/ (106 g/mol xylene)  (8 mol C/mol xylene) 
(R  1.86)  (Y  0.305)  (22.6 g/mol CH1.4O0.4N0.2)  8.32 mg biomass.
Change the yield factor from 0.305 to 0.5, explain the results?
ANSWER: degradation starts later, final biomass becomes 8.32  0.5 / 0.305 
13.6 mg/L.
Change the yield factor back to 0.305 and make the initial biomass 100 mol/L in the input file, compare
the results.
ANSWER: degradation starts immediately with zero order rate.
Include the inhibition factor S2 / kHD and change m0, compare the results.
ANSWER: Result is the same as in Figure 10.23.
What will be the effect of zero sorption of xylene (Kd  0)?
ANSWER: solute xylene is exhausted quicker; final biomass  8.32 / 1.86  4.47 mg/L.
And if you have bacterial decay (kBd  1  106)?
ANSWER: xylene degradation starts later, biomass decreases when xylene is gone.
Model the retardation as sorption reaction, cf. Example 10.3.
Estimate the initial number of active bacterial cells in Schirmer’s experiment?
ANSWER: (1.7106 g/L) / (22.6 g/mol) / (1014 mol C/cell)  7.5106 bacteria/L.
Schirmer expressed the yield factor as mg biomass/mg xylene. Find his Y?
ANSWER: 0.305  (8 mol C/mol xylene)  (22.6 g/mol CH1.4O0.4N0.2) / (106 g/molxy-
lene)  0.52
He also expressed k ⁄ and kHD in mg xylene/L. Derive his values?
1
2

ANSWER: 7.45103  106  0.79 mg/L; kHD 0.865  106  91.7 mg/L
And lastly, obtain Schirmer’s max?
ANSWER: 2.49105/s, does not change.

Multiplicative Monod equations

In the experiments of Schirmer et al. (Example 10.8) oxygen was available in excess for the xylene
degradation reaction, but this may not be the case in plumes of BTEX-chemicals in an aquifer. Here,
oxygen must diffuse and disperse into the plume and its concentration will limit the breakdown rate.
This dependency is modeled with the multiplicative Monod equation, in which the concentration of the
electron acceptor appears as a multiplier in the same form as the electron donor (MacQuarrie et al., 1990;
Schirmer et al., 2000; Prommer et al., 2000; Rittmann and McCarty, 2001). For example for oxygen:

dS  k S   k max,O2 mO2  (10.53)

   max   
dt k ⁄  S   k ⁄ O  mO 
1
2
1 ,
2
2 2

where k ⁄ ,O is the half-saturation concentration for oxygen (mol/L) and kmax,O is the maximal rate-
1
2 2 2

multiplier for oxygen (). In the experiment of Schirmer et al. the whole term was included in kmax.
If another electron acceptor is present such as NO 3 , another rate equation is written for the same sub-
strate using NO 3 instead of oxygen in the multiplicative part. If moreover, the oxidants are known to
be utilized stepwise, yet another multiplier can be added which limits the reaction as long as the
other oxidant is present. The equations then tend to become rather formidable (Essaid et al., 1995,
2003; Van Cappellen and Wang, 1996; Boudreau, 1996; MacQuarrie and Sudicky, 2001) but the
logic is clear if simple steps are followed (Rittmann and McCarty, 2001). On the other hand, the
equations seem to have lost the touch of the microbiological system in which different groups of
bacteria utilize different substrates, each of them operating at a different pace that is probably dic-
tated by the number of bacteria, and their growth and death rates.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

526 Pollution by organic chemicals

Empirically, the usual oxidants in groundwater have large differences in their kmax,i, undoubtedly
because the various bacterial groups are present in optimal form for the specific conditions of the
aquifer. The large difference shows up in the persistence of nitrate when oxygen has been depleted,
and of sulfate when nitrate has been utilized already completely. The multiplier that limits the reac-
tion by the oxidant concentration (Equation 10.53) then is no longer necessary. Based on the degra-
dation rates of organic carbon in aerobic soils and of nitrate reduction by organic carbon in aquifers,
the following combination was obtained for the oxidant multiplier (Appelo and Parkhurst, 1998):

 mO 0.01 mNO 6.4  105 mSO2 

rOxidants   2
 3
 4
 (10.54)
 2.94  104  m 1.55 104  m 1 104  m 
 O2 
NO3 SO4
2


where mi is molality. The degradation of natural organic carbon can now be calculated with:

dSOC
 k1 SOC rOxidants (10.55)
dt

where SOC is organic carbon content (mol/kg soil), and k1 is the first order decay constant/s. With
k1  1.57109/s and mO2 0.3 mM, Equation (10.55) gives a first order degradation rate of 0.025/yr,
or 2.5% of organic carbon degrades each year, in accordance with estimates for aerobic soils in a tem-
perate climate. The value for k1 is for organic carbon in soils, and may be changed for older organic car-
bon (Equation 9.61). Note that Equation (10.54) is purely empirical, it does not account for the number
of bacteria that process the 3 oxidants, which will differ from place to place, but it may give a first guess.

Monod rate coefficients

Values for rate coefficients have been noted in the literature, but these require careful consideration
before comparisons can be made. Clearly, we would like to know the number of bacteria or the
microbial mass and the yield factor Y since the rate is directly depending on it. However, these
parameters are often not available, were assumed constant, or included in the value of kmax which is
the easiest factor for describing the system. Moreover, the yield factor is usually expressed simply
as mg dry biomass/mg substrate, with neglect of the different formula weights of the two (cf. Questions
with Example 10.8). Finally, the oxidant species and its concentration will influence the overall rate,
but usually these are only accounted for in models. A few values for max (or kmax) and k ⁄ (for the 1
2

organic compound) are noted in Table 10.4 together with the experimental conditions.

Table 10.4. Monod coefficients for degradation of organic chemicals in aquifers.

Chemical max k⁄
1
2
Reference
(s1) (mol/L)

BTX 5105 2.2104 Goldsmith and Balderson, 1988, batch with contaminated soil, oxygen
xylene 1.06104 1.4104 Kelly et al., 1996, batch with contaminated soil, oxygen
BTX 5.6105 2104 Kelly et al., 1996, batch with contaminated soil, oxygen
4.35106 2104 Eckert and Appelo, 2002, contaminated aquifer, 2.5 mM sulfate. kmax (in
mol/L/s) decreases by 100 in less contaminated parts of the aquifer
toluene 5.69106 7.1106 MacQuarrie et al., 1990, Column with pristine sediment, oxygen
m-xylene 2.49105 7.45106 Schirmer et al., 1999. Batch with pristine sediment, oxygen (cf. Example
10.8)
phenol 1.16107 1.81105 Bekins et al., 1998. Inoculated sediment, methanogenic (cf. Example 10.7)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Transformation reactions of organic chemicals 527

It has been noted that the various BTX compounds (benzene, toluene and xylene) show similar
degradation rates under identical conditions (Kelly et al., 1996). A conclusion that can be drawn
from Table 10.4 is that k ⁄ is small (as expected from Michaelis-Menten kinetics), and tends to be
1
2

smaller in uncontaminated sediments, which indicates that the biomass is more prone to devour even
small concentrations of the compound when nutrients are restricted as is usually the case in
an uncontaminated environment. Where max is given, the Monod rate must be multiplied with B/Y
(the active biomass (mol C/L) divided by the yield factor (mol C in biomass/mol C in S)), to obtain
the rate in mol carbon in S/L/s. To do so in a general sense the number of active bacterial cells
(number/L) is multiplied with the average carbon content per cell (1014 mol C/cell), and divided by
the yield factor in mol C per mol C in the nutrient. In soils, the living biomass is about 5% of organic
carbon content, or about 1010 cells per g soil with 3% OC. In lake sediments and shallow aquifers the
concentration of cells was estimated to range from 105/g to 107/g, and in deep aquifers as many as
104–106 culturable organisms per gram sediment were obtained (Barns and Nierzwicki-Bauer,
1997). Of all the cells, only a part will be active in a specific process. Thus, for an aquifer,
B  (108 cells/kg)  (1014 mol C/cell)  (6 kg/L for b/)
 (0.1 for an estimated active part)  6107 mol CH1.4O0.4N0.2/L.
Then, for methanogenic degradation of phenol,
kmax  (1.16107/s, Table 10.4)  (6107 mol CH1.4O0.4N0.2/L)/
(Y  0.05 (mol CH1.4O0.4 N0.2/mol C in phenol)  (6 mol C/mol phenol))
 (94  103 mg/mol phenol)  (86400 s/day)  1.9 g phenol/L/day.
The degradation of phenol in anaerobic aquifers is indeed slow (Thornton et al., 2001).

QUESTION:
Compare kmax calculated above with the value of Bekins et al. (1998). Hint: recalculate B from 10 mg/L
into mol C/L.
ANSWER: B  0.01 / 22.6  4.4104 mol C/L, gives in the inoculated sediment
kmax  1.4 mg phenol/L/day.

The yield factor

The yield factor Y in Equation (10.50) relates the growth of the bacteria with the reaction, and we
noted that Y is small for methane producing reactions and increases in an aerobic milieu. Can we
explain that, at least in a qualitative sense? For growth, the microbes need energy and chemicals for
the synthesis of the organic molecules of the cell. Most experimental microbiological studies add
nutrients and assure that the supply of chemicals is sufficient. With plenty of nutrients, it is the reac-
tion energy released with substrate degradation that determines the energy available to the cells and
thus the growth that can be sustained. The synthesis of a bacterial cell from pyruvate (C3H4O3) as
substrate, requires about 75 kJ/mol CH1.4O0.4N0.2 (Rittmann and McCarty, 2001).
Since more energy is released when oxygen is used as electron acceptor than in the case of fer-
mentation (Table 9.2; Figure 9.15), the yield factor for aerobic systems should be higher than for
anaerobic conditions (Lensing et al., 1994; Rittmann and McCarty, 2001; Watson et al., 2003). If we
compare the breakdown of pyruvate by fermenting and aerobic bacteria, the aerobic ones gain
1138 kJ/mol pyruvate and their fermenting friends only 115.6 kJ/mol pyruvate, or 9.8 times less
(reaction energies from Rittmann and Mcarthy, 2001). Thus, also 9.8 times more aerobic bacteria can
form. If the two microbial groups are present together and the reaction rates are comparable for the
breakdown of pyruvate, then clearly, the oxygen reducers will quickly outcompete the fermenters in
number and use up all the available pyruvate and oxygen in the proportion given by the reaction

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

528 Pollution by organic chemicals

stoichiometry. First when oxygen is depleted, the fermenters stand a chance to consume the pyruvate
substrate. Apparently, we have growth of the fittest who can snatch the energy (cf. Question).
The weak point in the theory is that the reaction rates of substrate breakdown by the different bac-
teria are variable and the parameters in the rates not exactly known. If the reaction rate is 9.8 times
higher for the fermentation reaction than for aerobic oxidation, the energy advantage would be egal-
ized and both groups grow with equal handicap. Various bacteria show only about two-fold growth
increase when nitrate is added to an anaerobic culture with glucose as substrate, and up to four-fold
increase when oxygen is added (Stouthamer, 1976). Thus, the observed growth-rate increase is
smaller than expected from pure energy consideration, which can be related to low reaction efficiency
or entropy loss (Heijnen and Van Dijken, 1992; VanBriesen, 2002). Moreover, the increase is twice
larger for oxygen than for nitrate, which can not be explained simply by energetics, since the two oxi-
dants provide nearly equal energy in the oxidation reaction. On the other hand, when chlorophenol,
nitrate or sulfite were provided as electron acceptors to a bacterium that was able to degrade chlori-
nated aromatic hydrocarbons, the yield factor doubled almost uniformly compared with an anaerobic
culture in pyruvate without electron acceptors (Mackiewicz and Wiegel, 1998). These oxidants give
about equal energy. Also, VanBriesen (2002) compared yield factors from experiment with estimates
based on reaction energies of various substrates with oxygen and found a good agreement (Figure
10.24). The observed yield (mol CH1.4O0.4N0.2/mol C in the chemical) varied from 0.57 for glycerol
(C3H8O3, the most reduced substance in the comparison) to 0.22 for glyoxylate (C2H2O3, the most
oxidized substance). Rittmann and McCarty (2001, Table 3.1) list yield factors that vary from 0.7 for
aerobic heterotrophic bacteria to 0.05 for methanogens and sufate reducers which use H2 as electron
donor, which shows the same trend of decreasing yield with smaller reaction energy.

0.65

0.6
Predicted theoretical yield (C-mol/C-mol)

0.55

0.5

0.45

0.4

0.35

0.3

0.25

0.2

0.15
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
Reported maximum yield (C-mol/C-mol)

Figure 10.24. Comparison of predicted theoretical and reported maximum yield for various substrates in
aerobic cultures. Triangles and circles are based on the methods proposed by Rittmann and McCarty (2001)
and Heijnen and Van Dijken (1992), respectively. Experimental points are given on the solid line with 1:1
correlation, with error bars for reported yields. The dotted lines indicate 15% above and below reported yields.
(VanBriesen, 2002).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetic complexation of heavy metals on organics 529

The small yield factor for anaerobic processes translates in the slow start of fermentation compared to
aerobic digestion in nature. Again, it is important to consider the timeframe when interpreting water
quality patterns in an aquifer. In Chapter 9, we have related the water quality transitions in the plume
below a waste site to oxidation reactions in the aquifer (Figure 9.18). Models typically attribute the water
quality patterns to sequential biodegradation reactions in the aquifer and assume a constant composition
of the leachate over time, that is given by the reduced, methanogenic water sampled directly below the
waste site (Brun et al., 2002; Barry et al., 2002). However, it is likely that some time must pass before
the leachate attains the methanogenic stage. The waste must be wet and sufficiently compacted so that
oxygen diffusion is inadequate for aerobic degradation. It takes time before the bacterial population
reaches the methanogenic stage, first a butyric acid stage must be ended that is characterized by high
concentrations of organic acids and accompanied by high alkalinity concentrations. (These alkalinities
are due to the organic acids, they are not from HCO 3 ). The spatial sequence in the aquifer has been
observed as a time sequence in waste percolates that leached from waste sites and never did contact the
aquifer below (Kjeldsen et al., 2002). The time sequence is related to the gas composition in the waste;
it may take a decade before the methanogenic stage is reached (Farquhar and Rovers, 1973). Thus,
instead of showing the reactions in the aquifer, the lateral concentration pattern in the plume probably
represents the time variations in the composition of the waste site leachate, the initial, aerobic percolate
having traveled farthest away from the site and the final, methanogenic stage being closest to the waste.

QUESTION:
Give the number of bacteria (n) after 24 hr for duplication times (d ) of 1 hr (for aerobic bacteria) and 9.8 hr
(for fermenting bacteria), starting with n0?
ANSWER: n / n0  2t/d yields 1.7107 and 5, respectively.

10.5 KINETIC COMPLEXATION OF HEAVY METALS ON ORGANICS

We have noted that natural organic matter and organic chemicals can complex heavy metals and sol-
ubilize them (Section 7.5). Examples of concern are americium from nuclear waste bound to humic
acid (Artinger et al., 1999) and gadolinium-pentatetic acid which has a medical application for mag-
netic resonance imaging (MRI) and can be traced in natural waters that receive effluents from hospi-
tals (Knappe et al., 1999; Kummerer and Helmers, 2000). Natural organic matter, even in the
dissolved state, forms large molecules with a charge-dependent potential which necessitates a special
correction term when calculating complexation (discussed in Section 7.5). Complexation of heavy
metals on smaller, artificial compounds such as NTA and EDTA which are used in industrial deter-
gents and domestic washing-powders can be calculated in a more straightforward manner, but the
kinetics of the complexation reactions are often slow (Hering and Morel, 1989). The kinetics of com-
plex formation can be important if only a specific complex is biodegradable (VanBriesen et al., 2000).
We will discuss an example of EDTA (ethylene diamine tetra acetate) in the Glatt river in Switzerland,
which receives EDTA as the FeEDTA complex from sewage treatment plants. The FeEDTA forms
because FeCl3 is added to precipitate and coagulate phosphate and remove it from the waste water. In
the river, iron is exchanged for Ca2 and Zn2, in a slow, kinetic exchange (Xue et al., 1995).
The equilibrium speciation of EDTA for the river conditions shows that ZnEDTA
and CaEDTA are the major complexes over the pH range 7.5–8.5 (Figure 10.25A). The total
EDTA concentration in this figure is 0.06 M (107.22 M). Zn is extremely strongly complexed
by EDTA and is able to displace Ca2 from the complex already at a Zn / Ca (total
concentration) ratio of 5105 (mZn  107 M in Figure 10.25B). Lastly, the effect of increasing
the EDTA concentration is illustrated in Figure 10.25C. The various complexes increase 1:1 with
EDTA, except for ZnEDTA, which levels off when it is limited by the total Zn concentration.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

530 Pollution by organic chemicals

7 7 6
ZnEDTA
CaEDTA
CaEDTA 7 ZnEDTA
8 8
ZnEDTA 8 CaEDTA
9 9
log (MEDTA)

log (MEDTA)

log (MEDTA)
9 CuEDTA
CuEDTA
10 10 MgEDTA 10
MgEDTA CuEDTA MgEDTA
11
11 11
FeOHEDTA FeOHEDTA
12
FeOHEDTA
12 12 FeEDTA
FeEDTA 13
FeEDTA
13 13 14
7.5 8 8.5 9 8 7.5 7 6.5 8 7.5 7 6.5 6
(A) pH (B) log (Zn) (C) log (EDTA)

Figure 10.25. Equilibrium speciation of EDTA computed for the conditions of Glatt river water. The concen-
trations of the complexes are plotted as function of pH (A), of total Zn concentration (B), and of EDTA con-
centration (C). The initial water composition is given in Example 10.9. (Xue et al., 1995).

The data in the figure were calculated with Microql, and can be calculated with PHREEQC with the
Minteq database. However, the constants must be adapted to fit the plotted concentrations and the
alkalinity must be crossed out, as shown in Example 10.9.

EXAMPLE 10.9. Speciation of EDTA in Glatt river water

The speciation calculation for Glatt river water detailed in the PHREEQC input file turns out to
be different from what is shown in Figure 10.25A, at pH  8.25. However, a reasonable agreement can be
obtained when a few complexation constants are adapted and the alkalinity is crossed out (apparently, the
complexes with HCO 3 are absent in Microql, used for Figure 10.25). Also, a complex of Cu
2
with DOC is
included in the input file as was noted by Xue et al., but with a modified constant.

DATABASE minteq.dat
SOLUTION_MASTER_SPECIES; Lcu Lcu-2 0 1 1 # DOC for Cu
SOLUTION_SPECIES
# Complex Cu with DOC...
Lcu-2 = Lcu-2; -log_k 0; Lcu-2 + Cu+2 = LcuCu; log_k 11.5 # Xue et al. used 13.6
# Modify Minteq database...
Zn+2 + EDTA-4 = ZnEDTA-2; log_k 17.04 # Minteq 16.44
Fe+3 + EDTA-4 = FeEDTA-; log_k 26.9 # Minteq 27.7
FeOH+2 + EDTA-4 = FeOHEDTA-2; log_k 21.49 # Minteq 21.99

SOLUTION 1 Glatt river water

pH 8.25; pe 12 O2(g) -0.68;
temp 25; # Alkalinity 3.97
Na 1.2; K 0.18; Mg 0.71; Ca 1.83
Zn 170e-6; Cu 31.5e-6; Fe 0.2e-3 Ferrihydrite -1.69
EDTA 60e-6; NTA 20e-6; Lcu 40e-6
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetic complexation of heavy metals on organics 531

which gives the speciation for EDTA (partial results):

-------------------------------Distribution of species-------------------------------
Log Log Log
Species Molality Activity Molality Activity Gamma
EDTA 6.000e-08
ZnEDTA-2 4.061e-08 2.940e-08 -7.391 -7.532 -0.140
CaEDTA-2 1.909e-08 1.382e-08 -7.719 -7.860 -0.140
CuEDTA-2 1.816e-10 1.314e-10 -9.741 -9.881 -0.140
MgEDTA-2 1.178e-10 8.526e-11 -9.929 -10.069 -0.140
FeOHEDTA-2 5.662e-12 4.099e-12 -11.247 -11.387 -0.140
FeEDTA- 9.949e-13 9.177e-13 -12.002 -12.037 -0.035

The differences in the databases are not unusual and should be considered when doing these
calculations. For example, Xue et al., use ferrihydrite with a solubility constant of 1038.8, while the
constant is 1037.11 in MINTEQ.DAT. Thus, ferrihydrite is more soluble in the minteq
database, which is mimicked in the input file by imposing subsaturation with SI  1.69.

QUESTIONS:
Check the numbers in the column Log Molality with Figure 10.25A?
Find the concentration of ZnEDTA2 and CaEDTA2 with the minteq constants?
ANSWER: cross out the complex definitions, log mZnEDTA2  7.648; log mCaEDTA2 
7.433

The driving force for displacing iron from the FeEDTA complex is the low solubility of iron-
hydroxide which gives a very small solute iron activity. The analyzed iron concentration in water can still
be substantial because small, colloidal hydroxides may pass even the 0.05 m filter that was used by
Xue et al. However, the breakup of the FeEDTA complex is slow, and at least part of the analyzed
iron will be present as the FeEDTA complex. The slow decline of the complex was followed
over time after 1.25 M FeEDTA had been added to a river water sample, as shown in Figure 10.26.

1.2
0
In((EDTA) /(EDTA) )
0

1.0 0.5
Photolabile EDTA (µmol/L)

1
1

0.8 1.5

2

0.6 2.5
7 27 47 67
Day of reaction
0.4

0.2

0.0
0 200 400 600 800 1000 1200 1400 1600
Reaction time (hr)

Figure 10.26. Concentration change of FeEDTA added to Glatt river water as a function of time. Inset shows
logarithmic (first order) plot of the data.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

532 Pollution by organic chemicals

The half-life of the first order reaction can be found to be 17.5 days from the inset at ln(0.5)  0.69.
In that time, the river may have flown a distance of 2000 km, flushing the complex out from Switzerland
down into the North Sea via the River Rhine (if it is not degraded in the meantime).
With Ca- and ZnEDTA as the major species in the Glatt river water, the exchange follows the
reactions:
Ca 2  FeOHEDTA 2 → CaEDTA 2  FeOH 2 (10.56a)

Zn 2  FeOHEDTA 2 → ZnEDTA 2  FeOH 2 (10.56b)

Ca 2  ZnEDTA 2 → CaEDTA 2  Zn 2 (10.56c)

Zn 2  CaEDTA 2 → ZnEDTA 2  Ca 2 (10.56d)

and similar for the other complexes (their concentrations are less important). The exchanged iron
precipitates:
FeOH 2  2H 2O → Fe(OH)3  2H (10.56e)

We can write kinetic equations for the formation of the EDTA complexes, for example for
ZnEDTA2, Reaction (10.56b):

d[ZnEDTA 2 ] / dt  k fw [Zn 2 ] [FeOHEDTA 2 ]  kbw [ZnEDTA 2 ] [FeOH 2 ] (10.57)

where kfw is the forward rate coeffcient (kfw  10/s, Xue et al., 1995) and kbw the backward rate
coefficient (s1). The backward rate coeffcient can be obtained from detailed balancing
(d[ZnEDTA2] / dt  0):

kbw  k fw K FeOHEDTA 2 / K ZnEDTA 2  10  1021.49 / 1017.04  28184/s (10.58)

where KZnEDTA2 is he association constant for the reaction:

Zn 2  EDTA 4 ↔ ZnEDTA 2; K  1017.04 ,

and similar for KFeOHEDTA2 (cf. Example 10.9). Already, we can note that the backward rate coeffi-
cient is much higher than the forward counterpart.
The reaction of Zn2 with CaEDTA2 (Reaction 10.56d) is added to (10.57) to obtain the total
change of [ZnEDTA2], and a rather formidable set of equations results. Some assumptions do per-
mit to simplify the system and to derive analytical solutions for parts of the reactants (Xue et al.
(1995), see also Atkins and de Paula (2002) for the basic explanations for doing these manipula-
tions). Here, we will calculate the system numerically with PHREEQC (Example 10.10).

EXAMPLE 10.10. Kinetic exchange of Fe(3)EDTA

Calculate the kinetic exchange of Ca2 and Zn2 with 6108 M FeOHEDTA2 added to Glatt river water.
Use kfw  10/s for Zn/Fe and Ca/Fe exchange, and 1200/s for Ca/Zn exchange (Xue et al., 1995).

ANSWER:
We make an input file in which rates are defined for the EDTA complexes according to Equations (10.57)
and (10.58). The rates for CaEDTA2 and ZnEDTA2 are explicitly written out, and the rate for
FeOHEDTA2 follows by mass balance. For the complex activity, the moles of kinetic reactant are used. The
mutual interaction of the complex reactions makes this a “stiff ” set to solve with very different rates for the

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Kinetic complexation of heavy metals on organics 533

two complexes. The Runge-Kutta procedure that is programmed in PHREEQC tends to take small timesteps
to achieve the required tolerance, but the calculations can be quickened with the Ordinary Differential
Solver that is implemented in PHREEQC-2.9. We note that the concentrations of EDTA are very low, so that
the tolerance for the kinetic calculation must be adapted (default tolerance is 108 moles, which is decreased
to 1012 moles). We use a simplified Glatt river composition, and precipitate the exchanged iron as
ferrihydrite.

PRINT; -status false

PHASES
Ferrihydrite; Fe(OH)3 + 3H+ = Fe+3 + 3H2O; log_k 4.891
RATES
CaEDTA
# E is short for EDTA...
# Reactions: FeOH-E + Ca = Ca-E + FeOH & Zn-E + Ca = Ca-E + Zn
#
# d(Ca-E)/dt = k_f1 * [FeOH-E] * [Ca+2] - k_b1 * [Ca-E} * [FeOH+2]
# + k_f2 * [Zn-E] * [Ca+2] - k_b2 * [Ca-E} * [Zn+2]
#
# k_f1 = 10 /s. k_b1 = K_FeOH-E/K_Ca-E * k_f1
# k_f2 = 1200 /s. k_b2 = K_Zn-E/K_Ca-E * k_f2
#
# parms: 1= k_f1. 2= K_FeOH-E. 3= K_Ca-E. 4= k_f2. 5= K_Zn-E.
-start
10 k_b1 = parm(1) * parm(2) / parm(3)
12 FeE = kin("FeEDTA")
20 rate = parm(1)*FeE*act("Ca+2") - k_b1 * m * act("FeOH+2")
30 k_b2 = parm(4) * parm(5) / parm(3)
32 ZnE = kin("ZnEDTA")
40 rate = rate + parm(4)*ZnE*act("Ca+2") - k_b2 * m * act("Zn+2")
50 moles = rate * time
60 save -moles
70 put(moles, 1)
-end

ZnEDTA
# Reactions: FeOH-E + Zn = Zn-E + FeOH & Ca-E + Zn = Zn-E + Ca
# d(Zn-E)/dt = ... as for CaEDTA
#
# parms: 1= k_f1. 2= K_FeOH-E. 3= K_Zn-E. 4= k_f2. 5= K_Ca-E.
-start
10 k_b1 = parm(1) * parm(2) / parm(3)
12 FeE = kin("FeEDTA")
20 rate = parm(1)*FeE*act("Zn+2") - k_b1 * m * act("FeOH+2")
30 k_b2 = parm(4) * parm(5) / parm(3)
32 CaE = kin("CaEDTA")
40 rate = rate + parm(4)*CaE*act("Zn+2") - k_b2 * m * act("Ca+2")
50 moles = rate * time
60 save -moles
70 put (get(1) + moles, 1)
-end

FeEDTA
# Initially from waste stream, otherwise by mass balance...
-start
10 save get(1)
-end

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

534 Pollution by organic chemicals

SOLUTION 1 Simplified Glatt river water

pH 7.9; pe 12 O2(g) -0.68;
Fe 0.2e-3 Ferrihydrite -1.69
# Run file with Ca or Zn crossed out, adapt step...
Ca 1.83
# Zn 91.8e-6

KINETICS 1
-step 432 in 50 # for Ca, 0.12 hr
# -step 21.6e6 in 50 # for Zn, 6000 hr
# -step 3.6e6 in 50 # for Zn and Ca, 1000 hr

CaEDTA; -formula Ca 1; m0 0; -parms 10 3.09e21 2.51e12 1200 1.1e17 6e-8

ZnEDTA; -formula Zn 1; m0 0; -parms 10 3.09e21 1.1e17 1200 2.51e12 6e-8
FeEDTA; -formula Fe 1; m0 6e-8
-tol 1e-12
-cvode true # available in PHREEQC-2.9
INCREMENTAL_REACTIONS true

EQUILIBRIUM_PHASES 1; Ferrihydrite -1.69

USER_GRAPH
-head hours FeEDTA CaEDTA ZnEDTA
-axis_titles Hours mol/L
-init false
-start
10 graph_x total_time / 3600
20 graph_y kin("FeEDTA"), kin("CaEDTA"), kin("ZnEDTA")
-end
END

The file should be run separately with either Ca or Zn in solution, and then with both ions present, with
different timescales for the three runs (“-steps” under keyword KINETICS). The results are displayed in
Figure 10.27. The exchange of Ca is complete in 0.1 hour, and follows for the most part a first order rate with
a half-life of 0.014 hr (0.8 minutes). The exchange with Zn is much slower and takes 5000 hrs to complete
(note the different time-axes in Figure 10.27). The difference is due to the 5105 times smaller concentration
(and activity) of Zn2 compared to Ca2. Also, the shape for the ZnEDTA curve is different from the one for
CaEDTA. It is more a second order reaction that is initially steeper as it depends both on the FeOHEDTA2-
and the Zn2-activity; the driving activity of Zn2 becomes less as it is taken up in the EDTA complex, while
the activity of Ca2 is barely changed by the small amount that is lost by the exchange reaction. Lastly, cal-
culating the exchange with Ca2 and Zn2 together in solution shows intermediate kinetics and needs about
1000 hours for its completion, which is in the same range as shown in Figure 10.26.

7.E08 7.E08 7.E08

6.E08 Ca Zn Ca  Zn
6.E08 6.E08
5.E08 CaEDTA ZnEDTA FeEDTA
5.E08 5.E08
4.E08 4.E08 4.E08 ZnEDTA
mol/L

mol/L

mol/L

3.E08 3.E08 3.E08

2.E08 FeEDTA 2.E08 2.E08 CaEDTA
FeEDTA
1.E08 1.E08 1.E08
0.E00 0.E00 0.E00
0.00 0.02 0.04 0.06 0.08 0.10 0 1000 2000 3000 4000 5000 0 200 400 600 800 1000
Hours Hours Hours

Figure 10.27. Kinetic exchange of Ca2 , Zn2 and both ions together, with FeOHEDTA2, calculated
with PHREEQC. Note the different timescales in the three figures.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Problems 535

The mutual influence of various ions in kinetic complexation reactions is a common feature that will
be noted when the driving force for the individual reactions is different because of concentration dif-
ferences (as in Example 10.10), or because the rate coefficients are much dissimilar. The kinetics of
complexation reactions are correlated with the water exchange rates of the water molecules in the
primary hydration shell around the ion. Water exchange rates for many ions are illustrated in Figure
10.28 and show differences that span orders of magnitude. Correlation with the data in Figure 10.28
can be used to estimate the timelength that is necessary for attaining equilibrium in laboratory exper-
iments in which complexation or sorption constants are to be determined (cf. Question).

Very fast
10 Pb2
Cs Tl3
2 Cu2
Li Ba Gd3 1 ns
Ca2 Cd2
Mn2 Zn2 La3
Fe2 U3
Co2
Mg2 1 µs
5 Ti3
Ni2 In3
Fe3
Be2 Ga3
log10 (kex  sec)

1 ms
V3
V2

Ru2
0 1 second
Al3
t1/2
Pt2

1 hour

5
Cr3
Ru3 1 day

Rh3

Very slow

Figure 10.28. Half-lives for water exchange from the bulk solution with the primary hydration shell of
various ions (Burgess, 1988).

QUESTION:
Estimate the half-life for sorption equilibrium of Fe2 on the strong sites of goethite, cf. Section 7.6.1)?
ANSWER: The water exchange rate of Fe2 is 1000 times smaller than of Pb2 (Figure
10.28), hence we expect t ⁄  1000  45 min  31 days for Fe2 (Figure 7.37).
1
2

PROBLEMS
10.1. Find KH of H2O at 20 and 10°C from the PHREEQC database, with keywords
REACTION_TEMPERATURE and USER_PRINT.
10.2. Find the different diffusion fluxes of H218O and H216O during evaporation, express the difference in the
form of
notation.
10.3. Model the naphthalene column experiment of Bayard et al., 2000, shown in Figure 10.13. Column
length 7 cm, longitudinal dispersivity L  0.7 cm, flow rate 72 mL/hr, pore volume 21.6 mL. The soil

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

536 Pollution by organic chemicals

has foc  0.014 with Koc  302 L/kg, b  1.05 g/cm3,   0.58. One third of the organic carbon reacts
instantaneously, the rest with a first order exchange rate k  0.26/hr. Injected were 85 PV’s with 9 mg
naphthalene/L followed by elution with napthalene-free water.
In the second experiment, 4 g coaltar was mixed with the soil, foc  0.38, Koc  2026 L/kg, hereof
sorbed only 10% instantaneously naphthalene, and 90% with a first order exchange rate k  0.24/hr.
Injected were 110 PV’s with 9 mg naphthalene/L, again followed by elution with naphthalene-free water.
10.4. Use the PHREEQC input file made for Problem 5.19 to estimate the half-life of dehydration of HCO
3
from 37 mM NaHCO3 solution at pH  8.3, when CO2 is reduced to equilibrium with air.
10.5. Calculate the kinetic precipitation of calcite in your tea kettle, considering kinetic loss of CO2(g) from the
heating water by the two-film diffusion model.
Data: The drinking water has temperature 10°C, is in equilibrium with calcite and PCO2 
0.01 atm. It degasses kinetically to the atmospheric pressure of 103.5 atm during heating of 1 L in 2 min-
utes to 100°C. Consider only the water-film resistance, correct the diffusion coefficient for the tempera-
ture dependency of the viscosity (cf. Equation. 3.42) using:
T  10^((1.37023 * (tc  20)  8.36104 * (tc  20)^2) / (109  tc))
where tc is temperature in °C.
Already 10 g calcite exists in the kettle, with surface area A  2.4 dm2/dm3. The specific weight of cal-
cite is 2.5 g/cm3.
10.6. Polycyclic aromatic hydrocarbons (PAH) were found in the soil below gasworks. Estimate the relative
mobility of naphthalene and tetracene with respect to water flow in sand (0.1% OC;   0.3) and peat
(70% OC;   0.6).
10.7. A waste-dump is situated as illustrated in the figure; Cd and TCE are leached from the
waste with precipitation-excess in concentrations of resp. 1.12 and 7.1 mg/L, and enter the aquifer. Fronts
and depths of leachates must be calculated, as well as a remedial pumping scheme, given the following
hydrological parameters and aquifer-properties: P  0.3 m/yr; porosity  0.36; bulk density  1.8 g/cm3;
dispersivity  10% of flowlength. The aquifer is homogeneous.
a) Calculate distance and depth of the front for a conservative solute (i.e. without decay or adsorption)
after 25 and 100 years; calculate the thickness of the plume at these fronts. Compare the distances,
covered after 25 and 100 years.
b) TCE has an adsorption-isotherm q  7.4 c0.4, where q is adsorbed TCE, mg/L groundwater, and c is
solute TCE, mg/L.
Estimate a linear distribution-coefficient for Cd2 from exchange equilibrium with Ca2, when
CEC  10 meq/kg, Ca2  10 mmol/L, and the exchange coefficient KCd\Ca  1.
c) Calculate the fronts for Cd2 and TCE for 25 and 100 years (distance and depth).
d) Calculate the distance over which 68% of the Cd-front has spread after 100 years as a result of disper-
sion, and give the concentration-profile in 3 points. Will the TCE-front be sharper or more spread-out?

