DIFFUSION

IN AND THROUGH
SOLIDS
BY
RICHARD M. BARRER
D.Sc, Sc.D.? Ph.D., F.R.I.C.
Professor of Chemistry, Aberdeen University; formerly
Research Fellow, Clare College, Cambridge

CAMBRIDGE
AT THE UNIVERSITY PRESS
1951
 PUBLISHED BY
THE SYNDICS OF THE CAMBRIDGE UNIVERSITY PRESS
London Office: Bentley House, N.W.I
American Branch: New York
Agents foir Gana54t,?IhdiaV^tid:Paki§tan5 M«*clhf]lan

FirpLprinted 1041
Al'Witfi, cortectians , MET

First printed in Great Britain at the University Press, Cambridge

Reprinted by offset-litho by Bradford & Dickens
 C O N T E N T S
PAGE
FOREWORD. By Professor E. K. RIDEAL ix
AUTHOR'S PREFACE xi

CHAPTER I. SOLUTIONS OF THE DIFFUSION

EQUATION 1

Differential Forms of Fick's Laws 1

Steady State of Flow 4
One Dimensional Diffusion in Infinite, Semi-infinite
and Finite Solids 7
Solutions of the Radial Diffusion Equation 28
Diffusion in Cylindrical Media 31
Diffusion Processes Coupled with Interface Reactions 37
Instantaneous Sources 43
Treatment of the Equation ^ - = ^ - 2 ) ^ - 47
A Note on the Derivation of Diffusion Constants from
Solutions of the Diffusion Equation 50
Some Cases for which there is no Solution of the Diffu-
sion Equation 50

CHAPTER II. S T A T I O N A R Y AND N O N - S T A T I O N A R Y

S T A T E S OF M O L E C U L A R F L O W IN C A P I L L A R Y
SYSTEMS 53
Types of Flow 53
Permeability Constants, Units, and Dimensions 60
Some Experimental Investigations of Gas Flow in
Capillaries 61
Flow of Gases through Porous Plates 65
Permeability of Refractories 69
Flow in Consolidated and Unconsolidated Sands 73
Flow through Miscellaneous Porous Systems 78
Separation of Gas Mixtures and Isotopes 78
Non-Stationary States of Capillary Gas Flow 82
VI CONTENTS

PAGE
CHAPTER III. GAS FLOW IN AND THROUGH
C R Y S T A L S AND G L A S S E S 91

Structures of some Silicates and Glasses 91

Diffusion of Water and Ammonia in Zeolites 96
Diffusion in Alkali Halide Crystals 108
Diffusion of Helium through Single Crystals of Ionic
Type 116
Permeability of Glasses to Gases 117
The Solubility of Gases in Silica and Diffusion Con-
stants within it 139

CHAPTER IV. GAS F L O W THROUGH METALS 144

Introduction 144
The Solubility of Gases in Metals 145
The Solubility of Gases in Alloys 158
The Measurement of Permeation Velocities 161
The Influence of Temperature upon Permeability 162
The Influence of Pressure on the Permeability of
Metals to Gases 169
Permeation Velocities at High Pressures 175
Some Mechanisms for the Process of Flow 178
The Behaviour of Hydrogen Isotopes in Diffusion and
Solution in Metals 183
The Influence of Phase Changes upon Permeability 191
The Influence of Pre-Treatment upon Permeability 192
Grain-Boundary and Lattice Permeation 197
Flow of Nascent Hydrogen through Metals 200

CHAPTER V. DIFFUSION OF G A S E S AND NON-

METALS IN METALS 207

Introduction 207
Measurement of Diffusion Constants in Metals 208
A Comparison of Methods of Measuring Diffusion
Constants 219
The Diffusion Constants of Various Elements in Metals 221
 CONTENTS Vll
PAGE
Degassing of Metals 226
Diffusion and Absorption of Gases in finely divided
Metals 230
The Influence of Impurity upon Diffusion Constants 234

CHAPTER VI. D I F F U S I O N OF I O N S IN I O N I C
CRYSTALS, AND THE I N T E R D I F F U S I O N OF
METALS 239
Introduction and Experimental 239
Types of Diffusion Gradient 245
The Structure of Real Crystals 247
Some Equilibrium Types of Disorder in Crystals 248
The Influence of Gas Pressure upon Conductivity 251
The Energy of Disorder in Crystals 254
The Influence of Temperature on Diffusion in Metals
and Conductivity in Salts 257
The Identity of the Current-carrying Ions 265
The Relation between Conductivity and Diffusion
Constants 268
Diffusion Constants in Metals and Ionic Lattices 272
Diffusion Anisotropy 276
The Influence of Concentration upon Diffusion Con-
stants in Alloys 279
Summary of Factors influencing Diffusion Constants 283
Models for Conductivity and Diffusion Processes in
Crystals 291

CHAPTERVII. S T R U C T U R E - S E N S I T I V E D I F F U S I O N 311
Types of irreversible Fault in real Crystals 311
Non-equilibrium Disorder in Crystals 313
Structure-sensitive Conductivity Processes 321
Structure-sensitive Diffusion Processes 327
CHAPTER VIII. M I G R A T I O N IN T H E S U R F A C E
L A Y E R OF S O L I D S 337
Introduction 337
Vlli CONTENTS
PAGE
Evidence of Mobility from Growth and Dissolution of
Crystals 337
Evidence of Mobility from the Condensation and
Aggregation of Metal Films 339
Measurements of Surface Migration in some Stable
Films 347
The Migration of other Film-forming Substances 368
A Comparison of the Data 370
The Variation of the Diffusion Constants with Surface
Concentration 371
Phase changes in Stable Monolayers 374
Calculation of the Surface Diffusion Constant 375
Applications of Surface Mobility in Physico-Chemical
Theory 377
CHAPTER IX. P E R M E A T I O N , S O L U T I O N AND D I F -
F U S I O N OF G A S E S I N O R G A N I C S O L I D S 382
Permeability Spectrum 382
Structures of Membrane-forming Substances 385
Permeability Constants of Groups A and B of Fig. 132 391
The Air Permeability of Group C of Fig. 132 406
The Solution and Diffusion of Gases in Elastic Polymers 411
Models for Diffusion in Rubber 422
CHAPTER X. P E R M E A T I O N OF V A P O U R S THROUGH,
AND D I F F U S I O N IN, O R G A N I C J S O L I D S 430
Water-Organic Membrane Diffusion Systems 430
The Permeability Constants to Water of Various
Membranes 438
Sorption Kinetics in Organic Solids 443
A Modified Diffusion Law for Sorption of Water by
Rubbers 445
The Passage of Vapours other than Water through
Membranes 447
Remarks on the Permeation and Diffusion Processes 448
AUTHOR INDEX 454
SUBJECT INDEX 460
 F O R E W O R D
The general theory of diffusion is based upon analogy to the
flow of heat through solid media, as is exemplified in the
classical treatments of Fourier and also of Lord Kelvin in the
Encyclopedia. In the actual process of diffusion of molecules
and ions through and in solids a whole set of new phenomena
is observed. For example, limitations on the magnitude of
the diffusion potential are much more frequently imposed on
these material systems by such factors as solubility or com-
pound formation than are observed in systems in which the
flow of heat alone is concerned. Again the flow of matter
through solids is frequently composite in character, the
various modes of transport being dependent on micro-
heterogeneity, as is exemplified by lattice diffusion and move-
ment along crystal boundaries, canals and capillaries, or even
on molecular discontinuity as is the case when migration
depends on the existence of molecular or ionic "holes" or
vacant lattice points in a crystal. In studying the movement
of material particles through a solid, we must consider the
nature of the interaction between diffusing material and
diffusion medium. We may in a somewhat general manner
note that either only dispersive or Van der Waals interactions
are involved, or that electronic switches have taken place
leading to chemi-sorption or chemical combination. Migra-
tion across a surface or through a solid medium may thus
involve movement across an energy barrier from one position
of minimum potential energy to another. At sufficiently high
temperatures in the higher energy levels activated surface
migration naturally merges into free migration, with a con-
sequent change in the temperature dependence of the
diffusion.
The mechanism of the transfer of material across phase
boundaries likewise presents a number of novel and interesting
problems. Here we have to consider firstly the abnormal
x FOREWORD
distribution of diffusing material at the phase boundary. We
note that the Donnan membrane equilibrium may apply to
electrolytic systems diffusing through a solid membrane, the
Gibbs relation for a non-electrolyte or the adsorption iso-
therm for a gas permeating a solid. Diffusion must then take
place into and out of the boundary layer from the homo-
geneous phases on both sides. Energies of activation are
involved which may differ considerably from those required
for the diffusion process in the homogeneous phases. Dr Barrer
has been interested in these problems for a number of years,
from the theoretical as well as from the experimental point of
view. I pointed out to him that they were the concern of
many who would find a monograph on the subject invaluable
in their work, and I experienced a deep sense of satisfaction
when my colleague assented to the suggestion of writing a
book.
E. K. RIDEAL

LABORATORY OF COLLOID SCIENCE

THE UNIVERSITY
CAMBRIDGE

12 February 1941
 A U T H O R ' S P R E F A C E
Diffusion processes are related to chemical kinetics on the one
hand, and to sorption and solution equilibria on the other.
There are available excellent surveys dealing with chemical
kinetics and also with sorption equilibria. No previous
text has attempted to correlate and summarise diffusion data
in condensed phases, save briefly, and in relation to one or other
of these fields. It is apparent that the study of diffusion touches
upon numerous aspects of physico-chemical research. There
are in general two states of flow by diffusion—the so-called
stationary and non-stationary states. From the former one
derives the permeability constant (quantity transferred/unit
time/unit area of unit thickness under a standard concen-
tration or pressure difference) and from the latter the diffusion
constant. The permeability constant, P, and the diffusion
constant, D, are related by

when dC/dx is a standard concentration gradient. One there-

fore has to do also with the solubility of the diffusing substance
in the solvent. The aim of this book has been to study the
permeability of materials to solutes, and the diffusion con-
stants of solutes within them, and not specifically the sorption
equilibria partly controlling the permeability. However,
wherever the permeability and diffusion constants are dis-
cussed it becomes essential to outline also the data on solubility,
and this has been done throughout the book.
In treating the data, the author has tried to keep a balance
between experimental methods and their mathematical and
physical interpretation; the listing of adequate numerical
values of permeability and diffusion constants which may
serve as reference material, and as starting points for further
investigations; and outlines of current theories of processes of
Xii INTRODUCTION
permeation, solution and diffusion in solids. Many problems
remain partially or completely unsolved, but if this book
directs attention to them, and produces further work in a very
fruitful field of research, it will adequately repay the time and
trouble involved in its preparation.
In Chapter i is given a number of solutions of the diffusion
equation suitable for treating the various diffusion problems
that may arise. The solutions are as explicit as possible, so that
they may be employed at once, or with the aid of tables of
Gauss or Bessel functions. The purpose of this chapter is to
provide ready-made integrals of the diffusion equation, thus
avoiding too long a search through a scattered literature for
suitable solutions. Not all experimentalists have the mathe-
matical training to derive at will solutions to suit the boundary
conditions of their particular problem. Chapter n is devoted
to a survey of the different types of gas flow in capillary systems,
from the study of which interesting methods of fractionating
gas mixtures have been evolved. The peculiar permeability of
some glasses to certain gases has been known for a con-
siderable time, and Chapter n i considers gas flow through
glasses, and in crystals. These diffusions are much more
selective than those considered in capillary systems, but they
are not specific as are the processes of diffusion of gases through
and in metals, discussed in Chapters iv and v. The uptake
and evolution of gases by metals is a subject of technical
importance concerning which a great body of experimental
evidence has been amassed. These chapters, however, show
that many problems remain unsolved.
Chapters vi and vn describe the phenomena of conductivit}7
and diffusion of ions and atoms in ionic lattices and metals.
The first of these chapters has to do with reversible diffusion
phenomena, and the second with irreversible diffusion pro-
cesses depending upon the past history of the system con-
cerned. Many of the problems discussed in Chapter vn have
to do with diffusion processes down grain boundaries or
internal surfaces, so that it is but a step to the treatment of
numerous and interesting surface diffusions, described in
 INTRODUCTION Xlll
Chapter VIII. Finally, in Chapters ix and x the problems of
gas and vapour flow through and diffusion in organic polymers
are considered. In these chapters a large number of perme-
ability and diffusion constants are collected for systems of
technical importance. As examples one may cite the passage
of gases and vapours through rubbers, proteins and celluloses,
and of water through leather, or insulating substances such
as vulcanite, ebonite, rubber, or guttapercha.

Among many colleagues I wish to thank Professor E. K.

Rideal, and especially Dr W. J. C. Orr, of the Colloid Science
Laboratory, Cambridge, who read the proofs and made many
valuable suggestions. I am glad to take this opportunity of
thanking the University Press for their painstaking work, and
numerous authors for permission to reproduce diagrams.
R.M.B.

March 1941
 CHAPTER I

SOLUTIONS OF T H E D I F F U S I O N EQUATION
DIFFERENTIAL FORMS OF FICK'S LAWS
No adequate compilation of those solutions of the diffusion
equation applicable to the diffusion of matter has as yet been
made. The author has become aware of the difficulty of
obtaining suitable solutions in a number of studies of diffusion,
and it is hoped that this chapter will provide a source of
reference for such solutions. Equations of chemical kinetics
can be used readily in differential or integral form, but the
equations of diffusion kinetics require for their treatment a
special and often laborious technique. Many of the cases which
can arise have not yet been solved rigorously, though the field
is a rich one both from the mathematical and the experimental
viewpoints. The present chapter aims at giving some solutions
of the diffusion equation in a form in which they may be applied,
together with cases of diffusion systems in which the boundary
conditions are those of the diffusion equation solved.
Two familiar differential forms of Fick's laws of diffusion
are: ^fi

dC
and Tt
Equation (1) gives the rate of permeation, in the steady state
of flow, through unit area of any medium, in terms of the
concentration gradient across the medium, and a constant
called the diffusion constant D. The second equation refers to
the accumulation of matter at a given point in a medium as
a function of time. That is, it refers to a non-stationary state
of flow. This second equation may be derived from the first,
by considering diffusion in the -f x direction of a cylinder of
unit cross-section. The accumulation of matter within an
element of volume dx bounded by two planes, I and 2, normal
2 SOLUTIONS OF THE DIFFUSION EQUATION

to the axis of the cylinder, dx apart, may be estimated as

follows:
The rate of accumulation is

+D
dx dx
which gives T t W
In two dimensions the above equation becomes
dC
^ ^ ^ , (3)
dt dx2 dy2
and in three dimensions
dC ^d2C ^d2C ^d*C
dt dx2 dy2 dz2
or ~ = DV2C (5)
dt
or ^ = D(divgrad)O, (6)

as it may alternatively be written. All the equations (3)-(6)

assume an isotropic medium, but if the medium is not isotropic
one may simply write

D
x
dt~ dx
and then make the substitution

to bring the equation to the form of (4):

dt -V\W W Wl'
This transformation involves then a change of co-ordinates,
after which methods suitable for the solution of (4) apply
also to (8).
In many cases the diffusion constant, D, depends itself upon
the concentration in the medium, e.g. the interdiffusion of
 SOLUTIONS OF THE DIFFUSION EQUATION 3
metals (Chap. VI), the diffusion of water in certain zeolites
(Chap. Ill), the surface diffusion of caesium, sodium, potassium,
or thorium on tungsten (Chap. VIII), or the diffusion of organic
vapours in media such as rubber which swell during the
permeation process (Chap. X). It is now necessary to solve
an equation of the form
dC_ i

Equation (4) may be expressed in spherical polar co-

ordinates r, 0, and 0, by means of the transformation
equations
x = r sin 6 cos <f>,
y = rsin#sin0, J- (10)
z = r cos 0.
Equation (4) then becomes

Equation (11), for a spherically symmetrical diffusion, re

duces to
dC JT&C 2dCl

for then dCjdd = 0 and d2C/d<f>2 = 0.

If one desires to express equation (4) in cylindrical co-
ordinates r, 6 and z, the transformation equations are

y = rsin#, j
so that the equation becomes
dC
\+d (ldC)+d (rdG

Equation (14) again takes simpler forms in certain cases. In

problems concerning the sorption and desorption of gases in
4 SOLUTIONS OF THE DIFFUSION EQUATION

and from long metal wires, for example, where end-effects

are small, the equation reduces to
dc _ri d

because dCjdd = 0, and d2C/dz2 = 0.

For examples of the use of the full equations (11) and (14)
text-books on the conduction of heat may be consulted (e.g.
Carslaw and Jaeger, Conduction of Heat in Solids, Oxford
University Press, 1947).
The equations (1)-(15) serve to show most of the differential
forms of the diffusion equation which may be encountered in
studies of diffusion kinetics.
. Various solutions taken from different fields of diffusion
kinetics may now be given. In order, the solutions given will
be for
A. The steady state of flow.
B. The non-stationary state of flow:
(1) Diffusion in one dimension—the infinite, semi-
infinite, and finite solid.
(2) Radial diffusion.
(3) Diffusion in cylinders.
(4) Surface reaction and diffusion simultaneously.
(5) Instantaneous point, surface, and volume sources.
(6) Diffusion when D depends on the concentration.
The solutions given cover a great variety of diffusion
systems, and with the aid of appropriate tables for error and
Bessel functions may be readily applied to practical problems.

A. THE STEADY STATE OF FLOW

The diffusion equation takes particularly simple forms when

the term dC/dt = 0, i.e. as much solute leaves a given volume
element as enters it per unit of time. The equations of flow then
become:
d2G
Through a plate: D ~ = 0. (16)
 THE STEADY STATE OF FLOW

Through a hollow spherical shell:

1 d2(rC)
r dr2 0. (17)

Through a cylindrical tube:

ldjrdC/dr)
= 0. (18)

The plate. Consider the diffusion of a gas through a plate

of thickness I. If the boundary conditions are
C = Cx at x = 0 for all t,
C = C2 at x = I for all J,
the solution C = ^4x+ B (19)
of equation (16) may be further elucidated by putting x = 0,
and a; = J, and eliminating ^4 and B. This procedure leads to
the expression n n
°-~ o i = ? (20)

where C denotes the steady state concentration at any value

of x. The flow through unit area of the plate is

where Cg denotes the concentration of the gas which has

diffused through the plate into a volume F a t a time t (the
condition C = C2 at x =• I for all I naturally makes it necessary
that Cg4, C2). Then the amount of gas which has diffused in
time t is f1 —P
VCg=D*^j^t. (22)

The hollow sphere., For the diffusion of a gas through a

hollow spherical shell of internal and external radii 6 and
a respectively, with boundary conditions
G = Ct at r = 6 for all t,
C = C2 at r = a for all J,
6 SOLUTIONS OF THE DIFFUSION EQUATION
the solution of (17) is
Cx-C a(r-b) (23)
Cj-Cg r(a-bf
where C is the steady state concentration at any value of r.
The outward flow of gas per unit time and per unit area of
the shell is then

Over the whole area of the shell the expression is

C2)^y (25)
rt
and the quantity diffused in time t, Q — geft? ie
Jo
~t. (26)

cylindrical tube. The solution of equation (18) is

C = A]nr + B,
and if the boundary conditions are
C = Cx at /• = 6 for all *,
C = C2 at r = a for all t,
where b and a are respectively the internal and external radii
of the tube, one may as for (20) and (23) eliminate A and B
and obtain n n n i r r* i
. (27)
lno-lna inb —
This leads to an expression for Q, the quantity which diffuses
through unit length of the wall of the cylinder in time t:
2nD{C2C)t^
v
* hia-ln6 '
The equations given will be found useful in evaluating the
permeability constants of various types of membrane. The
diffusion constants may also be evaluated when the absolute
values of Cx and C2 are known. The permeability constant
 THE STEADY STATE OF FLOW 7
may conveniently be defined as the quantity of gas diffusing
per unit of time through unit area of the outgoing surface of
a plate, cylinder, or hollow sphere when unit concentration
gradient exists at that surface. Other equivalent definitions
may be found more practical, and will be used in later
chapters (e.g. Chap. II). The permeability constant as defined
above is numerically equal to the diffusion constant. The
dimensions are, however, different.