P  0.3 m/yr

50 525 m
50 m

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 537

10.8. Model denitrification in an aquifer, using the rate Organic_C in phreeqc.dat. Input NO
3 is 6 mM, Organic
C is 0.1%, b  1.8 g/cm3, w  0.3. Downward percolation velocity is 1 m/yr. At what depth is
NO3  0.4 mM (25 mg/L, the drinking water limit)? Consider the effect of dispersion, L  1 m?

REFERENCES
Ahmad, R., Kookana, R.S., Alston, A.M. and Skjemstad, J.O., 2001. The nature of soil organic matter affects
sorption of pesticides. 1. Relationships with carbon chemistry as determined by 13C CPMAS NMR
Spectroscopy. Env. Sci. Technol. 35, 878–884.
Alexander, M. and Scow, K.M., 1989. Kinetics of biodegradation in soil. In B.L. Sawhney and K. Brown (eds),
Reactions and movement of organic chemicals in soils. Soil Sci. Soc. Am. Spec. Pub. 22, 243–269.
Appelo, C.A.J. and Parkhurst, D.L., 1998. Enhancements to the geochemical model PHREEQC-1D transport
and reaction kinetics. In G.B. Arehart and J.R. Hulston (eds), Proc. 9th Water Rock Interaction Symp.,
873–876, Balkema, Rotterdam.
Arnold, W.A. and Roberts, A.L., 2000. Pathways and kinetics of chlorinated ethylene and chlorinated actetylene
reaction with Fe(0) particles. Env. Sci. Technol. 34, 1794–1805.
Artinger, R., Buckau, G., Geyer, S., Fritz, P., Wolf, M. and Kim, J.I., 1999. Characterization of groundwater
humic substances: influences of sedimentary organic carbon. Appl. Geochem. 15, 97–116.
Atkins, P.W. and de Paula, J., 2002. Atkins’ Physical chemistry, 7th ed. Oxford Univ. Press, 1149 pp.
Balkwill, D.L., Leach, F.R., Wilson, J.T., McNabb, J.F. and White, D.C., 1988. Equivalence of microbial
biomass measures based on membrane lipid and cell wall components, adenosine triphosphate, and direct
counts in subsurface aquifer sediments. Microb. Ecol. 16, 73–84.
Barns, S.M. and Nierzwicki-Bauer, S., 1997. Microbial diversity in modern subsurface, ocean, surface envi-
ronments. Rev. Mineral. 35, 35–79.
Baedecker, M.J., Cozzarelli, I.M., Siegel, D.I., Bennett, P.C. and Eganhouse, R.P., 1993. Crude oil in a shallow
sand and gravel aquifer. Appl. Geochem. 8, 569–586.
Barry, D.A., Prommer, H., Miller, C.T., Engesgaard, P., Brun, A. and Zheng, C., 2002. Modelling the fate of
oxidisable organic contaminants in groundwater. Adv. Water Resour. 25, 945–983.
Bayard, R., Barna, L., Mahjoub, B. and Gourdon, R., 2000. Influence of the presence of PAHs and coaltar on
naphthalene sorption in soils. J. Contam. Hydrol. 46, 61–80.
Bedient, P.B., Rifai, H.S. and Newell, C.J., 1994. Groundwater contamination. Prentice-Hall, Englewood Cliffs,
541 pp.
Bekins, B.A., Warren, E. and Godsy, E.M., 1998. A comparison of zero-order, first-order, and Monod biotrans-
formation models. Ground Water 36, 261268.
Berg, J.M., Tymoczko, J.L. and Stryer, L., 2002. Biochemistry, 5th ed. Freeman, New York, 974 pp.
Boudreau, B.P., 1996. Diagenetic models and their implementation. Springer, Berlin, 414 pp.
Broholm, M.M., Tuxen, N., Rügge, K. and Bjerg, P.L., 2001. Sorption and degradation of the herbicide
2-methyl-4,6-dinitrophenol under anaerobic conditions in a sandy aquifer in Vejen, Denmark. Env. Sci.
Technol. 35, 4789–4797.
Brun, A., Engesgaard, P., Christensen, T.H. and Rosbjerg, D., 2002. Modelling of transport and biogeochemical
processes in pollution plumes: Vejen landfill, Denmark. J. Hydrol. 256, 228–247.
Burgess, J., 1988. Ions in solution. Ellis Horwood, Chichester, 191 pp.
Chapelle, F.H., 2001. Groundwater microbiology and geochemistry, 2nd ed. Wiley and Sons, New York, 477 pp.
Chiou, C.T., Peters, L.J. and Freed, V.H., 1979. A physical concept of soil-water equilibria for non-ionic organic
compounds. Science 206, 831–832.
Chiou, C.T., Kile, D.E., Brinton, T.I., Malcolm, R.L., Leenheer, J.A. and MacCarthy, P., 1987. A comparison of
water solubility enhancements of organic solutes by aquatic humic materials and commercial humic acids.
Env. Sci. Technol. 21, 1231–1234.
Clausen, L. and Fabricius, I., 2001. Atrazine, isoproturon, mecoprop, 2,4-D and bentazone adsorption onto iron
oxides. J. Env. Qual. 30, 858–869.
Conant, B.H., Gillham, R.W. and Mendoza, C.A., 1996. Vapor transport of trichloroethylene in the unsaturated
zone: field and numerical modeling investigations. Water Resour. Res. 32, 9–22.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

538 Pollution by organic chemicals

Curtis, G.P., Reinhard, M. and Roberts, P.V., 1986. Sorption of hydrophobic compounds by sediments. In
J.A. Davis and K.F. Hayes (eds), Geochemical process at mineral surfaces. Adv. Chem. Series 323, 191–216.
Danckwerts, P.V., 1970. Gas-liquid reactions. McGraw-Hill, New York, 276 pp.
Eberhardt, C. and Grathwohl, P., 2002. Time scales of organic contaminant dissolution from complex source
zones: coal tar pools vs. blobs. J. Contam. Hydrol. 59, 45–66.
Eckert, P. and Appelo, C.A.J., 2002. Hydrogeochemical modeling of enhanced benzene, toluene, ethylbenzene,
xylene (BTEX) remediation with nitrate. Water Resour. Res. 38, DOI 10.1029/2001WR000692.
Essaid, H.I., Bekins, B.A., Godsy, E.M., Warren, E., Baedecker, M.J. and Cozzarelli, I.M., 1995. Simulation of
aerobic and anaerobic biodegradation processes at a crude oil spill site. Water Resour. Res. 31, 3309–3327.
Essaid, H.I., Cozzarelli, I.M., Eganhouse, R.P., Herkelrath, W.N., Bekins, B.A. and Delinc, G.N., 2003. Inverse
modeling of BTEX dissolution and biodegradation at the Bemidji, MN crude-oil spill site. J. Contam.
Hydrol., DOI 10.1016/S0169-7722(03)00034-2.
Farquhar, G.J. and Rovers, F.A., 1973. Gas production during refuse decomposition. Water Air Soil Poll. 2,
483–495.
Feenstra, S. and Guiguer, N., 1996. Dissolution of dense non-aqueous phase liquids (DNAPLs) in the sub-
surface. In J.F. Pankow and J.A. Cherry, op. cit., 203–232.
Garbarini, D.R. and Lion, L.W., 1986. Influence of the nature of soil organics on the sorption of toluene and
trichloroethylene. Env. Sci. Technol. 20, 1263–1269.
Gan, J., Yates, S.R., Becker, J.O. and Wang, D. 1998. Surface amendment of fertilizer ammonium thiosulfate to
reduce methyl bromide emission from soil. Env. Sci. Technol. 32, 2438–2441.
Geller, J.T. and Hunt, J.R., 1993. Mass transfer from nonaqueous phase organic liquids in water-saturated
porous media. Water Resour. Res. 29, 833–845.
Gillham, R.W. and O’Hannesin, S.F., 1994. Enhanced degradation of halogenated aliphatics by zerovalent iron.
Ground Water 32, 958–967.
Goldsmith Jr., C.D. and Balderson, R.K., 1988. Biodegradation and growth kinetics of enrichment isolates on
benzene, toluene, and xylene. Water Sci. Technol. 20, 505–507.
Grathwohl, P., 1990. Influence of organic matter from soils and sediments from various origins on the sorption of
some chlorinated aliphatic hydrocarbons: implications on Koc correlations. Env. Sci. Technol. 24, 1687–1693.
Haderlein, S.B. and Schwarzenbach, R.P., 1993. Adsorption of substituted nitrobenzenes and nitrophenols to
mineral surfaces. Env. Sci. Technol. 27, 316–326.
Heijnen, J.J. and Van Dijken, J.P., 1992. In search of a thermodynamic description of biomass yields for the
chemotrophic growth of microorganisms. Biotechnol. Bioeng. 39, 833–858.
Hering, J.G. and Morel, F.M.M., 1989. Slow coordination reactions in seawater. Geochim. Cosmochim. Acta 53,
611–618.
Hines, M.E., Crill, P.M., Varner, R.K., Talbot, R.W., Shorter, J.H., Kolb, C.E. and Harriss, R.C., 1998. Rapid
consumption of low concentrations of methyl bromide by soil bacteria. Appl. Env. Microbiol. 64,
1864–1870.
Hussein, M. Jin, M. and Weaver, J.W., 2002. Development and verification of a screening model for surface
spreading of petroleum. J. Contam. Hydrol. 57, 281–302.
Jellali, S., Benremita, H., Muntzer, P., Razakarisoa, O. and Schäfer, G., 2003. A large-scale experiment on mass
transfer of trichloroethylene from the unsaturated zone of a sandy aquifer to its interfaces. J. Contam.
Hydrol. 60, 31–53.
Jin, Y. and Jury, W.A., 1996. Characterizing the dependence of gas diffusion coefficient on soil properties. Soil
Sci. Soc. Am. J. 60, 66–71.
Kan, A.T. and Tomson, M.B., 1990. Effect of pH concentration on the transport of naphthalene in saturated
aquifer media. J. Contam. Hydrol. 5, 235–251.
Karickhoff, S.W., 1981. Semi-empirical estimation of sorption of hydrophobic pollutants on natural sediments
and soils. Chemosphere 10, 833–846.
Karickhoff, S.W., 1984. Organic pollutant sorption in aquatic systems. J. Hydraul. Eng. 110, 707–735.
Kelly, W.R., Hornberger, G.M., Herman, J.S. and Mills, A.L., 1996. Kinetics of BTX biodegradation and
mineralization in batch and column systems. J. Contam. Hydrol. 23, 113–132.
Kile, D., Wershawand, R. and Chiou, C.T., 1999. Correlation of soil and sediment organic matter polarity to
aqueous sorption of nonionic compounds. Env. Sci. Technol. 33, 2053–2056.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 539

Kjeldsen, P., Barlaz, M.A., Rooker, A.P., Baun, A., Ledin, A. and Christensen, T.H., 2002. Present and long-term
composition of MSW landfill leachate: A review. Crit. Rev. Env. Sci. Technol. 32, 29–336.
Knappe, A., Sommer-von Jarmersted, C., Pekdeger, A., Bau, M. and Dulski, P. 1999. Gadolinium in aquatic
systems as indicator for sewage water contribution. In H. Ármannsson (ed.), Geochemistry of the earth’s
surface, 187–190, Balkema, Rotterdam.
Kummerer, K. and Helmers, E., 2000. Hospital effluents as a source of gadolinium in the aquatic environment.
Env. Sci. Technol. 34, 573–577.
Lensing, H.J., Vogt, M. and Herrling, B., 1994. Modeling of biologically mediated redox processes in the
subsurface. J. Hydrol. 159, 125–153.
Liss, P.S., 1973. Processes of gas exchange across an air-water interface. Deep Sea Res. 20, 221–238.
Liss, P.S. and Slater, P.G., 1974. Flux of gases across the air-sea interface. Nature 247, 181–184.
Lyman, W.J., Reehl, W.F. and Rosenblatt, D.H., 1990. Handbook of chemical property estimation methods.
Am. Chem. Soc., Washington.
Mackiewicz, M. and Wiegel, J., 1998. Comparison of energy and growth yields for Desulfitobacterium dehalo-
genans during utilization of chlorophenol and various traditional electron acceptors. Appl. Env. Microbiol.
64, 352–355.
MacQuarrie, K.T.B., Sudicky, E.A. and Frind, E.O., 1990. Simulation of biodegradable contaminants in ground-
water. 1. Numerical formulation in principal directions, Water Resour. Res. 26, 207–222.
MacQuarrie, K.T.B. and Sudicky, E.A., 2001. Multicomponent simulation of wastewater-derived nitrogen and
carbon n shallow unconfined aquifers 1. Model formulation and performance. J. Contam. Hydrol. 47, 53–84.
Madigan, M.T., Martinko, J.M. and Parker, J., 2003. Brock biology of microorganisms, 10th ed. Prentice Hall,
Englewood Cliffs, 1019 pp.
Madsen, L., Lindhardt, B., Rosenberg, P., Clausen, L. and Fabricius, I., 2000. Pesticide sorption by low organic
carbon sediments. J. Env. Qual. 29, 1488–1500.
Mayer, K.U., Blowes, D.W. and Frind, E.O., 2001. Reactive transport modeling of an in situ reactive barrier for the
treatment of hexavalent chromium and trichloroethylene in groundwater. Water Resour. Res. 37, 3091–3103.
McWhorter, D.B. and Kueper, B.H., 1996. Mechanics and mathematics of the movement of dense non-aqueous
phase liquids in porous media. In J.F. Pankow and J.A. Cherry, op. cit., 89–128.
Miljøprojekt, 1998. Status for lukkede boringer ved almene vandværker. Danish EPA.
Miller, C.T., Poirier-McNeill, M.M. and Mayer, A.S., 1990. Dissolution of trapped nonaqueous phase liquids:
Mass transfer characteristics, Water Resour. Res. 26, 2783–2796.
Pankow, J.F. and Cherry, J.A. (eds), 1996. Dense chlorinated solvents and other DNAPLs in groundwater.
Waterloo Press, Guelph, Ont.
Powers, S.E., Abriola, L.M., Dunkin, J.S. and Weber Jr, W.J. 1994. Phenomenological models for transient
NAPL-water mass-transfer processes, J. Contam. Hydrol. 16, 1–33.
Prommer, H., Barry, D.A. and Davis, G.B., 2000. Numerical modelling for design and evaluation of ground-
water remediation schemes. Ecol. Model. 128, 181–195.
Rittmann, B.E. and McCarty, P.L., 2001. Environmental biotechnology. McGraw-Hill, Boston, 754 pp.
Schirmer, M., Butler, B.J., Roy, J.W., Frind, E.O. and Barker, J.F. 1999. A relative-least-squares technique to
determine unique Monod kinetic parameters of BTEX compounds using batch experiment. J. Contam.
Hydrol. 37, 69–86.
Schirmer, M., Molson, J.W., Frind, E.O. and Barker, J.F., 2000. Biodegradation modelling of a dissolved gaso-
line plume applying independent laboratory and field parameters. J. Contam. Hydrol. 46, 339–374.
Schwarzenbach, R.P. and Westall, J.C., 1985. Sorption of hydrophobic trace organic compounds in groundwa-
ter systems. Water Sci. Technol. 17, 39–55.
Schwarzenbach, R.P., Gschwend, P.M. and Imboden, D.M., 1993. Environmental organic chemistry. Wiley and
Sons, New York, 681 pp.
Stouthamer, A.H., 1976. Yield studies in microorganisms. Meadowfield Press, Durham, England, 88 pp.
Tiktak, A., Van den Berg, F., Boesten, J.J.T.I., Van Kraalingen, D., Leistra, M. and Van der Linden, A.M.A.,
2000. Manual of FOCUS PEARL version 1.1.1. RIVM Rapport 711401008.
Thornton, S.F., Lerner, D.N. and Banwart, S.A., 2001. Assessing the natural attenuation of organic contami-
nants in aquifers using plume-scale electron and carbon balances: Model development with analysis of
uncertainty and parameter sensitivity. J. Contam. Hydrol. 53, 199–232.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

540 Pollution by organic chemicals

Troeh, F.R., Jabro, J.D. and Kirkam, D., 1982. Gaseous diffusion equations for porous materials. Geoderma 27,
239–253.
VanBriesen, J., 2002. Evaluation of methods to predict bacterial yield using thermodynamics. Biodegradation
13, 171–190.
VanBriesen, J., Rittmann, B.E., Xun, L., Girvin, D.C. and Bolton, H., 2000. The rate-controlling substrate of
nitrilotriacetate for biodegradation by Chelatobacter heintzii. Env. Sci. Technol. 34, 3346–3353.
Van Cappellen, P. and Wang, Y., 1996. Cycling of iron and manganese in surface sediments. Am. J. Sci. 296,
197–243.
Verschueren, K., 2001. Handbook of environmental data on organic chemicals, 4th ed. Wiley and Sons,
New York, 2416 pp.
Wanninkhof, R. and McGillis, W.R., 1999. A cubic relationship between air-sea CO2 exchange and wind speed.
Geophys. Res. Lett. 26, 1889–1892.
Watson, I.A., Oswald, S.E., Mayer, K.U., Wu, Y. and Banwart, S.A., 2003. Modeling kinetic processes con-
trolling hydrogen and acetate concentrations in an aquifer-derived microcosm. Env. Sci. Technol. 37,
3910–3919.
Weaver, J.W., Charbeneau, R.J. and Lien, B.K., 1994. A screening model for nonaqueous phase liquid transport
in the vadose zone using Green – Ampt and kinetic wave theory. Water Resour. Res. 30, 93–105.
Weber, W.J., LeBoeuf, E.J., Young, T.M. and Huang, W., 2001. Contaminant interactions with geosorbent
organic matter: Insights drawn from polymer sciences. Water Res. 35, 853–868.
Weber, J.B. and Miller, C.T., 1989. Organic chemical movement over and through soil. In B.L. Sawhney and
K. Brown (eds), Reactions and movement of organic chemicals in soils. Soil Sci. Soc. Am. Spec. Pub. 22,
305–334.
Xue, H., Sigg, L. and Kari, F.G., 1995. Speciation of EDTA in natural waters: Exchange kinetics of Fe-EDTA
in river water. Env. Sci. Technol. 29, 59–68.
Yabusaki, S., Cantrell, D., Sass, B. and Steefel, C., 2001. Multicomponent reactive transport in an in situ zero-
valent iron cell. Env. Sci. Technol. 35, 1493–1503.
Yeats, P.A. and Measures, C.I., 1998. The hydrographic setting of the second IOC contaminants baseline cruise.
Mar. Chem. 61, 3–14.
Zachara, J.M., Ainsworth, C.C., Felice, L.J. and Resch, C.T., 1986. Quinoline sorption to subsurface materials:
Role of pH and retention of the organic cation. Env. Sci. Technol. 20, 620–627.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

11

Numerical Modeling

We like to address now how we can predict the fate of chemicals that leach from waste sites and
appear to vanish in the invisible subsurface environment. We use numerical models since the com-
bination of transport and chemistry often yields complicated concentration patterns which are diffi-
cult or impossible to calculate otherwise. Numerical solutions of the advection–reaction–dispersion
(ARD) equation are approximative and may be incorrect if certain stability criteria are violated.
These criteria are explained with 1D examples that illustrate how to calculate diffusion of seawater
and advective flow in laboratory columns. They also apply in multidimensional models.
For modeling aquifers, the pattern of groundwater flowlines and travel times must be studied
first. Inhomogeneities in an aquifer can be unraveled with environmental tracers and we consider an
example of groundwater age dating with tritium/helium concentrations. Subsequently, cases are pre-
sented where hydrogeochemical modeling is used to learn what process is determining the concen-
tration of organic and inorganic pollutants in groundwater. Often, chemical parameters are unknown
and need to be estimated and optimized using literature data. Essentially, this chapter shows how the
combination of modeling and predicting can help in improving our understanding.
In Chapter 3 we have combined the processes in one equation to model changes in concentration
due to transport, reaction and dispersion:

 ∂c   ∂c   ∂q   ∂2 c 
   v       DL  2  (3.62)
 ∂t  x  ∂x t  ∂t  x  ∂x t

where c is the solute concentration (mol/L), v is pore water flow velocity (m/s), DL is hydrodynamic
dispersion coefficient (m2/s) and q is the concentration in the solid (mol/L of pore water).
Analytical solutions for Equation (3.62) are simple when q is a linear function of c, which per-
mits replacement of (#q / #t) by K d (#c / #t) so that we obtain

 ∂c   ∂c   ∂2 c 
(1  K d )    v    DL   (11.1)
 ∂t  x  ∂x t  ∂x 2 t
or:

 ∂c   ∂c   ∂2 c 
   v    DL  2  (11.2)
 ∂(t /R )  x  ∂x t  ∂x t

where (1  Kd) is replaced by the retardation R, which defines the retarded time t /R. If the retarda-
tion is constant (i.e. the distribution coefficient Kd is independent of the concentration), then
Equation (11.2) is identical with the simpler advection–dispersion equation (3.64). All the solutions
for that equation are valid when the retarded time t /R is used: retarded chemicals show exactly the

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

542 Numerical modeling

same amount of spreading at all points as a conservative chemical is doing, but arrive later depend-
ing on the retardation. For example, the formula which describes front spreading in an infinite
column (Equation 3.65) becomes, with retardation:

 
c(x,t)  ci  1⁄2 (c0  ci ) erfc  x  vt / R  (11.3)
 4D t / R 
 L 

We apply the usual simplification DL  Lv, and vt / R  x0.5 to get:

 
c(x,t)  ci  1⁄2 (c0  ci ) erfc  x  x0.5  (11.4)
 4 x 
 L 0.5 

Thus, transport in an aquifer of a chemical that shows linear retardation can be calculated very sim-
ply, and completely analogous to a conservative compound. Let us compare the analytical solution
with results from a numerical model in Example 11.1.

EXAMPLE 11.1. Calculation of aquifer pollution by waste site leachate

Figure 11.1 shows results of numerical calculation of pollutants leached from a waste site in an aquifer
profile (Pickens and Lennox, 1976). Effects of dispersivity and different retardations are illustrated by
iso-concentration lines (contours) of 0.1, 0.3, 0.5, 0.7, and 0.9 (fraction of final concentration). Compare the
spread of the front with an analytical guess. Data: vt  415 m; L  10 m; b  1.8 g/cm3;   0.3; Kd  0,
0.1 and 1 mL/g. Initial concentration ci  0 mg/L, the waste site is a continuous source with c0  1 mg/L in
the leachate.

ANSWER:
Assume that a flowline, originating at the downstream end of the waste site is horizontal, and that
transverse dispersion is absent. Travel distance of the c / c0  0.5 concentration is x0.5  vt / R.
The concentration c / c0  0.1 is at x0.1, a distance further ahead that can be obtained from Equation (11.4):
c(x0.1, t)  0.1  1⁄2 erfc[(x0.1  x0.5) / (4L x0.5)  1⁄2 erfc[ f(x)].

c /c0
0.9
0.7
0.5
0.3
0.1

Waste disposal site

Ground surface Kd  0
Water table
0.9

0.5

0.1

Stream

Contaminant input zone

Kd  0.1
0.9
0.5
0.1

50
Meters

Kd  1
25
0
50
100

Meters

Figure 11.1. Effects of dispersivity and varying retardations on waste site leachate transport in an aquifer.
Numerical results from Pickens and Lennox (1976). Approximate longitudinal spreading of 0.1 / 0.9, and
0.3 / 0.7 contour lines as calculated here is indicated by the bar.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 543

If we know the inverse complementary error function of 0.2 (from Table 3.3), we obtain f(x), and can calcu-
late x0.1. Thus, f(x)  inverfc[0.2]  0.906  (x0.1  x0.5) / (4L x0.5), which gives x0.1 as function of L, vt
and R. The concentration c / c0  0.9, at x0.9, is similarly found from 0.906  (x0.9  x0.5) /(4L x0.5). For
c / c0  0.3 and 0.7 the procedure is identical, with the value of inverfc[0.6]  0.371. The results are
summarized as:

|xc/c  (vt / R) | (m)

0
vt / R
R (m) 0.1 / 0.9 0.3 / 0.7
Kd  0 1 415 117 47.8
Kd  0.1 1.6 259 92 37.8
Kd  1.0 7 59 44 18.1

11.1 NUMERICAL MODELING OF TRANSPORT

Simple functions may be unrealistic when geochemical problems must be tackled. For example, if
precipitation of calcite is to be modeled along a flowline, then qCa is clearly a function of pH,
[HCO 3 ], and also of aqueous complexes and activity-coefficients. It is, in principle, possible to
write out the transport equation (3.62) for each component, include equations for reactions among
components, and solve the whole set with a differencing scheme. This approach has been developed
mainly by Lichtner and co-workers (Lichtner, 1985; Miller and Benson, 1983; Lichtner and Seth,
1996; Steefel and Yabusaki, 1996 and Mayer et al., 2002). In another approach, transport (the A and
D part of Equation 3.62) and chemistry (the R part) are solved separately in each timestep. The first
method is the most general and may become the routine when equation solvers have advanced and
still faster pc’s become available. The second approach is less stringent in memory requirements
(Yeh and Tripathi, 1989; Steefel and MacQuarrie, 1996) and is implemented in 1D in PHREEQC
and recently also in 3D versions (PHAST, Parkhurst et al., 2004, or based on MODFLOW and
MST3D in Alliances, and in PHT3D, Prommer, 2002).

11.1.1 Only diffusion

Let us split up the problem in three parts, corresponding to the three terms of the right-hand side of the
advection-reaction-dispersion Equation (3.62). If we separate the dispersion term from the advection
part, it is more appropriately termed diffusion. The change in concentration as result of diffusion is:
 ∂c   ∂2 c 
   Df  2  (11.5)
 ∂t  x  ∂x t

where we want to evaluate the change of c with time at some location x. It requires calculation of the
second derivative of c with x. An approximation is obtained by combining two Taylor expansions of
the function f(x). First in forward sense:

f ′′(xn ) f 3(xn )
f (xn  x )  f (xn )  f ′(xn )x  (x )2  (x )3  … (11.6)
2 6
where f  is the first derivative, f 2 is the second derivative, etc., and x is the step-size. A similar
Taylor expansion in backward sense gives:
f ′ (xn ) f 3(xn )
f (xn  x )  f (xn )  f ′(xn )x  (x )2  (x )3  … (11.7)
2 6

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

544 Numerical modeling

Adding Equations (11.6) and (11.7) gives the finite difference approximation of the second derivative:
f (xn  x )  2f (xn )  f (xn  x )
 f 2(xn )  O(x )2 (11.8)
(x )2
The approximation of the second derivative of f(x) at xn is obtained by combining a backward and a
forward step, and it is therefore termed a central difference equation. The error term O(x)2 can be
calculated with formulas derived in standard books on numerical techniques; here it is of the order
of the square of the step-size x (Gerald and Wheatley, 1989).
The f (x) used in the Taylor formula is nothing but the concentration (c)tx, i.e. the concentration at
x, at time t. Thus, the second derivative in Equation (11.5) is approximated as:

 ∂2 c  ctx11  2ctx1  ctx11

 2   f 2(x )t 1  (11.9)
 ∂x t 1 (x)2

We can also difference the first derivative with respect to time:

 ∂c   c  ct 2  ctx1
     x (11.10)
 ∂t  x  t x t

which amounts to taking steps forward in time. The solution for the diffusion Equation (11.5) is then
obtained with a scheme that is central in space, and forward in time by combining Equations (11.9),
(11.10) and (11.5):
D f t
ctx2  ctx1  (ct 1  2ctx1  ctx11 )
(x )2 x1 (11.11)
 mixf ctx11  mixf ctx11  (1  2mixf ) ctx1

where mixf is Df t /(x)2.

The parameter mixf is called a mixing factor since, according to Equation (11.11), diffusion con-
sists of mixing of neighboring cells. When mixf  1⁄2, Equation (11.11) simplifies to:

ctx11  ctx11
ctx2  (11.12)
2
On the other hand, when mixf ! 1⁄2, a negative concentration will be obtained in the central cell,
cxt2  0, if the concentrations in the neighboring cells are zero, c t1
x1  0. This is impossible of course,
and numerical instability sets in, leading to oscillations in the results. Sometimes the instabilities occur
already for 1⁄3  mixf  1⁄2, especially when the concentrations change sharply from cell to cell.
Figure 11.2 shows such a case, with the central cell having the concentration ct1 x  1, surrounded
by cells with ct1 x1  0. Diffusion in time, of course, smears out the sharp transition and gives a
smooth, Gaussian distribution. However, concentrations oscillate when calculated at t  t2 and t3
with mixf  1⁄2. When the mixing factor is smaller than 1⁄3, such oscillations are prevented since
never more is mixed out of a cell, than remains behind. This stability criterion thus requires that:

D f t (x)2
mixf  ' 1⁄3, or t ' (11.13)
(x )2 3Df
It is called the Von Neumann condition for the maximal size of the timestep (actually, Von Neumann
specified mixf 0 1⁄2). In case the simulation time tn is larger than t, mixing is repeated a number of
times (Nmix) until t  tn.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 545

mixf  1/ 2 mixf  1 / 3

t1 0 0 0 1 0 0 0 0 0 0 1 0 0 0

t2 0 0 1/ 2 0 1/ 2 0 0 0 0 1/3 1/ 3 1/3 0 0

t3 0 1 /4 0 1/ 2 0 1/ 4 0 0 1/9 2 /9 3 /9 2 /9 1/9 0

t4 1/8 0 3 /8 0 3 /8 0 1/ 8 1 / 27 3 / 27 6 / 27 7/ 27 6 / 27 3 / 27 1 / 27

x
3
2
1 0 1 2 3
3
2
1 0 1 2 3

Figure 11.2. Stability in a central difference scheme depends on the mixing factor mixf; the scheme is stable
when mixf  1 / 3.

QUESTION:
Calculate the concentrations in Figure 11.2 at t3 for mixf  1?
ANSWER: 0 1 2 3 2 1 0, note that the oscillations augment with time, but “mass
balance” is conserved.

Boundary conditions may demand slightly different formulas for the mixing factor. When the concen-
tration is a fixed c0 at one boundary (for example when seawater diffuses into a fresh water sediment),
a better approximation is obtained if we fix the concentration at that point. This can be arranged as
illustrated in Figure 11.3. We halve the first cell, and calculate the concentration c1 from central dif-
ferences with c1 1⁄2 x and c11⁄2 x. Clearly, c11⁄2 x  c0 (Figure 11.3). The concentration c11⁄2 x can
be approximated as an average of two extremes. The first extreme is the midpoint-concentration
between the first two cells (which is 1⁄2(c1  c2)), and the second one is the expected concentration
when the line c0  c1 is extrapolated (which gives c1  (c0  c1)). The average of the two is:
c11⁄2 x  c2/ 4  c0 / 2  5c1/4 (11.14)

Once more we use Equation (11.11) to calculate the concentration c1 at the next timestep from c1,
c1 ⁄ x and c1 ⁄ x, and obtain:
1
2
1
2

c1t2  1⁄2 mixf *c0t1  1⁄4 mixf *c2t1  (1 3⁄4 mixf *)c1t1 (11.15)

c1x / 2

Cell 1 Cell 2 Cell 3

c0
Concentration

c1
c2 c3
+

c1x / 2 +

x x
——
2

Figure 11.3. When concentrations are fixed at one boundary, the difference formulas at that point must be
adapted.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

546 Numerical modeling

The mixing factor for the halved cell is mixf *  Df t / (1⁄2 x)2  4mixf, and rewriting Equation
(11.15) in terms of the mixing factor for the full cell yields:

c1t2  2mixf c0t1  mixf c2t1  (1 3mixf )c1t1 (11.16)

We note that this boundary condition requires for generally stable calculations that
(1  3mixf ) ! 1⁄3, or mixf  2⁄9.

EXAMPLE 11.2. A Pascal code to model Cl-diffusion from seawater into fresh-water sediment
Pascal is a popular programming language somewhat similar to the C language. The Turbo Pascal compiler,
which operates under DOS, can be downloaded from www.xs4all.nl/appt, together with the programs in
the examples in this Chapter. The following program calculates 100 years diffusion of seawater
(Cl  0.56 mol/L) into a 10 m thick fresh water sediment (Cl  0.0 mol/L). The lower end is a closed
boundary (a mirror for Cl-ions).

Program diffuse; { brackets enclose comments }

{ define variables...}
const Ncel = 10;
var i, j, Nmix : integer;
Totx, Tott, Delx, De, mixf : real;
Ct1, Ct2 : array[0..Ncel] of real;
begin
{ set variables and initial concentrations...}
De := 1e-5 {cm2/s}; Totx := 1000.0 {cm}; Tott := 100 * 365 * 24 * 60 * 60.0 {s};
Delx := Totx/Ncel;
for i := 1 to Ncel do Ct1[i] := 0.0; Ct1[0] := 0.56;
{calculate the mixing factor ...}
mixf := De * Tott/(Delx*Delx);
Nmix := 1 + trunc(mixf * 2/9); {trunc(..) gives integer part }
mixf := mixf/Nmix;
{ now diffuse...}
for i:= 1 to Nmix do
begin { mix into Ct2-cells. first cell...}
Ct2[1] := 2 * mixf * Ct1[0] + mixf * Ct1[2] + (1 - 3 * mixf) * Ct1[1];
for j := 2 to Ncel - 1 do { inner cells... }
Ct2[j] := mixf * (Ct1[j-1] + Ct1[j+1]) + (1 - 2 * mixf) * Ct1[j];
Ct2[Ncel] := mixf * Ct1[Ncel-1] + (1 - mixf) * Ct1[Ncel]; { .. end cell }
for j := 1 to Ncel do { copy back into Ct1-cells... }
Ct1[j] := Ct2[j];
end;

writeln(' Depth below seabottom (m); Cl (mol/L) after', Tott/(31536e3):8:2,' years');

writeln(' 0 ;',Ct1[0]:8:4, Nmix:5, ' timesteps');
for i:= 1 to Ncel do
begin
writeln(' ', (Delx * (i - 0.5)/100):9:3, Ct1[i]:8:4);
end;
end.

Figure 11.4 shows a comparison of the results of this program with the analytical solution:
 x 
cx ,t  0.56 erfc   (11.17)
 4Det 

The agreement is excellent, also for the first cell near the seawater boundary (usually this one gives some
deviations).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 547

0.6

0.5

0.4

mol/L
0.3 CI – analytical
Num. model
0.2

0.1

0
0 2 4 6 8 10
Depth below sea bottom (m)

Figure 11.4. 100 years of seawater diffusion in fresh water sediment; comparison of central difference and
analytical solution.