B. THE NON-STATIONARY STATE OF FLOW

(1) Solutions of the linear diffusion law

Solutions of the Fick law for diffusion in the ^-direction
may be divided into those for infinite, semi-infinite, and finite
solids. The infinite solid extends to infinity in + and —x
directions, the semi-infinite solid extends from a bounding
plane a t # = O t o # = +oo, and the finite solid is bounded
by planes at x = 0, x = I and sometimes x = (1 + h). All these
solutions are exemplified by physical systems, which will be
illustrated in the text.
(a) The infinite solid.
We have to solve the equation
dC_ d*C
dt " dx2'
given that C = C(x91), when t > 0,
C =/(«), wheiU = 0,
f(x) can be differentiated when t > 0.
The solution may be written as
C = f{x)F(t), (29)
which allows one to separate the variables in Fick's law.
<j)(x) and F(t) take the forms
<j>(x) = A cos kx + B sin kxA . x
v (uU)
F(t) =akiDl j
8 SOLUTIONS OF THE DIFFUSION EQUATION
where A, B, a and k are real constants. Any sum of solutions
is also a solution, and a new solution is therefore obtained
when the equations (30) are integrated over all values of k:

C = f °° [g(k) cos kx + h(k) sin kx] e~k2Dtdk. (31)

Jo
In (31) g(k), h(k) are functions chosen to fulfil the condition
C = f(x) where t = 0, and Fourier showed that these functions
took the forms

rt*) = ; , _
(32)
' +1
h(k) = - f + °°/(a;') sin (*'&)<&:'.
7TJ - o o '
Thus when the equations (32) are substituted in (31) one has
as a general solution

7T J 0 J —oo

This equation, by an integration process, can be brought to

the form
1
u
f+Q0,,x ,xe x f ( ^ - ^ ) 2 1 axT , 4
~ ^~TZTnT\ J\ ) P AT\* • W)

Case 1. Suppose we have diffusion of a solute across a sharp

boundary at x — 0 from a solution into a solvent. The solute
may be a salt dissolved in water; it may be radioactive lead
dissolved in lead diffusing into pure lead: or it may be an ion
in an ionic lattice diffusing into another lattice, e.g. silver
diffusing from silver sulphide into copper sulphide, while
copper passes in the opposite direction. For the equation (34)
to hold we know that D must not be a function of C; and that
the amounts of solution and solvent must be great enough,
the diffusion slow enough, or the time short enough, so that
no appreciable amount of solute diffuses from the far extremity
of the solution, or reaches the far extremity of the solvent
(Fig. 1). In these systems, at t = 0, C = Co for x< 0, (7 = 0
 THE NON-STATIONARY STATE OF FLOW 9
for x>0. For all positive values of x, the concentration at
time t and at a point x becomes

exp - v ' \dx' (35)

Co
(36)
*/2V(«)
(37)

0 1 2 3 4 5 6 7 ^ 9 10

-10 -9 -3 -7 -6 -5 -4 -3 -2 -1 0

Distance from Boundary

Fig. 1. Diffusion across a boundary from a solution into a solvent.

The second term is the Gaussian error function, erf (y), and
may be evaluated from mathematical tables. Equation (37)
may be expanded as a series:

~" 2 L V^ I 2 yl(Dt)"~ 3.1! {2 <J(Dt)}* + \

(38)
7.
The solution for x < 0 is correspondingly
=
~2

a:5

__^ ?+ n (39)
10 SOLUTIONS OF THE DIFFUSION EQUATION

Case 2. I t is necessary to have a readily applicable solution of

Fick's law for systems in which two diffusion media are present.
In many instances, when metals interdiffuse for example, a
number of solid phases occur, as alloys, with sharp phase
boundaries. This is true when molybdenum diffuses into
iron(i) at a temperature of 1300° C , for there is a diffusion of
molybdenum dissolved in a molybdenum-iron alloy formed in
the outer layer of the iron, and a diffusion of molybdenum also
in the purer iron at the core of the rod. This diffusion system
is complicated further by a movement inwards of the alloy-iron
phase boundary. The simpler case of a stationary phase
boundary can be easily treated by the methods already
outlined (2).
The boundary conditions are

(i) = Dl*£ for the first solvent; for

dx*
the second solvent.

Z2 ^ ^ 7" 0 ^
Fig. 2. Diffusion in two phases I and II where k = 3 and D1 = /) 2 .
\f5F-0

-5 -) 0 / " 2
Fig. 3. Dift'usion in two phasrs I and 11 where k — J and Dl = D2.
 THE NON-STATIONARY STATE OF FLOW 11
(ii) Cx = Co for t = 0 and x < 0; C2 = 0 for t = 0 and x > 0.

= ZU-—I at the interface between the

solvents for all t.
/»0 /•»
( iv ) (Co - ci)dx = ^2^=the total
amount diffused.
J-oo J0

(v) i = ^k, the partition coefficient, in the equilibrium

state.
The solutions of the problem for x > 0 and x < 0 are
2
(40)

-J -2 -1 o • 1 z—~*
Fig. 4. Diffusion in I and II where k = 1 and X^ = 4Z>,.

Fig. 5. Diffusion in two phases I and II where k — 3, Dx — 4D2, and for the curves
jD2t) = 1, V W ) = 2» w n i l e for t h e
curves (b) <J(D2t) = 0-5, V(A0 = !•
12 SOLUTIONS OF THE DIFFUSION EQUATION

t ic I'D I 2 Cxl2y/{Dt)
2

Some solutions of the equations (40) and (41) are illustrated

graphically(3) in Figs. 2-5.

(b) The semi-infinite solid.

The medium now extends from x = 0 to x — + oo, and the
general equation takes the form

Case 1. At the plane x = 0 one may generate a supply of

solute, in which the rate of supply is any function of time.
Instances where the supply is constant will be of the greatest
practical significance. Examples would be the sorption of
gas at constant pressure in a large amount of a solid * (oxygen
in silver, hydrogen in palladium, nitrogen in iron, ammonia
in analcite) or a crystal in contact with a quiescent solvent
liquid.
When the supply at the plane x = 0 is constant the boundary
conditions are
C = Co at x = 0 for all t,
C = 0 at x > 0 and t = 0,
C = C(x, t) at x > 0 and t > 0,
and the solution giving the concentration C at a point x and
a time t is

or —
— a*
where y =-

* Neglecting for the moment the possibility of a rate controlling process at

the interface gaa-solid.
 THE NON-STATIONARY STATE OF FLOW 13
(c) Finite solid.
We have been concerned hitherto with so much solvent that
its amount for practical computations of D can be reckoned
as infinite. Such a supposition limits the applicability of the
equations given, for it will be much more usual to work with
small amounts of diffusion media. The solutions now to be
given will concern themselves with problems such as the
following. A solute diffuses from a solution bounded between
the planes x = 0, x = h into a solvent bounded between the
planes x = h and x = I. The concentration-distance-time
curves may be measured readily enough and have now to be
interpreted so that the diffusion constants, D, may be
evaluated. The new solutions of Fick's law will apply to
numerous cases of the interdiffusion of metals, and salts, so
long as D does not depend on the concentration, and whenever
the amount of metal or salt is limited.
Another group of solutions will be given for the problems
involving diffusion into, through, and out of a slab bounded
by the planes x = 0 and x — Z. This group has to do with
the diffusion of gases or liquids through membranes, organic
(rubber, cellulose, gelatin, plastics) or inorganic (metals,
crystals, and glasses). It is necessary in applying the solu-
tions that the surface processes should be sufficiently rapid
not to interfere with the internal diffusion process. This
is true when gases diffuse through organic membranes, or
inorganic glasses and crystals, but it is not true of gas-metal
systems such as H2-palladium, to which solutions of the
diffusion equation must be applied with caution.
The general solution for diffusion within parallel boundaries
which obeys the conditions

(i) ^Tt=D^> (ii

) at
' = °> C=f(x) for
(iii) G = 0 at x = 0 and x = I for all t,
may be written as follows:
O = f sV<»»">'-sin^f /(*')sin^*'. (45)
* n=l ^ JO *
14 SOLUTIONS OF THE DIFFUSION EQUATION

Case 1. Diffusion from one layer to another may be treated

by regarding the system as being a single layer with im-
permeable boundaries in which the distribution at time t = 0
is as follows:
f(x) = Co for 0 < x < h,
f(a>) = 0 for h < x < I, (l> h).
The solution for this particular case is
O - C o (^£ V < » ^ c o S ^ i n ^ ) . (46)
For any numerical ratio of h/l, equation (46) may now readily
be expanded as a series, each term of which may be given a
numerical value. These numerical values have been worked
out when \l = h by Stefan(4) and Kawalki(5) whose tables
may be consulted.
Case 2. For some purposes it may be necessary to evaluate
D from measurements as a function of time of the total solute
which has diffused across the boundary. This quantity, Q, is
given by the equation
(47)
JXSL*-
From equation (46) one has, at x = h,

Thence, substituting (48) in (47) and integrating,

n I " 1 . 2nnhf, I Dn27rH\l

Equation (49) may be evaluated for various ratios of hjl. For

instance, when hjl = J, equation (49) reduces to
8 Z)(2m+1)2
V-J_-exD! ^ l l (50)
which may again be expanded as a series and readily employed.
The various possibilities which arise when a solute at a
constant concentration diffuses into a slab may now be con-
 THE NON-STATIONARY STATE OF FLOW 15
sidered. The slab, as before, is bounded by the planes x = 0
and x = I, and the boundary conditions are
C = Cx at x = 0 for all t,
C = C2 at x = I for all J,
C = f(x) at £ = 0 for 0 < x < L.
The general solution of this problem is (6)
2
(71 . nnx F
C0T + S s i n'sin ex [ Dn nHl
J
ySsin-y-exp ^— f{x')sin—j-dx'. (51)

3. When the slab is initially free of solute, and the

concentrations of solute at the faces are Cx (at x = 0) and
C2 (at x = I), one has the example of diffusion of gas at
constant pressure into a membrane of solid free of gas. Here
f (x') = 0 and the solution is

(52)
If instead of the membrane being initially free of gas there is
an initial uniform concentration, Co, of gas in the membrane,
the solution is

. (53)
TT m r 0 (2rn+l) Z
Again, if C^ = 0 for all t, and O0 = 0 at t = 0, the solution is

Also, when Cx = C2 for all ^, and Co = 0 at ^ = 0, the solution

becomes
4 l
C c\\ y zm{2m+1)nZcxJ, '
1 ^m?o2m+l Sm I exp
\ » (55)
16 SOLUTIONS OF THE DIFFUSION EQUATION
The reader may work out other examples from the general
equation (51), simply by giving Cl9 C2 and f(x') their appropriate
initial values. Special cases of equations (52)-(55) follow with
equal ease. For instance, in equation (55) the concentration
at the mid-plane of the slab (x = \l) is given by putting
x = U and so getting

All the equations (52)-(56) have to do with the flow of gas or

other solute into a membrane. The use of these equations
presents no difficulty. They may be expanded term by term,
and all terms save the first two or three can be neglected, be-
cause the series converges rapidly. In this form the equations
become of special interest in the field of sorption kinetics.
Case 4. The desorption of gas from a membrane is an
equally important example of the application of equation (51).
Suppose at the time t = 0 the membrane contains a uniform
concentration Co of solute throughout, from x = 0 to x — I and
Cx = C2 = 0. The solution in this case by substitution in (51)
becomes

C = -— S s m - ^ e x p -=— sin — dx
l
l I I L J J o *<
4C0 °° 1 . (2m+l)7rx f D(2m+l)27T2tl
y V^ _ oifi ^ 1 own I 2 I

-Vw?o(2m+T)8in I eXP
L I2 I
(57)
At the midpoint of the membrane (x = |Z) equation (57)
reduces to

Should the gas desorb from the slab not into a vacuum, as in
(58), but into a gas atmosphere where at x = 0 and x — I the
concentration is Cx, the solution of (58) becomes

C - ^ + (60-^)- ^ _ ^ _ - _ e X p ^ ? J.
(59)
 THE NON-STATIONARY STATE OF FLOW 17

Case 5. In the study of sorption kinetics one frequently

requires an expression for the amount of gas (or other solute)
left in a membrane at any time t, or alternatively, for the
quantity which has diffused out of the membrane. The total
amount of sorbed gas at the time t is given by the expression