QUESTIONS:
What are the values of Nmix and the timestep t?
ANSWER: Nmix  15, t  6.67 years
What are Nmix and t in a 3 times finer grid (Ncel  30, x  0.333 m)?
ANSWER: 130, 0.77
Program the problem in a spreadsheet.
Refine the grid twice and compare the results (note how much easier this is done in Pascal or another pro-
gramming language).
Calculate the problem with PHREEQC, refine the grid.
After seawater diffusion, continue the PHREEQC model with 50 years diffusion of fresh water.
Calculate the fractionation of H216O and H218O during 100 years diffusion. The initial
18O  10‰,

18O  0‰ in the boundary solution, cf. Chapter 3.
Draw/model the c/x curve for retarded chemicals, 1) q  c; 2) q  c2; 3) q  c0.5

We used the (known) concentrations at time t1 to calculate next timestep concentrations, but why not
use these immediately in the central difference formula? Equation (11.11) would become:
D f t
ctx2  ctx1  (ct 2  2ctx2  ctx21 )
(x )2 x1 (11.18)
(1  2 mixf )ctx2  ctx1  mixf ctx21  mixf ctx21
Both the right hand side of this equation and (1  2mixf ) are positive, which means that we will
always obtain a positive concentration c t2
x , irrespective of the size of the timestep. Apparently,
Equation (11.18) is unconditionally stable, and we can increase the timestep as much as we like.
The ct2
x ’s can be found by solving a set of linear equations. If the equations for the case of
Example 11.2 are written out, we obtain:
1c0t 2 0c1t 2 0c2t 2 0c3t 2 0
 c0t 1

2 mixf c0t 2 (1  3mixf )c1t 2 mixf c2t 2 0c3t 2 0
 c1t 1

(11.19)

0c0t 2 mixf c1t 2 (1  2 mixf ) c2t 2 mixf c3t 2 0
 c2t 1

… … … …

…
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
548 Numerical modeling

in which we recognize a matrix form:

 1 0 0 0 0 L  ct 2   c0t 1 
  0   
  ct 1 
0 L   ct 2 
2mixf (1  3mixf ) mixf 0
    
1 1 (11.20)
 
 0 mixf (1  2 mixf ) mixf 0 L   c2t 2   c2t 1 
 
 L    
L L L L L 
L  L 
which can be written as:

A  C t2  C t1 (11.21)
There are as many equations as unknown ct2 t2
x ’s and the vector C can be obtained as follows. The
matrix of the coefficients A is split up in a lower triangular L and an upper triangular U matrix, with
the property that L U  A. For example, with 4 cells in total and mixf  2,
L U  A
 1 0 0 0  1 0 0  0  1 0 0 0 
    
 4   
7 0 0   0 1 2/7 0  4 7 2 0  (11.22)
 
   
 0 2 0   0 14/31  0 2 5 2 
 31/7 0 1 
    
 0 0 2 65/31   0 0 0 1   0 0 2 3 

We now have to solve

L U C t2  C t1

which is easy because the matrices are triangular. First, we define U C t2  y, and solve y in L y  C t1,
starting in the first row and substituting consecutively in the following rows. Then, C t2 is calculated from
U C t2  y, now starting in the last row and working upwards to the first row. Gerald and Wheatley
(1989) and Press et al. (1992) present calculation schemes for what is called LU decomposition.

EXAMPLE 11.3. Implicit calculation of diffusion

We redo the calculation of the Cl diffusion profile of Example 11.2, using only next-timestep concentrations
and compare results for different timesteps. Only the 3 non-zero elements of each row of the coefficient
matrix are stored.
Program diffuse2;
const Ncel = 10;
var i, j, Timesteps : integer;
Totx, Tott, Delx, Delt, De, mixf, x : real;
Ct1, Ct2, y : array[0..Ncel] of real;
A : array[0..Ncel, 1..3] of real;
begin
{ set timesteps and initial concentrations...}
Timesteps := 1; { ... also for Timesteps := 5, 15 }
De := 1e-5 {cm2/s}; Totx := 1000.0 {cm}; Tott := 100 * 365 * 24 * 60 * 60.0 {s};
Delx := Totx/Ncel;
for i := 1 to Ncel do Ct2[i] := 0.0; Ct2[0] := 0.56;
{ calculate the mixing factor ...}

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 549

Delt := Tott/Timesteps;
mixf := De * Delt/(Delx * Delx);
{ fill coefficient matrix A, put the diagonal elements in A[.., 2] }
{ boundary cells ...}
A[0, 2] := 1; A[0, 3] := 0;
A[1, 1] := -2 * mixf; A[1, 2] := 1 + 3 * mixf; A[1, 3] := -mixf;
A[Ncel, 1] := -mixf; A[Ncel, 2] := 1 + mixf;
{ inner cells ... }
for i := 2 to Ncel - 1 do
begin
A[i, 1] := -mixf; A[i, 2] := 1 + 2 * mixf; A[i, 3] := -mixf;
end;
{ decompose A in LU: store L in A[.., 1-2] and U in A[.., 3] ...}
for i := 1 to Ncel do
begin
A[i-1, 3] := A[i-1, 3]/A[i-1, 2];
A[i, 2] := A[i, 2] - A[i, 1] * A[i-1, 3];
end;
{ now diffuse...}
for i:= 1 to Timesteps do
begin
for j := 0 to Ncel do Ct1[j] := Ct2[j];
{ solve Ct2 in A.Ct2 = L.U.Ct2 = Ct1 ... First, find y in L.y = Ct1 }
y[0] := Ct1[0];
for j := 1 to Ncel do y[j] := (Ct1[j] - A[j, 1] * y[j-1])/A[j, 2];
{ Now obtain Ct2 in U.Ct2 = y ...}
Ct2[Ncel] := y[Ncel];
for j := Ncel - 1 downto 1 do Ct2[j] := y[j] - A[j, 3] * Ct2[j+1];
end;
{ writeln(‘ Depth .... etc. as in Example 11.2 }
end.

Figure 11.5 compares the results for various time discretizations. With 15 timesteps the accuracy of
Example 11.2 is nearly reached. For 1 and 5 timesteps, the results are not as good, but they are smooth and
without oscillations, and of course, obtained in less time.

0.6
1 timestep
5 timesteps
0.5
15 timesteps
0.4
Cl (mmol/L)

0.3

0.2

0.1

0
0 2 4 6 8 10
Depth (m)

Figure 11.5. Diffusion of seawater Cl in a fresh sediment. Implicit calculation using different timesteps.

QUESTIONS:
Write the program with separate arrays for L and U, use L[0..Ncel, 1..2], U[0..Ncel].
Write the program, storing y in C t2 (this may be useful when computer memory is an issue).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

550 Numerical modeling

Rather conspicuous in Figure 11.5 is that the concentrations are underestimated when few timesteps
are used. The reason is that the gradient, at the end of a timestep, has decreased as result of diffusion.
Accordingly, the flux is less than when the initial gradient is employed as in Example 11.2. For a
better approximation, the two gradients may be combined, and a weighting factor can be applied:

 D f t 
ctx2  ctx1   2  (
(1   ) (ctx11  2ctx1  ctx11 )  (ctx21  2ctx2  ctx21 ) ) (11.23)
 (x ) 
where  is the weighting factor. When   0, the known concentrations at t1 are used, and the calcula-
tions are explicit (Example 11.2). When   1, the concentrations to be calculated are part of the algo-
rithm, and the calculation scheme is implicit (Example 11.3). When   0.5, the two gradients are given
equal weight, which makes the scheme central in time, commonly denoted as Crank-Nicholson and very
often applied. Following the earlier discussion, the maximal timestep in the Crank-Nicholson scheme is:

2(x )2
t ' (11.24)
3 f

or twice larger than in the explicit scheme (note that it is common to use t 0 (x)2 /D, but this
may give oscillations for initially abrupt concentration fronts, cf. Figure 11.2). Actually, for a
given number of cells and a fixed timestep, the most accurate results are obtained with the explicit
method. However, for very precise calculations in which the timestep integration is varied depend-
ing on the accuracy of the computed result, the implicit method is often the most rapid. Variable
timestepping can be done with Runge-Kutta routines that use intermediate steps to evaluate the
accuracy of the integration and automatically decrease or increase the timestep to achieve a prede-
fined exactness.

QUESTION:
Calculate the maximal timestep in a central-in-time scheme for   0.45.
ANSWER: t  (x)2 / (1.65D)
Program the seawater diffusion problem with the Crank-Nicholson scheme, using the outline from
Example 11.3.

11.1.2 Advection and diffusion/dispersion

The second step in our modeling effort is to include transport due to advection:

 ∂c   ∂c 
   v   (11.25)
 ∂t  x  ∂x t

The spatial derivative (#c /#x) can be approximated by weighting concentrations from neighboring cells:

∂c [(1  )cx  cx1 ]  [(1  )cx1  cx ]

 (11.26)
∂x x
where  is the spatial weighting factor.
If we take   0, then (#c /#x)  (cx  cx1) /x, which gives in Equation (11.25):

vt t 1
ctx2  ctx1 
x x
(
c  ctx11 ) (11.27)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 551

x may become negative when vt ! x. Thus, in any case the timestep
The equation indicates that ct2
should be limited by
x
t ' (11.28)
v
which is known as the Courant condition.
Let’s take t such that vt  x. In that case is:

ctx2  ctx1  (ctx1  ctx11 )  ctx11 (11.29)

Thus, we simply move along, pouring concentrations from one cell into the next one. Fronts with dif-
ferent concentrations on both sides, move neatly with the grid boundaries, and remain sharp.
Such sharpness is blurred when the front movement does not correspond with the grid boundaries,
i.e. when vt  x. In this case the mixing of old and new concentrations in a cell leads to a grad-
ual smoothening of transitions, termed numerical dispersion. Figure 11.6 illustrates the process.

 t x  t  ½x

t1 1 0 0 0 0 1 0 0 0 0

t2 1 1 0 0 0 1 1/2 0 0 0

t3 1 1 1 0 0 1 3 /4 1/4 0 0

t4 1 1 1 1 0 1 0.875 0.5 0.125 0

x 1 2 3 4 5 1 2 3 4 5

Figure 11.6. Numerical dispersion as result of time discretization.

Numerical dispersion can be counteracted by decreasing the gridsize (i.e. decrease x). However,
this is a costly affair since smaller gridsize automatically decreases the timestep according to
Equation (11.28), and thus increases the number of timesteps as well. Numerical dispersion can also
be used to mimick physical (field) dispersion by choosing the discretization steps t in relation to
x in such a way, that numerical dispersion is just equal to the physical dispersion which must be
modeled (Van Ommen, 1985). However, front retardation as a result of reactions must also be taken
into account. It can be shown (Herzer and Kinzelbach, 1989, Notodarmojo et al., 1991), that the
numerical dispersivity amounts to:

x vt
 num   (11.30)
2 2R
where num is numerical dispersivity (m), and R is the retardation.
The retardation can be calculated as the relative change of concentrations through reactions
(sorption, precipitation, dissolution, etc.) within a cell:
q cx, new  cx, newreact
R  1  1 (11.31)
c cx, newreact  cx, old

where cx,old is the old concentration in a cell, cx,new is the concentration after the transport step, and
cx,newreact is the concentration after reactions.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

552 Numerical modeling

We recall here that (field) dispersivity forms part of the hydrodynamic dispersion coefficient
(Chapter 3):
DL  De   L v (3.73)

and must be included in the calculation of the mixing factor when advective flow is modeled in addi-
tion to diffusion. The actual value of L, to be comprised in DL, can be corrected by subtracting the
numerical dispersivity as result of transport and reaction:

 x vt 
 mixf   L     (11.32)
 2 2R 

and the mixing factor becomes:

De t  mixf vt
mixf   (11.33)
(x )2 (x )2

Noting that mixf should be larger than zero, Equations (11.32) and (11.33) provide an estimate of
maximum x, and hence of the minimum number of cells that is necessary for a numerical disper-
sion-free calculation. Take De  0, and vt  x (the latter is part of the sensible strategy to maintain
vt as large as possible, for efficient calculations and also to keep mixf positive in Equation (11.32),
which means that in Equation (11.33), mixf /x ! 0. We combine with Equation (11.32), and obtain:
 L 2R
mixf ! 0, if x  (11.34a)
R 1
or with Ncell  Totx /x:

Totx  1  1 / R 
mixf ! 0, if Ncell !   (11.34b)
L  2 

where Totx is the length of the (laboratory) column or flowline.

We have gained a measure for the maximal timestep (from (t)A 0 x / v) and the minimal grid
spacing (Equation (11.34) for numerical stability and arithmetical accuracy. But, what happens with
the timestep criterion for diffusion ((t)D 0 (x)2 / (3De), Equation 11.13)? That one relates the
maximal timestep to the square of the cellsize, implying that when both timestep criteria are met for
a given x, a grid refinement of 2 dictates that the advective timestep becomes twice smaller, while
the diffusive timestep should be four times smaller. This conflict can be solved by operator splitting,
taking multiple dispersion timesteps such that
(t ) D  (t ) A (11.35)

Thus, the mixing factor is defined as

DL (t ) A
mixf  (11.36)
n(x )2

where n is the smallest positive integer that keeps mixf 0 1⁄3 (or 2⁄9 for a fixed boundary condition).
The dispersive timestep is then (t)D  (t)A / n, and n mixes are performed during an advective
timestep. With the discussion so far, we are in the position to write a program for advective/disper-
sive transport with linear retardation.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 553

EXAMPLE 11.4. Model the linear retardation of -HCH in a 5 cm long, and 5 cm diameter laboratory
column with a sandy aquifer sample. The sand is low in organic carbon (0.05%), has porosity of 30%, and
dispersivity is 7 mm. Injected in the column are 450 mL water containing 20 g -HCH/L at a rate of
10 mL/hr. Assume that diffusion is negligible.
Preliminary calculations provide the following data. With Koc  2  103 (Table 10.1) for -HCH, the
distribution coefficient Kd  5  104 2  103  1 L/kg, and the normalized distribution coefficient
Kd  (b /  1.8 / 0.3 kg/L)(1 L/kg)  6. Hence the retardation R  7. Pore volume (Porv) in the column
is $r2L    29.5 mL; pore water flow velocity (veloc) is 1.7 cm/hr.
From Equation (11.34) follows that Ncell ( Ncel) ! (5 / 0.7  (7  1) / 14)  3.1, hence Ncel gets the
integer value  4. Then x ( Delx)  Totx / Ncel  1.25 cm.
The maximal value of t ( Delt) is when vt  x, hence t  (1.25 cm/1.7 cm/hr)  0.735 hr. Total
injection time of 45 hr is subdivided in 45/0.735  (integer value)  62 timesteps ( Nshift). Thus
t  45/62  0.726 hr. Finally, we calculate mixf  (1.7 cm/hr  0.7 cm)  (0.726 hr) / (1.25 cm)2  0.553.
Hence, we take 2 dispersive timesteps (Nmix  2) for each advective timestep, with Mixf  0.553 / 2  0.276.

20

R1 R7
16

Analytical
12
4 cells, corrected for numerical
g/L

dispersion
4 cells
8
10 cells, no correction

0
0 3 6 9 12 15
Pore volumes

Figure 11.7. Breakthrough curves for conservative element, and retarded -HCH; comparison of finite
difference and analytical solutions.

We subdivide our program into a number of subroutines:

dspvty : calculates initial mixing factors mixf and dispersive timesteps.
dspcor : corrects the mixing factors for mixing and retardation.
dffdsp : mixes to mimick diffusion and dispersion.
flush : performs advective transport.
distri : distributes the chemical over solid (q) and water (c).
analyt : calculates the analytical solution.
{MAIN} : sets the variables, and calls the subroutines in the right order.
Program Columnperc;
Const Nel = 1; Ncelmax = 80;
Var Numdisp : boolean;
i, j : integer;
Ncel, Nshift, Ish, Nmix, Mi : integer;
Porv, Veloc, Tott, Delt, Totx, Delx : real;
Mixf, D, Disp, R, Rmax, Cort : real;

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

554 Numerical modeling

Cin : array[0..Nel] of real;

Ct1, Ct2, Ctold, Q, Mix : array[0..Nel, 1..Ncelmax] of real;
Pv : array[1..200] of real;
Cout : array[0..Nel+2, 1..200] of real;
file1 : text;

procedure dspvty; { calculate mixing factor and dispersive timesteps... }

begin
Cort := 1 + 2/Ncel; { Cort corrects for lack of end-cell mixing... }
Mixf := (D + Disp * Veloc) * Delt/(Delx * Delx) * Cort;
Nmix := 1 + trunc(3.0 * Mixf);
Mixf := Mixf/Nmix;
{ correct mixing factors for conservative element...}
for j := 1 to Ncel do
Mix[0, j] := Mixf - (1 - Veloc * Delt/Delx)/2/Nmix * Cort;
end;

Procedure dspcor; { correct mixing factors ... }

Var Cdif : real;
begin { Retardation is calculated as R ... }
for j := 1 to Ncel do for i := 1 to Nel do
begin
{ Ct2 ( = C_x, new) from flush. Ct1 ( = C_x,new+react ) from distri. }
Cdif := Ct1[i,j] - Ctold[i,j];
if abs(Cdif) > 1e-9 then R := (Ct2[i,j] - Ctold[i,j])/Cdif;
if R < 1 then R := 1.0;
if Rmax < R then Rmax := R;
Mix[i,j] := Mixf - (1 - Veloc * Delt/R/Delx)/2/Nmix * Cort;
end;
end;

procedure dffdsp; { diffuse and disperse... }

var Mixup, Mixdn : real;
begin
{ mix inner cells ... }
for j:= 2 to Ncel-1 do for i:= 0 to Nel do
{ average mixing factors of neighboring cells .. }
begin
Mixup := (Mix[i, j-1] + Mix[i, j])/2;
Mixdn := (Mix[i, j] + Mix[i, j+1])/2;
if Mixup < 0 then begin Mixup := 0; Numdisp := true; end;
if Mixdn < 0 then Mixdn := 0;
Ct2[i, j] :=
(1.0 - Mixup - Mixdn) * Ct1[i, j] + Mixup * Ct1[i, j-1] + Mixdn * Ct1[i, j+1];
end;
{ mix boundary cells ... }
for i:= 0 to Nel do
begin
Mixdn := (Mix[i, 1] + Mix[i, 2])/2;
if Mixdn < 0 then Mixdn := 0;
Ct2[i, 1] := (1 - Mixdn) * Ct1[i, 1] + Mixdn * Ct1[i, 2];
Mixup := (Mix[i, Ncel-1] + Mix[i, Ncel])/2;
if Mixup < 0 then Mixup := 0;
Ct2[i, Ncel] := (1 - Mixup) * Ct1[i, Ncel] + Mixup * Ct1[i, Ncel-1];
end;
for j:= 1 to Ncel do for i:= 0 to Nel do Ct1[i, j]:= Ct2[i, j];
end;

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 555

procedure flush; { advect... }

var x_in_cell : real;
begin
x_in_cell := Veloc * Delt/Delx;
for i:= 0 to Nel do
begin
for j:= Ncel downto 2 do
begin
Ctold[i, j] := Ct1[i, j]; { ... keep old conc’s for dspcor }
Ct1[i, j] := x_in_cell * Ct1[i, j-1] + (1 - x_in_cell) * Ct1[i, j];
Ct2[i, j] := Ct1[i, j];
end;
{ also the first cell... }
Ctold[i, 1] := Ct1[i, 1];
Ct1[i, 1] := x_in_cell * Cin[i] + (1 - x_in_cell) * Ct1[i, 1];
Ct2[i, 1] := Ct1[i, 1];
end;
end;

procedure distri;
var Tot : real;
begin
for j := 1 to Ncel do
begin
Tot := Q[1, j] + Ct1[1, j];
Ct1[1, j] := Tot/R;
Q[1, j] := Tot - Ct1[1, j];
end;
end;

procedure analyt;
var P, a1, a2, a3, a4, a5, s, e, er1, er2 : real;

function erfc : real;

begin
e := 1/(1 + P * abs(s));
erfc:= (e *(a1 + e *(a2 +e *(a3 + e *(a4 + e *a5))))) * exp(-s * s)
end;

begin
P := 0.3275911; a1 := 0.254829592; a2 := -0.284496736;
a3 := 1.421413741; a4 := -1.453152027; a5 := 1.061405429;
for i := 1 to Nshift do for j := 1 to 2 do
begin
if j = 1 then R := 1 else R := 7;
s := (1 - Pv[i]/R)/2/sqrt(Disp/Totx * Pv[i]/R);
if s > 0 then
er1 := erfc else er1 := 2 - erfc;
s := (1 + Pv[i]/R)/2/sqrt(Disp/Totx * Pv[i]/R);
er2 := erfc * exp(Totx/Disp);
Cout[Nel+j, i] := (er1 + er2) * Cin[j-1]/2;
end;
end;

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

556 Numerical modeling

{Main}
begin
{ define column data... }
D := 0.0 {cm2/s}; Disp := 0.7 {cm}; R := 7; Totx := 5.0 { cm };
Ncel := 1 + trunc(Totx/Disp * (1 - 1/R)/2);
if Ncel < 10 then Ncel := 10;
Delx := Totx/Ncel { cm };
Porv := pi * (2.5 * 2.5) * Totx * 0.3 { mL };
Veloc := 10/Porv/3600 * Totx { cm/s };
Nshift := 1 + trunc(450/Porv * Ncel);
Delt := 45 * 3600/Nshift { s };
{ set initial concentrations ... }
for j := 1 to Ncel do for i := 0 to Nel do
begin
Ct1[i, j] := 0; Q[i, j] := 0;
end;
Cin[0] := 20; Cin[1] := 20 {ug/L};
{ find mixing factors... }
dspvty;
{ now the experiment ...}
for Ish := 1 to Nshift do
begin
{ advect, distribute and correct mixing factors... }
flush; distri; dspcor;
{ disperse and distribute... }
for Mi := 1 to Nmix do
begin dffdsp; distri; end;
{ find pore volume, assign concentrations... }
Pv[Ish] := (Ish * Delt * Veloc/Delx + 0.5)/Ncel;
for i := 0 to Nel do Cout[i, Ish] := Ct1[i, Ncel];
end;
{ Calculate analytical Cout... }
analyt;
{ write results to file... }
assign(file1, ‘ex11_4’); rewrite(file1);
writeln(file1, ‘ PV R = 1(num) R = 7(num) R = 1(ana) R = 7(ana)’);
for j := 1 to Nshift do
begin
write(file1, Pv[j]:6:3);
for i := 0 to Nel + 2 do write(file1, Cout[i,j]:9:3); writeln(file1);
end;
close(file1);
end.

You may note that the subroutine “dspcor” could have been called only once since in this example the mixing
factor correction is constant (for a constant retardation). However, the same procedures are used in Example
11.5 where the retardation is variable, and we like to keep the algorithms general. The cellsize x also deter-
mines resolution of the outflow-profile, and it may be useful to include a statement that Ncel must have a
lower value of e.g. 10 as in the program. Output of the example is shown in Figure 11.6 for Ncel of 4 and 10,
and the figure includes the analytical solution (Equation 3.66):

    vL   
ci  L  vt / R  L  vt / R 
cL, t  erfc   exp   erfc 
2  4D t / R   DL   4 D t / R  (11.37)
  L   L 

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 557

In the procedure analyt, the arguments are all divided by L( Totx) so that distance is expressed in pore vol-
umes vt / L which are injected (or eluted).
Element 0 is a conservative substance that is transported by advection and dispersion only (e.g. Cl).
The numerical calculation with only 4 cells shows already excellent agreement with the analytical solution.
When HCH-elution is not corrected for numerical dispersion, an outflow-profile is calculated that is too
diffuse compared with the analytical solution. Decreasing the grid-size by increasing Ncel to 10, gives
a steeper curve and a better approximation, but even better results are obtained when the calculations
with Ncel  4 are corrected for numerical dispersion. The correction is, of course, highly profitable in
terms of execution time compared to the other somewhat blunt (but easy) possibility of decreasing the
cellsize.
There is a point with respect to boundary conditions that should be noted. Concentrations are calculated
for the midpoint of each cell, including the two end-cells (a “cell-centered grid”). A “shift” moves concen-
trations in, and out of the column, without considering dispersion over the distance 1⁄2 x to the column
boundary. This neglect gives some deviation when few cells are used, but can be compensated for by multi-
plying the model mixing factor with (1  2/Ncel). Furthermore, the concentrations in the last cell need to
be transported from the cell midpoint to the column end which requires a half shift, from the shifts.

QUESTIONS:
Include diffusion in the model, estimate the effective diffusion coefficient (cf. Equation 3.47).
Write a program that uses the implicit method for this column experiment (  1). Is the Courant condition
essential when   0?

For the overview, we write down the general finite difference solution for the ARD equation:

ctx2  qtx2  ctx1  qtx1 

 D t vt 
(1   )  e 2  ctx11  2ctx1  ctx11    ctx11  (1  2)ctx1  (1  )ctx11   
 
 (x ) x 
 De t t 2 v t 
   cx1  2ctx2  ctx21    ctx21  (1  2)ctx2  (1  )cxt 21  
    
 (x )
2 x 
(11.38)

where the weighting factors  and  signify different calculation schemes, with respect to
time   0, explicit
  1⁄2, central in time, Crank-Nicholson
  1, implicit
and with respect to
distance   0, upwind when v ! 0
  1⁄2, central in space
  1, upwind when v  0.

A similar equation can be formulated for more dimensions (Zheng and Bennett, 2002). We have
noted the pitfalls of the discretization, resulting in numerical oscillations and dispersion and inaccu-
rate results for coarse grids. Also in more dimensions, the program to be used should be checked for
consistency by calculating a straightforward 1D example and comparing the results with the analyt-
ical solution of the problem and the effects of grid refinement.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

558 Numerical modeling

11.1.3 Non-linear reactions

The third step is to extend the model structure of Example 11.4 to more complicated chemical sys-
tems. Whenever a model is available for calculating the distribution of the chemicals over solution
and solid in a batch system, it can be introduced in the procedure “distri” and then applied to model
transport along a flowline. Let us consider an example of non-linear sorption, described by the
Freundlich equation:
q  K f cn (11.39)

where the exponent n determines the non-linearity of the sorption equilibrium. Newton’s iterative
method can be used to distribute the chemical over water (c) and solid (q) when the total quantity of
a chemical (T) is known. Newton’s method uses derivatives of the distribution function to obtain suc-
cessive better approximations of the variables that determine the function. In our case we seek:
f (c )  c  q  T  c  K f c n  T  0 (11.40)

At a certain initial estimate c1, the derivative of f(c) can be approximated by:
f (c2 )  f (c1 )
 f ′(c1 ) (11.41)
c2  c1

Now, the value of c2 should be such that f(c2)  0; hence from Equation (11.41):
c2  c1  f (c1 ) / f ′(c1 ) (11.42)

This simple formula shows generally rapid convergence for smooth, continuous functions. The
derivative f (C ) can be obtained analytically, or numerically as in Example 11.5.

EXAMPLE 11.5. Effect of the Freundlich exponent on breakthrough curves from a column
Determine the effect for a range of n, varying from n  1 to 0.2, all with identical Kf  1.
The program of Example 11.4 needs only small extension for this problem. The procedure “distri” now cal-
culates the distribution of the chemical over solid and solution iteratively, and stops when the mass-balance
of Equation (11.40) is solved to be better than 106. A function “frndl” calculates the amount of q for each
concentration c. Since the value of the retardation used to correct numerical dispersion may vary, a prelim-
inary run with 3 cells is made to determine its actual maximal value. This maximal value is then used to esti-
mate the number of cells for a dispersion-free calculation. The parameter “numdisp” is set to “true”
whenever numerical dispersion occurs, i.e. when averaging of two mixing factors would give a negative
result. It proved necessary to increase this first estimate of Ncel by 1 (probably due to small concentrations
which have a very high retardation when n  0.2, and run farther ahead than is calculated in the 3-cell
model).

Program Columnperc2; { ... copy from Example 11.4, add Var ... }
Var Kf, nf : array[0..Nel] of real;
procedure dspvty; { ... }
Procedure dspcor; { ... }
procedure dffdsp; { ... }
procedure flush; { ... }
function frndl (Kf, nf, C : real) : real;
begin
if C > 1e-10 then frndl := Kf * exp(nf * ln(C)) else frndl := 0;
end;

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Numerical modeling of transport 559

procedure distri;
{ distribution of Q and C according to Freundlich-isotherm .. }
Var C1, C2, T, Fc1, Fc2 : real;
ic, ie, it : integer;
begin
for ic := 1 to Ncel do for ie := 1 to Nel do
begin
T := Ct1[ie, ic] + Q[ie, ic];
{ .. T(otal) is redistributed over Ct1 and Q with Newton’s method ... }
C2 := T/(Kf[ie] + 1);
it := 0;
repeat
C1 := C2;
Fc1 := frndl(Kf[ie], nf[ie], C1) + C1 - T;
C2 := C1 + 1e-9;
Fc2 := frndl(Kf[ie], nf[ie], C2) + C2 - T;
if abs(Fc1 - Fc2) > 0 then C2 := C1 - Fc1/((Fc2 - Fc1)/1e-9);
if C2 < 1e-9 then C2 := 0.0;
if C2 > T then C2 := T;
it := it + 1;
if it > 20 then begin
writeln(‘FREUNDLICH Eqn, iteration > 20’); halt; end;
until (abs(Fc1)  = 1e-6) or ((C1 + C2) = 0);
Ct1[ie, ic] := C2; Q[ie, ic] := (T - C2);
end;
end;

procedure fill;
begin
Delx := Totx/Ncel;
for i := 0 to Nel do for j := 1 to Ncel do
begin
Ct1[i,j]:= 0; Q[i,j]:= 0; { .. or, if Ct1>0, then Q = frndl(.,.,.)}
end;
Cin[0]:= 1; Cin[1]:= 1 {ug/L};
end;

{Main}
begin
D := 0.0 { cm2/s }; Disp := 0.5 { cm }; Totx := 5.0 { cm };
Porv := pi * (2.5 * 2.5) * Totx * 0.3 { mL };
Veloc := 10/Porv/3600 * Totx { cm/s };
Tott := Totx/Veloc * 5; { Time for injection of 5 PV, s};
Kf[1] := 1; nf[1] := 0.5;
{ initial run for estimating minimum no. of cells ...}
Ncel := 3; Delt := Totx/Veloc/3 { s }; fill;
Rmax := 1; dspvty; Numdisp := false;
for Ish := 1 to 3 do begin flush; distri; dspcor; dffdsp; distri; end;
if Numdisp = true then Ncel := 2 + trunc(Totx/Disp * (1 - 1/Rmax)/2);
{ Now start realistic simulation ...}
fill; Rmax := 1; Numdisp := false;
Nshift := 1 + trunc(Tott * Veloc/Delx);
Delt := Tott/Nshift {s};
dspvty;
for Ish:= 1 to Nshift do

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

560 Numerical modeling

begin
{ ...cf Example 11.4... }
end;
{ write to file1... }
end.

Output of the program is shown in Figure 11.8. The results have been calculated for 5 pore volumes entering
the column with the chemical at a concentration Cin  1, and 10 pore volumes flushing the column. Note
that the fronts of increasing concentration are sharper when the Freundlich exponent n is smaller; a smaller
n gives initially a more rapid elution, but leads to longer tails (cf. Section 3.4).

0.8

Kf  1; n1
0.6
Kf  1; n  0.6
g/L

Kf  1; n  0.2
0.4
Kf  0
(cons.)
0.2

0
0 2 4 6 8 10 12 14
Pore volumes

Figure 11.8. Effect of the exponent n in the Freundlich equation on column breakthrough curves.

PROBLEM:
Write a program for cation exchange of Na and K in a laboratory column (cf. Figure 6.26). The column
length is 8 cm, dispersivity is 0.2 mm, the flow velocity is 100 m/yr, CEC  1.1 mM, ci,Na  1 mM,
ci,K  0 mM, c0,K  1.1 mM, c0,Na  0 mM.

11.2 EXAMPLES OF HYDROGEOCHEMICAL TRANSPORT MODELING

This section illustrates how transport models can help to bring order in the intricate geochemical
reactions in aquifers and discusses practical examples from the literature in a step-by-step manner.
First, an example of groundwater age dating with tritium/helium, showing the effects of differences
in diffusion behavior and dispersion that must be understood to interpret vertical concentration
profiles in aquifers. Second, the breakdown reactions of a NAPL (toluene) in an aquifer, and
the reactions that occur when a site is remediated by adding a surplus oxidant to the system.
Third, a pollution case of acid mine drainage from a uranium mine, where the retardation of the heavy
metals determines the impact on the environment. Fourth, a system of in-situ iron removal where we
learn to optimize a parameter from experimental data. Fifth, a closer look at the origin of the high As lev-
els in Bangladesh and Bengal groundwaters in which we develop a concept that connects the concentra-
tion differences with geographical location. Sixth, a demonstration of isotopic modeling with
PHREEQC.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 561

DFG1 DFG5 DFG6

m.a.s.l.
20

10

10
Aquiclude

20
0 0.5 1 km

Age (years) TU
0 10 20 30 0 20 40 60 80
0 0
Depth (m below g.w.t.)

10 10
3 3
H He
20 Homogeneous 20
aquifer

30 30

40 40

Figure 11.9. Regular flowlines in a phreatic aquifer (top) are contradicted by the age profile calculated from
tritium/helium measurements in borehole DFG1 (modified from Leuchs, 1988).

11.2.1 Tritium-Helium age dating

The flowlines in a phreatic aquifer may follow the pattern shown in Figure 11.9 (modified from Leuchs,
1988, cf. Figure 9.14 for a chemical profile upstream in the same aquifer). Tritium and helium analyses
of water from depth-specific sampling points were used for age-dating the groundwater and to obtain
flow properties of the aquifer (Schlosser et al., 1988, 1989). To interpret the analyses, first consider the
homogeneous case in which the age of water varies with depth according to (cf. Chapter 3):

De w  D 
t  ln   (3.10)
P Dd

where t is time (yr), D is thickness of the aquifer (here
35 m), w is water-filled porosity (
0.3), P
is recharge (
0.3 m/yr) and d is depth (m). Travel time through the unsaturated zone is about 1 year.
In 1986, the year of tritium/helium sampling, the 1963 tritium peak is expected to be at 16.3 m depth
below the groundwater table. In Figure 11.9, the tritium peak is located at 3 m, and the maximum of
tritium  helium lies at 13.3 m depth. Schlosser et al. (1989) assumed that P / w  0.375 m/yr
which would place the maximum at an intermediate depth of 7.3 m. However, a precipitation surplus
(P) of (0.375 m/yr  0.3)  0.1 m/yr is unlikely small for the local conditions. Thus, there are dis-
crepancies that must be explained.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

562 Numerical modeling

We can calculate the age of the water according to:

t  ln(3H / (3H  3He)) / 0.0558 (yr) (3.20a)

The water in the borehole is about 20 years old over the 10–22 m depth range (Figure 11.9). The age
is reasonable for P  0.3 m/yr and the average depth, but the large spread suggests that waters of
various ages are mixed over depth by dispersion. To derive the dispersion data, we follow Schlosser
(1989) and calculate 1D concentration profiles for the variable tritium input in rain from 1950–1985,
with 3H to 3He decay (Equation 3.20), a flux boundary condition for 3H and either a flux or a con-
stant concentration boundary for 3He, and finally, a 4 times higher diffusion coefficient for 3He than
for 3H. All these conditions can be included with little effort in the program given in Example 11.4
and we will discuss the results step-by-step (download tr_he.pas from the website).
First, to check the program, diffusion and dispersion are set to zero and the water age is cal-
culated with depth following Equation (3.20a). The results are exactly identical to the analytical
depth/time solution. Second, tritium and helium can be given the same diffusion coefficient of
109 m2/s and flux boundary conditions. Now the calculated ages deviate slightly from the ideal
(hydraulic) age (Figure 11.10). Above the tritium peak the ages are higher than the hydraulic age and
below the peak they are lower. The deviations are due to the interplay of mixing (a linear process)
and radioactive decay (an exponential process). As the result, the apparent age of a mixture of waters
will come closer to the age of the tritium peak. Third, the diffusion coefficient of 3He can be made 4
times larger than of 3H (cf. Jähne et al., 1987 and data in Table 3.5). The generated 3He now diffuses
faster away from the peak than 3H, and consequently, the 1963 water will appear to be younger while
the surrounding water will be calculated to be older. In this case, the 1963 water is assigned a 5 years
younger age, while the water at 23 m depth is aged by 13 years.

TU Age (years)
0 250 500 750 0 10 20 30 40 50
0 0
Depth (m below g.w.t.)

10 10
D He  4 D H

3H 3
He Ideal
age
20 20
D He  D H D He  4 D H

D He  D H
30 30

Figure 11.10. Tritium/Helium profiles and ages in a homogeneous phreatic aquifer with diffusion. 3He has
4 times higher diffusion coefficient than 3H which reduces the apparent age of water from the 1963 peak.

The analyzed 3H and 3He profiles are more disperse than in Figure 11.10 and as the fourth step
dispersion is introduced in the model. A dispersivity   6 m reduces the maximal tritium concen-
tration to the analyzed 46 TU (Figure 11.11). However, the peak depths are different in Figures 11.9
and 11.11. Also, the model calculates a 3He concentration of 30 TU at the groundwater table
where a concentration of zero was analyzed. It suggests that 3He may escape into the unsaturated
zone, and as the fifth step, a constant boundary condition is defined for 3He, c 3He,d0  0. The
profile now loses appreciable amounts of 3He and the calculated age is considerably reduced.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 563

TU Age (years)
0 25 50 75 100 0 10 20 30 40 50
0 0
Flux b.c. Flux b.c.
Depth (m below g.w.t.)

10 c  0 b.c.
10

3He
3H
20 20
c  0 b.c. Ideal
age

30 30

Figure 11.11. Tritium/Helium profiles and ages in a homogeneous phreatic aquifer with diffusion and disper-
sion. A flux boundary condition at zero depth for 3H and 3He is compared with a constant boundary condition
for 3He.

With the large dispersion, the difference in diffusion coefficients of 3H and 3He has no effect on the
profiles. (This neatly shows the distinction between diffusion, that acts to separate the isotopes of a
species, and dispersion, which mixes them).
How well do the model parameters represent the field conditions? A transverse vertical disper-
sivity of 6 m would imply a longitudinal dispersivity of 600 m (Chapter 3). However, borehole DFG1
lies about 2.3 km from the divide, which suggests that the longitudinal dispersivity should be 230 m
or smaller.
We can also compare mass balances. The mass flow of tritium in a borehole is obtained by inte-
grating over depth:
d D
FTU  ∫ c3 H 3 He H
2O
d(d ) (11.43)
d 0

In the homogeneous aquifer, the flow velocity is the same at all depths:
P x
H  (3.7)
2 O, x Dew

where x is the distance between the borehole and the divide. The flow velocity increases linearly with x.
Thus, in a homogeneous aquifer we can integrate the 3H and 3He concentrations over depth, multiply
with vH2O,x and divide by x, and the result is identical for all boreholes.
In the real aquifer the flow velocity probably varies with depth. Leuchs (1988) indicated that a
5 m thick gravel layer is present at 11 m depth in the otherwise sandy aquifer. If the gravel layer has
a 6 times higher conductivity than the sand, it transmits 50% of the water in the profile.
Consequently, it has a large share in the mass balance of Equation (11.43) and, actually, the mass bal-
ance is difficult to calculate if the variations in flow velocity are unknown. The gravel layer also
explains the disperse distribution of tritium and helium in DFG1 since it mixes waters of various
ages. The mixed water is spread out over depth again when the gravel layer ends, as sketched in
Figure 11.12.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

564 Numerical modeling

DFG1 DFG5 DFG6 m.a.s.l.

20

10

10

Aquiclude 20

0 0.5 1 km

Figure 11.12. Water in the aquifer (Figure 11.9) mixes in a gravel layer and spreads out over depth when it ends.

With a few simple formulas we can estimate the flow properties of the gravel layer that may explain
the observed age profile in DFG1. For example, if half of the profile (17.5 m) is homogenized, then
the gravel layer has a conductivity that is (D – z) / z larger than the rest, z being the thickness of the
gravel layer. Mixing in the layer requires a flowpath L that can be estimated with Equation (3.56),
neglecting the cross terms which increase xy:

 2  2DL t  2(De t  T L) (3.56a)

in which t and L are connected through:

De w  x 
t  ln   (3.10c)
P  x  L

For example, with De  31010 m2/s, x  2200 m, T  0.005 m, and   z / 2  2.5 m, L  605 m.
These relations are useful for initiating a hydrological model in which irregularities are tried out and
distributed over the section to simulate observed age patterns. However, it will be difficult to find a 2D
or 3D hydrological model in which helium and tritium can be given different diffusion coefficients
and/or different boundary conditions at the water table.
Neglecting the difference in the transport properties of tritium and helium, we can calculate
the profile of Figure 11.12 with PHAST (Parkhurst et al., 2004, files tr_he.chem.dat and
tr_he.trans.dat from www.xs4all.nl/appt). Figure 11.13 shows calculated tritium and tritium/
helium age profiles at 300, 1900 and 2300 m distance from the divide for 1986. The 300 m profile
is for the homogeneous aquifer and would be seen everywhere in case the gravel layer were absent.
The 1900 m profile crosses the gravel layer at 11–16 m depth. Note that the peak concentration is
much less than at 300 m and that the integrated concentration over depth is only half of what is found
at 300 m. However, the fluxes calculated with Equation (11.43), divided by x, are identical. Lastly,
the 2300 m profile is similar to the observed one in borehole DFG1 (Figure 11.9).
Inspection of the PHAST output files shows that the hydraulic heads in the cross section are down
to the mm scale invariant over depth, implying that the flow velocity at x  1900 m changes according
to the hydraulic conductivity variation. Thus, when calculating the tritium flux according to Equation
(11.43), the tritium peak concentrations between 11 and 16 m depth are multiplied by the higher veloc-
ity that is in direct proportion with the higher hydraulic conductivity of the gravel layer. We can also
conclude that the velocity variations in the aquifer are difficult to discern from hydraulic head meas-
urements which have only a mm-scale resolution. They can only be found by doing tracer tests.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 565

TU Age (years)
0 25 50 75 100 0 10 20 30 40 50
0 0
Depth (m below g.w.t.)

10 10

1900 m 300 m
20 20
Ideal
2300 m (DFG1) age
2300 m (DFG1) 300 m 1900 m
30 30

Figure 11.13. Tritium and tritium/helium age profiles in the aquifer of Figure 11.12 at varying distance from
the divide, calculated in 2D with PHAST. The profile at 1900 m crosses a gravel layer and has a much smaller
tritium content than at 300 and 2300 m (but the flux FTU remains the same). The profile at 2300 m is for bore-
hole DFG1 (Figure 11.9).

Finally, the modeled gravel layer is not the unique factor for explaining the analyzed profile,
it may well be possible to find better matching results for the analyses of DFG1 with other hydro-
geological configurations. Indeed, many profiles with environmental tracers such as tritium, tri-
tium/helium and CFC’s or SF6 have been reported and were compared with flow model results, but
it appears difficult to generate a satisfactory model from many possible alternatives (Schlosser et al.,
1988, 1989; Engesgaard et al., 1996; Szabo et al., 1996; Sheets et al., 1998; Shapiro et al., 1998;
Shapiro, 2001; Zoellmann et al., 2001; Weissmann et al., 2002). Discrepancies are invariably attrib-
uted to improper conceptualization or non-uniqueness of the system. However, statistical optimiza-
tion of the hydraulic conductivity variations using these environmental tracers has not been
performed yet. On the other hand, the results presented in this section illustrate that water quality is
rapidly homogenized over depth after a few km flow, which marks the problem of finding exact
flowpaths as ill-posed from the start. We’ll never find the exact answer.
How then, can we explain the presence of chemical variations with depth in many (saturated)
profiles presented in previous chapters (Figures 1.7, 2.18, 3.2, 8.22, 9.14, 9.18, 9.20, 9.26, 9.44,
9.45, etc.)? Reinspection shows that the largest variety is found close to the point of infiltration
where dispersion has not yet gained a dominant influence (Figures 3.2 and 9.18, DFG6 in Figure
9.14 is close to the divide, cf. Figure 11.9). In case a conservative parameter like Cl is already
mixed over depth, the variations result from a chemical reaction with the aquifer (Figures 2.18, 8.22,
9.26 and 9.44). The changes can also indicate a diffusive source from another water, or from sedi-
ments with a different origin (Figures 3.28 and 9.45). Thus, it is the position in the flowfield of the
aquifer, the reactivity of the aquifer with respect to the water quality, and the different properties of
adjoining formations that determine contrasts in groundwater chemistry. And, that’s why hydraulics,
geology and geochemistry must be integrated when explaining water quality variations in aquifers.

11.2.2 Toluene degradation in an aquifer

Toluene, C7H8, and other BTEX compounds (benzene, toluene, ethyl-benzenes and xylene)
have been spilled at thousands of sites and have contaminated groundwater. Yet, the actual
spread of these contaminants is usually limited because the chemicals are degraded in the aquifer.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

566 Numerical modeling

P  200 (mm/year)

15
Residual source: 1 vol% C7H8(I)
Elevation (m)

10 Vapor plume Source zone: 10 vol% C7H8(I)

5 hw  9 m hw  7.5 m
Dissolved plume
0
0 100 200 300 400
Distance (m)

Figure 11.14. Cross section through a model aquifer polluted by toluene. The depicted plume extension is for
conservative behavior of toluene. Note for the hand calculations in the text, that the left boundary (x  0) is at
50 m from the divide (Mayer et al., 2002).

Breakdown occurs via microbially mediated oxidation with dissolved oxygen, nitrate, sulfate, and
with sedimentary iron-oxyhydroxides as electron acceptors (MacQuarrie et al., 1990; Hutchins et al.,
1991; Lovley et al., 1994; Chapelle et al., 1996; Schmitt et al., 1996; Barbaro and Barker, 2000;
Cozarelli et al., 2001, Example 9.11). An interesting point is that the dissolved oxidants need to
be mixed with the contaminant plume by dispersion and diffusion, while the solid and sorbed oxi-
dants (and the associated bacteria which are fixed on the sediment particles) may react when
the reducing pollutant flows by. In addition, BTEX can also be degraded by fermentation. The
different degradation paths lead to various reaction products located distinctly within or around
the plume.
The outline of an example case modeled by Mayer et al. (2002) is shown in Figure 11.14. The soil above
a uniform, phreatic aquifer was polluted with toluene and a plume developed in the groundwater body.
The concentrations of various chemicals, 10 years after onset of the pollution, are presented in Figure
11.15; we follow the numerical simulation of Mayer et al. with simple hand calculations. First, we
calculate concentration and distance traveled by toluene, assuming conservative behavior. The effective
solubility of toluene used by Mayer was 1.1 mM (this is less than the actual solubility to mimick
the time-dependent dissolution from blobs, cf. Chapter 10). In 10 years time, c0.5  0.55 mM is expected
to have traveled from 100 to 229 m for the conditions given in Figure 11.14, an average hw  8.5 m and a
water-filled porosity w  0.38 (hw  D in Equation (3.9); note that x  100 m in Figure 11.14 is 150 m
from the divide). The position of c0.16  0.18 mM would be at 100  129 
(2  129  0.1)  234 m
for L  0.1 m. This plume extension fits the numerical result in Figure 11.14. However, Figure 11.15
shows that toluene has not advanced as far because of degradation.
The oxidation reaction of toluene,
C7H8  14H 2O → 7CO 2  36H  36e (11.44a)

requires that 36 electrons per mol toluene are taken up by oxygen:

36e  9O 2  36H → 18H 2O, (11.44b)

or by iron-oxyhydroxide:
36e  36FeOOH  108H → 36Fe 2  72H 2O, (11.44c)

or by methanogenic reduction of toluene:

36e  1.8C7H8  36H → 12.6CH 4 (11.44d)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 567

15
C7H8 (aq)
Elevation (m)

10 1.11003

2.41004
5.11005
5 1.11005
(mol/L)
0
50 100 150 200 250 300
Distance (m)
Elevation (m)

15 15
P CO2 8.41002 PCH4 8.01001
5.61002 5.41001
 
10 10
2.91002 2.71001
5
1.11003 5 8.01003
0 0 (atm)
(atm)
50 100 150 200 250 300 50 100 150 200 250 300
Elevation (m)

15 15
pH 8.2 Ca 3.51003
2.41003
 
10 7.6 10
7.1 1.21003
5 6.5 5 3.81005
0 () 0 (mol/L)
50 100 150 200 250 300 50 100 150 200 250 300
Elevation (m)

15 15
9.91006 1.51005
 Goethite 6.61006
 Siderite 1.01005
10 10
3.41006 5.01006
5 9.91008 5
1.51007
0 0
50 100 150 200 250 300 () 50 100 150 200 250 300 ()
Elevation (m)

15 1.01005 15
C7H8 oxidation (O2) C7H8 oxidation (goethite) 1.01007
1.01006 1.01008
 
10 10
1.01007 1.01009
5 1.01008 5 1.01010
0 (mol/L/d) 0 (mol/L/d)
50 100 150 200 250 300 50 100 150 200 250 300
Distance (m) Distance (m)

Figure 11.15. Simulation results of biodegradation of toluene in groundwater. Dissolved concentrations are in
mol/L, gas pressures in atm, solids as volume fraction [], and reaction rates in mol/L/day. (Mayer et al., 2002).

Protons are balanced when Reaction (11.44a) is combined with (11.44b) or (11.44d), so that the pro-
duced CO2 acidifies the water. As a result, the pH decreases, calcite (when available) dissolves and the
Ca2 concentration increases (Figure 11.15). On the other hand, Reaction (11.44c) appears to consume
considerable amounts of protons, but most of the released Fe2 will precipitate in the form of siderite:

36e  36FeOOH  72H  36HCO

3 → 36FeCO3  72H 2O (11.44e)
which still consumes acidity. Thus, the development of pH can indicate which oxidation reaction
takes the lead.
The model results in Figure 11.15 show that pH is lower in the plume than in the surrounding
groundwater, which indicates that oxidation by oxygen or methanogenic fermentation is the most
important degradation reaction. It can be verified indeed in Figure 11.15, that the rate of oxidation
by O2 has been assumed to be 100 times larger than by goethite. Most of the oxidation takes place in
the unsaturated zone where gaseous diffusion of oxygen is much more effective than in the aquifer,
where only a small rim of the plume is reached by aqueous oxygen. Mayer et al. (2002) have con-
sidered various other reactions in their model, such as CH4 oxidation, and they included also the less
crystalline ferrihydrite, which reacts quicker than goethite.

QUESTIONS:
Recalculate the volume fractions of 105 goethite and 1.5105 siderite in Figure 11.15 to mmol/L
pore water? sid  3.96 kg/L, go  4.3 kg/L, w  0.38.
ANSWER: 1.27 mmol go/L, 1.35 mmol sid/L.
Calculate the residence of water in the unsaturated zone of the toluene case, Figure 11.14, when w  0.1?
ANSWER:
3.5 yr.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

568 Numerical modeling

Use PHREEQC to calculate the pH, the CO2 pressure and the Ca2 concentration when 105 mol
C7H8/L/d oxidizes with O2 in the unsaturated zone? Calcite dissolves to equilibrium, temperature is 25°C.
ANSWER: pH  5.78, PCO2  100.36  2.29 atm, mCa  13.0 mM.
CO2 escapes from water in the unsaturated zone, therefore repeat the calculations limiting [PCO2] to 8.4102.
ANSWER: pH  6.69, mCa  3.6 mM.
Use the result of the previous calculation as starting point for toluene polluted groundwater, and calculate
the composition when goethite oxidizes 107 mol C7H8/L/d over 10 yrs in 1 yr steps. Allow for precipitation
of siderite. Plot the resulting pH, PCO2, Ca2 and Fe concentrations with user_graph as a function of dis-
tance (use Equation (3.9), note that x  0 is 50 m from the divide) and compare with Figure 11.15.
ANSWER: SOLUTION 1; -temp 25
EQUILIBRIUM_PHASES 1
Calcite; O2(g) -0.68; CO2(g) -1.075 0
REACTION
C7H8 1; 0.0128
SAVE solution 2
END
USE solution 2
EQUILIBRIUM_PHASES 1
Calcite; Siderite 0 0; Goethite
REACTION
C7H8 1; 3650e-7 in 10
USER_GRAPH
-head x pCO2/10 Ca Fe*100 pH
-start
10 graph_x 150 * exp(0.2 * step_no /(8.5 * 0.38)) - 50
20 graph_y 10^SI(“CO2(g)”)/10, tot(“Ca”), tot(“Fe”) * 100
30 graph_sy -la(“H+”)
-end
END

Write the redox equation for toluene oxidation with SO42 as electron acceptor?
ANSWER: 36e  4.5SO42  45H l 4.5H2S  18H2O
In the PHREEQC input file, add 1 mM SO42 to the initial, polluted groundwater and allow for siderite and
pyrite precipitation. Plot the sulfate concentration and the amounts of siderite and pyrite that precipitate.
Which phase is more insoluble, pyrite or siderite?

11.2.3 Remediation of a BTEX polluted site

The previous modeling example showed that a 2- or 3D model is necessary for evaluating the reac-
tions at the plume’s rim where water with oxygen or other oxidants admixes. Internally in the plume
a 1D model may be applied along a flowline. The advantage of a 1D model is that simulated con-
centrations can be easily compared with observations in a single picture. A site below a former gas-
works in Düsseldorf (Germany) was remediated by injecting water with nitrate, and the
concentrations in the aquifer were measured in multilevel sampling points (Figure 11.16). The com-
parison of model and observed concentrations allowed to define a comprehensive reaction scheme
in which the reactions of both mineral and organic compounds were combined (Eckert and Appelo,
2002).
Figure 11.17 shows a selection of observed concentrations and model results for conser-
vative transport of the compound (grey lines) and with reactions (black lines) in the plume at
2 m distance from the injection well. The decrease of injected NO 3 is initially accompanied
by an increase of the SO2 4 concentration due to reaction of FeS. This FeS formed earlier when
the BTEX compounds reacted with SO2 4 from groundwater (Wisotzky and Eckert, 1997).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 569

m.a
10 m KNO3 solution
s. l.
40

38 12 Multi-level-
filter element
36
14 Well screen
34 6 6 Finite-
7 Excavation
7 difference-net
32 8 8 area
9 9
30 Injection-
10 10 plume
11
28 12
12
26 Quaternary 14 Groundwater flow 13
14
Tertiary
24

Figure 11.16. Cross section showing injection, abstraction and observation wells below a former gasworks
site (Modified from Eckert and Appelo, 2002).

The oxidation of FeS was proved with
34S, which is low in reduced sulfur compounds and indicates
the source of dissolved SO 2
4 . When FeS is exhausted after about 50 days, the still present BTEX
compounds give a small reduction of the NO 3 concentration.

2.0 2.0

NO3
2 1.5
1.5 SO4
mmol/L
mmol/L

1.0 1.0

0.5 0.5

0.0 0.0
0 20 40 60 80 0 20 40 60 80
Time (days) Time (days)

0.2 7.6
2 7.4 pH
Fe
mmol/L

7.2
pH

0.1
7.0
0.0 6.8
0 20 40 60 80 0 20 40 60 80
Time (days) Time (days)

Figure 11.17. Selected concentrations in groundwater during remediation of a BTEX site with nitrate
amended water. Squares: observed data; grey lines: conservative transport of injected water; black lines: model
with reactions (Eckert and Appelo, 2002).

The data show the combined presence of NO 3 and Fe

2
in the groundwater after 20 days, which is
thermodynamically not feasible, and points to a kinetic oxidation of Fe2 assisted by microbes or
perhaps by solid surfaces. The kinetic reaction was modeled assuming that the rate was additive to
the a-biological oxidation of Fe2 with O2 (cf. Example 4.12):
dmFe2 / dt  mFe2 (2.91109  1.331012 [OH ]2 (773 mO  f1 mNO)) (11.45)
2 3

where t is time (s) and mi is the molality of i. The dependency of the oxidation rate on nitrate was
added as (f1 mNO3 ) and calibrated to f1  200, or 4 times lower than for O2. Actually, we don’t know
anything about the oxidation mechanism except that it happens in soils and with microbes.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

570 Numerical modeling

Another interesting feature is the pH change that results from the redox reactions and which can be
used for deciphering the reactions. When the oxidation is incomplete (Fe2 is not oxidized), the
reaction of NO
3 with FeS increases the pH:

NO
3  ⁄8FeS.  H
5 
→ 0.5N2  5⁄8SO2
4  ⁄8Fe
5 2
 0.5H2O (11.46)

but when the oxidation is complete, the pH decreases:

NO
3  ⁄9FeS  ⁄3H2O →
5 1 1
4  ⁄9FeOOH  ⁄9H
⁄2N2  5⁄9SO2 5 1 
(11.47)

Whether or not iron-oxyhydroxide forms in the reaction does not really matter, the hydroxy-complexes
of Fe3 are already acidifying.
The fraction of iron that should oxidize for pH to remain neutral (as observed, Figure 11.7), and
the amount of FeS that reacts can be calculated (cf. Question):

NO
3  ⁄16FeS  ⁄4H2O →
9 1 1
⁄2N2  9⁄16SO 2
4  ⁄2FeOOH  ⁄16Fe
1 1 2
(11.48)

or 1 / 9 of Fe from FeS should remain Fe2, which is close to observed in the initial 20 days (com-
pare the reaction amounts of SO 2
4 and Fe
2
in Figure 11.7). Furthermore, initially the pH increases
rather sharply due to the reaction of (Fe,Ca)CO3:

x/5NO
3  Fex Ca1x CO3 7x/5H2O  (1  9x/5)H

→
x/10N2  xFeOOH  (1  x)Ca2  HCO (11.49)
3

which increases pH when x  5⁄9. The siderite formed earlier together with FeS by reaction of BTEX
with iron-oxyhdroxide, cf. Figure 11.15.

QUESTIONS:
In Example 4.12, the rate of Fe2 oxidation is expressed with PO2 as variable. Find the conversion factor
to use mO2, check with Equation (11.45)?
Include the empirical kinetic oxidation reaction of Fe2 with NO 3 in PHREEQC (use Example 9 from the
PHREEQC manual as template). Is the reaction with NO 3 more or less acid than with O2?
ANSWER: less acid
Find the concentration of NO 3 for which the Fe
2
oxidation rates with O2 and NO 3 are equal?
ANSWER: mNO3  773 mO2/200
Compare the changes in oxidation rates with time for O2 and NO 3 when the initial rates are the same?
ANSWER: NO 3 maintains a larger rate because the reaction is less acid and the rate
depends on OH.
Derive a pH-neutral oxidation reaction of pyrite with nitrate, similar to Equation (11.48). Hint: find the
coefficients a and x in the reaction:
NO
3  aFeS2  (8a  2x  3)H2O →
1
⁄2 N2  2aSO2
4  (ax)Fe
2
 xFeOOH
by balancing on H and charge.
ANSWER: a  6 / 14, x  1 / 14

11.2.4 Acid drainage from a Uranium mine

The former uranium mine “Königstein” is located South of Dresden (Germany), at a depth of
about 250 m in Cretaceous deposits shown in cross section in Figure 11.18. The deposits consist
of alternating (consolidated) sand and clay layers in which four aquifers are discerned.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 571

Langenhennersdorf Panoramahöhe
N

2309
Elbe River

2318
Nikolsdorfer

Leupoldishain
S

2308
mNN Wände

Kleiner Bärenstein

Großer Bärenstein

Teufelsschlucht

2349
2346
2468
1716/1936
400

2331

1161
2363
Kehllochweg

Höllengrund

2492
1694
Thörmsdorf

2082
Kohlgrund

2160
1528/1522
Waldbach
Brosche
1

1770
1424
1665
1422
419
2

2615

1838
BI.490
300

2022
1235
1874
BI.446

2606
2630
2628/89

2619
1078
3
BI.2424/81
200
4

100
mined strata
0 1000 m
0

100
Quaternary, in recent valleys
Granodiorite
Sandstone, fine / coarse
Oxidation effects
Sandstone, fine / loamy / clayey notable in sandstone
Claystone, fine sandy, glauconitic Uranium ores

'Pläner', sequence of lime- and claystones

Limnic and riverine deposits

1-4 Main aquifers

Figure 11.18. N–S cross section over the Königstein mine (after Tonndorf, 2000).

The fourth and deepest aquifer contains locally minable uranium, probably emplaced in the strata
when U(6), advected by groundwaterflow, was reduced by organic matter and sulfides, to
precipitate as uraninite (UO2) and coffinite (USiO4) in the form of rollover fronts (Tonndorf, 2000).
The fourth aquifer is overlaid by an aquiclude and by the third aquifer, which under natural flow con-
ditions is waterfilled for 50 m at the site of the mine.
However, the mine was pumped during operation, which created a pumpage cone in the third
aquifer and oxidizing conditions in the mine, in which sulfides oxidized. In addition, sulfuric acid
leaching of U formed part of the mining operations and the mine water still has, 10 years after clo-
sure, an extremely low pH and high concentrations of U and heavy metals. The mine water has been
pumped and treated above ground so far, but the costs are excessive. When pumping ceases, the mine
water will enter the third aquifer via a fault zone and flow towards the Elbe River where it will seep
up in and near the river bed. We want to predict the concentrations of heavy metals in the plume and
the seepage zones near the river, and to model the attenuation processes (Walter et al., 1994; Bain
et al., 2001). The flowtime for conservative transport from the mine to the Elbe River is estimated to
be about 15 years (data in Bain et al., 2001).
Ideally, we like to have column experiments to verify the fast sorption reactions of heavy metals
and protons. The slow dissolution and precipitation reactions of various minerals that may occur over
15 years flowtime can only be extrapolated by using published kinetic rate data. The slow reactions of
the resident alumino-silicates, sulfides, and other minerals are important, since they change the pH
and pe of the mine water which in turn affect sorption and precipitation of the heavy metals.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

572 Numerical modeling

Exchange and sorption of U, Cd2 and Zn2 as a function of pH

So far, the literature on the Königstein mine has neglected the sorption reactions that attenuate trans-
port of the heavy metals, but we may apply the exchange and surface complexation models from
Chapters 6 and 7 to this case. For lack of data, we use the default constants from the WATEQ4F data-
base and assume the exchange capacity of the quartz rich sediments, CEC  50 mmol/L, and the
surface complexation on ferrihydrite, Hfo  10 mmol/L. The dominant clay mineral in the sedi-
ments is kaolinite (or dickite), which has a PZC at pH
4.5. Accordingly, in the low-pH mine water
proton exchange on X is important and included in the model. Using the mine water composition
of Bain et al., we can estimate distribution coefficients (Kd  q / c) as a function of pH. We make an
input file with the solution, add exchange and surface, and use NaOH as reactant to increase the pH
from 2.3 in the mine water to about 10 (Table 11.1).

Table 11.1. PHREEQC input file for calculating distribution coefficients of U, Zn2 and Cd2 in mine water
(Figure 11.19).

PRINT; -reset false

SOLUTION 1 # Mine water from Bain et al., 2001
-temp 10; pH 2.3
Na 23.8; K 0.1; Mg 2.0; Ca 11.6
C 1.7e-4; Cl 13.0; P 0.08; S(6) 52.8
Al 6.5; Cd 0.01; Fe(3) 10.7; Fe(2) 0.27;
U(6) 0.18; Zn 1.5;

SURFACE 1
Hfo_w 2e-3 600 0.89; Hfo_s 5e-5; -equil 1 # (0.012 % FeOOH_am  0.01 mol HFO/L)

EXCHANGE_SPECIES
H+ + X-  HX; log_k 1.0; -gamma 9.0 0.0 # kaol has PZC  4.6, sorbs protons
EXCHANGE 1
X 50e-3; -equil 1 # (CEC  50 meq/L)
REACTION 1; NaOH 1; 105e-3 in 100
INCREMENTAL_REACTIONS

USER_GRAPH # plot q/c...

-head pH U Zn Cd
-start
10 graph_x -la(“H+”) # find sorbed amounts...
20 q_U  mol(“Hfo_sOHUO2+2”) + mol(“Hfo_wOUO2+”)
30 q_Zn  mol(“ZnX2”) + mol(“Hfo_sOZn+”) + mol(“Hfo_wOZn+”)
40 q_Cd  mol(“CdX2”) + mol(“Hfo_sOCd+”) + mol(“Hfo_wOCd+”)
# plot the distribution coefficients ...
50 graph_y q_U / tot(“U”), q_Zn / tot(“Zn”), q_Cd / tot(“Cd”)
END

The results are shown in Figure 11.19. The distribution coefficient for Cd2 is about 0.2 over most
of the pH range, but increases for pH ! 9.5. The sorbed concentration is entirely due to the
exchange reaction, except at very high pH where surface complexation on Hfo becomes important.
The distribution coefficient for Zn2 is 1 in the low pH range. It increases to 2.5 at pH 8 by sorption
on Hfo, and diminishes again at higher pH where aqueous complexes redistribute Zn2 back into the
solution.
At low pH, Zn2 is only retarded by cation exchange, like Cd2, but why is the distribution coef-
ficient 5 times higher for Zn2 than for Cd2? Inspection of the database shows that both metals have
the same exchange constant. However, for ZnX2 an activity correction is defined, but not for CdX2.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 573

Distribution coefficient, q/ c (–)

4

U
2
Zn

Cd
0
2 4 6 8 10
pH

Figure 11.19. Distribution coefficients of trace metals as function of pH. The mine water composition is given
in Table 1.1. The pH is increased by adding NaOH.

This explains most of the difference, while another part is due to aqueous complexes which are
stronger for Cd2 than for Zn2 in this water. The distribution coefficient for U, which is only bound
to Hfo (at least according to the WATEQ4F database), displays the usual competition among com-
plexes in solution and on the surface that is characteristic for heavy metal sorption. At low pH, all U
remains in solution and the distribution coefficient is 0. At intermediate pH, Kd increases to 4, but it
dwindles again to 0 at higher pH. Kohler et al. (1996) noted that sorption of U(6) occurs on quartz,
but it is small compared to Hfo.

Retardation in a multicomponent solution

In the previous section, the retardation of heavy metals was estimated from the distribution coeffi-
cient. Actually, the slope of the isotherm should be considered (Chapter 3), which is influenced by
concentration changes of all the elements. For example for uranium:
 ∂c   ∂c   ∂2 c     ∂c     ∂q   ∂c  
∂q
 U   v  U   D  U    U   U  
    ∂c
U   Zn   

 ∂t  ∂x  ∂x t  ∂cU  x, Cd,Zn,…  ∂t  x, U,Cd,… ∂t  x 
2
x t  x   Zn 
(11.50)
The slopes (#qU / #cj)x,i4j are termed flushing factors for U. A higher flushing factor means higher
amounts sorbed, and thus more pore volumes for flushing the chemical. A flushing factor of 0 indi-
cates zero effect on sorption and no effect on transport. For cj  cU, the largest effect is expected on
qU, and indeed, 1  #qU /#cU is equal to the retardation for U, as was defined in Chapter 3.
We can calculate the flushing factors for the Königstein example numerically, chang-
ing the concentration of an element in the initial solution a tiny bit and comparing the
exchangeable and surface complexed concentrations before and after the addition. The result for U
and Zn2, when changing the U concentration (adding 108 mol/L), is shown in Figure 11.20.
The flushing factor is markedly different from the distribution coefficient for U, and the flushing fac-
tor for Zn2 by U can even become negative. What is happening?
The distribution coefficient is the ratio of sorbed and solute concentrations, and it is assumed
constant as if we have a linear isotherm. However, with the flushing factor we have calculated
the slope of the isotherm. Apparently, the slope #q / #c is larger than q / c for U, indicating that we
have a concave isotherm at a fixed pH (the slope of the isotherm lies below the curve).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

574 Numerical modeling

10

∂q / ∂c U or q/c (–)
6
∂q U / ∂c U
4

2 q U /c U q Zn /c Zn

0
∂qZn / ∂cU
2

4
2 4 6 8 10
pH

Figure 11.20. Flushing factors (slopes (#qi / #cU)) for U(6) and Zn2 as function of pH (thick lines). The
distribution coefficients q / c from Figure 11.19 are drawn for comparison (dotted lines).

Consequently, the U(6) front will be sharpening when concentrations decrease. On the other hand,
for Zn2 the slope (#qZn /#cU) i5U is negative in the pH range where U competes with Zn2 for sorp-
tion sites. The negative flushing factor for Zn2 by U indicates that Zn2 is desorbed by U. It can be
seen in Figure 11.20 that the displacing effect of U can be even larger than of Zn2 itself, which sug-
gests that the flushing factor for Zn2 might become negative. However this does not happen. When
the sorbed concentration of Zn2 diminishes, also the flushing factor tends to zero and the retarda-
tion becomes 1 (retardation cannot be smaller than 1).
We can calculate a concentration profile with PHREEQC and see how these retardations apply
along a flowline. Going from the mine to the Elbe River at 2000 m, Figure 11.21 shows the concen-
trations of SO2 2
4 , Zn , U(6) and pH when the mine water has traveled 1500 m. Sulfate shows almost
conservative behavior (it is only very slightly sorbed by Hfo) and indicates the position of the acid
front. Zinc is retarded by 2.1, in agreement with the calculated distribution coefficient of 1.1 when
the pH is below 5 (Figure 11.19). In the low pH range, surface complexation of Zn2 on Hfo is
minor, and the retardation is not affected by U. The concentration of U shows an increase as if a heap
of snow is pushed up in front of a snowplough. The typical shape is due to the pH variation in the
front; uranium is sorbed at neutral pH, but as the pH decreases to below 4 in the acid mine water, U
is desorbed again, and the decreasing concentration gives a sharp front.

0.3 8

7
U(6) pH
0.2 6
mmol/L

2
SO4 /300
pH

0.1 Zn/10 4

0 2
0 500 1000 1500 2000
Distance (m)

Figure 11.21. Calculated concentrations of SO42, Zn2, U(6) and pH along a flowline from the Königstein
mine to the Elbe River. The acid front of the mine water is at 1500 m.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 575

The flushing factors are needed in transport models when correcting for numerical dispersion
(Example 11.4), or when correcting for reactions in implicit schemes. Kirkner and Reeves (1988)
suggest to calculate the factors by numerical differentiation as we did with U. However, in general
this is not easy since the difference between activity (which determines sorption) and concentration
(which is transported) may influence the results. Furthermore, the flushing factors will show numer-
ical roundoff errors which affect the charge balance of the transported solution after the correction.
This may lead to non-convergence of the combined transport and geochemical model.
In most cases, the retardation (1  #qi /#ci) will have the largest influence on transport of i,
and can be calculated easily from changes in the solute concentration over time in a cell using
Equation (11.31). Zysset et al. (1994) calculated its value after a half-time transport step, and
corrected transport with it in the next half-time step in a procedure called Strang-splitting. After
the correction, the model may iterate until the newly calculated retardation is equal to the previous
one within a given tolerance (Yeh and Tripathi, 1989; Engesgaard, 1991). By neglecting the cross-
terms (#qi / #cj), possible chromatographic peaks and troughs (cf. Chapter 6) are disregarded,
but these are smeared out anyhow over the cell domain for which the geochemical model calcu-
lates the reactions. The code HydroGeoChem by Yeh et al. (2004) does consider the cross-terms
(Yeh, pers. comm.).
An interesting point to consider is the retardation of Si, when the solution is in equilibrium with
quartz, and pH is below 10. In this case, dcSi  0, and the retardation is infinite by Equation (11.31).
The concentration of H4SiO 04 is invariable when quartz is present, and we don’t even need to calcu-
late transport and reactions. This holds for other minerals as well, each mineral fixing the concen-
tration of one component by the Phase Rule (Saalting et al., 1999).

QUESTIONS:
Calculate the distribution coefficient for mine water with activity correction for CdX2 at low pH?
ANSWER: 0.8
What will be the flushing factor of U by Cd2?
ANSWER: minimal, Cd2 is not much sorbed by Hfo
Calculate the flushing factors for Cd2, Zn2 and U by U, Zn2 and Cd2 for the Königstein case? Hint:
in the input file in Table 11.1, put qi and ci for each reaction step in a memory location, “60 put(q_U, 1,
step_no); 70 put(tot(“U”), 5, step_no)”, etc. Then, in a subsequent simulation, add 108 mol U/L to the solu-
tion, calculate again the qi’s, and find qi /cU. Repeat, adding 108 mol/L of Zn2 or Cd2.