1== [lC{x,t)dx
Jo
(2m+l)nx [~ D(2m +
I eXp
L T*
(60)
When the integration is carried out one gets

Corresponding to equation (61), the amount of gas which has

been desorbed is

The corresponding problem of the quantity of gas which lias

been sorbed in the membrane after a given time t appears
from either of the two relationships

when the boundary conditions in the most general practical

case are those which led to equation (53). The solution when
Co = 0 is simply equation (98), if u1 and uz are the constant
concentrations at x — 0 and x = I respectively. The amount
dissolved for the general case (concentrations Cx at x — 0, C2 at
x = I, and C = Co at t = 0) is

(63)
n°-»£o (2m + 1)2 M P jj'
18 SOLUTIONS OF THE DIFFUSION EQUATION

Case 6. If one is measuring the rate of flow of a gas (or any

other solute) through a membrane in which the gas dissolves,
there will be an interval from the moment the gas comes into
contact with the membrane until it emerges at a constant rate
on the other side. By analysing stationary and non-stationary
states of flow it is possible to measure the diffusion constant,
the permeability constant, and the solubility of the gas in
the membrane. Once more one may employ equation (53) to
determine the intercept (L), in terms of D, I, and C, which
the pressure-time curve makes on the axis of time.
The boundary conditions are
C = Cx at x = 0 for all t,
C = C2 at x = I for all t,
C = Co for 0<x<l and at t = 0,
giving as solution
Ci . nnx F Dn2nH~
ism—exp|
4C0 " 1 . (2m+l)nx T D(2m
L (53)
Equation (53) may be differentiated, and one obtains at x = 0
C2-C1 2

. (64)

But if the gas flows through a membrane into a volume V,

the flow of gas is given by

Substitute (64) in (65) and integrate between the limits 0 and t,

and so obtain
r nc*-ci* + 2l v

J)- (66)
 THE NON-STATIONARY STATE OF FLOW 19
Equation (53) as £->oo approaches the line

Had there been no time lag, Gg would be expressed by

C9 = ffWt-Ci)t, (68)
so that the intercept, L, on the time axis is given by
L i rq*1, w ojn
(C2-C1)l6D+'3D~^Dj' (69)

In the case where Co = 0 at t = 0

and where Co = 0 and Cx is very small (~ 0)

^ =~ . (696)
In equations (69), (69a) and (696), one is provided with an
easy means of measuring D.

(d) Finite solids with surface concentration a function of time.

Case 1. In many experiments, e.g. the uptake of gases by
solids, the surface concentration is a function of time if
sorption occurs at constant volume or variable pressure. The
following considerations will apply to systems such as O2-Ag,
H2-Cu, N2-Mo, NH3-, H2O-zeolite, where sorption occurs
in a sheet of metal or crystal, provided there is no rate
controlling interface reaction. Then the boundary conditions
are for the most general case:
C = <j)x(t) a t x = 0,
C = <f>2(t) a,tx = l,
C = f(x) when t = 0,
20 SOLUTIONS OF THE DIFFUSION EQUATION

and the solution becomes (7)

n 2« r Dn2nHl . nnxl f* .

^ ^

When the slab is initially gas free and ^(t') — ^ 2 (O ^ e con-

centration becomes
n _± ™ . (2m+l)7Tx(2m+l)Dn
L> — 7 2J sin z j

x
Jo exp L— ^—-J#(*)*,
and Q, the amount sorbed in unit area at time t, is

Q=\ C(x,t)dx
Jo
LrSwaai*. (72)
^ m=0j0
If in the interval t' = Ototit is possible to represent <p(t') by

where Co is the initial surface concentration and A and B are

constants, one obtains
00 1

^4 2J5^ 2 5 / A 2B\ ."] 7

- + 2- + I1 + -2-)e~ai \, 72a)
a a a2 \ a a2 J J v ;
where a=

It is possible, however, to solve the diffusion equation

completely for this type of sorption problem, without recourse
to empirical expressions for <j){tr), as the method of case (2)
will show.
Case 2. If one has a plate enclosed between the planes x — 0
and x = I with the face x = 0 impermeable, and an initial
 THE NON-STATIONARY STATE OF FLOW 21

concentration Co of solute in it; and if another plate enclosed

between the planes x = I and x = h -f I is made of the same
material with an initial concentration zero, then the solution
of the problem of flow from one plate into the other is given
by equation (46).
Suppose, however, that the solute in the plate between
x = I and x = h +1 could be kept at the same concentration
throughout by some process of stirring. The concentration in
this plate would then be a function of time only. The diffusion
equation in this instance is to be solved for the important case
of desorption from a slab into a constant gaseous volume.
Thus the desorption of water, ammonia, or another gas from
a plate of a zeolitic crystal would be examples of diffusion
systems of the kind postulated, save that the distribution
_ . C (in the solid) ,
coefficient k = --pf—. = -^ is not unity. Another example
C (m the gas)
where k = 1 is the diffusion of urea from a layer of aqueous
gel into a stirred aqueous layer.
The adaptation of the solution which will be obtained for
the case where k ^ 1 is very simply made, by supposing the
actual plate between x — I and x = (I + h) replaced by a plate
between x = I and x = (I + h/k) in which the partition coefficient
is now unity.
The outline of the solution for this interesting system may
now be given. The equation — = D -^ has to be solved for
the boundary conditions
(i) dC/dt = 0 at x = 0, t > 0,
(ii) C =f(x) for 0 < z < Z a n d * = 0,
(iii) contact between solid and gas gives at x = I
C(l,t) = kCg(t)t
where k is the distribution coefficient and Cg(t) denotes the
concentration at time t in the space between x = I and x — I + h.
First take k = 1 and write the solution as the sum of two
terms
C=
 THE NON-STATIONARY STATE OF FLOW 23
and this value of A must be used in evaluating Bt and fit\ and
also Cg(oo) in equation (74) becomes

c
The constants pi and Bt are obtained in terms of D by means
of the equations (77)-(81). The /?/s are found from the positive
roots of the transcendental equation
tanz + Az = O, (77)
which are easily obtained graphically by plotting the curves
y = — Az and y = tanz. The solutions are the points of inter-
section of the curves.
One may now evaluate the constants Bi} fiv In equation (74)

& = -»f2, (78)

A==
when ZF
The constants Bt may be evaluated from the relationships

B>
- <80)
and the equations
B'o = -0-099A + 0-644 -.
B[ = -0-146A + 0-894-ei,
B'2 = - 0-157A + 0-946 - e 2 ,
(81)
£3 = - 0-159A + 0-958 - e 3 ,
and for all B't where i > 3
B\ = -0-162A + 0-972-e 3 .,
The quantities et are then read from Fig. 6 for all values of
A between 0 and 5. Then in the final expression D disappears
from the ratio EBJfi^
One now sees that it is easily possible to evaluate D, the
diffusion constant, from desorption experiments at constant
volume, provided 1 < A < 5, and that no rate controlling inter-
face reaction intervenes. The advantage of this method over
the method outlined in equations (72)-(72a) is that one does
24 SOLUTIONS OF THE DIFFUSION EQUATION

not need to represent a <j)(t) — t curve by means of an empirical

equation, in order to obtain an expression for Q, the amount
sorbed by the plate. Accordingly, the method of March and
Weaver (8) has been treated fully as far as practical details
for determining the constants in the solution of the diffusion
equation are concerned. This method, however, applies only
to systems where the diffusing molecule exists in the same
molecular state in both phases. For instance, it cannot be
applied to systems such as H2-Pd, where in the gas phase one
has molecules, and in the solid phase, atoms. The earlier
method of equations (72)-(72a) is, however, still available.

03

I
f i

w s sVs
Wf
uFA
01 # s,

1I 1 1
i "7 /
1 \mj \
11 m/
F L
s
\m s
ooKrJ#1J 3
Fig. 6.

Case 3. It is easy to set up a diffusion cell so that one has

a porous membrane separating two well-stirred solutions, or
two gas volumes. The gas or solution on either side of the
membrane contains the same constituents. This system is an
important one biologically, as well as chemically. The mem-
brane may be organic or inorganic if it separates two stirred
solutions or two gas phases; or it may also be a crystal if it
separates two gas phases only.
 THE NON-STATIONARY STATE OF FLOW 25
The exact interpretation of the data to give diffusion
constants may now be given (9). The general method of solution
is similar to that of March and Weaver given in the previous
section. One has a membrane between the planes x = 0 and
x = I; and well-stirred solutions of concentrations gC0, hC0 are
in contact with the planes x = 0 and x = I respectively. The
solution in contact with the plane x = 0 extends from x = 0
to x = — g\ the solution in contact with x — I extends from
x = Z to x = l + h. Then the boundary conditions for which
the equation
d_C__ dC
dt - dx2
is to be solved are
(i) C = G0(x) at t = Ofor
(ii) G(O,t) = gC(t); C(l,t) = hC(t) for t>0.
(iii) C = gC0a,tt = OforO<x< -g,a,ndC = hCotitt = Ofor
I <x<l + h.
D dC
Vi
(
' dt g\dx)x^ dt
For this system Barnes (9) found solutions in the form
A.
(82)

(83)
Si
where C^ denotes the uniform final concentration in the
solution or gas phases. When the distribution coefficient
between membrane and solution is unity,

C^g + l + h) = (gC0)g+ Cco(x)dx + (hCo)h. (84)

Jo
Also, where t — 0,

U Si
26 SOLUTIONS OF THE DIFFUSION EQUATION

There are three more relationships available for determining

the A's, JB'S and £'s which are

/I

(86)

and

(87)
)'

where zt = ~ - . From equations (86) one may form the

equations
- cotz,/z,. >
Ax — cot zJzS (88)

and z2— (Ai + Agjzcotz — A1A2 = 0. (89)

By finding the points of intersection of the graphs y = z2 — AXA2

and y= (A1 + A2)zcotz, one can readily determine the z's,
for which real positive values only are taken. From the infinite
set of linear equations (87), whose form is greatly simplified
when C0(x) = Co (i.e. when there is a uniform concentration
within the membrane at t — 0), and (88) and (85), one may
determine Bi and AL in terms of zt and D. In the ratios
ZAJ£i9 EBil^i which appear before the exponential terms of
(82) and (83), the D cancels, since it is a simple multiplier
in At, Bt and £,t. As with all diffusion problems, at large values
of t, all the exponential terms save the first become small,
so that if one plots log(gC—G^) against t one obtains a
curve such as that in Fig. 7, in which the curve approaches
asymptotically a straight line of slope — £t. In any case
even at small values of t it will not be necessary to take many
terms of the exponential series, in solving for Ai and Bt.
 THE NON-STATIONARY STATE OF FLOW 27
The complete solutions of (82) and (83) for the case when
g = h (i.e. ^ = A2 = A, k = 1, C0(x) = 0 = gG0) are

>t 1 2
1
1 1
1

0-8

0-6 -
\

\
0-4
Fig. 7. A typical diffusion rate curve of log (gC - C*>) against time.
The units of ordinates and abcissae are arbitrary.

which reduce as illustrated in Fig. 7 for small values of A and

large t to

(90a)

(91a)
28 SOLUTIONS OF THE DIFFUSION EQUATION

When the distribution coefficient of solute between the

membrane and the solvent is not unity, but k = -77-1 ,
J
Cm gas
the problem can be treated identically by writing
1
W
h (92)
lk\
and using for C^ the equation (93) instead of (84),

CJg + kl + h) = (gC0) g+C C0(x) dx + (,,C0) A. (93)

Jo
If the diffusing solute exists in the membrane in atomic
form, and in the gas phase (or in solution) in molecular form
(e.g. H2-Pd), then the treatment of Barnes must be further
modified.