Other effects on transport of heavy metals

Overall, Figures 11.19 and 11.20 show that retardation of the heavy metals will be small when
acid mine drainage enters a silicate aquifer, but that it may become significant when the pH
increases to 6. There are at least four mechanisms imaginable that can increase the pH, viz. sorption of
protons, silicate dissolution (the concentration of Si is quite high in the mine water, cf. Bain et al., 2001),
dilution by mixing with local recharge water, and reaction with calcite in confining layers. The assumed
CEC and concentration of Hfo in the Königstein aquifer can take up 10 mmol H/L mine water at
pH  2.3, and hence can retard the acidity by a factor of 1  (0.01 / 102.3)  3 (cf. Figure 11.21
which shows that pH  2.3 has only traveled 500 m). Equilibrium with goethite, pyrite and kaolinite
will bring the pH to 4.6 and perhaps even higher when the iron-oxyhydroxide in the aquifer is more
soluble than pure goethite (cf. Questions), but the reactions are kinetic. To assess the effect of dilution
by mixing with local recharge water requires detailed hydrogeological information, and finally, the reac-
tion with calcite may be feasible since large amounts of calcite are present in the confining layers
(Tonndorf, 2000).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

576 Numerical modeling

3 5
pH

Saturation index
0 Pyrite
4

pH
3
Goethite
3
6

9 2
0 5 10 15 20 25 30
Years

Figure 11.22. Kinetic reaction of pyrite and goethite with mine water increases the pH. The saturation indexes
of pyrite and goethite indicate the time course of the reactions.

Kinetic mineral reactions

The reactivity of the kinetically reacting minerals in the aquifer with mine water can be explored
with PHREEQC in a step-by-step approach starting with pyrite/goethite and then adding the silicate
minerals (cf. Glynn and Brown, 1996). The major variables are the pH and the saturation indices of
the minerals, which change during the reaction progress and end up at zero when final equilibrium
is reached (Figures 11.22 and 11.23).
The mine water is supersaturated with respect to goethite, but undersaturated for pyrite. Dissolution
of a tiny amount of pyrite lowers the pH and leads to subsaturation for goethite, which starts to dissolve:

FeS2  14FeOOH  26H → 15Fe 2  2SO 24  20H 2O (11.51)

The kinetic dissolution rates for pyrite were discussed in Section 9.4.2, and for goethite given in
Equation (9.82). Figure 11.22 shows how the pH increases concomitantly with the dissolution of
goethite. Pyrite reaches saturation after approximately 12 years in the model, goethite needs 22 years
and the pH is then 4.6.
The mine water is also subsaturated for kaolinite, a mineral that reacts quicker than goethite:
Al 2Si 2O5 (OH)4  6H → 2Al3  5H 2O  2SiO 2 (11.52)

where the released silica precipitates as cristobalite.

For kaolinite, we look up the rates defined by Sverdrup and Warfvinge (1995) and find:
Kaolinite
# Rate from Sverdrup and Warfvinge, 1995, Am Mineral. 31, 485.
# r in 1/6 kmol kaol/m2/s (is 1 kmol H+/m2/s)
-start
10 A  3150 * m0 # m2 for 10 m2/g
20 f_Al  1 + tot(“Al”)/4e-6
30 r  10^-15.1 * act(“H+”)^0.7 * f_Al^-0.4 + 10^-17.6 * f_Al^-0.2
40 moles  A * r/6 * 1e3 * (m/m0)^0.67 * (1 - SR(“Kaolinite”)) * time
50 if moles  0 then moles  moles / 10 # 10 times smaller precipitation rate
60 save moles
-end

Sverdrup and Warfvinge (1995) defined the rate in terms of proton buffering of 6 mol H per mol
kaolinite, and the rate was reformulated to moles of kaolinite. The rate contains inhibition factors
that account for the decrease in rate by dissolved Al3 as discussed in Chapter 8. Furthermore, the
rate for precipitation was (arbitrarily) assumed to be 10 times smaller than for dissolution.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 577

3 6
Kaolinite
5

Saturation index
0
Pyrite
pH

pH
3 4
Goethite

6 3

9 2
0 5 10 15 20 25 30
Years

Figure 11.23. Kinetic reaction of pyrite, goethite and kaolinite. Initially, the pH increases about twice faster
than without kaolinite, but the pH is lower after 30 years.

When kaolinite is added to the assemblage, the pH increases initially more rapidly to 4 after 16 years.
However, the increase slows down when kaolinite reaches saturation. Kaolinite even becomes super-
saturated because the pH continues to rise as goethite dissolves and it starts to precipitate (Figure
11.23). The precipitation of kaolinite buffers much of the pH increase from the goethite reaction and
the approach to equilibrium of goethite is therefore retarded.
The calculations suggest that the pH will remain below 4.5 during the 15 years flowtime from the
mine to the Elbe River. Thus, the pH buffering by mineral dissolution is inadequate to enhance sorp-
tion of the heavy metals.

QUESTIONS:
Calculate the pH of mine water in equilibrium with goethite, pyrite and kaolinite (WATEQ4F
database)?
ANSWER: 4.6
And if the iron-oxyhydroxide in the aquifer has higher solubility, SIgoethite  2.0?
ANSWER: 5.5
Find the distribution coefficient for U after the reaction with pyrite? Why is it low?
ANSWER: zero, surface complexation of U(4) is undefined in the WATEQ4F database.
Model the kinetic reaction of pyrite and goethite in mine water with PHREEQC over 30 yr. Use 100 mM pyrite
(A / V  0.3 m2/mol/L) and 20 mM goethite (A / V  2000 m2/mol/L).
Add the kinetic reaction of kaolinite, as discussed before. The aquifer contains 5% kaolinite, surface area
10 m2/g.
Add the kinetic reaction of K-feldspar (Table 8.7). The aquifer contains 5% K-feldspar and surface area
0.4 m2/g.
Simulate an experiment with mine water used as influent in a 0.1 m column with pristine sediment in
which pyrite and goethite react kinetically. Flowtime in the column is 1 day. The pristine water has
[PCO2]  102 and is in equilibrium with calcite, pyrite and goethite (SIgoethite  2).

A 2D transport model for Uranium

So far, we have assumed that uranium is present as U(6) (mainly UO2SO4 and UO2
2 ) in the mine
water. However, it can be reduced to UOH3 by H2S in the aquifer:
UO 22  0.25H 2S  0.5H → UOH3  0.25SO 24
(11.53)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

578 Numerical modeling

The reduced form of U precipitates readily in various minerals, of which coffinite (USiO4) is one of
the most insoluble, and commonly observed in the ore at Königstein:

UOH3  H 4SiO 4 → USiO 4  H 2O  3 H (11.54)

Actually, it is probably this reaction scheme that emplaced the ore at Königstein (Tonndorf,
2000). When operating in the (oxic) mine water, it decreases the concentration of uranium by orders of
magnitude from 0.1 mM to less than 0.1 M. However, if the mine water enters directly the oxidized
zone that was created during pumping of the mine, uranium is retarded only by surface complexation. In
this case, only mixing by dispersion with water from the reduced confining layers may start the reduc-
tion reaction and precipitate coffinite. Let’s look at the combination of transverse dispersion and chem-
ical reaction with Alliances, an MT3D/PHREEQC combination developed at Andra, France.
A rectangular cross-section of the aquifer, 50  2000 m above the mine was discretized in
40  100 cells (Figure 11.24). Acid mine water enters the section from the lower left corner and is
accompanied by an equal flow of natural water from the upper left corner. The water leaves the sec-
tion in the middle right. The lower right quarter, from the mine until the outflow cell contains pyrite
and reduced water, the rest of the cross-section is oxidized and contains water with 0.3 mM O2.

Oxidized
zone
50 m

Out

Mine
water
Reduced zone

2000 m

Figure 11.24. Outline of the model aquifer. The reduced zone contains pyrite.

The solute concentrations of U after 20 years indicate the loss towards the reduced zone where coffi-
nite precipitates, while at the front the snowplough effect appears (Figure 11.25). The contact with
pyrite is determined by the transverse dispersivity, here taken to be 1 m. As we note before, the trans-
lation towards the reduced zone enriched the strata with uranium in the geological past, and they will
be fixing uranium again according to the model.

U (mol/L) USiO4 (M)

2.0e04 0.012

1.5e04 0.009

1.0e04 0.006

5.0e05 0.003

0.0e00 0.000

Figure 11.25. Modeled concentrations of U(6) in acid mine drainage entering an aquifer (left) and
precipitation of USiO4 (coffinite, right) at the border of the reduced zone.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 579

11.2.5 In-situ iron removal from groundwater

The Fe2 concentration of pumped groundwater can be lowered when first a volume of oxygenated
water is injected in the aquifer. This process of in-situ iron removal is beneficial since above-ground
treatment can be reduced or even omitted and the concentrations of other trace elements such as
manganese and arsenic are diminished as well. The oxygen oxidizes Fe2 to Fe3, which precipitates
as iron-oxyhydroxide in the aquifer, but the reaction scheme and the contact mode of O2 and Fe2
have been disputed. At present, an intermediate reaction involving the desorption of Fe2 from
exchange and surface sorption sites seems the most plausible (Appelo et al., 1999). The desorbed
Fe2 reacts with O2 according to:

O 2  4Fe 2  10H 2O → 4Fe(OH)3  8H (11.55)

When groundwater is pumped, Fe2 from groundwater sorbs again to the exchange sites that lost
their iron. The iron concentration increases as the exchange sites approach saturation with Fe2 and
when a limiting concentration is reached, pumping may be interrupted to start a new cycle by inject-
ing another volume of oxygenated water. Figure 11.26 shows the iron concentration during the first
and the seventh cycle in Schuwacht, the Netherlands. The arrival of Fe2 occurs later in the 7th than
in the 1st cycle. The delay is related to the transient nature of the process, and to the increase of the
sorption capacity on iron-oxyhydroxide.

8
No reaction PHREEQC Observed
7
1 Run 1
6 1
1
1
1
5 1
Fe (mg/L)

7 Run 7
1
4 1
1 7
3 R1 1
7 7
1 7 7 7
2 7
7
1 7 7
7 7
0 1 1 1
0 2 4 6 8 10 12 14
V/Vi

Figure 11.26. Concentrations of Fe during the 1st and 7th cycle of in-situ iron removal at Schuwacht. The
x-axis gives the ratio of pumped to injected volume (Appelo and De Vet, 2003).

To understand what happens, consider the sequence of oxygen and iron profiles in Figure 11.27.
During injection of 1000 m3 aerated water, groundwater with dissolved iron is displaced into
the aquifer. If sorption sites were absent, Fe2 would simply move along with groundwater, and
the reaction between oxygen and Fe2 would be limited to small amounts at the front where mixing
takes place. However, cations from the injected water exchange for sorbed Fe2 and the oxidation
of this iron consumes oxygen. Thus, the oxygen front lags behind the injected water front.
When the operation is switched to pumping, first the injected volume is withdrawn, with some front
spreading as a result of dispersion. Then, groundwater can be pumped with a reduced iron concen-
tration, because Fe2 is lost to the exchange sites when groundwater flows through the oxidized
zone. After a fixed time, or when the iron concentration exceeds a limit, another volume of aerated
water is injected and the next cycle of in-situ iron removal begins.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

580 Numerical modeling

1
Water with oxygen is O2 Fe2
injected in a reduced 0.8
aquifer.
Fe2 is exchanged from 0.6 Injecting

c/c0
exchange complex and 1000 m3
dissolves in injected water. 0.4
Oxygen lags behind injection Injected water front
front as it reacts with Fe2. 0.2
Oxidized Fe3 precipitates.
0
0 10 20 30 40 50 60
Distance from injection well (m)

0.8 Fe2
Injected water
is withdrawn.
Solute Fe2 is sorbed on 0.6
Pumping
c /c0

exchanger when 1000 m3

groundwater is pumped. 0.4
Injected water
0.2

0
0 10 20 30 40 50 60
Distance from injection well (m)

0.8 Fe2
Efficiency depends on
local conditions:
0.6 Pumping
c /c0

oxidant amount in injected

3000 m3
water and exchange 0.4
capacity of the aquifer.
0.2

0
0 10 20 30 40 50 60
Distance from injection well (m)

Figure 11.27. Oxygen- and iron-concentration profiles at three stages of an in-situ iron removal cycle.

The efficiency of the process can be calculated if we neglect dispersion and limit the sorption reactions to
the zone where oxygen has penetrated, and then consider how much iron can be sorbed in that part of the
aquifer. First, the position of the oxygen front at the end of the injection stage must be located. This posi-
tion can be found from the Reaction (11.55) of oxygen with ferrous iron, in which all Fe2 comes from
sorbed iron, each mole consuming 1/4 mole of dissolved oxygen. Thus, the retardation of oxygen is:
RO  1  qFe / 4 mO (11.56)
2 2

where qFe is the difference in sorbed iron over the front (mol/L pore water) and mO2 the difference
in dissolved oxygen (mol/L). Sorbed iron in Equation (11.56) is in exchange equilibrium with dis-
solved iron in groundwater and is zero when oxygen is present. Likewise, oxygen is zero where
Fe2 is present. These reactions result in a sharp front that migrates through the aquifer.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 581

For the case of linear flow, the distance between the oxygen front and the injection well is:
xO  xinj / RO (11.57)
2 2

where xinj is the distance traveled by the injected water. The fraction of injected water from which
oxygen is consumed is:
f inj  (xinj  xO ) / xinj  1  1 / RO (11.58)
2 2

Note that Equation (11.58) implies that oxygen may not be reacting at all when the retardation equals
1, i.e. when sorbed iron is zero (qFe  0 in Equation 11.56). On the other hand, all oxygen is used
up when sorbed iron is infinite. Thus, the efficiency of in-situ iron removal depends on the sorption
capacity of the aquifer for iron, and it will be low in a coarse, gravelly aquifer.
Now during pumping, if native groundwater with dissolved iron returns and flows along the
emptied sorption sites, Fe2 sorbs again, and iron is retarded with respect to groundwater flow. The
retardation equals (we assume a sharp front, but this depends on the counter-ions):
RFe  1  qFe / mFe (11.59)

Equation (11.59) allows to calculate the volume of groundwater that can be pumped until the sorption
sites are filled and the iron concentration of the native groundwater arrives at the well. The volume of
water between the injection well and xO2 is:
Vinj (1  f inj )  Vinj / RO (11.60)
2

and the volume of groundwater that can be pumped is:

Vgw  (Vinj / RO )  RFe (11.61)
2

Note again, that Equation (11.61) implies that iron arrives immediately at the well when the ratio of
sorbed and dissolved iron is very small, and, of course, that it will never arrive when the concentra-
tion of dissolved iron is zero. Thus, the efficiency of the process also depends on the iron concen-
tration in the groundwater.
We can define the efficiency of in-situ iron removal as:
E  Vgw / Vinj (11.62)

which, according to Equation (11.61), equals:

E  RFe / RO (11.63)
2

For a hand calculation, we can assume that qFe in Equation (11.59) is the same as was used
for calculating the retardation of oxygen (Equation 11.56). If sorption on the precipitated iron-
oxyhydroxide is known, it can be used to increase qFe in Equation (11.59).

Sorption of Fe2 on ferrihydrite

For modeling the process, we need the surface complexation constant of Fe2 to iron-oxyhydroxide.
The constant can be estimated with linear free energy relations (LFER) given by Dzombak and
Morel (cf. Table 7.5). We can also derive a value from experiments reported by Liger et al. (1999),
who published the sorption edge of Fe2 on amorphous iron-hydroxide vs pH (Figure 11.28). For
fitting the sorption edge, we use a general non-linear parameter program (PEST) together with
PHREEQC. PEST is popular among hydrologists for optimizing groundwater model parameters.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

582 Numerical modeling

It fits model values with observations by varying one or more parameters in the input file of the sim-
ulation model. The output file generated by the simulation program is read by PEST and compared
with the observations, and the parameters are optimized using the Marquardt algorithm. A shareware
DOS version of PEST can be downloaded from www.xs4all.nl/appt, together with guidelines for
modeling the sorption edge shown in Figure 11.28.

Fraction of Fe2 sorbed 0.8

Optimized
0.6
LFER est
0.4

0.2

0
5 6 7 8 9
pH

Figure 11.28. Sorption edge of Fe2 on ferrihydrite in 0.1 N NaNO3. Experimental points from Liger et al.,
1999; fitted curves from model with surface complexation constants estimated with linear free energy relations
(LFER) and optimized with PEST and PHREEQC.

Modeling an in-situ experiment

Let us try to model the iron concentrations shown in Figure 11.26. We consider a flowline of 15 m,
inject 10 m3 during 1 day, then pump at the same rate during 3 days and repeat the cycle three times.
The water compositions are given in Table 11.2.

Table 11.2. Water compositions in a model for in-situ iron removal.

Temp, °C pH pe Ca2 Fe2 Alkalinity Cl

Groundwater 10 7.0 0(?!) 3.0 0.1 mmol/L 6.2 mmol/L 0.001

Injected water 10 7.0 PO2  0.2 3.0 6.0 mmol/L 0

The aquifer contains iron-oxyhydroxide (100 ppm Fe, or 10.7 mM Goethite, SI  3.0), the CEC is
60 meq/L. How shall we set up the input file? Here is an outline:

SOLUTION_SPECIES
H2O + 0.01e-  H2O-0.01; log_k -9. # For program convergence
EQUILIBRIUM_PHASES 1-15 # Define sediment
Goethite 3.0 10.7e-3 # Goethite equilibrium, SI, moles
SOLUTION 1-15 # Define groundwater, 15 cells
-temp 10; pH 7.0; pe 0.0 Goethite 3.0
Ca 3; C(4) 6 charge; Fe 0.1; Cl 1e-3 # Cl added as tracer
END

EXCHANGE 1-15; X 0.06; -equil 1

SURFACE 1-15
Hfo_w Goethite 0.2 5.3e4 # Coupled to Goethite, proportion, m2/mol

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 583

Hfo_s Goethite 1e-3 # Coupled to Goethite, proportion

-equil 1
END

PRINT; -reset false

SOLUTION 0 # Inject oxygenated water
-temp 10; pH 7.0; pe 14.0 O2(g) -0.68
Ca 3; C(4) 6 charge
TRANSPORT
-cells 15; -length 1 # Flowtube is 15 x 1  15 m.
-disp 0.1 # Dispersivity, m
-shifts 10 1 # Inject 10 m3
-time 8640 # 1 shift  2.4 hours residence time
-punch_cells 1
END

TRANSPORT # Pump 30 m3
-shifts 30 -1 # No of displacements, direction backward
USER_GRAPH
-heading days Tracer Fe
-axis_scale x_axis 0 3; -axis_scale y_axis 0 1.0; -axis_titles Days mmol/l
-plot_concentration_vs time
-start
10 graph_x (sim_time + 4320)/(3600*24)
20 graph_y tot(“Cl”)*1e6, tot(“Fe(2)”)*5e4
-end
END

TRANSPORT; -shifts 10 1; END # 2nd cycle

TRANSPORT; -shifts 30 -1
END
TRANSPORT; -shifts 10 1; END # 3rd cycle
TRANSPORT; -shifts 30 -1
END

This simple example will produce the basic features of Figure 11.26. The increase of Fe2 is retarded
with respect to Cl (added as a tracer for conservative transport). The retardation increases in the
second cycle because Fe2 has not refilled all the exchange sites after 3 days of pumping ground-
water. In subsequent cycles, the retardation increases because more sorption sites become available
as more and more iron-oxyhydroxide precipitates.
A closer look at the actual data in Figure 11.26 shows that the Fe2 concentrations increase
quicker and more smoothly with V / V0 than the model simulates. This discrepancy may be due to the
heterogeneity of the aquifer (coarse, gravelly parts with a low exchange capacity have smaller retar-
dation than loamy-sandy parts), to the kinetics of the sorption/exchange reaction, or perhaps to the
oxidation-kinetics of Fe2 to Fe3.

QUESTIONS:
What will be the effect of doubling the Ca2 concentration?
ANSWER: Sorbed qFe2 is approximately halved, the retardation of iron is reduced,
cf. Equation (11.59).
What is the effect, having Na instead of Ca2 as major cation?
ANSWER: the Fe2 front during pumping is sharpening, it will steepen.
In-situ iron removal does not function when pyrite is present in the aquifer. Include pyrite in
EQUILIBRIUM_PHASES 1–15 and explain the concentration patterns?

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

584 Numerical modeling

Behavior of Arsenic
It has been observed during in-situ iron removal (Appelo and De Vet, 2003) that the concentration of
As increases temporarily above the pristine groundwater concentration in the initial pumping stage
(Figure 11.29). Can we unravel the cause with PHREEQC?

15
11
1 1

10 1
As (g/L)

77
1
7
1
1 7 7
5 77
1 1
7
7 1 1
1

0 1
0 2 4 6 8 10 12 14
V / Vi

Figure 11.29. Observed concentrations of Arsenic in cycles 1 and 7 during in-situ iron removal (Appelo and
De Vet, 2003).

Add 1 M As to the groundwater (SOLUTION 1–15) and run again while displaying pH, As(3) and As(5).
PHREEQC calculates that the As concentrations in pumped groundwater have decreased, but it is all As(5),
while in groundwater the dominant species is As(3). It seems that As(3) is oxidized to As(5) which sorbs on Hfo.
Let’s uncouple the redox species to have only As(3) in the simulation, assuming that the oxidation of As(3) is
slow and does not take place on the timescale of the experiment.
Make As(5) negligible with …
SOLUTION_SPECIES
H3AsO4 + 2H+ + 2e- = H3AsO3 + H2O; log_k 180. # log_k = 18.897 in database

The As(5) concentration indeed becomes zero, but the As(3) concentration remains steady at 0.81 M in the
first cycle and is 0.79 M in the second cycle. Apparently, the solute concentration is buffered by the surface
complex on ferrihydrite. Appelo and De Vet did not analyze the speciation of As, but Rott et al. (1996) noted
that As(3) in groundwater was almost completely oxidized during in-situ iron removal. Maybe we should model
that As(3) is all oxidized and no longer present.
Make As(3) negligible with …
SOLUTION_SPECIES
H3AsO4 + 2H+ + 2e- = H3AsO3 + H2O; log_k -180.

Now the concentration of As(5) decreases initially, and then increases to 1.17 M. The decrease coincides with
the pH decrease that is associated with the precipitation of Fe(OH)3 at the oxygen front. Inspection of the sur-
face complexes of As(5) shows the dominant presence of Hfo_wOHAsO43. This complex is enhanced when the
surface gains positive charge with the pH decrease (hence, the As concentration in water decreases), and is
released when the pH increases again. Thus, it seems that reactions involving redox and surface complexation
yield the concentration peaks, while the competitive reactions with various other ions may also play a role (cf.
Appelo and De Vet, 2003).
Usually, several knobs are tried and turned to improve the fit of the model with observed concen-
trations. The number of surface sites for ferrihydrite may be increased, or the K values may be varied,
assuming that the natural oxyhydroxides are different from the laboratory synthesized material
(Stollenwerk, 2003). The variety of bonding sites that can be discerned with spectroscopic techniques
is indeed huge (Foster, 2003). On the other hand, the database contains K values that were selected
based on experience in various laboratory and field situations and they provide a good starting point.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 585

QUESTIONS:
Explain the distribution of As(3) and As(5) in the pumping stage of in-situ iron removal, with As(3) and
As(5) present according to the database speciation?
ANSWER: Initially As(5) is dominant, but As(3) turns up when Fe2 appears (cf. the
redox in Figures 9.5 and 9.8).
Why is the Fe2 front retarded when only As(5) is present?
ANSWER: Sorption of the negative As(5) complex decreases the positive charge on
ferrihydrite, which enhances sorption of Fe2.
What is the effect of adding 10 M P to the groundwater?
ANSWER: Sorption of As is reduced, the As concentrations diminish more quickly during
pumping.

11.2.6 Arsenic in Bangladesh groundwater

A relatively large part of the human intake of arsenic stems from drinking water (Table 1.1) and con-
centrations not much above the WHO recommended limit of 10 g/L may already be detrimental for
health. High concentrations of As have been found in many aquifers around the world (Welch et al.,
2000; Smedley and Kinniburgh, 2002), the saddest incidence being Bangladesh where millions of
people are using groundwater with an excessive As content (BGS and DPHE, 2001; Kinniburgh et
al., 2003). In the past, the Bangladesh people relied for drinking water on the great Himalayan rivers
(Ganges, Brahmaputra and Meghna) which were reasonably clean until the population escalated in
number and the natural clean-up proved insufficient for removing the increased viral and bacterial
contamination to a harmless level. Groundwater was promoted as an alternative to surface water by
various international organizations and the local government, and millions of shallow tubewells were
installed. The change in the source of drinking water is believed to have been partly responsible for
the reduction in infant mortality in Bangladesh since 1980 (Kinniburgh et al., 2003). However, it was
not known that the groundwater in the area commonly contains As in concentrations that may far
exceed the safe drinking water limit. It has taken some time before illnesses like skin and bladder
cancer developed and the problem was diagnosed correctly, first in West Bengal (Chakraborty and
Saha, 1987, cited by Smedley, 2003).
Based on the hydrological situation and confirmed by groundwater age determinations with
tritium, the high arsenic concentrations stem from natural sediment-water interactions. This is
rather enigmatic and a cause for intense debate since the solid concentrations lie in the common
range of 1–10 mg/kg for sediments. One theory relates the release to oxidation of pyrite which
may contain arsenic, in response to groundwater drawdown that facilitates the access of oxygen
(Figure 11.30b). However, pyrite is not present everywhere and the sulfate concentrations in
groundwater are low and indicate reduction and possibly pyrite formation, rather than oxidation
(Kinniburgh et al., 2003).
Another theory relates the liberation of As to reductive dissolution of iron-oxyhydroxide that
contains As in sorbed or occluded form (Bhattacharya et al., 1997; Nickson et al., 2000, Figure
11.30a). Oxalate extractions of the sediments, by which ferrihydrite is dissolved, show a clear corre-
lation of As and Fe which makes iron-oxyhydroxide indeed a plausible source (Kinniburgh et al.,
2003). In addition, As and Mg are correlated in the sediments, which suggests that biotite may also
be important. Apparently, the iron-oxyhydroxide forms during weathering of primary minerals, and
it takes up As in the soils in the upland or during erosion and transport with surface runoff. The As
concentrations in river water can at times be relatively high (5–20 g As/L).
One problem of the reductive dissolution theory is how the reductant contacts the
oxyhydroxide. Organic matter is the likely reductant, but iron-oxyhydroxide is expected mostly in

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

586 Numerical modeling

A) Reductive dissolution of Fe(OH)3

Fe(OH)3  As
Pump
In aquifer:
CH2O  H2O → CO2  4H  4e

Peat, CH2O Fe(OH)3As  3H  e → Fe2  3H2O  As

B) Oxidation of pyrite

O2 O2
In unsaturated zone:

FeS2  As FeS2(As)  15 4O2  7 2H2O → 2SO2  4H  Fe(OH)3  As

4

C) Displacement of As by HCO3

Rain
Rain
Pump
A As, mg /L As, mg /L
no CO2cplx with CO2cplx
lcite
B in soil: CO2  ca A A  river water  Hfo: 1 1
B
B B B  A  CO2(g)  Calcite  Hfo: 4 142
C  B  CH2O: 47 189
Peat
C C

Figure 11.30. Mechanisms for explaining the release of As from sediment into groundwater in Bangladesh.
(A) iron-oxyhydroxide is reduced by organic matter and releases sorbed As; (B) groundwater drawdown facili-
tates access of oxygen to pyrite that contains As which dissolves; (C) change in HCO
3 concentration displaces
As from iron-oxyhydroxide.

oxidized sediments where organic matter is sparse or refractory, meaning that organic matter
must be transported with groundwater or by diffusion to the iron-oxyhydroxide. On the other
hand, the As problem seems mostly confined to wells in Holocene aquifers, which may contain
iron-oxyhydroxide and still reactive organic matter that were jointly sedimented in the
river bed.
Let us calculate concentrations according to the reductive dissolution theory and compare these
with correlations of As with HCO 3 and Fe
2
found by Nickson et al., 2000 (Figure 11.31).

Make an input file with average river water, equilibrate with atmospheric oxygen and ferrihydrite, assume that
only 1% of the analyzed goethite has the surface complexation properties of ferrihydrite and that it is 1000
times more soluble than goethite. Add CH2O in steps, first reducing dissolved oxygen, then 1 mM ferrihydrite
in 5 steps (i.e., add 0.05 mM CH2O in these steps).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 587

300 300

As (µg/L)
As (µg/L)

200 200

100 100

0 0
0 10 20 30 40 0 200 400 600 800
Fe (mg/L) HCO3 (mg/L)

Figure 11.31. Correlations of As with Fe and HCO

3 in Bangladesh groundwater (Nickson et al., 2000).

SOLUTION_SPECIES; H2O + 0.01e-  H2O-0.01; -log_k -9 # ... speed up convergence

PHASES; Ferrihydrite; FeOO2H3 + 3H+  Fe+3 + 3H2O; log_k 2.0
SOLUTION 1 Ganges river water
pH 8.39; pe 14 O2(g) -0.68
-units mg/l
Na 19.3; K 4; Mg 14.1; Ca 55.7; Si 16 as Si; P 0.1 as PO4
Cl 9.6; Alkalinity 260 as HCO3; S(6) 4.4 as SO4
As 1e-3 # ...As is usually 10 times higher
EQUILIBRIUM_PHASES 1
# 0.1% Fe in solid  18 mmol/kg  0.1 mol
# FeOOH/L pore water
Ferrihydrite 0 1e-3 # assume only 1% has the properties of ferrihydrite
SAVE solution 1
END

SURFACE 1 # iron hydroxide in riverbed

Hfo_w Ferrihydrite 0.2 5.34e4; Hfo_s Ferrihydrite 0.5e-2; -equil 1
END

EXCHANGE 1 # clay minerals in riverbed

X 0.1; -equil 1
USE solution 1 # for plotting initial conc’s..

USER_GRAPH
-axis_titles "mg HCO3/L", "mg / L", "pH"; -connect; -head Alk As*10 Fe S(6) pH
-start; 10 graph_x Alk * 61e3
20 graph_y tot("As") * 74.92e4, tot("Fe(2)") * 55.85e3, tot("S(6)") * 96e3
30 graph_sy -la("H+"); -end
END

# reduction of O2 and ferrihydrite by organic

# matter in the aquifer...
USE surface 1; USE solution 1; USE exchange 1
EQUILIBRIUM_PHASES 2
Ferrihydrite 0 1e-3; # Siderite 0 0
REACTION 2 # reduce O2, ferrihydrite in 5 steps, then some SO4
CH2O 1e-3; 0.2674 0.05 0.05 0.05 0.05 0.05 0.05 0.05
INCREMENTAL_REACTIONS
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

588 Numerical modeling

9 9
8 pH 8.5
7 8
6 7.5
5 7

mg/L
2

pH
SO4
4 6.5
3 6
2 As  10 5.5
1 Fe 5
0 4.5
260 270 280 290 300


HCO3 (mg/L)

Figure 11.32. Release of As by reductive dissolution of iron-oxyhydroxide, with pH and Fe concentration as

function of the alkalinity.

The result is shown in Figure 11.32. Overall, the As concentration increases as expected but there are less obvi-
ous aspects in the plot. In the first step, O2 is reduced, and the pH decreases. Thereafter ferrihydrite is reduced
and the pH increases again. Remarkably, the Fe concentration does not increase to 55.85 mg/L, expected when
1 mM ferrihydrite is reduced. Of course, part of Fe2 will be taken up by the exchanger, but inspection of the
PHREEQC output file reveals that ferrihydrite is not completely reduced. The remaining ferrihydrite will have
As on its surface, not all available As is released yet into solution. Furthermore, sulfate is being reduced and
also some methane is formed. Apparently, ferrihydrite is too stable to be reduced (cf. Section 9.6).
Decrease the stability of ferrihydrite in the input file by increasing SI to 3.0:

EQUILIBRIUM_PHASES 2
Ferrihydrite 3.0 1e-3 # Siderite 0 0

Indeed (inspect the PHREEQC graph and output), now the As and Fe2 concentration reach a plateau value, all fer-
rihydrite is dissolved in the end. The As concentration becomes as high as 1190 g/L. Only in the last but one step
sulfate starts to be reduced, which is later than before and indicates that it has become more stable relative to ferri-
hydrite. Note also that the concentration of As increases more slowly because the proportion of As(5) is larger than
in the previous case and because As(5) is sorbed more strongly than As(3) under the model conditions.
The concentration of Fe2 is rather high (22 mg/L) and the solution is supersaturated with respect to
siderite, so let’s precipitate that mineral.
Uncomment the mineral Siderite in EQUILIBRIUM_PHASES 2, and run again.
The calculations show (cf. the PHREEQC graphs) that the Fe2 concentration is low, that the alkalinity is
smaller and that the pH increases to more than 9.0. All sorbed As is released into solution because ferrihydrite
is converted completely to siderite. The lower Fe2 concentration diminishes the stability of ferrihydrite, and
practically the same results are obtained with SIferrihydrite  0 (check!).
The release of As is in agreement with the reductive dissolution theory, but the simulated water composi-
tions not altogether resemble the observed groundwaters qualities which have a pH slightly below 7, dissolved
Fe2 of 1–10 mg/L (in approximate equilibrium with siderite), and a much higher alkalinity together with a
much higher CO2 pressure and a much lower CH4 pressure. It can be checked with PHREEQC that the reaction
of organic matter with ferrihydrite cannot reproduce these concentrations. Apparently, organic matter reacts
with O2 to produce large amounts CO2, i.e. the reaction takes place in the soil which, anyhow, is the source of
most groundwater. So, let’s have production of CO2 and dissolution of calcite in the soil before the reduction of
ferrihydrite starts.
Replace the last simulation of the input file with:
# CO2 production and calcite dissolution in the soil...
USE surface 1; USE solution 1; USE exchange 1
EQUILIBRIUM_PHASES 2
Ferrihydrite 0 1e-3; Calcite; O2(g) -0.68

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 589

12 9
8
10 pH
7
8 6
Ferrihydrite reduction
5
mg/L

pH
6
4
As  10
4 3
2
2 CO2 production 1
0 0
200 300 400 500 600 700

HCO3 (mg/L)

Figure 11.33. Release of As from ferrihydrite by displacement with HCO

3 followed by reductive dissolution
of ferrihydrite.

REACTION 2
CH2O 1; 10e-3 in 10
INCREMENTAL_REACTIONS
SAVE solution 2; SAVE surface 2; SAVE exchange 2;
END # water B in Fig 11.30C
# reduction of O2 and FeOOH by organic matter...
USE surface 2; USE solution 2; USE exchange 2
EQUILIBRIUM_PHASES 3
Ferrihydrite 0 1e-3; Calcite; Siderite 0 0
REACTION 3 # reduce O2, ferrihydrite in 5 steps, then SO4...
CH2O 1e-3; 0.2674 0.05 0.05 0.05 0.05 0.05 0.05 0.05
END # water C in Fig 11.30C

Remarkably, the concentration of As now increases to more than 600 g/L when the alkalinity increases by the
reaction of CO2 and calcite (Figure 11.33). The increase is due to displacement by HCO 3 which is sorbed by
ferrihydrite (Appelo et al., 2002). Thus, iron-oxyhydroxide, transported by the river and in equilibrium with the
low As and HCO 3 concentration in surface water, is sedimented in the delta and loses part of its As when con-
tacted by groundwater with a high HCO 3 concentration (Figure 11.30c). Phosphate, which also has a higher
concentration in groundwater than in surface water, has a similar effect. A further increase in As takes place
when ferrihydrite is reduced (Figure 11.33). The Fe2 concentration in groundwater indicates that reduction
occurs, but the As lost by displacement alone is sufficient to explain the observed concentrations. Mixing of
waters from different source areas with different CO2 pressures, in combination with the different retardations
of As, Fe2, and HCO 3 , leads to the observed variations in groundwater compositions.