(2) Solutions of the radial diffusion equation

We may restrict ourselves to considerations of difiFusions such
that the spherical surfaces of constant concentration are con-
centric; in this case the equation of diffusion is (12):

(12)

Examples of diffusion in spheres may then be treated in a very

simple manner by making the substitution
u = Cr,
,. , . du n 8 2 w
which gives -5- = D^-^. (94)
ot or*
The methods used in the previous pages for all the linear cases
may then be employed, and analogous solutions obtained.
Some examples of these solutions of the diffusion equation
will now be given.
Suppose one has a sphere of radius a, and containing solute
at an initial concentration O — f(r). The surface of the sphere
is kept at a constant concentration C2. The substitution
 THE NON-STATIONARY STATE OF FLOW 29
u = Cr in equation (12) leads to equation (94) which must be
solved for the boundary conditions:
ux = 0 at r = 0 for all t,
u2 = aC2 at r = a for all t,
u = rf(r) at t = 0 and for 0 < r < a.
These are the conditions for diffusion in a plate of thickness I
and with surface concentrations at x = 0 and # = Z of 0 and
aC2. The solution is therefore (51) in the form (51a)
u»r 2 ™Uo cos nn ; rarr f Dn2n2f]
-1-+-Yi— sin — exp —
2
a 7T i n a |_ a j

Case 1. The amount of absorption or desorption in spheres

can easily be derived from (51 a) for some important examples.
For instance, when f(r) = Co throughout the sphere at t = 0,
equation (51a) becomes

Equation (95) corresponds to absorption with C2>CQ, and

desorption with Co > 02, and from it by means of the equation

the quantity Q absorbed or desorbed per unit area may be

found. The necessary differentiation and integration leads
to (96): )a( 6 £ 1 f" Dn27T2f]\
a
(96)
3 \ IT* n==\n* L J/
2. In the problem of diffusion into or out of the wall
of a hollow spherical shell of inside radius b and outside
radius a, one may proceed analogously. The solution of
du/dt = Dd2u/dr2 is to be found for the boundary conditions:
u = ux = bCx at r — b for all t,
u = u2 — aC2 at r = a for all t,
and Co = 0 for b<r<a and at £ = 0.
30 SOLUTIONS OF THE DIFFUSION EQUATION

When one also makes the substitution r = b + x, the problem

reduces to the case of flow into a plate of thickness I = (a — 6),
and with constant concentrations ux and u2 at the faces x = 0
and x = L Thus the solution is that given by equation (51),
with appropriate substitutions:
.x 2™UoCosnn — u, . nnx r — Dn2n2tl
u = u^^-u,)-^ -S-*— ^m —exp^—-—J,
(526)
where ux — bCx at x = 0, u2 = aC2 at x = (a — 6) = I, x = r — 6,
^ = r/(r) = 0 at t = 0 (6 < r < a).
The quantity Q which has flowed into the spherical shell at
any time t is given by

(98)
Equation (98) for the case when C2 = C^ reduces to

1 ^ - ^ - ) j . (99)

The equations 526 and 98 and 99 are useful both to show

the distribution of gases or vapours in shells of plastics or of
metals, and also the total quantities sorbed, under the given
boundary conditions. It is necessary for their quantitative
application that the diffusion coefficient D should not depend
on the concentration of the diffusing species, and to realize
that in some cases this condition is not fulfilled (Chap. X,
p. 443). Further solutions involving spherical shells have been
given.by Barrer (9a).
 THE NON-STATIONARY STATE OF FLOW 31
Case 3. The intercept L upon the £-axis of the curve, where
Cg (the concentration built up in the gas phase by permeation
of gas through the hollow spherical shell of Case 2) is plotted
against t, is given by the theory of equations (63)-(696). One
has only to make the substitutions of Case 2 in equation 526,
to determine [^I and to evaluate —4na2D l^\ dt, in
order to apply this theory to obtain L:
Cx (q-6)« G2 a(a-br ( .
(Cx-Ctf 6D + ( 6 \ - C 2 ) - UD " {
'
When in addition C = Co between 6 < r < a (9 a)
p g
T = (( p ii - o ) (a-6) 2 ( g 2 - Q a(a-6) 2 .

In equation (101) when C2 = Co, or in equation (100) when

C2 ~ 0 (which will be the case if diffusion occurs into a vacuum
or near vacuum on the outgoing side of the spherical shell), the
lag L reduces to , _ , v2
L=4^, (102)
which is just the same as for a plate. The equations (101) and
(102) provide a very easy method of evaluating Z>, the diffusion
constant.

(3) Diffusion in cylindrical media

Examples of diffusion problems in wires are fairly common.
Of practical importance is the outgassing of metal filaments,
and the converse process of sorption in wires. It is necessary
to know how much gas remains in the filaments, or how much
has diffused away under vacuum conditions. The diffusion of
metals such as thorium into or out of tungsten filaments has
a profound effect upon the thermionic and photoelectric
emission of the filament, and this in its turn is of importance
in various types of valve. The quantitative expression for
the flow of thorium can be obtained by integration of the
differential equation for flow in a cylinder, using appropriate
boundary conditions. Sometimes a supply of a solute may be
32 SOLUTIONS OF THE DIFFUSION EQUATION

generated chemically or physically at the surface of a wire at

a concentration which is a function of time. This provides a
more complex example of diffusion into a cylinder. For
example, a supply of carbon may be generated by the de-
composition of hydrocarbons at the surface of an iron wire and
the carbon may then diffuse into the wire. Often analogous
problems will arise in which gases are sorbed in, desorbed from,
or diffuse through the walls of cylindrical tubes.
The equation for radial flow in a cylinder is

and by making the substitution C — ue~DctH, the equation (15)

is transformed to
+ oiu = O, (103)
v
dr r dr '
which is Bessel's equation of zero order. Solutions of Bessel's
equation may be obtained in terms of the appropriate Bessel
functions whose choice is governed by the boundary conditions.
Case 1. A circular cylinder of radius r = a is the diffusion
medium, at its surface a constant concentration Ct is main-
tained, and the medium is initially free of solute. In this
example the solution may be given in terms of Bessel's function
of the first kind and of zero order J0(x) and its differential
Jf0(x). The solution is

i±?^\)
« i ocnJo(ana)
(104)
where an is the nth root of the equation J0(<xna) = 0. The first
four roots of Jo(ocna) = 0 are
2-405 5-520 8-654 11-7915 , ^
«i = - — , oc2 = -—, a 3 = - — , a4 = — . (105)
II Hi U> U/

These roots give four exponential terms in an infinite series,

and it will be found as a rule that these terms are adequate to
express the diffusion process. Indeed, for larger values of the
 THE NON-STATIONARY STATE OF FLOW 33
time, t, a single term will be sufficient. The functions J0(x)
and J'0(x) are given by the series

(106)

and their values for any values of x are given in tables.*

For most purposes it will be more important to know the
mean concentration in the cylinder, or alternatively the
quantity Q which has diffused into the cylinder, per unit
length. C, the mean concentration in the cylinder, is given by

C = — .2 -iTtrCdr^-A CCrdr
TO Jo a2 Jo
(107)
and Q = 2n I Crdr, per unit length.

The integration of (107) in which (104) has been substituted

leads to the equations

(108)

where the first four values of an are given by equation (105).

Equations (108) follow easily from the relation
ra a
- rJo(ocnr)dr = —J'0(ana)9
Jo ocn
and by substitution of (105) lead to a value for the mean
concentration C in the filament:

i)(8-654)2q \
x exp{ — j (8-654) exp
2 - a* ) — • ; •
(108a)
* References to suitable tables are given at the end of this Chapter.
BD 3
34 SOLUTIONS OF THE DIFFUSION EQUATION

Case 2. If the cylindrical diffusion medium is of radius a and

the boundary conditions are
Cx = 0 at r = a for all t,
C = f(r) for a < r < 0 at t = 0,
the general solution is
2 . \\f(r)J0{*nr)dr

When /(r) = Co, the solution is

C = - * *£ 1 e - ^ ^ ^ 1 . ( i09a)
a xan /(aa)
By using the relations
C = ~ f a Crdr and Q = 7Ta2C0-2n("Crdr, (107a)
a* Jo Jo
where Q is now the quantity of solute which has diffused out
of the cylindrical medium per unit length, one obtains the
solutions for C and Q just as for Case 1, equations (108):

(110)

Since the a s are given by (105) it is an easy matter to write

the first four terms of equations (110), and so to employ them
in practice.
Case 3. If in the cylindrical medium of radius a the boundary
conditions are
C = C1 at r = a for all t,
C = f(r) for a < r < 0 and t = 0,
the general solution is

2- , J/ttr)M«nr)dr
 THE NON-STATIONARY STATE OF FLOW 35
lff(r) = Co for a < r < 0 and at t = 0 (which would correspond
to the sorption equilibrium of a gas in a wire before admitting
another dose of gas at constant concentration C^), equation
(111) reduces to

In turn, equation (112) gives for the mean concentration C

in the cylinder, or the quantity Q of gas which has diffused
into or out of the cylinder per unit length,
V __ Q __ Q

where the values of a1? a2, a 3 and a4 are given by (105), and
permit four exponential terms of the series to be used.
Case 4. If the surface concentration of solute is C = At,
where A is a constant, and there is an initial solute concen-
tration Co in the cylinder of radius a at t = 0, the general
solution is(96)

Z>L* V a?ot%Jo(ana)

)
and so one finds for Q and C the expression

=C1=
(115)
5. If one has a hollow cylinder of external and internal
radii 6 and a respectively, the initial concentration in the
cylinder is f(r) for a<r<b9 and the concentrations at the
surfaces at r = b and r = a are zero for all time t, the general
solution for the concentration as a function of time(io) is

)P
Jo
(116)
3-2
36 SOLUTIONS OF THE DIFFUSION EQUATION

In this equation the a n 's are the roots of the equation

Jo(aa)H^(ab)-Jo(ab)HP(oca) = 0,
f)
in which H& (ocr) denotes a Bessel function of the third kind,
for U0{ar) = J0(oir)H%\od))-J0(ab)Htf(ar), and with the
above meaning of <xn, U0(ana) = UQ(anb) = 0, satisfying the
boundary condition that at r = a and r — b the concentration
is zero.
When/(r) = Co, equation (116) reduces* to

\anr)

L
From this equation by means of the relation
Q = n{b2 - a2) Co - 2/r C Crdr
Ja
one may find Q, the amount of material which has diffused
out of unit length of the walls of the cylindrical tube, as a
function of time,

(118)
The roots of the equation U0(anb) — 0 = U0(ana) are com-
puted from the expression (ii)
nn (p-1) riOOQi)»-l) 1 l(p-l)3
n + +
* (p-1) 8p(nn) ]
55! )
r 32(1073)(p - +
3 ( ) ( p 5 0 ( ^ - 1 )
L
L 5(8p)Hp-
* This integration makes use of the relationships

Ul{anr)=-l!p&,
1V n
' o(anr)
rdU0(ccnrn _^J}
L dr Jr=b n'
a/* Jr=a
the first of which will be found in Watson <ioa) and the others in Carslaw(io).
 THE NON-STATIONARY STATE OF FLOW 37
where p — b/a. The values of Jo(ocnb) and J0(<xna) may then
be found from tables of Bessel functions.
Case 6. If one has a hollow cylinder of internal and external
radii a and b respectively, and there is a concentration C1 at
r = a and C2 at r = b, and a concentration Co in the wall of
.the cylinder, one may treat equation (116) in the same way
as was done in equations (63)-(696) to find the rate of approach
to the steady state of flow through the wall of the cylinder.
The intercept L made on the time axis by the asymptote to
the curve of the quantity diffused plotted against the time is
now given by
n*a log a/6 » Jl(ocna)
L
~ 2D(C^Q f Jl(anb)- Jl(ana)^o^]r=a, (119)
if
rb r tin .
= rUo(ocnr) C 0 - p -
Ja LI loga-logft
when Co — 0, Cx = 0 the most important experimental case
arises, and the equation above reduces to *
2 ° ^ J0(ccna)J0(ocnb) b

where an is as before the nth positive root of

Uo(ocnb) = 0 = U0(ana),
and one has once more a useful method of measuring the
diffusion constant (cf. equations (69), (69a), (696) for plates,
and equation (101) for a hollow spherical shell).

(4) Diffusion processes coupled with interface reactions

Sometimes there are slow processes at the surface of the
diffusion medium which may sensibly alter the rate at which
the diffusing substance leaves or enters the medium. For
instance, under certain conditions the diffusion of hydrogen in
palladium is fast compared with its rate of entry into the solid
from the sorption layer. In this case it will be necessary to
* Jaeger ( l l a ) has recently shown that the eq. (120) is precisely equivalent
to the result: T« « , / 2 , i 2 \ i /».
4:D In alb
making the time lag method for the hollow cylinder very simple in application.
38 SOLUTIONS OF THE DIFFUSION EQUATION
include a term to allow for the leaving or entering of hydrogen
at the surface. In general two cases will have to be considered:
(a) the solute is dissociated on entering the diffusion medium,
and (b) the molecular condition is the same inside as outside
the diffusion medium. The first condition will be encountered
in systems such as O2-silver, H2-palladium, N2-iron or molyb-
denum ; and the second when ammonia, water or gases diffuse
into an alkali halide lattice, or a zeolite lattice. It will be
possible to attempt a solution for the conditions of (6).
Case 1. The diffusion of ammonia gas occurs from a sphere
of a zeolite (analcite). The initial ammonia concentration
in the sphere is f(r); the radius of the sphere is a, and the
concentration at r = a is maintained very low (by condensing
evolved ammonia in liquid air). One has thus to solve the
equation dC ,d%c 2dC\ r ^
-*- =D K - 2 + - - ^ - forO<r<a, (a)
ct \dr2 r dr)
when C=/(r)att = 0, (6)
dC
and D y - + K7 = 0 a t r = a. (c)

The substitution u = Cr gives

•gr = 2) g-^ (0 < r < a), (d)

•=- + (A — I/a) w = 0 a t r = a when A = k/D, (e)

u = r/(r) at t = 0. (/)
The problem is analogous to the process of cooling of a sphere
with radiation at its surface (12), the radiation corresponding to
the surface desorption. Proceeding therefore along analogous
lines to those adopted in the problem of the cooling of the
earth (13), one obtains

i; r'f(r')smanr'dr' I s i n a / e - ^ ^ ,
(121)
where a f t is t h e nth r o o t of
oca
tan aa =
I-ah'
 THE NON-STATIONARY STATE OF FLOW 39
When/(r) = Co the equation reduces to

The quantity Q which has been desorbed from the sphere

is then
Q = %na*C0 - 4TT f "cV 2 dr = Qo - 4TT 1 "
Jo Jo

A - A (123)
v

When one plots In ° against £, the curve for large values

of ^ approaches a line of slope — Daf, for which the intercept
on the axis of t is
T6A2 1 a 2 a 2 + (aA-l) 2 . 2 "1
La 2 <4oc{a2 + ah(ah-l) x
J
These two equations can be solved for D and k with the aid
of the equation
tan (oca) = -——,

i.e. tanz + Az = 0, where A = —- and oca — z. The roots

(ah— 1)
of this equation (8) are given by
1 3A-1

_ 1575A3+1575A2 + 483A + 45

39690A4 + 52920A3 + 24696A2 + 3834A +' 9 _

In Fig. 8 are shown the first four roots of the equation

tanz + Az = 0 for A between 1 and 5(8), i.e. when ak/D lies
between 2 and 1-2. The radius a may of course be varied.
When the desorption process at the surface depends on the
square of the concentration, the problem cannot be solved
along analogous lines.
40 SOLUTIONS OF THE DIFFUSION EQUATION
Case 2. If desorption into a vacuum occurs at the surfaces
x = 0 and x = I of a slab, the conditions of Case 1 become (M)

(i) _
dt
(ii)
(iii) - —- + he = 0 at x = I, where h = k/D,
dx
(iv) C =f(x) = 0.

hi

Bl
s m mt m ms
=
20

1-9
\
% N Ml
if
b m • S
v Mi
= !B m m,
• K S
= =
9

H Si ^,
^

1-6 JS
s • • m. m Is SI ••
S
s 9 IB mm H s

«E 91
m
m

5
Fig. 8. This figure gives the first four roots, z0, zv z2, zz, of the equation
tanz+Az=0, for values of A ranging from 1 to 5. For the curve z0 the
vertical scale reads as shown. For zl9 the values read from the curve should
be increased by 3: for z2, add 6: for z3, add 9.

The expression e~Dcfil(A cos OLX + B sin ocx)

may be used to develop the solution, for it satisfies (i), and
also (ii) and (iii), when
-otB + hA = 0
and a(B cos ad — A sin od) + h(B sin od + A cos od) = 0,
which in turn give tan ad = 2och (124a)

A_B
(1246)
CL h'
 THE NON-STATIONARY STATE OF FLOW 41
The roots of the first expression are the points of intersection
of the curves y = ^ and y = \ - \ .
The second curve is a hyperbola; the roots of the equation
(124 a) are all real, not repeated, and only positive values are
to be taken.
The solution of the problem is
n - 2 y

/(^(a^ (125)
Jo
which, if f{x') = Co, reads
n cos * n

a
J
and gives for Q, the quantity desorbed per unit area,

Thus^ at large values of ^ one finds in plotting In—^—

against t that the curve approaches the line of slope — Daf,
and the intercept upon the f-axis is
, — In cos <xxl -f h]2

TJ^e method outlined is inapplicable when the desorption

velocity depends on the square of the concentration at the
surface.
Case 3. Barrer(i4a) treated the problem of diffusion through
a slab of thickness Z, when in addition to pure diffusion two
other rate-controlling surface processes are important:
(i) The passage of the sorbed atom from the surface to the
interior of the plate. Velocity constant kv
42 SOLUTIONS OF THE DIFFUSION EQUATION

(ii) The passage of a dissolved atom from the plate to the

surface. Velocity constant k2.
It was assumed in order to treat the problem that the
sorption and desorption, with velocity constants k3 and &4
respectively, in the gas and the surface layer occurred so
rapidly that the surface concentrations were defined by the
adsorption isotherm. The distribution of Fig. 9a was sub-
divided into the two distributions of Figs. 96 and c. In Fig. 96
the gas desorbs into a vacuum; in Fig. 9c one has the steady
state of flow through the plate. The complete solution for
Fig. 9a was then (employing the notation of Fig. 9)
v = u + w.
The solution u=(Ax + B) (128)

for the stationary state of Fig. 9 c between x = 0 and x = I

was given when the constants A and B took the values

A •—

(129)

(130)
One may work out the different simple types of behaviour
that may be encountered. In the equations (129) and (130) vs
denotes the saturation concentration at the surface, and u8
the saturation concentration in the metal. It was observed
that very great concentration discontinuities may occur at
 THE NON-STATIONARY STATE OF FLOW 43
the interfaces when diffusion is rapid compared with rates of
transport from the surface to the interior of the plate and
vice versa.
0
V;

^
1

\7?/A 77
j>o x-l. x*o x-lDistance x'° &1
Fig. 9.

The complete solution for the non-stationary state of flow

of Fig. 9a is given by
v = u -f w,
where u is defined by equations (128)-(130), and w is given by

w _

x (v0 —u)(oc cos ocnx-rh sin octlx)dx, (131)

Jo
2och Ic
where an is the nth positive root of tail od = -^—j-v and h = -=|.

(5) Instantaneous sources

A number of problems on diffusion into solids require solutions
of the diffusion equation for conditions described below.
A definite quantity Q of matter is deposited at a point on the
44 SOLUTIONS OF THE DIFFUSION EQUATION

surface of a solid, and left to diffuse into it, or over its surface.
Important examples of these diffusion systems are:
(i) Diffusion of caesium along tungsten filaments (15).
(ii) Diffusion of thorium around a tungsten strip (16).
(iii) Diffusions of sodium and potassium into, and over,
tungsten strip(17,18).
It will be seen from the above examples, and in Chap. VIII,
that most of our quantitative data on surface diffusions are
based on treatments of Fick's law for instantaneous sources.
The solution for an instantaneous plane source in an infinite
solid may be derived from equation (34), in which is given
the general solution
!
D [ > ' <34)
4. +U .. W _3C
to the equation —- = X> -—-
ot ox*
for the boundary conditions
C = f(x) when t = 0,
C = f(x} t) when t > 0.
Suppose in the infinite solid of equation (34) there exists
initially a concentration C — Co between the planes x = — \h
and x = + \h. Equation (34) becomes for the new conditions

and if h tends to zero while Coh remains constant and equal

to Q, one finds
L
-2j(nDt)e • (133)

In Fig. 10 are given(19) a set of curves showing the progress

of diffusion according to (133) from a thin layer into a solvent.
The curves of Fig. 10 may be compared with those of Fig. 128,
Chap. VIII, for sodium diffusing into tungsten.
 THE NON-STATIONARY STATE OF FLOW 45
When the instantaneous source, quantity Q, is deposited
at the plane x = 0 of a semi-infinite solid, the solution corre-
sponding to (133) is
Q (134)
c
The variation of C with time at x = 0 is given by the equation

(135)
&
obtained by putting x = 0 in (134). Equation (135) gives a
very simple method of obtaining Z>.