The HCO
3 displacement theory predicts that As concentrations will be low in river water that infiltrates via
the river bed into the aquifer and keeps its low alkalinity. Thus, bank filtration may be an option in
Bangladesh to obtain safer groundwater with a lower As concentration.

QUESTIONS:
Show with reaction equations that the pH decreases and increases when CH2O reduces O2 and FeOOH,
respectively?
ANSWER: O2  CH2O → H  HCO 3;
4 FeOOH  CH2O  H2O → 4 Fe2  HCO 3  7 OH


Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

590 Numerical modeling

Show with a reaction equation that the pH increases more when siderite precipitates on reduction of
FeOOH with CH2O (for the same Fe2 concentration).
ANSWER: 4⁄3 CH2O  16⁄3FeOOH → 4⁄3 FeCO3  4 Fe2  8 OH
What is the maximal As concentration in the model system, if 2 mM ferrihydrite were present initially?
ANSWER: 1190  2  2380 g/L
Calculate the distributions (1  Kd) and retardations (1  dqi /dci) for As, Fe2 and HCO 3 for the model
of Figure 11.33, at the last CO2 production step (681.5 mg HCO 3 /L) and the fourth reduction step
(680.57 mg HCO 3 /L).
ANSWER:

(1  Kd) R
(HCO
3) As Fe 2
C As Fe2 C
(681.5) 1.75 1 1.01 1.38 1 1.0019
(680.6) 1.35 3.81 1.004 1.45 3.87 1.0009

(Note that the retardation of As is smaller than of Fe, and that RAs and RC are interdependent)
Why is it impossible to have both goethite and ferrihydrite in equilibrium with a solution?
ANSWER: Try it out with PHREEQC

11.2.7 Fractionation of isotopes

Most simply, isotopic fractionation considers the ratio of two isotopes in two species or two phases
at equilibrium. The ratio may grow and change 1) in separation processes where the product is
removed from the reactant (Rayleigh fractionation), 2) with the progress of chemical reactions and 3)
during diffusion. Rapidly, the calculations become complicated when these processes are conjuncted,
but fortunately, PHREEQC can bring relief. It will be shown how to calculate oxygen fractionation
during condensation of water vapor, and carbon fractionation in precipitating calcite. More exam-
ples are given by Appelo (2003).
When dN molecules of light H216O condense, they are accompanied by dNi molecules of heavy
18
H2 O, with the ratio given by

dN i N
 i (2.7)
dN N

We have calculated the isotopic ratio in rain with the Rayleigh formula:

R  N (1)
   (2.10)
R0  N0 

for a constant fractionation factor  (Example 2.1). To calculate the actual process, where vapor cools
and condenses and  changes with temperature, requires numerical integration of Equation (2.7).
The kinetic integrator of PHREEQC can do this in several ways, and one is given in the following
input file. We calculate the condensation of 1 g H216O vapor over the temperature range from 25
to 20°C as a kinetic process (rate “condense”), which provides stepwise values of d(16O). For each
step, the amount d(18O) is calculated in another rate (“Frn_18O”) according to Equation (2.7). We
take care to make the timestep large (“-step 6”) to gain equilibrium among liquid and vapor H216O in
the rate “condense”. The result is shown in Figure 11.34.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 591

Rainout and fractionation of 18O

0 30

Rain
10 20

Temperature (ºC)
Ice

18O (‰) 20 10
Vapor

30 Temperature  0ºC 0

40 10

50 20
0 0.2 0.4 0.6 0.8 1
Fraction of vapor remaining

Figure 11.34. Fractionation of 18O during cooling of water vapor from an initial temperature of 25°C to 20°C
with condensation of rain and subsequently of ice. Initial
18Ovapor  10‰. The line without circles indicates
temperature. Note the jump of 18O in the transition rain/ice at 0°C, where the fractionation factor changes.

# Calculate 18O fractionation during condensation

RATES
Condense
# Condense 1 g water during cooling from 25 to -20 C
-start
10 P_eq  SR(“H2O(g)”) # the vapor pressure
20 n_eq  P_eq * 43.54165 / 0.082057 / TK # n  PV/RT
30 d16O  (m - n_eq) * time
40 save d16O # ... integrate
50 put(d16O, 1) # store d16O in temporary memory
-end

Frn_18O
# d 18O/d 16O  aa_l_v * 18O/16O
-start
10 if TK > 273.15 then c  -2.0667e-3 else c  1.0333e-3
20 aa_l_v  exp(1.137e3/TK^2 - 0.4156/TK + c) # Majoube, 1971
30 d18O  aa_l_v * m / kin(“Condense”) * get(1)
40 put(aa_l_v, 2)
50 save d18O # ...integrate
-end

KINETICS 1
Condense; -formula H2O 0; m0 55.506e-3
Frn_18O; -formula H2O 0; m0 0.110177e-3 # d18O  -10 permil.
-steps 6
INCREMENTAL_REACTIONS
REACTION_TEMPERATURE
25 -20 in 46 # change temp in 1 degree steps
SOLUTION 1

PRINT
-reset false
USER_GRAPH
-headings frac liq/ice vap temp
-axis_titles “Fraction of vapor remaining” “delta 18O” “temp, C”
-axis_scale y_axis -50 0

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

592 Numerical modeling

-start
2 if get(1)  0 then goto 60
10 vap  (kin(“Frn_18O”)/kin(“Condense”)/2.005e-3 - 1)*1e3
# in rain...
20 liq  vap + 1e3*log(get(2))
30 graph_x kin(“Condense”)/55.506e-3 # ...Fraction of vapor remaining
40 graph_y liq, vap
50 graph_sy TC # ... temp, C
60 end
-end
END

One particular aspect of the condensation problem is that actual reactions are never calculated, the
rates add H2O with a stoichiometric coefficient of zero as indicated with “-formula H2O 0” under
keyword KINETICS. The integration could have been done as quickly with MATLAB or any other
integrator, although it is helpful that PHREEQC has a database from which the water content of the
vapor as a function of temperature can be extracted (lines 10 and 20 in “Condense”).

QUESTIONS:
How many moles are in 1 g H2O?
ANSWER: 55.506e-3 (cf. KINETICS 1)
What is the volume of 1 g H2O as saturated vapor at 25°C?
ANSWER: 43.54165 L (cf. line 20 in the rate “Condense”)
Add fractionation of deuterium to the input file, use ln 2Hl/g  52.612  103  76.248 / K  24844 / K2,
K is temperature in Kelvin, initial
2H  80‰.

Fractionation of Carbon during precipitation

The previous example could be done with any integrator, but PHREEQC is particularly useful for
including the effects of a changing hydrogeochemistry in isotope calculations. For example, frac-
tionation of 13C in calcite that precipitates due to degassing of CO2 is a function of pH because 13C
fractionates among the solute carbonate species and into CO2 gas. During degassing, the pH
increases. Again, this is a Rayleigh problem that needs numerical integration. The following input
file calculates the isotopic composition for a system where CO2 and calcite are both removed from
solution (do not react back with the solution).

# Degas CaHCO3 solution, precipitate calcite, find d13C...

SOLUTION 1
Ca 5 charge; C(4) 10
RATES
CO2(g) # d(CO2) / dt  -k * (P_aq - P_gas)
# parms 1  target log(P_CO2)_gas. 2  reaction factor relative to calcite rate
-start
10 P_aq  sr("CO2(g)")
20 dCO2  parm(2) * (10^parm(1) - P_aq) * time
30 save dCO2
40 put(dCO2, 5)
-end

Calcite
# lines 10 to 200 from PHREEQC.DAT database, add line
210 put(moles, 4) # store amount of calcite precipitate
-end

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Examples of hydrogeochemical transport modeling 593

C13
# Distribute 13C among solute species 1  CO2aq, 2  HCO3-, 3  CO3-2
# and in calcite and CO2 gas.
-start # Define alpha’s...
10 aa_1_2  exp-(9.866/TK - 24.12e-3) # Mook et al., 1974
20 aa_3_2  exp(-0.867/TK + 2.52e-3) # Mook et al., 1974
30 aa_cc_2  exp(-4.232/TK + 15.1e-3) # Mook, 1986
40 aa_g_2  exp-(9.552/TK - 24.1e-3) # Mook et al., 1974
# Find H12CO3- and H13CO3-...
50 aH  act(“H+”)
60 K1  aH * act(“HCO3-”) / act(“CO2”) / act(“H2O”)
70 K2  aH * act(“CO3-2”) / act(“HCO3-”)
80 m13C  m / (aa_1_2 * aH/K1 + 1 + aa_3_2 * K2/aH)
90 m12C  (tot(“C(4)”)) / (aH/K1 + 1 + K2/aH)
# and the ratio 13C/12C in HCO3-...
100 R2  m13C / m12C
# Fractionate 13C into calcite...
110 d13C  aa_cc_2 * -get(4) * R2
# and in CO2(g)...
120 d13C_g  aa_g_2 * -get(5) * R2
# Integrate...
130 save d13C + d13C_g
# for printout and graphics...
140 put(R2, 2); 180 put(d13C, 10); 190 put(d13C_g, 20)
-end

C13_cc; -start; 10 save -get(10); -end # ... store 13C in calcite

C13_g; -start; 10 save -get(20); -end # ... store 13C in gas

KINETICS 1
CO2(g); m0 0; -parm -3.5 0.001 # parm(2)  0.1 for equilib
Calcite; m0 0; -parm 60 0.67
C13; -formula C 0; m0 0.1112463e-3 # d13C  -10
C13_cc; -formula C 0; m0 0
C13_g; -formula C 0; m0 0
-step 1e-5 3 5 10 15 20 30 45 60 100 150 300 600 1e3 5e3
INCREMENTAL_REACTIONS
USER_GRAPH
-head cc d13C_cc _HCO3- _sol_tot _gas_tot pH
-axis_titles "mmol Calcite" "d13C" "pH"
-axis_scale x_axis 0 4 1 0.5
-axis_scale y_axis -20 0 5 2.5
-start
2 if step_no  0 then goto 50
10 Rst  0.011237
20 graph_x kin("Calcite")*1e3
30 graph_y (kin("C13_cc")/kin("Calcite")/Rst - 1)*1e3, (get(2)/Rst - 1)*1e3,\
(kin("C13")/tot("C(4)")/Rst - 1)*1e3,\
(kin("C13_g")/kin("CO2(g)")/Rst - 1)*1e3
40 graph_sy -la("H+")
50 end
-end
END

Degassing of CO2 is defined to be a function of the difference in CO2 pressure of the solution and of
the gas phase in the rate “CO2(g)”. It provides stepwise values for the amounts of 12CO2 lost to the
gas phase. Furthermore, the rate for calcite dissolution and precipitation from the PHREEQC database
is copied in the input file to render the amounts of 12C that go into calcite. Like in the condensation

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

594 Numerical modeling

problem, the values for d(12C) are used to fractionate d(13C) in the gas and the solid. First, the
ratio 13C / 12C in HCO3 is calculated by fractionation over the solute species (Equation 5.13). Next,
d(13C) is obtained from the fractionation factors for gas-HCO 
3 and calcite-HCO3 . Figure 11.35 shows
some results for varying relative rates for CO2 lost to the gas-phase and the calcite.

(A) 0 12

2.5 HCO3
11
5
Calcite 10
7.5
δ13C

pH
9
10 Solution pH
8
12.5
15 Gas
7

17.5 6
0 1 2 3 4
Precipitated (mmol) calcite

(B) 0 12

HCO3
2.5 11
5
Calcite 10
7.5
δ13C

Solution

pH
pH 9
10
8
12.5
15 Gas 7
Gas from case A
17.5 6
0 1 2 3 4
Precipitated (mmol) calcite

(C) 0 12

2.5 HCO3
11
5
Calcite 10
7.5
δ13C

pH

9
10
Solution pH 8
12.5
15 7
Gas
17.5 Gas from case A 6
0 1 2 3 4
Precipitated (mmol) calcite

Figure 11.35. 13C fractionation during degassing of 5 mM Ca(HCO3)2 solution. Initial pH  7, solution

13C  10‰. In (A) the rates for degassing and for calcite precipitation approach equilibrium at the same
pace. In (B), the rate for CO2 degassing is 100 fold greater, in (C), the rate for calcite precipitation is 100 fold
greater than in (a) which leads to subtle changes in the isotope compositions. In (B) and (C) the trace of
13C in
gas from (A) is shown with a light-grey line for comparison.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 595

For Figure 11.35A the two rates were set such that overall equilibrium with the final CO2 pressure
(103.5 atm) and with calcite was reached at the same time. In this case, the pH increases rapidly
to about 8 due to loss of CO2, and subsequently more gradually because the precipitation of calcite
keeps pace with the CO2 loss. The
13C of the solution increases because the gas phase has low

13C  15‰. The average
13C of the solution, the gas phase and of calcite all increase during the
approach to final equilibrium when the lighter isotope is fractionated into the gas phase which
increases in mass and is removed from the system. Initially,
13C of the bicarbonate ion lies above
the solution average because the lighter CO2(aq) forms a substantial part of total inorganic carbon
(TIC), but as pH increases to 8.3, HCO 3 becomes the major component of TIC and
C of the solu-
13

tion and of HCO3 coincide. The line for calcite shows the change of
C in the average solid, which
13

increases from 7.5‰ to 5.4‰. However, the actual precipitate follows the line for bicarbonate
and in the successive layers of the solid,
13C ranges from 7.5‰ to 0.7‰.
Subtle changes take place when the rate for CO2 loss is much greater than for calcite precipita-
tion, which is probably the most common condition for field situations (Figure 11.35B). With a 100-
fold increase of the rate “CO2(g)”, equilibrium among solution and the gas-phase is virtually
attained at all stages (the factor entered as parm(2) is increased from 0.001 to 0.1). Now, the pH
increases initially to 8.9, but then decreases as protons are released by the precipitation of calcite.
The greater initial loss of carbon to the gas phase increases
13C in solution and in calcite. Therefore,
a slightly lighter solution and gas are found in the end.
Conversely, if the rate of calcite precipitation is increased 100-fold, the pH decreases initially
(Figure 11.35C). Again,
13C in HCO 3 increases, this time because the fraction of CO2(aq) increases.
As a result, the gas phase becomes heavier, and the final solution and calcite are lighter than in the
two previous cases. Thus, although the final, chemical composition of the solution is identical for all
the paths, the isotope composition differs and may be a tracer for the actual pathway.

PROBLEM:
Include degassing of CO2 as a time dependent process with the two-film model (cf. Chapter 10). Calculate
the fractionation of 2H and 18O during evaporation with the two-film model.

REFERENCES
Alliances, couplage chimie et transport avec MT3D/PHREEQC. Under development at Andra, France. Contact
[email protected]
Appelo, C.A.J., 2003. Calculating the fractionation of isotopes in hydrochemical (transport) processes with
PHREEQC-2. In H.D. Schulz and A. Hadeler (eds), GEOPROC 2002, 383–398, Wiley-VCH, Weinheim.
Appelo, C.A.J. and De Vet, W.W.J.M., 2003. Modeling in situ iron removal from groundwater with trace ele-
ments such as As. In A.H. Welch and K.G. Stollenwerk, op. cit., 381–402.
Appelo, C.A.J., Drijver, B., Hekkenberg, R. and De Jonge, M., 1999. Modeling in situ iron removal from
ground water. Ground Water 37, 811–817.
Appelo, C.A.J., Van der Weiden, M.J.J., Tournassat, C. and Charlet, L., 2002. Surface complexation of ferrous
iron and carbonate on ferrihydrite and the mobilization of arsenic. Env. Sci. Technol. 36, 3096–3103.
Bain, J.G., Mayer, K.U., Blowes, D.W., Frind, E.O., Molson, J.W.H., Kahnt, R. and Jenk, U., 2001. Modelling
the closure-related geochemical evolution of groundwater at a former uranium mine, J. Contam. Hydrol. 52,
109–135.
Barbaro, J.R. and Barker, J.F., 2000. Controlled field study on the use of nitrate and oxygen for bioremediation
of a gasoline source zone. Bioremediation J. 4, 259–270.
BGS, DPHE, 2001. Arsenic contamination of groundwater in Bangladesh. In D.G. Kinniburgh and P.L.
Smedley (eds), British Geological Survey (Technical Report, WC/00/19, 4 Volumes). British Geological
Survey, Keyworth.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

596 Numerical modeling

Bhattacharya, P., Chatterjee, D. and Jacks, G., 1997. Occurrence of arsenic-contaminated groundwater in allu-
vial aquifers from Delta Plains, eastern India. Int. J. Water Resour. Devel. 13, 79–92.
Chapelle, F.H., Bradley, P.M., Lovley, D.R. and Vroblesky, D.A., 1996. Measuring rates of biodegradation in a
contaminated aquifer using field and laboratory methods. Ground Water 34, 691–698.
Cozarelli, I.M., Bekins, B.A., Baedecker, M.J., Aiken, G.R., Eganhouse, R.P. and Tucillo, M.E., 2001.
Progression of natural attenuation processes at a crude-oil spill site: 1. Geochemical evolution of the plume.
J. Contam. Hydrol. 53, 369–385.
Eckert, P. and Appelo, C.A.J., 2002. Hydrogeochemical modeling of enhanced benzene, toluene, ethylbenzene,
xylene (BTEX) remediation with nitrate. Water Resour. Res. 38, DOI 10.1029/2001WR000692
Engesgaard, P., 1991. Geochemical modelling of contaminant transport in groundwater. Ph.D. thesis, Lyngby,
144 pp.
Engesgaard, P., Jensen, K.H., Molson, J., Frind, E.O. and Olsen, H., 1996. Large-scale dispersion in a sandy
aquifer: Simulation of subsurface transport of environmental tritium. Water Resour. Res. 32, 3253–3266.
Foster, A.L., 2003. Spectroscopic investigations of arsenic species in solid phases. In A.H. Welch and K.G.
Stollenwerk, op. cit., 27–66.
Gerald, C.F. and Wheatly, P.O., 1989. Applied numerical analysis, 4th ed. Addison-Wesley, Reading, Mass., 679 pp.
Glynn, P. and Brown, J., 1996. Reactive transport modeling of acidic metal-contaminated groundwater at a site
with sparse spatial information. In P.C. Lichtner, C.I. Steefel and E.H. Oelkers (eds), Rev. Mineral. 34,
377–438.
Herzer, J. and Kinzelbach, W., 1989. Coupling of transport and chemical processes in numerical models.
Geoderma 44, 115–127.
Hutchins, S.R., Downs, W.C., Wilson, J.T., Smith, G.B., Kovacs, D.A., Fine, D.D., Douglass R.H. and Hendrix,
D.J., 1991. Effect of nitrate addition on biorestoration of fuel-contaminated aquifer - field demonstration.
Ground Water 25, 571–580.
Jähne, B., Heinz, G. and Deitrich, W., 1987. Measurement of the diffusion coefficients of sparingly soluble
gases in water. J. Geophys. Res. 92, 10767–10776.
Kinniburgh, D.G., Smedley, P.L., Davies, J., Milne, C.J., Gaus, I., Trafford, J.M., Burden, S., Ihtishamul Huq,
S.M., Ahmad, N. and Ahmed, K.M., 2003. The scale and causes of the groundwater arsenic problem in
Bangladesh. In A.H. Welch and K.G. Stollenwerk, op. cit., 211–257.
Kirkner, D.J. and Reeves, H., 1988. Multicomponent mass transport with homogeneous and heterogeneous
chemical reactions: effect of the chemistry on the choice of numerical algorithm. 1. Theory. Water Resour.
Res. 24, 1719–1729.
Kohler, M., Curtis, G.P., Kent, D.B. and Davis, J.A., 1996. Experimental investigation and modeling of ura-
nium(VI) transport under variable chemical conditions. Water Resour. Res. 32, 3539–3551.
Leuchs, W., 1988. Vorkommen, Abfolge und Auswirkungen anoxischer Redoxreaktionen in einem pleistozänen
Porengrundwasserleiter. Bes. Mitt. dt. Gewässerk. Jb. 52, 106 pp.
Lichtner, P., 1985, Continuum model for simultaneous chemical reactions and mass transport in hydrothermal
systems. Geochim. Cosmochim. Acta 49, 779–800.
Lichtner, P.C. and Seth, M.S., 1996. User’s Manual for MULTIFLO: Part II—MULTIFLO 1.0 and GEM 1.0
Multicomponent-Multiphase Reactive Transport Model. CNWRA 96–010, San Antonio, Texas.
Liger, E., Charlet, L. and Van Cappellen, P., 1999. Surface catalysis of Uranium(VI) reduction by iron(II).
Geochim. Cosmochim. Acta 63, 2939–2955.
Lovley, D.R., Woodward, J.C. and Chapelle, F.H., 1994. Stimulated anoxic biodegradation of aromatic hydro-
carbons using Fe(III) ligands. Nature 370, 128–131.
MacQuarrie, K.T.B., Sudicky, E.A. and Frind, E.O., 1990. Simulation of biodegradable contaminants in ground-
water. 1. Numerical formulation in principal directions. Water Resour. Res. 26, 207–222.
Mayer, K.U., Frind, E.O. and Blowes, D.W., 2002. Multicomponent reactive transport modeling in variably sat-
urated porous media using a generalized formulation for kinetically controlled reactions. Water Resour. Res.
38, 1174, DOI 10.1029/2001WR000862.
Miller, C.W. and Benson, L.V., 1983. Simulation of solute transport in a chemically reactive heterogeneous sys-
tem: Model development and application. Water Resour. Res. 19, 381–391.
Nickson, R.T., McArthur, J.M., Ravenscroft, P., Burgess, W.G. and Ahmed, K.M., 2000. Mechanism of arsenic
release to groundwater, Bangladesh and West Bengal. Appl. Geochem. 15, 403–413.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 References 597

Notodarmojo, S., Ho, G.E., Scott, W.D. and Davis, G.B., 1991. Modeling phosphorous transport in soils and
groundwater with two-consecutive reactions. Water Res. 25, 1205–1216.
Parkhurst, D.L., Kipp, K.L., Engesgaard, P. and Charlton, S.R., 2004. PHAST—A program for simulating
ground-water flow and multicomponent geochemical reactions.
http://wwwbrr.cr.usgs.gov/projects/GWC_coupled/phreeqc/index.html
Pickens, J.F. and Lennox, W.C., 1976. Numerical simulation of waste movement in steady groundwater flow
systems. Water Resour. Res. 12, 171–180.
Press, W.H., Flannery, B.P., Teukolsky, S.A. and Vettering, W.T., 1992. Numerical recipes in Pascal. Cambridge
Univ. Press, 759 pp.
Prommer, H., Barry, D.A. and Davis, G.B., 2000. Numerical modelling for design and evaluation of groundwa-
ter remediation schemes. Ecol. Model. 128, 181–195.
Prommer, H., 2002. A reactive multicomponent transport model for saturated porous media. www.pht3d.org
Rott, U., Meyerhoff, R. and Bauer, T., 1996. In situ treatment of groundwater with increased concentrations of
iron, manganese and arsenic (in German). Wasser-Abwasser 137, 358–363.
Saalting, M.W., Ayiora, C. and Carrera, J., 1998. A mathematical formulation of reactive transport that elimi-
nates mineral concentrations. Water Resour. Res. 34, 1649–1656.
Schlosser, P., Stute, M., Dörr, H., Sonntag, C. and Münnich, K.O., 1988. Tritium/3He dating in shallow ground-
water. Earth Planet. Sci. Lett. 89, 353–362.
Schlosser, P., Stute, M., Sonntag, C. and Münnich, K.O., 1989. Tritiogenic 3He in shallow groundwater. Earth
Planet. Sci. Lett. 94, 245–256.
Schmitt, R., Langguth, H.R., Püttmann, W., Rohns, H.P., Eckert P. and Schubert, J., 1996. Biodegradation of
aromatic hydrocarbons under anoxic conditions in a shallow sand and gravel aquifer of the Lower Rhine
Valley, Germany. Org. Geochem. 25, 41–50.
Shapiro, A.M., 2001. Effective matrix diffusion in kilometer-scale transport in fractured crystalline rock. Water
Resour. Res. 37, 507–522.
Shapiro, S.D., Rowe, G., Schlosser, P., Ludin, A. and Stute, M., 1998. Tritium-helium 3 dating under complex
conditions in hydraulically stressed areas of a buried-valley aquifer. Water Resour. Res. 34, 1165–1180.
Sheets, R.A., Bair, E.S. and Rowe, G.L., 1998. Use of H-3/He-3 ages to evaluate and improve groundwater flow
models in a complex buried-valley aquifer. Water Resour. Res. 34, 1077–1089.
Smedley, P.L., 2003. Arsenic in groundwater - South and East Asia. In A.H. Welch and K.G. Stollenwerk, op.
cit., 179–209.
Smedley, P.L. and Kinniburgh, D.G., 2002. A review of the source, behaviour and distribution of arsenic in nat-
ural waters. Appl. Geochem. 17, 517–568
Steefel, C.I. and MacQuarrie, K.T.B., 1996. Approaches to modeling reactive transport in porous media. In P.C.
Lichtner, C.I. Steefel and E.H. Oelkers (eds), Rev. Mineral. 34, 83–125.
Steefel, C.I. and Yabusaki S.B., 1996. OS3D/GIMRT, Software for multicomponent-multidimensional reactive
transport: User’s Manual and Programmer’s Guide. PNL-11166, Richland.
Stollenwerk, K.G., 2003. Geochemical processes controlling transport of arsenic in groundwater: a review of
adsorption. In A.H. Welch and K.G. Stollenwerk, op. cit., 67–100.
Sverdrup, H.U. and Warfvinge, P., 1995. Estimating field weathering rates using laboratory kinetics. In A.F.
White and S.L. Brantley (eds), Rev. Mineral. 31, 485–541.
Szabo, Z., Rice, D.E., Plummer, L.N., Busenberg, E. and Drenkard, S., 1996. Age dating of shallow groundwa-
ter with chlorofluorocarbons, tritium helium 3, and flow path analysis, southern New Jersey coastal plain.
Water Resour. Res. 32, 1023–1038.
Tonndorf, H., 2000. Die Uranlagerstätte Königstein. Sächsisches Landesamt für Umwelt und Geologie,
Dresden, 208 pp.
Van Ommen, H.C., 1985. The mixing cell concept applied to transport of non-reactive and reactive components
in soils and groundwater. J. Hydrol. 78, 201–213.
Walter, A.L., Frind, E.O., Blowes, D.W., Ptacek, C.J. and Molson, J.W., 1994. Modelling of multicomponent
reactive transport in groundwater. 2. Metal mobility in aquifers impacted by acidic mine tailings discharge.
Water Resour. Res. 30, 3149–3158.
Weissmann, G.S., Zhang, Y., LaBolle, E.M. and Fogg, G.E., 2002. Dispersion of groundwater age in an alluvial
aquifer system. Water Resour. Res. 38 (10), Art 1198, 14 pp.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

598 Numerical modeling

Welch, A.H. and Stollenwerk, K.G. (eds) 2003. Arsenic in groundwater, geochemistry and occurrence. Kluwer,
Boston, 475 pp.
Welch, A.H., Westjohn, D.B., Helsel, D.R. and Wanty, R.B., 2000. Arsenic in ground water of the United States:
occurrence and geochemistry. Ground Water 38, 589–604.
Wisotzky, F. and Eckert, P., 1997. Sulfat-dominierter BTEX-Abbau im Grundwasser eines ehemaliges
Gaswerksstandortes. Grundwasser 2, 11–20.
Yeh, G.T. and Tripathi, V.S., 1989. A critical evaluation of recent developments in hydrogeochemical transport
models of reactive multichemical components. Water Resour. Res. 25, 93–108.
Yeh, G.T., Sun, J., Jardine, P.M., Burgos, W.D., Fang, Y., Li, M.-H. and Siegel, M.D., 2004. HYDROGEOCHEM
5.0: A Three-Dimensional Model of Coupled Fluid Flow, Thermal Transport, and hydrogeochemical
Transport through Variably Saturated Conditions. ORNL/TM-2004/107.
Zheng, C. and Bennett, G.D., 2002. Applied contaminant transport modeling, 2nd ed. Wiley and Sons, New
York, 621 pp.
Zoellmann, K., Kinzelbach, W. and Fulda, C., 2001. Environmental tracer transport (H-3 and SF6) in the satu-
rated and unsaturated zones and its use in nitrate pollution management. J. Hydrol. 240, 187–205.
Zysset, A., Stauffer, F. and Dracos, T., 1994. Modeling of chemically reactive groundwater transport. Water
Resour. Res. 30, 2217–2228.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

APPENDIX A

Hydrogeochemical Modeling with PHREEQC

PHREEQC is used for simulating a variety of reactions and processes in natural waters or laboratory experi-
ments. PHREEQC needs an input file in which the problem is specified via KEYWORDS and associated data-
blocks. The keywords are summarized here in get-going sheets. A full description of many alternatives for input
and the mathematical backgrounds can be found in the manual of the program by Parkhurst and Appelo (1999).
PHREEQC was developed for calculating “real world” hydrogeochemistry and it does help to keep the con-
nection with the physical situation palpable and concrete. Thus, we may start imagining a laboratory table with
flasks with solutions, bottles with chemicals, minerals and exchangers, gascontainers, pipettes, a balance and a
centrifuge with accessories, where we want to do some experiments. We use the following KEYWORDS:

SOLUTION m–n for the composition and quantity of solutions in flasks m to n. We can USE a
specific solution, or MIX fractions of solutions in one flask.

Various reactants can be added to the flask:

EQUILIBRIUM_PHASES for a combination of minerals and/or gases which react reversibly to a
prescribed equilibrium;
EXCHANGE for the capacity and composition of an exchanger;
SURFACE for the capacity and composition of surface complexers;
REACTION for stepwise adding or removing chemicals, minerals or water;
KINETICS for chemicals which react depending on time and composition of the solution;
GAS_PHASE for a combination of gases in a specified volume or at a given pressure;
SOLID_SOLUTIONS for adding solid solutions of minerals;
REACTION_TEMPERATURE for changing the temperature of the flask.

END is the signal for PHREEQC to calculate the composition of the solution and the reactants in what is termed
a simulation. The compositions can be stored in computer memory with SAVE solution no, SAVE exchange no,
etc., to be used later on in the same computer run (the same input file) with MIX or USE solution no, USE
exchange no, etc., where no is a number or a range of numbers. The results can also be printed, punched in a
spreadsheet file, or plotted in a graph.

For defining the output we use:

PRINT to limit printout to specific items and to suspend/resume print options;
SELECTED_OUTPUT to obtain results in spreadsheet type format;
USER_PRINT, USER_PUNCH and USER_GRAPH for defining tailor-made, specific output.
Flow and transport in the field and in laboratory columns can be modeled with keyword TRANSPORT for 1D
dispersive/diffusive transport including mobile/immobile zones. For 3D transport, the code PHAST can be used
from the US Geological Survey website.

Lastly, the chemical reactions which led to a given water quality can be recovered using keyword
INVERSE_MODELING.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

600 Appendix A: Hydrogeochemical Modeling with PHREEQC

PHREEQC Get-going sheet #

Getting Started 1

SOLUTION 2

Mineral Equilibration 3

PHREEQC Output 4

REACTION 5

MIX 6

EXCHANGE 7

Surface Complexation 8

Convergence; Titrating 9

TRANSPORT 10

KINETICS 11

SOLID_SOLUTION 12

GAS_PHASE 13

Databases 14

INVERSE_MODELING 15

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 1 601

Getting Started
The program PHREEQC, the manual and various example files can be downloaded via links in:
www.xs4all.nl/appt. Download Vincent Post’s windows version and run psetup.exe to install the program
and databases in c:\Program Files\phreeqc. Click on the PHREEQC icon to start. You’ll be located in the input
window.
Various options can be set, click Edit, Preferences (or ALTe, e) and make your choices. The directories
and the database file can be selected, click Calculations, Files (ALTc, f).
To the right of the input window is a frame with PHREEQC keywords and BASIC statements. Click
on  to expand the keywords, click again on  to expand keyword SOLUTION to see the identifiers (sub-
keywords) temp, pH, pe, and more. Double clicking on a keyword or identifier will copy it to the input window.
Example files are in c:\Program Files\PHREEQC\Examples. Open U_GEX1, click Files, Open
(ALTf, o), U_GEX1. The file appears in the input window:

# Graphs Ca vs F concentration in equilibrium with fluorite

#
SOLUTION 1 # distilled water
END
EQUILIBRIUM_PHASES 1 # equilibrium with Fluorite, CaF2
Fluorite
REACTION # Change Ca concentration for plot
Ca(OH)2 1
0.05 0.1 0.2 0.5 1 2 3 5 7 9 12 14 mmol
USE solution 1
USER_GRAPH # various chart definitions are omitted here
10 graph_x tot("Ca")*1e3
20 graph_y tot("F")*1e3
END

• Run the file, click the calculator icon, or Calculations, Start (ALT c, s).
• Quickly, the progress window reports “Done”. Press Enter, and you’ll be transferred to the output tab. It con-
tains a complete listing and many details of the calculated solution compositions.
• Click the Database tab. The basic data for calculating a speciation model appear, listed under keywords. The
first one is keyword SOLUTION_MASTER_SPECIES which defines the elements and, among other things,
their formula weights.
• Click the grid tab. It contains rows with the Ca and the F concentrations in equilibrium with fluorite as writ-
ten by USER_GRAPH.
• Click the Chart tab. It shows a plot of the F vs Ca concentration in equilibrium with fluorite.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

602 PHREEQC Get-going sheet 2

SOLUTION
The aqueous concentrations in a solution are defined with SOLUTION. The symbols for the elements are listed
in the first column of SOLUTION_MASTER_SPECIES in the database.

# Information following a # is not read

SOLUTION 1 Speciate an analysis, calculate saturation indices

temp 25. # temperature in degrees Celcius, default = 25°C
pH 7.0 # default pH = 7.
pe 4.0 # pe = -log(electron activity), default = 4.
-units mmol/kgw # default units mmol/kg water
Ca 3.0 # total dissolved Ca
Na 1.0
Alkalinity 3.8 # mmol charge / kgw
S 1.0 # total sulfur, mainly sulfate, S(6), at pe = 4
N(5) 0.2 # nitrogen in the form of nitrate, N(5)
Cl 1.0
water 1 # kg water, default = 1 kg
density 1 # density, default = 1 kg/L
END

# Example 2. define range of solution no's, and use mg/L units

#
SOLUTION 2-4 # solutions 2, 3 and 4
-units mg/L
Ca 40 # mol. weights listed under SOLUTION_MASTER_SPECIES
Alkalinity 122 as HCO3 # HCO3 is used for recalculating to mol charge
END

# Example 3. Adjust concentration or pH to charge balance

#
SOLUTION 4 # default pH = 7, temp = 25, -units mmol/kgw
Na 2.0
Cl 1.3 charge # adapt Cl to obtain charge balance
END

# Example 4. Adjust pe to equilibrium with atmospheric oxygen

#
SOLUTION 6 # see EQUILIBRIUM_PHASES (sheet # 3) for more info
pe 10.0 O2(g)-0.68 # add PHASES name, SI = -0.68. P_gas = 10^-0.68
END

# Example 5. Enter a suite of samples in spreadsheet format

# The data are tab delimited
#
SOLUTION_SPREAD
-units mg/L
-temp 11
Ca Cl S(6) Temp Alkalinity pH Li
mg/L as HCO3 ug/L
87.7 22.7 94.4 13 233.4 7.3
78.6 35.3 87.4 13.1 221.7 7.1 4
75.0 24.9 83.2 221.3 6.9

Note
The default temperature of 11°C will be given to the third sample.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 3 603

Mineral Equilibration
A solution can be equilibrated with minerals assembled together in EQUILIBRIUM_PHASES, or the initial
concentration of a component can be adapted when defining the SOLUTION.

# Example 1. Equilibrate water with calcite and CO2

#
SOLUTION 1 # default pH = 7, pe = 4, temp = 25, units = mmol/kgw
Ca 3.0
Alkalinity 4.0
S(6) 1.0
EQUILIBRIUM_PHASES 1
Calcite 0.0 10.0 # Mineral from database, SI = 0.0, amount = 10 mol
CO2(g) -2.0 # SI = - 2.0 or [P_CO2] = 10^-2.0, amount = 10 mol
END

# Example 2. Equilibrate with Fluorite, adding gypsum.

#
SOLUTION 2
F 1.5 mg/kgw # drinking water limit for F = 1.5 mg/L
EQUILIBRIUM_PHASES 2
Fluorite 0.0 Gypsum # obtain SI = 0. by adding Ca from gypsum
END # note the high SO4 concentration in the output...

# Example 3. Equilibrate by changing an initial concentration

#
SOLUTION 3
Ca 3.0 Calcite # adapt Ca to equilibrium with calcite
Alkalinity 4.0
S (6) 1.0
F 0.1 Fluorite -1.0 # adapt F to SI = -1.0 for fluorite
END

Notes
The mineral names must be spelled exactly as entered in the database under PHASES (you can copy the name
after selection with CTRLc, then paste with CTRLv).
Gas equilibria. The gas pressure can be varied, setting SI. The SI is equal to log(gas pressure/1 atm).
Example 4 illustrates the different effects of equilibrating by adjusting the initial concentration, and by the
reversible dissolution reaction that is simulated with keyword EQUILIBRIUM_PHASES.

# Example 4. Adjust the CO2 gas pressure to 10^-22.5 = 0.00316 atm

#
SOLUTION 1
pH 7.0
C(4) 1.0 CO2(g) -2.5 # Only TIC is adapted, pH remains 7.0
END

SOLUTION 2
pH 7.0
EQUILIBRIUM_PHASES 2 # pH changes when carbonic acid dissolves
CO2(g) -2.5 10.0
END

Solution 1 obtains a large electrical imbalance, while solution 2 remains electrically neutral.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

604 PHREEQC Get-going sheet 4

PHREEQC Output
Using PRINT, you can modify the print to the output tab by setting output block identifiers true or false. These
identifiers can be clicked and copied to the input tab after expanding keyword PRINT in the right window. The
option -reset false is rather essential when doing transport simulations.

PRINT
-reset false # suspend all print output
-totals true # will print Solution composition, Description of solution
-status true # shows calculation progress

To write specific output in spreadsheet format (“punch” to a file), use SELECTED_OUTPUT. The selected
information is written to the file after each simulation.

SELECTED_OUTPUT
-file stine1.csv # name of output file. OBLIGATORY
-totals Fe Fe(2) Hfo_w # total conc. in mol/kgw, MASTER_SPECIES element name
-molalities Fe+2 CaX2 # molalities of species
-activities Fe+2 H+ # log activities of species
-eq Calcite # equilibrium_phases, note abbreviation -eq

Notes
Element names given under -totals should comply with the MASTER_SPECIES element names (first column
in the databases).
Species names given under -mol and -act and mineral names must match the database names.
In transport calculations, -punch_cells and -punch_frequency can be used to further limit the punchout to
the selected_output file.

USER_PUNCH, USER_PRINT and USER_GRAPH

The USER_ keywords give access to the BASIC compiler which is built in PHREEQC. You can punch out your
own variables to the selected_output file with USER_PUNCH, print to the output file with USER_PRINT, and
plot in PHREEQC for Windows with USER_GRAPH.

USER_PUNCH
-headings mg_Na/l P_CO2 # column heading in file
-start # start of BASIC lines
10 punch tot("Na")*22.999e3
20 punch 10^(SI("CO2(g)"))
-end # end of BASIC lines
END

USER_GRAPH # plot F and pH vs Ca concentration

-start
10 graph_x tot("Ca")*1e3 # mmol/L
20 graph_y tot("F")*1e3
30 graph_sy -la("H+") # pH on secondary y-axis
-end
END

The BASIC lines can be seen as spreadsheet formulas, with special functions which give access to the aqueous
model. These functions are explained under keyword RATES in the PHREEQC manual, and listed in the
PHREEQC BASIC statements at the right of the input window.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 5 605

REACTION
The keyword REACTION is used to stepwise add reactants or amounts of minerals to the solution.

# Example 1. Add 5 and 10 mmol NaCl to 1 liter distilled water.

#
SOLUTION 1 # Define start solution, just water
REACTION 1
NaCl 1.1 # Reactant formula, stoichiometric coefficient in the reaction
5e-3 10e-3 # Add 1.1*5e-3 mol NaCl and 1.1*10e-3 mol NaCl
# 10e-3 moles in 2 steps # alternate way of writing the steps
END

# Example 2. Add Calcite and CO2 to water

#
SOLUTION 2
REACTION 2
Calcite 1.0 CO2(g) 1.3 # mineral from database, stoichiometric coefficient
2e-3 # Add 1*2e-3 = 2e-3 mol Calcite and 1.3*2e-3 = 2.6e-3 mol CO2
END

# Example 3. Evaporate 99% of a HCl solution in 5 steps

#
SOLUTION 3
pH 3.0 charge
Cl 1.0
water 1.0 # 1 kg water = 55.51 mol
REACTION 3
H2O -1.0
54.951 moles in 5 steps
END

Notes
In example 1, the program recourses to the initial solution when adding the reaction steps. With keyword
INCREMENTAL_REACTIONS true, the reaction steps are added cumulatively, thus giving 5  103 and
15  103 mol NaCl.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

606 PHREEQC Get-going sheet 6

MIX
The effects of mixing of two or more groundwaters on the speciation and saturation state with respect to
minerals can be calculated with keyword MIX. This keyword is also useful for simulating laboratory batch
experiments.

# Example 1. Calculate mixing corrosion in limestone water

#
SOLUTION 1 Calcite saturated, high CO2
pH 7.1 charge
Ca 2.0 Calcite
C(4) 4.0 CO2(g) -1.5
SOLUTION 2 Calcite saturated, low CO2
pH 7.1 charge
Ca 2.0 Calcite
C(4) 4.0 CO2(g) -2.5
END
MIX 1
1 0.4 # use fraction 0.4 of solution 1
2 0.6 # use fraction 0.6 of solution 2
EQUILIBRIUM_PHASES 1
Calcite # find how much calcite dissolves in the mixture
END

# Example 2. Evaporate 90% of 1 liter NaCl water, restore to 1 liter

#
SOLUTION 2
Na 10
Cl 10
REACTION 2
H2O -1.0
49.955 # Evaporate 49.955 moles (= 0.9 * 55.506)
SAVE solution 3 # leaves 5.5506 mol H2O = 0.1 liter
END
MIX 3
3 10 # Take fraction 10 of solution 3,
# is 10*5.5506 = 55.506 mol = 1 liter
END

# Example 3. Displace exchangeable cations with 1 M NaCl

#
SOLUTION 0 1M NaCl
Na 1e3
Cl 1e3
END
EXCHANGE 1 Fill centrifuge tube with 1 g clay = 1 mmol CEC
KX 0.02e-3; MgX2 0.09e-3; CaX2 0.4e-3
MIX 1
0 0.02 # Add 20 ml of 1M NaCl solution
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 7 607

EXCHANGE
The keyword EXCHANGE is used to define exchange properties and to calculate the composition of the
exchanger.

# Example 1. Calculate composition of an exchanger

#
SOLUTION 1
Na 1.0
K 0.2
Cl 1.2
EXCHANGE 1
X 30e-3 # 30 mmol X-
-equil 1 # equilibrate with solution 1, --> 15 mmol NaX/L,
# 15 mmol KX/L
END

# Example 2. Input measured exchanger

#
EXCHANGE 1
CaX2 20e-3 # 20 mmol exchangeable Ca
NaX 10e-3 # 10 mmol exchangeable Na

# Example 3. Exchanger composition is calculated after reactions

#
SOLUTION 1-5
Na 2.0
C(4) 2.0 charge
EQUILIBRIUM_PHASES 1-5 # mineral equilibria for cells 1-5
Calcite 0.0 0.01 # 0.01 mol Calcite, SI = 0.0
SAVE solution 1-5 # save equilibrated solutions
END

EXCHANGE 1-5 # exchangers for cells 1-5

X 0.1 # 0.1 mol X
-equil 1
END

Notes
EXCHANGE defines moles of X, but the concentration depends on the amount of solution (default is 1 kg
water). In example 1, the concentration is 30 mmol X/kg water.

In example 3, the exchanger is calculated with a mineral equilibrated solution, after SAVE and END. The cal-
culation sequence is important here, the results are different when EXCHANGE 1–5 is included before the first
END. Any PHREEQC simulation starts with equilibration of the exchanger (if defined). Next, the reactions are
calculated together with the exchanger. Here, the initial solution contains only Na, and the exchanger would
be pure NaX. If the solution is then equilibrated with calcite, Ca2 displaces Na from the exchanger, and the
Na concentration in solution 1–5 will become higher than 2 mM.
The moles of an exchanger may be coupled to minerals (eq) or kinetic reactants (kine).

EXCHANGE 1-5
X Calcite eq 0.05 # X is coupled to calcite (an equilibrium phase)
# in proportion 0.05 mol X / mol calcite

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

608 PHREEQC Get-going sheet 8

Surface Complexation
PHREEQC contains the Dzombak and Morel database for surface complexation of heavy metal ions on hydrous
ferric oxide (ferrihydrite, hfo). Ferrihydrite binds the metal ions and protons on strong and weak sites and devel-
ops a charge depending on the ions sorbed. The keyword SURFACE is used to define the surface complexation
properties of sediments in much the same way as EXCHANGE, but since the surface may become charged,
some precautions are in order.

# Example 1. Calculate the composition of a surface complex

#
SOLUTION 1
Na 1.; Cl 1.; Zn 1e-3 # semicolon separates lines
SURFACE 1
Hfo_w 2.4e-3 600 1.06 # 2.4e-3 mol weak sites, sp. surface = 600 m2/g
# gram ferrihydrite = 1.06
Hfo_s 6e-5 # strong site, 6e-5 mol sites
-equil 1 # equilibrate with solution 1
END

# Example 2. Couple a surface with a mineral (eq) or kinetic reactant (kine)

#
SOLUTION 2 # adjust pe to prevent (too much) Hfo dissolution
pe 14 O2(g) -0.68 # pe is for P_O2 = 10^-0.68 = 0.21 atm
Na 1.; Cl 1.; Zn 1e-3
EQUILIBRIUM_PHASES 2
Fe(OH)3(a) 0.0 1e-3
SAVE solution 2 # Save solution 2 after equilibration
END
SURFACE 2
Hfo_w Fe(OH)3(a) eq 0.2 5.34e4 # weak sites, coupled to ferrihydrite,
# proportionality constant = 0.2,
# sp. surface = 5.34e4 m2/mol
Hfo_s Fe(OH)3(a) eq 0.005 # strong site, coupled to ferrihydrite,
# proportionality constant = 0.005
-equil 3
END

Notes
The default properties of Hfo are: 89 g/mol, 600 m2/g, 0.2 mol weak sites/mol, 0.005 mol strong sites/mol.
Make sure to have reactions calculated for a solution, to save it and to END the simulation, before equili-
brating a surface as is done in example 2 (see also discussion in get-going sheet EXCHANGE).
When Hfo is coupled to a mineral, the specific surface area is given in m2/mol.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 9 609

Two Tips
● In rare cases, mainly with redox problems with very low concentrations of the metal ions, PHREEQC has
difficulties finding a solution. The first trick is to include a redox buffer in the input file:

SOLUTION_SPECIES
H2O + 0.01e- = H2O-0.01; log_k -9.0

● Titration to a fixed pH is illustrated in the following example.

# Determine Cd sorption edge on Hfo

PHASES
Fix_ph # Define pH fixing mineral
H+ = H+; log_k 0 # pH = 10^(-SI)

SOLUTION 1
pH 4.0; Na 100; Cl 100 charge; Cd 1e-3

SURFACE 1
Hfo_w 0.2e-3 600 89e-3
Hfo_s 0.05e-3
-equil 1
END

USER_GRAPH
-head pH %_cd_sor titrans
-start
10 sor = mol("Hfo_sOCd+") + mol ("Hfo_wOCd+") # 2 Cd surface complexes
20 graph_x -la("H+") # pH on x-axis
30 graph_y 100 * sor/(sor + tot("Cd")) # % sorbed Cd on y-axis
# 40 graph_sy 1e6 * tot("Br") # (to plot mol titrans on secondary y axis)
-end
END

USE solution 1; USE surface 1

EQUILIBRIUM_PHASES 1
Fix_ph -4.01 Na(OH)0.999999Br0.000001 # NaOH for increasing pH, with tracer Br
END
USE solution 1; USE surface 1
EQUILIBRIUM_PHASES 1; Fix_ph -4.5 Na(OH)0.999999Br0.000001
END
# Etc. for pH up to 8...

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

610 PHREEQC Get-going sheet 10

TRANSPORT
Keyword TRANSPORT is used for modeling 1D flowlines and laboratory columns. The flowline is divided in
a number of cells 1–n for which SOLUTION 1–n must be defined. With advective flow, a shift moves water
from each cell into the next, higher- or lower-numbered cell, depending on the flow direction.
# u_gex3 -- Transport and ion exchange: CaCl2 flushes column with (Na,K)NO3
#
SOLUTION 1-16 Initial solution in column
Na 1.0; K 0.2; N(5) 1.2 # default pH, etc, note line separator ;
EXCHANGE 1-16
X 1.1e-3; -equilibrate 1
END
SOLUTION 0 Displacing solution, CaCl2
Ca 0.6; Cl 1.2
END
PRINT; -reset false # suspend output
TRANSPORT
-cells 16
-length 0.005 # cell length 0.005 m, column length 16*0.005 = 0.08 m
-dispersivity 0.002 # m
-shifts 40 # 40/16 = 2.5 pore volumes
-flow_direction forward
-time_step 720 # 1 shift = 720 s. total time = 40*720 s
-boundary_conditions flux flux # at column ends
-diffusion_coefficient 0.3e-9 # m2/s
-punch 16 # only punch/graph cell 16
USER_GRAPH
-heading PV NO3 Cl Na K Ca; -axis_titles Pore_Volumes mol/L
-axis_scale x_axis 0 2.5; -axis_scale y_axis 0 1.5e-3
-plot_concentration_vs time
-start
10 graph_x (step_no + 0.5)/16 # transfer cell centered conc to column end
20 graph_y tot("N"), tot("Cl"), tot("Na"), tot("K"), tot("Ca")
-end
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 11 611

KINETICS
For kinetic calculations a rate equation must be programmed in BASIC using keyword RATES. The rate is
called with keyword KINETICS which can also pass parameters to the rate. The BASIC functions are explained
under keyword RATES in the PHREEQC manual.
# Kinetic quartz dissolution
#
RATES
Quartz # rate name
-start
#1 rem dQu/dt = -k * (1 - Quartz). k = 10^-13.7 mol/m2/s (25 C)
#2 rem parm(1) = A (m2), parm(2) = V (dm3) recalculate to mol/dm3/s
10 moles = parm(1) / parm(2) * (m/m0)^0.67 * 10^-13.7 * (1 - SR("Quartz"))
20 save moles * time # integrate. save and time must be in rate definition
# moles count positive when added to solution
-end
KINETICS # Sediment: 100% qu, grain size 0.1 mm, por 0.3, rho_qu 2.65 kg/dm3
Quartz # rate name
-formula SiO2
-m0 102.7 # initial moles of quartz
-parms 22.7 0.162 # parameters for rate eqn. Here:
# Quartz surface area (m2/kg sediment), water filled porosity (dm3/kg sediment)
-step 1.58e8 in 10 steps # 1.58e8 seconds = 5 years
-tol 1e-8 # integration tolerance, default 1e-8 mol
INCREMENTAL_REACTIONS true # start integration from previous step
SOLUTION 1
USER_GRAPH
-heading time Si; -axis_titles years mmol/L
-axis_scale y_axis 0 0.12 0.02; -axis_scale x_axis 0 5
-start
10 graph_x total_time/3.1536e7
20 graph_y tot("Si")*1e3
-end
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

612 PHREEQC Get-going sheet 12

SOLID_SOLUTION
The solid solution option of PHREEQC can be used for modeling trace element behavior. The solubility of the
trace component in a solid solution is generally much smaller than when the trace element forms a pure solid.
Consequently the concentrations of the minor component are much reduced in solution, and the mobility of the
trace element is diminished.

SOLUTION 1
pH 7.5 Calcite # K = 10^-8.48
C(4) 4.0
Ca 1.0
Cd 1 ug/kgw Otavite # Otavite = CdCO3, K = 10^-12.1
SOLID_SOLUTION 1 # Note decrease of Cd concentration
CaCdCO3
-comp Calcite 0 # Ideal solid solution, initial amount = 0
-comp Otavite 0
END

USE solution 1
SOLID_SOLUTION 1 # Regular solid solution with ideal properties
CaCdCO3
-comp1 Calcite 0
-comp2 Otavite 0
# -distri D1 D2 x1 x2 # D1=D2=D_ideal = K_cc/K_ot for all x
-distri 4168.69 4168.69 0.1 0.6
END

Ideal solid solutions can have as many components as desired in PHREEQC, regular solid solutions have
only two components. The properties of regular solid solutions can be defined by various means (in the exam-
ple via distribution coefficients).
This keyword is most useful for modeling liquid solutions (BTEX etc., see Chapter 10).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 13 613

GAS_PHASE
This keyword is used to calculate equilibrium with a mixture of gases in a bubble. The bubble can have fixed
volume and varying pressure, or it may have fixed total pressure and varying volume.

# Calculate CH4 gas formation under peat

#
SOLUTION 1
pH 7.0 charge
Ca 2.0 Calcite
C(4) 2.0 CO2(g) -2.0 # P_CO2 = 0.01 atm
S(6) 1.0
N(0) 2.0 N2(g) 0.299 # P_N2 = 1.99 atm

REACTION 1
CH2O 1; 0.01 # decompose 10 mmol organic carbon

GAS_PHASE
-fixed_pressure
-pressure 2.0 # 2 atm, at 10 m depth
-volume 1e-8 # negligible initial volume, 1e-8 liter
-temperature 25 # default 25°C
CH4(g) 0.0 # gas name from PHASES, partial pressure
H2S(g) 0.0
CO2(g) 0.005
N2(g) 0.995
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

614 PHREEQC Get-going sheet 14

Databases
Currently 4 databases are provided with PHREEQC, PHREEQC.DAT is the smallest. WATEQ4F.DAT has addi-
tional data for heavy metals, MINTEQ.DAT has a few more organic chemicals, and LLNL.DAT is a huge data-
base with many minerals and large-range, temperature dependent equilibrium constants. The databases contain
lists under keywords.

SOLUTION_MASTER_SPECIES # always needed

#element species alk gfw_formula element_gfw
H H+ -1.0 H 1.008
H(0) H2 0.0 H
H(1) H+ -1.0 0.0
E e- 0.0 0.0 0.0
O H2O 0.0 O 16.0
O(0) O2 0.0 O
O(-2) H2O 0.0 0.0
# ...the above elements must always be present in the database
Ca Ca+2 0.0 Ca 40.08
Fe Fe+2 0.0 Fe 55.847 # Fe+2 is primary redox species
Fe(+2) Fe+2 0.0 Fe
Fe(+3) Fe+3 -2.0 Fe
# The primary species reaction is repeated in SOLUTION_SPECIES, with log_k = 0
# etc...
SOLUTION_SPECIES
H+ = H+ # H+ primary species
log_k 0.0
-gamma 9.0 0.0 # params for activity coefficient
Fe+2 = Fe+2 # Fe+2 is the primary redox species for Fe
log_k 0.0
-gamma 6.0 0.0
Fe+2 = Fe+3 + e- # new species defined on the right of = sign
log_k -13.02
delta_h 9.680 kcal # temp. dependency from Van 't Hoff eqn
-gamma 9.0 0.0
# etc...
PHASES # Minerals
Calcite
CaCO3 = CO3-2 + Ca+2
log_k -8.48
delta_h -2.297 kcal
-analytic -171.9065 -0.077993 2839.319 71.595 # log_k vs temp
# etc...
EXCHANGE_MASTER_SPECIES
X X- # element, species
EXCHANGE_SPECIES
X- = X- # reaction for master species with log_k = 0
log_k 0.0
Na+ + X-= NaX
log_k 0.0
-gamma 4.0 0.075
# etc...
SURFACE_MASTER_SPECIES # element, species
Hfo_s Hfo_sOH
Hfo_w Hfo_wOH
SURFACE_SPECIES
Hfo_sOH = Hfo_sOH # reaction for master species with log_k = 0
log_k 0.0
Hfo_sOH + H+ = Hfo_sOH2+
log_k 7.29 # = pKa1,int
# etc...
RATES
# etc...
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 PHREEQC Get-going sheet 15 615

INVERSE_MODELING
Inverse modeling aims to decipher the reactions and mixtures which have led to a given water quality.
# Find the reactions which produced solution 2 from solution 1.
#
PRINT; -reset false; inverse true
SOLUTION 1
Cl 1; Na 1
# evaporate 90% of solution 1, then add 2 mM gypsum
SOLUTION 2
Cl 10; Na 10; Ca 2; S(6) 2
END

INVERSE_MODELING
-solutions 1 2 # Balance solution 2 from 1
-phases # Using reactants from PHASES
Gypsum
Water # Water not in PHASES in database, is defined below
-balances # Any elements not in phases
Cl # Add 0 0.05 uncertainty for absolute balance of Cl
Na
Alkalinity 1 # Permitted uncertainty 100%

-uncertainty 0.05 # Default uncertainty, 5%

-range true

PHASES
Water # Define water
H2O = H2O
log_k 0 # log K not used
END

With the identifier -phases the possibly reacting phases are indicated. All reactants must have been defined
either as EXCHANGE_SPECIES or under keyword PHASES. These “phases” are used only for mole balancing,
their log K is unimportant when they are used in INVERSE_MODELING.
Elements which are not a part of the indicated phases, but are present in the solutions, must be listed with -
balances, as done for Na, Cl and alkalinity in this example. Numbers after the elements indicate the permitted
uncertainty for the solutions (in the order entered with identifier-solutions). A positive number indicates frac-
tional uncertainty, a negative number means that the uncertainty, in mol/kgw, may increase up to the absolute
value of the number. In the example, the permitted uncertainty for the alkalinity is set to 100% of the input con-
centration. The alkalinity is very small for these two solutions, and a balance can not be obtained without allow-
ing for a large uncertainty.
Isotopes can be included in inverse models, the
values are entered under SOLUTION.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

APPENDIX B

ANSWERS TO PROBLEMS

CHAPTER 1

1.1a. Na K Ca2 Mg2  Cl HCO3 SO42 

2.70 0.002 4.05 0.74 7.49 4.74 3.00 0.0 7.74 meq/L

The difference: 7.497.74  0.25 meq/L  1.7%

This is a fairly good analysis.

1.1b. Total Hardness  4.05 Ca2  0.74 Mg2  4.79 meq/L

expressed as mg/L CaCO3: 240 mg/L
expressed as °d (German Hardness): 13.5°d
Alkalinity (3.00 meq/L) equals 150 mg/L CaCO3.
Hence, non-carbonate hardness is 90 mg/L. It derives from other sources than calcite-dissolution and will
remain in the water when pH increases and calcite precipitates.

1.2.1.   7.4 meq/L,   6.0 meq/L.

Comparison of these values with the measured EC of 750 S/cm suggests that there is an anion-
shortage. This shortage is probably due to a too low HCO
3 value, since the calculated [PCO2]  10
3.7
,
which is below the atmospheric level.

1.2.2.   3.6 meq/L,   1.5 meq/L.

In this case there is a cation-surplus and if we look at the EC (150 S/cm), it seems that Ca2 determi-
nation must be wrong since the contribution of measured Ca2 concentration to the EC would be

280 mho/cm. At the measured pH, the Al content is also rather high.

1.2.3.   4.5 meq/L, (Fe included as Fe2),   4.5 meq/L.

Seems allright (EC  450 S/cm), but NO 2
3 cannot occur together with Fe . If a brown Fe(OH)3 pre-
cipitate is present in the non-acidified sample, the anions must be wrong. Remember that the concen-
tration of HCO 3 decreases when Fe
2
becomes oxidized in the non-acidified sample:

Fe2  1⁄4 O2  2 HCO3  1⁄2 H2O → Fe(OH)3  2 CO2

Thus, alkalinity is best determined in the field!

1.3b. A  St. Heddinge waterworks C  Haralds mineralwater

B  Maarum waterworks D  Rainwater Borris

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

618 Appendix B: Answers to problems

Concentrations in mmol/L

Sample A B C D

Na 0.783 19.139 0.391 0.313

K 0.161 0.358 0.026 0.016
Mg2 0.579 1.193 – 0.032
Ca2 3.019 1.048 – 0.035
NH4 0.001 0.194 – 0.068
 (meq/L) 8.134 24.173 0.417 0.531

Cl 0.818 5.669 0.480 0.338
HCO3 4.867 17.748 – 0.000
SO42 0.635 0.000 – 0.047
NO3 0.565 0.032 0.081 0.073
 (meq/L) 7.52 23.450 0.561 0.505

EC (S/cm) 750 1900 88

ad A: E.B.  (8.134  7.520) / (8.134  7.520)  100  3.9%, which is reasonable but not perfect.
Since EC / 100  750 / 100  7.50 is in best agreement with , one of the cation concentrations must
be a bit high, possibly Ca2.
ad B: E.B.  (24.173  23.450) / (24.173  23.450)  100  1.5%. Based on the E.B., this is a rather
good analysis. However, EC / 100  19.00 is not in good agreement with the sum of cations or anions,
but we are at the upper limit where the simple relation    (in meq/L)  EC / 100 is justified.
ad C: The analysis is incomplete and the E.B. cannot be calculated. The hardness in german degrees is
1°dH  17.8 mg CaCO3 / L  0.18 mmol/L (Ca2  Mg2)  0.36 meq/L. The  becomes then
0.78 meq/L  78 S/cm EC, which is in reasonable agreement with the measured EC.
ad D: E.B.  (0.532  0.505) / (0.532  0.505)  100  2.6%.

1.3c.

Concentrations in mg/L.

Rain contribution A B C

Concentration factor 2.42 16.75 1.42

Na 17.40 120.60 10.20
K 1.52 10.55 0.89
Mg2 1.89 13.07 1.11
Ca2 3.38 23.45 1.98
NH4 2.97 20.60 1.74
Cl 29.00 201.00 17.00
SO42 10.83 75.04 6.35
NO3 10.92 75.71 6.40
PO43 0.00 0.00 0.00

ad A: The contribution of rainwater is for most components small, except for Na which apparently is
derived mainly from rainwater.
ad B: The concentration factor of 16.75 is much too high for Danish conditions and the high Cl
concentration indicates a marine influence.
ad C: Most of the dissolved contents in this sample originates from rainwater.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 2 619

1.3d. ad A: The high Ca2 and HCO 3 contents indicate the presence of CaCO3 in the aquifer and the high
NO3 content suggests agricultural activity in the recharge area.

ad B: The high Na content, which is balanced by HCO3, indicates that ion exchange has influenced
this water composition.
ad C: This sample must originate from a reservoir with little reactive material, since most of the dis-
solved content is derived from rainwater.

CHAPTER 2
2.1. x  I 0 / (1.5 E)  (0.025 m H2O  4 m/s  3.15107 s/yr)/(1.5  0.5 m H2O/yr)  4200 km.
2.2. Integrate Equation (2.2) from 0 to 2500 km:
2.5 106

∫ E x dx
E (2.5106 )2 / 2
RPtot  0
  0.22
2.5 106 I 0 2.5106  0.5E (2.5106 )2 / 2
∫ ( I 0  0.5 E x ) dx
0

for I0  3.15106 m H2O  m/yr and E  0.5 m H2O/yr.

2.3a. In the following table, the concentrations of cations and anions are expressed in eq/L. The estimated
parameters are underlined.

H NH4 Na K Ca2 Mg2  Cl NO3 SO42 

Vlissingen 100 49 265 8 78 68 568 297 92 170 559

Deelen 112 119 30 3 32 10 306 37 90 179 306
Beek 26 143 23 11 100 10 313 34 87 192 313

Vlissingen: E.B.  0.8%

2.3b. The seawater-contribution to the total concentration of ion i may be calculated from ci,rain 
(ci,sea / cCl,sea)  cCl,rain.

Vlissingen Deelen Beek

Total Sea %Sea Total Sea %Sea Total Sea %Sea

H 100 0 0 112 0 0 26 0 0
Na 265 254.5 96 30 31.7 !100 23 29.1 !100
K 8 5.6 70 3 0.7 23 11 0.6 6
Mg2 68 57.7 85 10 7.2 72 10 6.6 66
Ca2 78 10.5 13 32 1.3 4 100 1.2 1
NH4 49 0 0 119 0 0 143 0 0
Cl 297 297 100 37 37 100 34 34 100
SO42 170 31 18 179 3.9 2 192 3.5 2
NO3 92 0 0 90 0 0 87 0 0

2.3c. – The very low H, NO3 and NH4 contents in seawater cannot contribute significantly to the rain-
water composition. Instead, their sources in rainwater are industrial processes and the evaporation of
ammonia.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

620 Appendix B: Answers to problems

– The seawater-contribution for the Na concentration is negative for both Deelen and Beek. This
might be an effect of industrial Cl production. In such cases Na is preferred as conservative ion.
– Dust and fertilizer are likely sources for K in rain.
– The seawater influence on the Ca2 concentration is low and cement industry seems to be the source
of high Ca2 concentrations in rain at Beek.
2.4. Substitute
for R:
 (
0  1000) f (1)  1000
differentiate:

d
/df  (
0  1000)(  1) f (2)

Hence:
d
2H (95  1000) (1.0876  1)
 0.9(1.091.01)  8.0
d
18O (12.2  1000) (1.0099  1)

The deuterium excess is (95  (1.0876  1)  1000)  8 (12.2  (1.0099  1))  1000)  10.4.
2.5. 18
O  0.11117:
18O  0.065‰; 18O  0.11118:
18O  0.025‰; from (0.05 / 103) Rst 16O 
5.56106 mol/L (i.e. 6 digits accurate).
2.6. From continuity for water flow, the stagnant, water filled porosity is st  rz  w  0.25 
0.089  0.161. The Cl mass balance is rzcrz  wcw  stcst. Hence, cst  (0.25  0.7 
0.089  0.19) / 0.161  0.98 mmol/L.
2.11. The answer follows the 3rd item in Example 2.1:
H O  80  91.3 f. f
0.2 at 0°C
2
(Figure 2.10),
H O
61.7‰.
2

CHAPTER 3
3.1. 3 m/yr.
3.2. The infiltration reach of 200 m is distributed proportionally over depth, and obtains a thickness of
200 / 2000  50  5 m. Mean depth is 1000 / 2000  50  25 m; travel time is for water infiltrated at
1000 m: t  34.7 yr, and for water infiltrated at 900 m: t  40 yr.
3.3. At 1 km is v  20, vD  6 m/yr. At 2 km is v  40, vD  12 m/yr. Travel time 34.7 yr. Age of water at
10, 20, 30, 40 m depth is 11.2, 25.5, 45.8, 80.5 yr. The formula t  d/ P gives 10, 20, 30 and 40 years.
3.4. x  x0 exp[Pt / (2 D)] and d  D(1  exp[Pt / (D)])
3.5. v  dx/dt  P(r2  x2) / (2 xD) gives ln[(x2  r2) / (x02  r2)]  Pt / (D).
3.6. 4 yrs.
3.7a. The gradient increases parabolically with distance (x2) in the upper, phreatic aquifer, and is invariant in
the lower aquifer.
3.7b. Flow is downward.
3.7c. A diverging bundle of lines in the D/x plane, with decreasing slope deeper down.
3.8. x0  333 m. Travel time above the clay layer is from 666 to 1666 m in the upper phreatic aquifer,
t  (2 Dw / (3P)) (ln(1666)  ln(666))  0.6Dw / P. The travel time below the clay layer is, from
mass balance, t  V / Q  (1000D/3)/(333P)  D/ P, or 1.67 times longer. However, the hydraulic
gradient increases at the end of the clay layer, and more water attempts to flow through the lower
aquifer. A numerical model gives a 1.16 times longer travel time through the lower clay layer.
3.9. b  2.12 g/cm3; 2 ppm  2  2.12 / 0.2  21.2 mg/L.
3.10.   0.26.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 4 621

3.11. c2 displaces c1: at PV  1, c shocks from 0.025 to 0.075 where dq / dc  0. c then shocks to the end con-
centration 0.175 after q / c  (0.9  0.2) / (0.17  0.075)  7.3 PV.
c1 displaces c2: c shocks from 0.17 to 0.05 after q / c  (0.9  0.4) / (0.17  0.05)  4.17 PV, then
a wave to c  0.025, which ends at dq / dc  8.
3.12.   25 and 79 cm.
3.13. Integrate from 0 to 100 cm:

100 100
10  x2 
∫ cx, t dx 
4$Dt
∫ exp   dx
 4 Dt 
0 0

Substitute x2  4Dt s2, and simplify:

100
100 4 Dt
10  100 
∫ cx, t dx 
$
∫ exp(s 2 )ds  5 erf  
 4Dt 
0 0

We find erf(100 /
  )  erf(0.89)  0.792. Hence 5  0.792  0.396 g NaCl resides in between
4Dt
0 and 100 cm.
3.15. DL  Df   v gives 5  103, 5  104 and 5.3104 cm2/s.
3.16. L  0.02 x and 0.06 x.
3.17. The temperature factor is 283  0.891 / (298  1.3)  0.65. Thus DNaCl  0.65  (1.33
109  2.03109) / 2  1.09109 m2/s. DCa(HCO )  0.65  (0.79109  1.18109) / 2  0.64
3 2

109 m2/s.

CHAPTER 4
4.1a. A: SI  log IAP  log K  11.08  10.57  0.51
B: SI  log IAP  log K  10.07  10.57  0.50
C: SI  log IAP  log K  8.78  10.57  1.79

4.1b. I F Ca IAP SI

A: 0.026 0.86 0.54 11.48 0.91
B: 0.057 0.81 0.44 10.61 0.04
C: 0.199 0.75 0.31 9.54 1.03

4.2a. 25°C log Kvilliaumite  0.49

4.2b. 10°C log Kvilliaumite  0.50
4.2c. log SIvilliaumite
A: 5.41
B: 4.45
C: 2.67
4.3a. SIgypsum  log IAP  log K  6.17  4.60  1.57. Log IAP decreases when CaSO04 is included in
the calculation.
4.3b. 2.56 mmol gypsum dissolves, [Ca2]  3.24 mmol/L and [F]  0.091 mmol/L.
4.4. Analysis A, but it requires that fluorite is present in the rock.

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

622 Appendix B: Answers to problems

4.5. O2(g) ↔ O2(aq) G0r  16.5  0  16.5 kJ/mol

G0r  RT lnK ⇒ K  [O2(aq)] / [PO2]  exp(16.5 / 2.48)  102.89

At 25°C:
[PO2]  100.68 → O2(aq)  0.27 mmol/L  8.6 mg/L.