-10 - 9 - 8 - 7 - 6 - 5 - 4 - 3 - 2 - 1 0 1 2 3 4 5 6 7 8 9 10

«- from
Distance Boundary
Fig. 10. Diffusion from a thin layer into a solvent.

The solution in two dimensions of the problem treated in

the equations (132)-(135) may be deduced in an exactly
analogous manner. The equation corresponding to (132)
becomes

and leads to equation (137) for a point source, diffusing over

an infinite plane surface:

C =
Q Q
4arDt exp - 4Dt
if r denotes the radial distance in the plane from the point
source.
46 SOLUTIONS OF THE DIFFUSION EQUATION

Similarly, for the point source diffusing in an infinite solid

the solution for G in three dimensions is

<138>
I t will be noted that in the 2- and 3-dimensional cases the
value of C a t r or R = 0 is

and (140)
^-idb'
respectively. Equation (139) should have ready application
in evaluating D in the case of surface diffusion. For
instance, one might by a molecular ray method deposit a
metal film on a crystal at a point, and heat the crystal. The
spreading of the metal could be followed photoelectrically
(Chap. VIII).
Equations (138) and (140) are not yet of particular interest
in the case of the diffusion of matter, because the conditions
of the solution are not capable of easy experimental realisation.
One might also have an instantaneous spherical surface source,
for which the boundary conditions are that a thin layer of
matter exists in the shell
a<r<a+h
at a concentration C. Then

and when, as in (132), we let h tend to zero while Q remains

constant, the solution of the general equation
a+h
,f i (r~r')2\
(141)
becomes
Q
C-
 THE NON-STATIONARY STATE OF FLOW 47
Should one have a spherical volume source of radius a,
where %7ra3C0 = Q, one can obtain the solution by integrating
equation (141) between the limits 0 and a, so obtaining (143):

f(r+o)/2V(») r(r-a)l2V(Dt)

Lp i p
i ~w~JJ- (143)
The equations (141)-(143) are more significant in considering
flow of heat than of matter, and give the form of temperature
waves spreading from areas or volumes through an infinite
surrounding medium. Equation (142) would apply to the
diffusion of solute from a sphere of aqueous gel into a sur-
rounding medium of solute-free gel. It might also apply to
the diffusion of a metal solute from a sphere of a solvent metal
embedded in a mass of the solute metal.

(6) Treatment of the equation -x- = ^-ID^—

dt dx\ dx
Very frequently the diffusion coefficient depends upon the
concentration, and it is therefore necessary to solve the
equation -=- = ^r—I Z>-=—I. The interdiffusion of metal-metal
^ dt dx\ dxj
pairs, the diffusion of ammonia or water in certain zeolites,
and the diffusion of vapours in media which swell during
sorption are typical examples where one must use the above
equation. The variation in D with concentration can be very
great indeed—a thousandfold or more in some metal-metal
systems (20).
The most favoured experimental procedure in following the
interdiffusion of metal pairs is to place a slab of each metal
in contact, and heat them. The diffusion can be followed by
X-ray or chemical analysis of thin layers near the origin, and
the slabs may be regarded as of infinite thickness if the diffusion
48 SOLUTIONS OF THE DIFFUSION EQUATION

is slow. The conditions for which a solution is sought are

now (21,22)
d
fi) - - lDd0\
(ii) C = Co for x = 0 to # = — oo at t = 0,
(iii) C = 0 for x = 0 to x = 4- oo at t = 0.
Boltzmann(2i) showed that the solution would contain C as
a function of a single variable A = x/<Jt. If this variable is
substituted in the diffusion equation it becomes

(144)

When some of the diffusing material has passed across the

plane x = 0, conservation of mass requires that

xdC=\ (~x)dC; or #dC = 0. (145)

J C(»-.) JO Jo
The interdiffusion of two metals in contact is more complicated
because each metal acts as both solute and solvent, and after
a time the original plane x — 0 will no longer satisfy the
condition above. I t is necessary to choose a new plane a t
x9 = 0 so that
re, rc(Jr-.)
x'dC= (-x')dC. (146)
Jv.) Jo
The zero plane may be chosen when the concentration-
distance curve is plotted at a given time. Integration with
respect to A then gives
ldA fc=Cl rc-c*
D(c^Ci)=~-^\ AdC, where XdC = 0. (147)

At any time t — a constant, the equation is

1 dx fc==c° rc=c 0
D{c==c ) = ^ = U xdC, where xrfC = 0 (148)
1
2taC Jc=Cx Jc=o
and Z) is easily evaluated, as a function of the concentration,
by graphical integration. Boltzmann's original assumption
used in treating this problem, that C is a function of A = x/<Jt
 THE NON-STATIONARY STATE OF FLOW 49
only, can be tested by plotting x against *Jt for a constant
value of C. If for each value of C the origin is chosen so that

J °xdC = 0, a series of straight lines must be found all

of which pass through the origin. Matano (20) showed that this
was true of data obtained on the metal pair Ni-Cu.
One can also obtain a solution of the problem with the
following boundary conditions:
(i) C = 0 at x = 0 for all t,
(ii) C = O0 at 0<#<oo and at t = 0.
If one writes C = dC/dX in equation (144), it may be re-
written in the form

so that In (DC) = - fA ^ + In a
Jo m
r "AAdAn (150)

or DC ••
Further integration gives

for which the boundary conditions (i) and (ii) above give
6 = 0 (from (i)) and
C
a =
T (152)

If there is an initial concentration at x = 0 of Cx, the integrated

equation gives
CxdX r r*AdAl
co-c Jo z>
Jo z >expL
J o aapp jj
L~Jo

Equations such as (151) and (153) are not of the same practical
importance as are (147) or (148).
50 SOLUTIONS OF THE DIFFUSION EQUATION

(7) A note on the derivation of diffusion constants

from solutions of the diffusion equation
The solutions which we have given can be expressed, as pre-
ceding examples indicate, as a converging series of exponentials:

If therefore log (Q& — Q)I{Q^ — Qo) is plotted against t a curve

of the type illustrated in Fig. 7 is obtained. This curve
approaches asymptotically to the line

and so the slope, and the intercept on the axis of log Q, gives
fx(A) and F±(b). The theory then gives D, and all the other
terms. This procedure is available for all converging series of
exponentials.
Special methods are available in other instances such as
the graphical integration method applied to equation (148).
Solutions involving semi-infinite or infinite solids all give

and so one can use the relationship between x and t for a fixed
value of C, to measure D. Other methods, or examples of the
application of these methods, will be indicated at various
points throughout the text (e.g. pp. 97 and 352 et seq).

(8) Some cases for which there is no solution of the

diffusion equation
Nearly all types of system in which the law -^- = D -x-^ applies
have now given an appropriate solution of the diffusion
equation. There are one or two outstanding and experi-
mentally important cases:
(i) Sorption or desorption occurs from a diffusion medium
into a gas phase where the gas exists as molecules, whilst it
exists in the medium as atoms (e.g. H 2 -Pd). The desorption-
 THE NON-STATIONARY STATE OF FLOW 51

sorption processes occur into a constant volume system, so

that the surface concentration is a function of time. This
problem has been solved for the case when the molecular
state is the same in gas and diffusion medium (p. 21); but it
would now be necessary to have a gas phase volume altering
with time also, owing to the existence of an equilibrium

in the gas phase. *

(ii) There is the analogous problem of flow through a
membrane from one constant volume system into another
which has been solved for the case when the molecular state
is the same in both gas phases and in the medium. When,
however, the gas exists in the medium in the atomic state,
consideration of an equilibrium such as

would require, if the same solution were to hold, that the

volume of the two gas phases should become functions
of time.
(iii) There exist also more complex instances of (i) and (ii)
above, when surface reactions occur simultaneously with
diffusion. These problems may be solved by the methods
outlined on pp. 37 to 43 as soon as the problems (i) and (ii)
can be solved.
REFERENCES
For tables of Bessel functions the following may be used:
Watson, G. Theory of Bessel Functions, Cambridge Univ. Press (1932).
Gray, A., Mathews, G. and MacRobert, T. Treatise on Bessel Functions,
Macmillan and Co. (1922).
Values of error functions may be obtained in:
Janke, E. and Emde, F. Funktionentafeln mit Formeln und Kurven,
Teubner, Berlin (1938).
Pierce, B. A Short Table of Integrals, Ginn and Co., Boston (1929).
Further examples of diffusion problems on solids may be worked
out by considering analogous problems in heat flow, as given in:
Carslaw, H. Conduction of Heat in Solids, Macmillan and Co. (1921).
Carslaw, H. and Jaeger, J. Conduction of Heat in Solids, Oxford
University Press (1947).
Ingersoll, L. and Zobel, O. Mathematical Theory of Heat Conduction,
Ginn and Co. (1913).
4-2
52 SOLUTIONS OF THE DIFFUSION EQUATION
(1) Jost, W. Die Chem. Reakt. in Festen Stoffen, p. 203, Steinkopf
(1937).
(2) The method for problems of this type involving spheres and rods
is outlined in Carslaw, H., Theory of Heat Conduction, pp. 206 et
seq., Macmillan and Co. (1921).
(3) Jost, W. Diffusion u\ Chem. Reakt. in Festen Stoffen, p. 17 (1937).
(4) Stefan, J. S.B. Akad. Wiss. Wien, n. 79, 161 (1879).
(5) Kawalki, W. Ann. Phys., Lpz., 52, 166 (1894).
(6) Carslaw, H. Theory of Heat Conduction, p. 67 (1921).
(7) Theory of Heat Conduction, p. 68 (1921).
(8) March, H. and Weaver, W. Phys. Rev. 31, 1081 (1928).
(9) Barnes, C. Physics, 5, 4 (1934).
(9a) Barrer, R., Phil Mag. 35, 802 (1944).
(96) Carslaw, H. Theory of Heat Conduction, p. 211 (1921).
(10) Theory of Heat Conduction, p. 127 (1921).
(10a) Watson, G. Theory of Bessel Functions, Cambridge (1932).
(11) Gray, A., Mathews, G. and MacRobert, T. Treatise on Bessel
Functions, Macmillan and Co. (1922).
(lla) Jaeger, J., Trans. Faraday Soc. 42, 615 (1946).
(12) Ingersoll, L. and Z obel, O. Mathematical Theory of Heat Conduc -
tion, p. 136, Ginn and Co. (1913).
(13) Carslaw, H. Theory of Heat Conduction, p. 136 (1921).
(14) Theory of Heat Conduction, p. 74 (1921).
(14a) Barrer, R. Phil. Mag. 28, 148 (1939).
(15) Langmuir, I. and Taylor, J. B. Phys. Rev. 40, 463 (1932).
(16) Becker, J. A. Trans. Faraday Soc. 28, 148 (1932).
(17) Bosworth, R. C. Proc. Roy. Soc. 150A, 58 (1935).
(18) Proc. Roy. Soc. 154A, 112 (1936).
(19) Williams, J. and Cady, L. Chem. Rev. 14, 177 (1934).
(20) Matano, C. Jap. J. Phys. 8, 109 (1930-3).
(21) Boltzmann, L. Ann. Phys., Lpz., 53, 959 (1894).
(22) Wiener, O. Ann. Phys., Lpz., 49, 105 (1893).
 C H A P T E R II