Van’t Hoff-equation:
log KT  log K25  (10,000 / 19.1)  (1 / 298  1/T).

At 15°C:
(288 K): log K  2.83 → O2(aq)  0.31 mmol/L  9.9 mg/L.

At 5°C:
(278 K): log K  2.76 → O2(aq)  0.36 mmol/L  11.5 mg/L.

4.6. For smithsonite (log K  10.0) and calcite:

------------------------------------------------User print------------------------------------------------
aCa  7.3848104 aZn  2.2293105 aCa / aZn  3.3126101
4 7
aCa  7.591010 aZn  2.314810 aCa / aZn  3.2793103
mCa / mCd  16 and 1600, respectively. The ZnCl complex is not very important.

4.8b. d(FeS2) / dt  k (FeSsurf)2 (Ssurf) PH2S

k  21020 cm6 atm/mol/L/sec

4.8c. The second, since the rate depends on the dissolved sulfide concentration.

4.9. First calculate moles of OC in 0.1  1  1 m3 soil: 0.1 m3  1400 kg/m3  0.4 kg OC/kg soil/
(12 g/mol C)  4.67 kmol.
At steady state, OC is constant: R  4.67  0.025  0.117 kmol/yr/m2. For the whole Netherlands:
0.117  3  1010  3109 kmol/yr.
Human emission is 10  103 kg/yr/person  16106 people/(12 kg/kmol)  1.31010 kmol/yr, or
about 5 times higher. Note that at steady state the soil absorbs as much C as it emits, thus increasing the
contribution of human emission to atmospheric CO2 even more.

4.10a. RATES; Organic_C

-start
10 mSO4 = tot("S(6)")
20 rate = 1.e-13 * mSO4 / (1.e-4 + mSO4)
30 moles = rate * parm(1) * m * time
40 save moles
-end
KINETICS 1; Organic_C
-formula CH2O; -m0 2.68; -parms 8e4; -step 2.16e6 in 25 #25 days
SOLUTION 1; S(6) 20; Na 40 charge

9
4.10b. 4 reduction, dOC /dt  k OC, k  810 /s  0.25/yr.
For SO2

CHAPTER 5
Laboratory exercise:
DATABASE phreeqc.dat
SOLUTION 1; EQUILIBRIUM_PHASES; CO2(g) -2.0

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 5 623

KINETICS 1
Calcite; -parms 110.3 0.67; -step 8640 in 10 # only 1/10 day
INCREMENTAL_REACTIONS
USER_GRAPH
-axis_titles "time / s" "Saturation Ratio"
-start; 10 graph_x total_time
20 graph_y sr("Calcite"); end
END

5.1. 1  0.90; 2  0.67. At pH  7 is [HCO 3 ]  1/0.9 times lower, 10

2.75
. Similarly, [CO32]  105.88
2.68
and TIC  10 . At pH  10, the ionic strength is above 19, for beyond the range of the Debye-
Hückel theory.
5.2. pH  6.81; [CO32]  106.38. pH  9.67; [CO32]  (1.7  1.3) / 2  0.2 mmol/L.
5.3. mCO 2   TIC; 1  1016.6 2[H]2  [H] 2 / (1 1010.3)  1.
3
At pH  10.5, 1  2.7105  0.47  1  1.47; mCO 2  1.70103.0 mol/L,
3
[CO32]  1.14103.
1 7 7
At pH  6.3,   14144; mCO 2  1.7710 ; [CO3 ]  1.1810 .
2
3
5.4. [H2CO3]  103.5; [H]  104.9  [HCO3 ]; [CO32]  1010.3.
5.5. Ca2  1.58 mmol/L.
5.6. PCO2  0.02 atm.
5.7. PCO2  102.55 atm.
5.8. k1 [H]
0; k2 [H2CO*3 ]  109.8; k3 [H2O]  107.0; Rfw  107.0.
[Ca2]eq.  0.5103; Rb  k4  2  [Ca2]2  Rfw; k4  0.2.
95% saturation, or mCa2  4.75104; t  9159 s  2.5 hr.
5.9a. [PCO2] can be calculated from log [PCO2]  7.8  log HCO
3  pH, which gives:
(1) 0.9 (2) 2.2 (3) 2.6 (4) 1.6
5.9b. Saturation with respect to calcite is calculated from:
H  CaCO3 ↔ Ca2  HCO 3 ; log K  1.8

In contact
log (IAP/Kcc) with calcite? Open/closed?

(1) 5.7 no –
(2) 0.5 yes Open
(3) 0.9 yes Partly closed
(4) 1.5 partly

5.9c. (1) acid rain; (2) ; (3) SO42 high; (4) K, NO3 high.
5.10a. CO2 in water is analyzed as H2CO3.
H2CO3 ↔ H  HCO3 ; K  106.3
[H]  106.3  66  103 / 6.38103  105.3; pH  5.3
5.10b. Alkalinity will not change if CO2 decreases. (Alkalinity  HCO  
3  2 CO3  OH  H ). CO2-
2

concentration will reach equilibrium with the atmosphere: H2CO3  105 mol/L. Hence:

[H ]  106.3  105 / 6.38103  109.1 ; pH  9.1

5.10c.   4.45 meq/L   7.78 meq/L; Probably Fe precipitated in the non-acidified sample.
5.10d. To maintain charge balance an equivalent amount (in meq/L) of HCO3 must be removed from the solution:
Fe2  2HCO3  1⁄4 O2  1⁄2 H2O ↔ Fe(OH)3  2CO2

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

624 Appendix B: Answers to problems

5.10e. Juvenile  first emergence at the Earth’s surface.

5.11.1. Constant; if H2CO3 dissociates (H2CO3 ↔ H  HCO  

3 ), H and HCO3 compensate each other’s
contribution to alkalinity.

5.11.2. Decrease; HCO  

3 reacts with added H : HCO3  H

↔ H2CO3

5.11.3. Decrease; Fe3 will take up OH.

5.11.4. Constant.

5.11.5. Decrease; Mn2 is oxidised by O2, and releases H: Mn2  H2O  1⁄2 O2 ↔ MnO2  2H.

5.11.6. Increase.

5.11.7. Increase.

5.11.8. Increase.

5.11.9. Constant.

5.12. 60 mg/L SiO2  1 mmol/L Si. At pH  10.0 :

[H3SiO
4 ] / [H4SiO4]  (10
0 9.1
) / (1010.0)  100.9  7.94
[H3SiO
4 ]  7.94 / 8.94  0.89 mmol/L
When Alkalinity is titrated, H3SiO4 associates with H, to become H4SiO40. The contribution to the
alkalinity is equivalent to the H3SiO4-concentration.
5.13. With Equation 5.24, we obtain Ca  1.36 mol/m3, or 0.41 mol/m2/yr leaves the area. This is 41 g
CaCO3, or 20 cm3/m2/yr. Denudation rate is therefore 20 mm/1000 yr.
5.14. PHREEQC calculates Ca  1.4 mM, which gives a flux of 0.14 mol/m2/yr or 14 g CaCO3 / m2/yr.
1 m3 sand  b /   1800 kg  360 kg calcite. Decalcifation rate is (1 m3/(360  103 g/14 g/m2/yr)
 1000 mm/m)  0.04 mm/yr.
5.15, 16. Download input files from www.xs4all.nl/appt
5.17. See PHREEQC manual, p. 43.
5.18. PRINT; -reset false; -user_pr true
SOLUTION 1; -temp 10
EQUILIBRIUM_PHASES 1; CO2(g) -1.5
RATES; Dolomite
-start; 10 save -1.2e-10 * log( SR("Dolomite") + 1e-5) * time
20 print SR("Dolomite"), total_time; -end
KINETICS 1; Dolomite; -step 1e8 in 100; INCREMENTAL_REACTIONS
END

For [PCO2]  103.5, t  2.6106 s. For [PCO2]  101.5, t  1.4107 s.

5.19. Download input file from www.xs4all.nl/appt
5.19a. 85 s at 10°C, 15 s at 25°C.
5.19b. 101.54
5.20. 53, 46 and 15pmc, respectively.
5.21. Download input files from www.xs4all.nl/appt
5.22. DATABASE phreeqc.dat
SOLUTION 1; KINETICS 1; Calcite; -parms 110.3 0.67; -step 864 in 10;
INCREMENTAL_REACTIONS; PRINT; -reset false; -user_print true
USER_PRINT; -start; 10 print total_time, "SR_cc", sr("Calcite"); end
END # (about 5 min are sufficient).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 6 625

CHAPTER 6
6.1. gfw  389.91 g/mol. One packet of 20 units (with 24 oxygens) has 2  0.7 charge over 2 basal planes,
i.e. CEC  2  0.7  1000 / 389.91 / 20  1000  180 meq/kg. Charge is 0.015 qe / Å2  0.24 C/m2.
6.2. gfw  368.22 g/mol. CEC  217 meq/kg. Charge  0.28 C/m2.
6.3. 10 times concentrated water: Na  0.0259; Mg  0.134; Ca  0.840.
seawater: Na  0.584; Mg  0.319; Ca  0.097.
6.4. 1  0.85; 2  0.53; KNa\Cd  0.286 and 0.339.
6.5a. Sample 1 is rainwater (with Na / Cl-ratio of seawater) in which some calcite has dissolved. Sulfate is
partly reduced. Sample 2 is slightly mixed with salt water (higher Cl). Cation-exchange is evident,
Ca2 is replaced by Na, causing undersaturation with respect to calcite. Renewed dissolution of calcite
gives high alkalinity. Sample 3 has a low Na/Cl-ratio, sulfate partly reduced. Sample 4 is Seawater.
6.5b. A: NaHCO3-water, interface moves downward. B: CaCl2-water, interface moves upward.

6.5c. Na K Mg Ca

Sample 1 0.0062 0.0014 0.0187 0.974
Sample 4 0.558 0.061 0.292 0.089

6.5d. Na-X K-X Mg-X2 Ca-X2 

Sample 1 0.4 0.09 1.1 58.4 60.0 meq/L
Sample 4 33.5 3.7 17.5 5.3 60.0 meq/L

6.6a. Evaporation concentrates Nile-water in the first place. Using a “conservative” species, Cl, we can
calculate the composition:
Concentration-factor Clgroundwater / ClNile  4.5 / .5  9.0

Na K Mg2 Ca2 HCO3 Cl SO42 pH

Nile-water .5 .1 .4 .7 2.2 .5 .1
9 concentrated 4.5 .9 3.6 6.3 19.8 4.5 .9
groundwater 8 .1 .2 .7 2.7 4.5 1.0 8.1
Difference 3.5 .8 3.4 5.9 17.1 – 0.1

The decrease of Ca2 and HCO

3 (and of some Mg) is an effect of calcite-precipitation. If we compare
K and IAP of the reaction:
CaCO3  H ↔ Ca 2  HCO
3; K  63; IAP  136,

the water is clearly oversaturated. The increase of Na is produced by cation exchange of Ca2 and
Mg2 for Na from the soil. The decrease of K is an effect of uptake by vegetation or illite.
6.6b. The very high Cl-concentration in water from borehole 121 can only be explained by salt water intrusion.
If we mix groundwater and seawater at the correct ratio, (contribution of seawater 0.473), we obtain:

Na K Mg2 Ca2 HCO

3 Cl SO42

Calculated 231 5 26 5.4 2.5 270 14

Analysed 180 2.5 35 15 7.5 270 1.5
Difference 51 2.5 9 10 5 – 12

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

626 Appendix B: Answers to problems

The decrease of Na is an effect of cation exchange. Salt water intrudes in a fresh aquifer, Ca2 will be
removed from the exchange complex and replaced by Na. The decrease of SO42 is an effect of reduction.

6.6c. Calculation of the composition in mg/L:

Na K Mg2 Ca2 HCO3 Cl SO42

4140 98 850 601 458 9571 144

The contribution of HCO  

3 to TDS  0.5  HCO3 (mg/L), since half of the HCO 3 originates from
CO2(g) (and would escape when the sample is evaporated to dryness). Hence: TDS  15,633 mg/L.

6.7a. Na-X  160.6; Mg-X2  283.18; Ca-X2  306.22 meq/L pore water.
6.7b. Total cations  14.66 meq/L. Na  13.28; Mg2  0.43; Ca2  0.26 mmol/L.

6.7c. Exchanger composition in equilibrium with injected water is: Na-X  56.35; Mg-X2  81.1;
Ca-X2  612.54 meq/L pore water.

6.7d. Assume that Na-X is flushed from 160.6 to 56.35 meq/L, which needs (160.6  56.35) /
(13.28  9.4)  26.9 pore volumes  6985 m3 water.

6.8a. Na-X  2.2, Mg-X2  16.1, Ca-X2  41.7 meq/L.

6.8b. Na  28.8, Mg2  50.2, Ca2  52.8 meq/L.

6.8c. The final composition of the exchanger is Na-X  25.8, Mg-X2  22.7, Ca-X2  11.5 meq/L. Ca-X2
requires (41.7  11.5) / (105.6  8.3)  0.31 pore volumes. Note that the concentration of Ca2 only
reaches up to 17 mM, because dispersion smoothens the peak.

2 K Na/I 2
KV 
6.9. Na/I 2
1   Na
K Na/I 2
KG 
Na/I 2
1   Na

6.11. Smectite with 0.09 C m2.

x: 0 10 50 100 Å

0: 0.393 0.284 0.078 0.015 V

6 4
n / n% 4.310 6.310 20.6 1.8
m 43,000 63 0.206 0.018 mol/L
(Note that the concentrations at 0 and 10 Å are physically impossible)
n / n% 2.3107 1.6105 0.05 0.55

6.12. SAR of irrigation water is 1.7; when 10 times concentrated, SAR changes to 5.5.
6.13. For illite (using activity of Ca2 and K in solution obtained with [i]  imi):

0.1 0.01 0.001 104N

KVCa\K 0.90 0.76 0.53 0.25 (Gapon)

KGCa\K 0.62 0.43 0.28 0.13 (Vanselow)
KC\K 1.09 0.81 0.55 0.25 (Gaines and Thomas)

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 7 627

and similarly for the resin:

0.1 0.01 0.001 104N

KVCa\K 2.68 2.41 2.37 1.80 (Gapon)

KGCa\K 1.53 1.26 1.20 0.90 (Vanselow)
KC\K 2.87 2.50 2.38 1.81 (Gaines and Thomas)

6.14. CEC  (IM i  MJ j  M K k  …) TEC

TEC  (I / i  J / j  K / k  …) CEC and
I CEC  i M
I TEC yield:

TEC M
I i
I  i M
I  M
CEC I i  M
J j  M
K k 

CEC I / i
M
I  I / i
M 
TEC I / i   j / j  K / k 

6.15. EXCHANGE_SPECIES; Ca+2 + 2X- = CaX2; log_k -10

Mg+2 + 2X- = MgX2; log_k -10 # remove database species
0.5 Mg+2 + X- = Mg0.5X; log_k 0.3
0.5 Ca+2 + X- = Ca0.5X; log_k 0.4
#or
Ca+2 + Mg0.5X = Ca0.5X + Mg+2; log_k 0.1; -no_check; -mole_balance Ca0.5X
SOLUTION 1; Mg 1; EXCHANGE 1; X 2e-3; -equil 1
REACTION 1; Mg -1 Ca 1; 1.999e-3 in 20
USER_GRAPH; -start; 10 graph_x tot("Ca"); 20 graph_y mol("CaX2") + mol("Ca0.5X")/2
-end
END (compare the isotherms by crossing out the species)

6.16. Use:
EXCHANGE_SPECIES; K+ + X- = KX; log_k 0.43; -gamma 3.5 0.015

6.17. mNa  1 M gives qNa  0.55 M, R  1.6. mNa  10 mM gives qNa  5.2 mM, R  1.5.

CHAPTER 7
7.1. / ln10  95. As  F2 / (95RT ln10)  1.7104 F/kg. Apparently (cf. Example 7.5), the capacitance
is not really constant at low ionic strength.
7.2. The ratio is 0.04, almost all sites are protonated, in agreement with Figure 7.21.

7.3. pKint  2.8; /ln 10  3 1; Kh-Al  3 103 give:

5
pKa

3
pKint

0 0.2 0.4 0.6 0.8 1.0

F

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

628 Appendix B: Answers to problems

7.4. SOLUTION 1; K 1 charge; Br 1; pH 5 # change Br concentration as necessary

SURFACE 1
Hfo_w 8.5e-3 600 3.8 # fitted sites...
-equil 1; -diff true
PRINT; -reset false; -user_print true
USER_PRINT
-start
10 print "mM Br = ", 1e3*tot("Br"), "R 5", 1 + edl("Br")/tot("Br")
-end
END
For a fixed pH, protons sorb more at high ionic strength because  decreases. Hence higher charge on
surface, and more Br needed in double layer for charge compensation. Charge increases approxi-
mately with √I. Hence, the distribution coefficient (Kd  R  1) decreases by 3 for each tenfold
increase of KBr concentration.
7.5. SOLUTION 0; K 1 charge; Br 1; pH 5
SOLUTION 1-5; K 4 charge; Cl 4; pH 5 # fitted initial KCl concentration...
END
SURFACE 1-5; Hfo_w 8.5e-3 600 3.8; -equil 1; -diff true
PRINT; -reset false; -warnings false
TRANSPORT; -cells 5; -length 0.02; -shifts 25 1; -disp 0.002; -punch 5
USER_GRAPH; -head PV Br; -axis_scale x_axis 0 5 1; -plot time
-start; 10 graph_x (step_no + 0.5)/cell_no; 20 graph_y tot("Br")/1e-3; -end
END

7.6. SURFACE 1; Hfo_w 5.6e-4 60 2.5; -equil 1 # 10% active sites

SOLUTION 1; pe 14 O2(g) -0.68; Na 100; Cl 100
REACTION 1; H4AsO4 1; 0.267e-3
PHASES; Fix_ph; H+ = H+; log_k 0; EQUILIBRIUM_PHASES 1; Fix_ph -7.0 NaOH
USER_PRINT; -start;
10 qAs5 = mol("Hfo_wH2AsO4") + mol("Hfo_wHAsO4-") + mol("Hfo_wOHAsO4-3")
20 qAs3 = mol("Hfo_wH2AsO3"); 30 print '% As5 sorbed:', qAs5/ (tot("As(5)") + qAs5)
40 print '% As3 sorbed:', qAs3 / (tot("As(3)") + qAs3)
50 print 'mmol As5/kg sorbed', qAs5 / 2.5e-6, 'mmol As3/kg sorbed', qAs3 / 2.5e-6
-end
END
SURFACE 1; Hfo_w 5.6e-4 60 2.5; -equil 1 # 10% active sites
SOLUTION 1; Na 100; Cl 100
REACTION 1; H3AsO3 1; 0.267e-3
EQUILIBRIUM_PHASES 1; Fix_ph -7.0 HCl
END
At high concentration of H4AsO4, the surface is rendered negative by Hfo_wHAsO4-, which dimin-
ishes sorption of the anion. At low concentration this effect is negligible; the higher chemical affinity
of As(5) for ferrihydrite then gives higher sorption of As(5) than of As(3).
7.7. The half-life increases by 100 (becomes 3 days).
7.9. Approximately 4.5 hr.
7.10. Hfo_sOPb  4.473104; Hfo_wOPb  3.132103.
7.11. cf. Figure 7.35.
7.14. qCd  5.7104 cCd /(30.9  cCd), both q and c in g/L pore water. For very small cCd, Kd  qtot /cCd 
5.7104 / 30.9  1844. The distribution is linear if cCd  3.09 g/L. 75% of qtot is reached when
cI/(KL  c)  0.75, or cCd  93 g/L.
7.15. SOLUTION 1; Ca 2; Zn 2e-4
EXCHANGE 1; X 73.5e-3; -equil 1
USER_PRINT; -start; 10 print 'Kd_Zn = ', mol("ZnX2") / tot("Zn"); -end
END
gives Kd s: a) 18.1; b) 18.1; c) 9; d) 36.2; e) 18.3
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
 Chapter 7 629

7.16. SOLUTION 1; Mg 2; Zn 2e-4; pH 7

SURFACE 1; Hfo_w 2e-4 600 0.088; Hfo_s 5e-6; -equil 1
USER_PRINT; -start
10 print 'Kd_Hfo', (mol("Hfo_wOZn+") + mol("Hfo_sOZn+"))/tot("Zn"); -end
END
Kd s: a) 12.6; b) 8.8 (strong sites are full already); c) 11.8; d) 25.2 (note to double the grams of Hfo and
weak and strong sites); e) 0.005

7.17. Mg2 is sorbed on the weak sites, which increases the potential, rejects Zn2. Increase of the ionic
strength increases the capacitance, decreases the surface potential. The chemical affinity becomes less
important, H displaces Zn2 from the strong sites. Ca2 is sorbed also on the strong sites and doubling
the concentration has a much stronger effect than Mg2.

7.18. SOLUTION 1; Pb 2; pH 3; Na 100; N(5) 100 charge

SURFACE 1; Hfo_w 20e-3 600 8.8; Hfo_s 5e-4; -equil 1
REACTION 1; NaOH 1; 28e-3 in 20
USER_GRAPH; -start; 10 graph_x -la("H+"); 20 graph_y (2e-3 - tot("Pb"))/2e-3;-end
END # and similar for Zn, Ca

7.19. SOLUTION 1; Ca 0.33; Ni 1e-5; pH 3; N(5) 0.33 charge

EXCHANGE 1; X 8.7e-3; -equil 1
EXCHANGE_SPECIES; Ni+2 + 2X- = NiX2; log_k 0.6; H+ + X- = HX; log_k 4.5
REACTION 1; NaOH 1; 10e-3 in 20
SURFACE 1; Hfo_w 2e-5 600 8.8; Hfo_s 5e-7; -equil 1
SURFACE_SPECIES; Hfo_sOH + Ni+2 = Hfo_sONi+ + H+; log_k 0.37
USER_GRAPH; -start; 10 graph_x -la("H+1");
20 q_Ni = mol("NiX2") + mol("Hfo_sONi+") + mol("Hfo_wONi+")
30 graph_y q_Ni / (q_Ni + tot("Ni")); -end
END
The model sorption edge is steeper than in Figure 7.18, 10% sorption at low pH cannot be
modeled, but see Bradbury and Baeyens (1999) who cancel the electrostatic contribution.

7.20. SOLUTION 1; Ca 1; Cd 2e-4; Cl 2

EXCHANGE 1; X 55.7e-3; -equil 1
EXCHANGE_SPECIES; Cd+2 + 2X- = CdX2; -log_k 0.8; -gamma 0 0
USER_PRINT; -start; 10 print 'X... Kd_Cd = ', mol("CdX2") / tot("Cd"); -end
END
gives Kd  24.1 L/kg

7.21. SOLUTION 1; -temp 10; pH 7 charge; Ca 1 Calcite; C(4) 1 CO2(g) -2.5; Mg 1 Dolomite
SURFACE 1; Hfo_w 8.4e-3 600 3.84; -equil 1
END
yields for carbonate R  1  3.6 / 3.9 (mM/mM)  1.9.

7.22. Download the files from www.xs4all.nl/appt

7.23 SOLUTION 1; pH 7.67; Na 2.93; Ca 0.59; Alkalinity 3.85; F 0.26; As 1.3e-3

EQUILIBRIUM_PHASES 1; Goethite 3.0 0; Fluorite 0 0; Calcite
SURFACE 1; Hfo_w Goethite 0.2 5400; Hfo_s Goethite 2.5e-3
REACTION; FeSO4 1e-3 NaClO4 0.167e-3; 10 in 10
USER_GRAPH; -head step As F
-axis_titles Step “ug As/L or mg F/L”
-axis_scale y-axis 0 5
-start; 10 graph_x step_no; 20 graph_y tot("As")*74.92e6, tot("F")*19e3; -end;
END
Already step 1 yields As  10 mg/L, and in step 3, F 2.5 mg/L. Further reaction doesn’t improve
much. To arrive at step 3, the pill should contain: 456 mg FeSO4, 61 mg NaClO4 nd 39 mg Calcite.
Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands
630 Appendix B: Answers to problems

CHAPTER 8
8.1. K-feldspar, quartz and gibbsite.

8.2. The base production by gibbsite dissolution (Al(OH)3) is:

[OH ]  3[Al3 ]  2[AlOH 2 ]  [Al(OH)

2]
[OH ]  3(61105 )  2(7.7106 )  6107  1.97104
[H ]initial  104.53  1.97104  2.26104 and pH initial  3.65

8.3. Combining Kgibbsite and Kjurbanite gives:

2 log[OH]  log[SO 24]  33.88  17.8  16.08

Combining with the dissociation of water yields:

2pH  log[SO 24 ]  11.92

8.4a. The water chemistry can be explained by:

dissolving 0.016 mmol/L NaCl; 0.015 mmol/L gypsum; 0.014 mmol/L biotite; 0.175 mmol/L plagio-
clase; 0.115 mmol/L calcite and precipitating 0.033 mmol/L kaolinite; 0.081 mmol/L Ca-smectite

8.4b. A greater weathering of plagioclase than of biotite or calcite is inconsistent with the usual weathering
sequence. However, Feth et al. (1964) stated rather explicitly that calcite is absent in the Sierra Nevada
source rocks, and calcite (or gypsum) should in fact not be included in weathering balances for the area.
Perhaps dry deposition is important.

8.5. Nasea  0.206 mM, Naalbite  0.137 mM.

PHREEQC finds 2 models:
Albite 1.443104; Biotite 4.212106; K-feldspar 6.293106; Anorthite 5.546105; Gibbsite
1.657105; Kaolinite 1.411104; CO2(g) 1.381103
and
Albite 1.360104; Biotite 4.601106; K-feldspar 5.904106; Anorthite 5.569105; Kaolinite
1.289104; CO2(g) 1.381103
The expected composition of the plagioclase is (Na0.71Ca0.29) (Al1.29Si3.7) O8.

CHAPTER 9
9.1. Find H2AsO4  0.58  As(5)  0.68 M with the speciation factor. Using Equation (9.32), pe  0.83,
Eh  0.05V.

9.2a. For example Fe2O3 / MnO2, Fe2O3 / MnCO3, FeCO3 / MnCO3, FeS2 / MnCO3.
For dissolved species Fe2 / Mn2, Fe3 / Mn2.

9.2b. For hematite, Fe2O3  2e  6H ↔ 2Fe2  3H2O, and pe  11.3  log[Fe2]  3pH

9.2c. A gradual decrease in pe or pH or both simultaneously.

9.2d. pH  7.16; [NO3 ]  110 mg/L  102.75 M; [NO

2 ]  2.1 mg/L  10
4.35
M.
   
From NO3  2 H  2 e ↔ NO2  H2O, log K  28.3
log[NO 
2 ]  log[NO 3 ]  2 pH  2 pe  28.3 → pe  7.79
From Mn  2 H2O ↔ MnO2  4 H  2 e,
2
log K  41.52
4 pH  2 pe  log[Mn2]  43.67, log[Mn2]  2.7

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 9 631

9.3. The PbO2 / PbO boundary: Combining reactions 1) and 3) yields:

PbO2  2 H  2 e ↔ H2O  PbO and
log K  2 pH  2 pe  log K1  log K3  49.2  12.7  36.5
The PbO/Pb boundary: Combining reactions 2) and 3) yields:
PbO  2 H  2e ↔ Pb  H2O and
log K  2 pH  2 pe  log K3  log K2  12.7  4.26  8.44
The PbO2 / Pb2 boundary: For reaction 1) and log[Pb2]  6:
log K1  log[Pb2]  4 pH  2 pe  49.2 and 4 pH  2 pe  49.2  6  55.2
The PbO / Pb2 boundary: For reaction 3) and log[Pb2]  6:
log K3  log [Pb2]  2 pH  12.7 and pH  1⁄2 (12.7  6)  9.35
The Pb(OH)3/Pb2 boundary: For reaction 4):
log K4  3 pH  log[Pb(OH)
3 ]  log[Pb ]  28.1
2

If log[Pb(OH)3 ]  log[Pb2] then pH  9.37. Accordingly, Pb(OH)

3 never becomes a dominant
dissolved species and is therefore not included in the diagram.
The Pb / Pb2 boundary: For reaction 2) and log [Pb2]  6:
log K  log [Pb2]  2 pe  4.26, and pe  1⁄2 (4.26  6)  5.13

20 O
H 2
2O
15

PbO2
10

Pb2
5
pe

0 HO
2 PbO
H
2

5

Pb
10

15
0 2 4 6 8 10 12 14
pH

9.4a. Combination of several equations gives: MnCO3  CO2  H2O ↔ Mn2  2 HCO 3 ; K  10
6.9
 6.9
In water in contact only with MnCO3: [HCO 3 ]  2[Mn ] ⇒ 4[Mn ] / [PCO2]  10 .
2 2 3

[PCO2]  101.5 → Mn2  1 mmol/L (i.e. 1 mmol/L MnCO3 will dissolve)

9.4b. Congruent dissolution: the mineral dissolves completely and no secondary phases are produced.
Incongruent dissolution: the stoichiometry of dissolved substances does not correspond to their ratio in
the dissolving phase, i.e. a secondary precipitate is formed.

9.4c. Rhodochrosite will dissolve according to 9.4a. Mn2 oxidizes in the presence of O2:
Mn2  H 2O  1⁄2 O 2 ↔ 2 H  MnO2 (pyrolusite)
meaning that rhodochrosite dissolves incongruently. The produced H reacts with HCO3 and increases


[PCO2]

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

632 Appendix B: Answers to problems

9.5. KMnO4  8H  5e ↔ Mn2  K  4H2O

(CH2O  H2O ↔ CO2  4H  4e) 5⁄4

KMnO4  3H  5⁄4 CH2O ↔ Mn2  K  5⁄4 CO2  11⁄4 H2O

9.6. Possible reactions are:

1. H2S(aq) ↔ HS  H; K  107

2. H2S(aq)  4H2O ↔ SO42  10H  8e; K  1040.6
 
3. S(s)  4H2O ↔ SO4  8H  6e ;
2
K  1036.2
 
4. ( 3  2) H2S(aq) ↔ S(s)  2H  2e ; K  104.4
5. ( 4  1) HS ↔ S(s)  H  2 e; K  102.6
6. ( 2  1) HS  4H2O ↔ SO42  9H  8e; K  1033.6

Precipitation of S is possible when the pH  6 a 7, and H2S-concentration ! 5 mmol/L

(at lower levels the stability-field of S will disappear).

2
SO
2
4
H2
S

0 2

SO2
4

4 3
2
pe

4

H2S
1
SO

5
S
2

6
4
S

HS 
HS 6
SO 
HS

2
4

8
2 4 6 8 10 12
pH

9.7b. Determine the fraction (f Hfo_wOH 2 ) with PHREEQC for the experimental pH’s (assuming that the
surface properties of goethite and ferrihydrite are the same):

SOLUTION 1; pH 1.5; pe 10 O2(g) -0.68; Fe 0.1 Goethite; Cl 1 charge

SURFACE 1; Hfo_w 1e-3 600 444e-3; -equil 1; END

Then regress, and find that the rate  1011.7 ( f Hfo_wOH 1.9 2
2 ) mol/m /s. (Note that the squared sum of

errors is higher than when regressing on H , Equation (9.82)).

9.8. For 2 electrons in the reaction H2 → 2H  2e, K increases by 10(2  3/5.708)  11, or H2 must be
11 times higher than at equilibrium.

9.9. Use Equation (4.29) and find K10 / K25  162. Hence H2 decreases by 1621⁄4  3.6

9.10. The reaction: 4H2(g)  2HCO

3 H

↔ CH3COO  4H2O; log K  25.28
add 8H  8e ↔ 4H2(g); log K  0

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

 Chapter 10 633

Write in the input file: SOLUTION_SPECIES; 8e  2HCO3  9H  CH3COO  4 H2O;
log_k 25.28. (Note to write redox equations in PHREEQC with electrons). Plot on the y-axis
mol(“CH3COO”)*1e8. (Note that acetate is thermodynamically quite unstable).

CHAPTER 10
10.1. With USER_PRINT, print 10^SI(“H2O(g)”) / (0.082 * TK) / (1000 / 18), which gives 1.73105 and
9.44106 L water/L air for 20 and 10°C
10.2. From Equation (3.61) find DH216O / DH218O  1.032. Hence
18Ovapor  32‰ for kinetic fractiona-
tion by diffusion. (Compare with equilibrium fractionation,
18Ovapor
10‰).
10.3. Download input file from www.xs4all.nl/appt
10.4. Download input file from www.xs4all.nl/appt; the half-life is about 105 s.
10.5. Download input file from www.xs4all.nl/appt
10.6. Naphthalene in peat: log Kd  3.0  log 0.7  2.85; Kd  700 L/kg; bulk density of peat is
b  1.0  0.4  0.4 kg/L. Hence Kd  700  0.4 / 0.6  467; velocity is 468 times smaller.
Tetracene in sand: b  1.855 kg/L; R  1  b /  631  3904.

10.7a. x  550 exp(0.3 t / 50 / 0.36):

t (years) x (m) Flowlength Depth (m) Thickness (m)

( x  x0)
25 834 284 17.0 3.0
100 2912 2362 40.6 0.9

Note that flowpath increases 9-fold, when time increases 4-fold.

10.7b. Assume CEC  [Ca-X2]  10 meq/kg. 1 kg  556 mL  200 mL H2O, hence CEC  50 meq/L 
25 mmol Ca2 / L
Kd(Cd)  [Cd-X2] / [Cd2]  [Ca-X2] / [Ca2]  2.5; R(Cd)  3.5.
For TCE, the slope dqdc decreases with c, hence we have a sharp front. RTCE 
1  q / c  1  (16.2  0) / (7.1  0)  3.3

10.7c. vH2O  R vi  R dxi / dt  P x / (D); x  550 exp(0.3 t / (R  50  0.36));

t (Years) x (M) Flowlength Depth (m) Thickness (m)

(x  x0)
25 620 70 5.6 4.0
100 885 335 18.9 2.8

10.7d. Flowlength after 100 yr  335 m;   10% of 335 m  33 m. 2  2Dt  2 x  2 

0.1  (335)2  22,110;   148 m. Conc. at 335  148  483 m is (16/100)c0  0.18 mg Cd/L. At 335 m:
(1/2)c0  0.55 mg/L. At 335  148  187 m is (84 / 100)c0  0.94 mg/L. The TCE-front will be sharper
since a convex isotherm gives small concentrations a higher slope dq/dc, and hence more retardation.
10.8. SOLUTION 1-40; -temp 10; pH 7.0 charge; Ca 1.0 Calcite; C(4) 2.0 CO2(g) -2.0
SOLUTION 0; -temp 10; pH 7.0 charge; pe 14.0 O2(g) -0.68; Ca 1.0 Calcite;
C(4) 2.0 CO2(g) -2.0
N(5) 6
END

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands

634 Appendix B: Answers to problems

KINETICS 1-40; Organic_C; -formula C; -m0 0.5

EQUILIBRIUM_PHASES 1-40; Calcite
END
PRINT; -reset false; -status false
USER_GRAPH; -headings depth Ca NO3 Alk
-start; 10 graph_x dist; 20 graph_y tot(“Ca”)*1e3, tot(“N(5)”)*1e3, Alk*1e3; -end
TRANSPORT; -cells 40; -shifts 30 1; -dispersivity 0.0; -diffc 0.0; -time_st 3.15e7;
-punch_frequency 30
END

For denitrifying 5.75 mM NO3, 7.19 mM C are needed. 0.1% C  500 mmol/L pore water. Reaction
time of C is t  1 / (5  104) ln((5  7.19) / 0.5)  28.96 yr, equals depth (m).

Copyright © 2005 C.A.J. Appelo & D. Postma, Amsterdam, the Netherlands