STATIONARY AND NON-STATIONARY STATES

OF MOLECULAR FLOW IN C A P I L L A R Y
SYSTEMS

TYPES OF FLOW

A number of types of flow have been found to occur in capillary

systems, each of which is observed in the appropriate region
of pressure difference, absolute pressure, and pore size. Each
kind of flow may be characterised by different permeability
constants, a fact which has often led to confusion in expressing
the data. The nature of gas flow in single capillaries teaches a
great deal concerning the more complex permeation processes
through porous plates, refractories, and consolidated and
unconsolidated sands, so that before discussing flow in these
systems the different types of capillary flow will be discussed.
The researches of Warburg(i), Knudsen(2), Gaede(3),
Smoluchowski(4), Buckingham (5), and others (6,7) have demon-
strated the properties of the following types of flow:
(1) Molecular effusion; (2) Molecular streaming, or Knudsen
flow; (3) Poiseuille or stream-line flow; (4) Turbulent flow;
(5) Orifice flow. These may now be considered in turn.

(1) Molecular effusion

If one has an orifice of area A in a thin plane wall such that
its diameter is small compared with the mean free path of the
gas the number of molecules N effusing up to it in unit time
is given by the well-known kinetic theory equations
N = ±AN0Cw
2kT\
nm ) (1)
ANoP

In these expressions, No is the Avogadro number, C

the concentration, w the mean velocity, k the Boltzmann
54 MOLECULAR FLOW IN CAPILLARY SYSTEMS
constant, m and M the mass of a molecule and of a gram-
molecule respectively. This equation may also be written

since p1v1 = (N/No) BT, where v± is the volume of gas in

gram-molecular volumes effusing per second at a pressure pv
One sees that the equations for molecular effusion may be
used to measure molecular weights, temperatures, or vapour
pressures. The emission through such an orifice obeys the
cosine law, so that by using two orifices in series, a molecular
beam may be defined, whose intensity is given by the effusion
equation coupled with Lambert's cosine law.

(2) Molecular streaming, or Knudsenflow

When the orifice is of considerable length molecules will collide
with its walls on their way through. If the collisions are elastic
the flow through a smooth tube will be identical with the
effusion velocity through the hole in the thin plane wall, since
no molecules will be turned back in the original direction by
collision with the wall. For such a tube therefore the equations
of flow are, as before,

(1)
Such a flow is independent of the length of the tube, which
Knudsen(2) showed was contrary to experiment. Knudsen
therefore supposed that of each N molecules striking the wall
a fraction / was emitted with random velocity distribution,
and a fraction (1 —/) was specularly reflected. Some molecules
are then returned in the direction from which they came, and
more return the greater the length of the tube. The number
of molecules striking unit area in unit time at the inlet and
outlet sides of the tube are respectively \NQCyW and lN0C2w.
The excess flow of molecules from the inlet to the outlet along
the axis x can be shown to be \B(dNjdx){§), in the stationary
state of flow, when B is a constant depending upon the shape
 TYPES OF FLOW 55
of the tube. N is a linear function of x, in the stationary state,
and the nett rate of flow in mol./sec. is given by
_ B __ ( . (2)

where L is the length of the tube. The derivation of the factor

(2 —/)// will be found in numerous places(6,2,3,4), and will not
be considered here. For tubes of circular cross-section the
constant B takes the value -1^3TT, where r denotes the radius
of the capillary.
(3) Poiseuille or stream-line flow
When an incompressible fluid flows down a tube without
turbulence, the volume of fluid passing through a cylindrical
tube in unit time(io) is
P)
(Poiseuille's law) (3)
12O7/1J

(where d = the pore diameter,

7] = the viscosity of the fluid,
and L = the tube length).
On the basis of the kinetic theory (10) the flow0in mol./sec. is
dn d2n3d2_dC
{ }
dt ~ 12 S2A dx'
where 3 Cj dx denotes the concentration gradient in the direction
of flow x, and the other terms have already been defined
(except A, the mean free path). These equations are valid for
an incompressible fluid throughout the pore length; but for
a compressible fluid obeying the gas law the equation must
be written ^ 2
=
at
which for an isothermal process, with v = a constant, becomes
dn dAn 1 p\—p\
== ( )
~di l2SL^^T~~¥~'
56 MOLECULAR FLOW IN CAPILLARY SYSTEMS
Equation (6) may take the form
dn d*n 1

where p = ^(Pi+p2) *s ^ n e mean pressure in the tube. Also

dn _ d*n p

since o ~^m — P> ^ n e niean density of the gas.

Z HI

A complete equation of flow.

Two corrections may be applied to Poiseuille's formula:
(i) Only part of the pressure difference is used in over-
coming friction; a fraction which must be subtracted from
the original pressure difference produces kinetic energy of
motion(io,n,i2).
(ii) In the boundary layer of gas, of thickness A, equal to
the mean free path, there may be specular reflection at the
surface, those molecules specularly reflected having the
streaming velocity component of the flowing gas. A fraction
/ only is emitted in random directions. The coefficient of
2-/
slippage is then —7-.4(i.e.).
The equation of flow then becomes
W , , 8(2-/M\r *& pl-pj M(dn\n
dt iL\
toiL\ f dJllGRT 2 n\dt)j9
in which expression the last term gives the correction due
to (i). It transpires, however, that this is only a small term;
and also when the pores are small, or the pressures are low,
2— fA
that —J*--J > 1. Thus the equation becomes
dn

which is the equation given for Knudsen's molecular streaming

save that the numerical constants are somewhat different.
 TYPES OF FLOW 57

(4) Turbulent flow

The type of flow known as viscous, stream-line or Poiseuille
flow changes when a certain limiting mass velocity W is
reached. We will define a quantity [i?](i3), called the " Reynold's
number", by
[Ml - * p . (ID
Al ±. cross-section . . A, „ ,
where rh denotes the ratio —; and is the so-called
n
periphery
hydraulic radius, which for circular tubes is one-quarter of
the diameter d, p and 7/ have their usual significance as fluid
density and viscosity respectively. Then for such circular
tubes it is found that Poiseuille's equation no longer applies
when [R] is greater than 580. In tubes of such a length that
any nozzle effect may be neglected, the differential equation
of flowis(i4,i5,ic^
J (12)
(
dx~ 2^' '
wherein /? is a constant. For smooth tubes of glass or steel(iT)

but is higher, and follows a different law for rough tubes. For
some tubes /? may be taken as independent of [R],
There is a formal connection between stream-line and
viscous-flow formulae which may be brought out as follows.
An alternative form of the above equation of flow is

1 ^ (15)

since the mass velocity W and the volume v of gas transfusing

in unit time are related by the expression v = (i^rd2) W. When
d(p2)/dx is constant down the length L of the tube, and since
58 MOLECULAR FLOW IN CAPILLARY SYSTEMS
for a gas at constant temperature pv or pv is also constant,
one may write

or (vp)2 = RT \{p\— pl)> (16a)

whilst the equation for Poiseuille flow gives

£=rf>. (17)
In the general case of isothermal high-pressure flow, whatever
the Reynold's number, we may thus say
(vp)»<x:(pl-Pl), (18)
and the value of n then determines whether stream-line or
turbulent flow is occurring. Never under any conditions have
values of n been found such that 2 < n< 1. Later in this chapter
will be given examples of gas flow in porous solids which are
partially turbulent (p. 73). The great importance of gas flow
or vapour flow in turbine design or in aeronautics can easily
be understood, and farther information on turbulent flow will
be found in references given earlier(14,15,16,17). Equations for
an adiabatic turbulent flow have been given by Stodola(is).

(5) Orifice flow

In designing permeameters for studying gas flow through
textiles, Buckingham (5) discussed another type of flow of a fluid
through frictionless jets, the flow being supposed to occur
adiabatically.
Let v, K, E denote the volume, kinetic energy and internal
energy of unit mass of fluid. For adiabatic flow
K^K = E-Ei+pv-p^. (19)
Let Aj be a cross-section of the jet at its nozzle, and A be a
cross-section farther upstream where the velocity is negligible.
Then K at A is negligible and
K1 = E-E1 + pv~p1vv (20)
 TYPES OF FLOW 59

If the gas is ideal,

pv = RT,
Cn = constant,
r (21)
E = TCV + constant,

iC and Gy are, respectively, the specific heats at constant

pressure and volume), which combined with the original
equation give g — C (T — T)

If S denotes the mean speed of unit mass over the cross-

section Al9 *i

i.e. S^^C^T-TJ}. (22)

As we have assumed the flow to be frictionless, these con-"
siderations are limited to short well-formed nozzles. Then one
may apply the adiabatic gas laws
pvy = constant, 1
\ (23)
T = p(r-i)/y x constant, J
whence -^ -

or Si = ^{2TCp(l - r^y-Wy)}. (24)

The volume passing the cross-section Ax in unit time is
Vx = A1Sli which under the initial conditions of pressure is
V = V^v/vj), and thus
V = A1—^j{2TC (1 -rty-Vly)}. (25)
p
V
l
When Ax is a circular cross-section of diameter d, and since
v/vx — r11?, one gets

V = ~ d? J{TCpr*'r( 1 - ^r-Wr)}. (26)

The conditions under which this type of flow is most accurately

obeyed are known to require short smooth nozzles, or sharp
perforations in a plate. The extent to which the law breaks
down as the nozzles increase in length has not been studied
60 MOLECULAR FLOW IN CAPILLARY SYSTEMS
experimentally, nor have its possible uses in determining y,
the ratio of specific heats, or Cp, the specific heat at constant
pressure. Reference to orifice flow will be made again in
considering the permeability of various paper and fibre-board
membranes, when it will be seen that it does not in general
occur with such membranes (Chap. IX).

PERMEABILITY CONSTANTS, UNITS, AND DIMENSIONS

By considering the flux of gas through porous membranes, as

the volume V measured at a pressure p, passing through the
membrane in time t, in relation to the equation governing the
flow (19,20), one can define permeability constants of various
types, and having various dimensions. The equation of flow
for an incompressible fluid in a pore system may be written:

j = APL(Pl-p2), (27)

where A is the cross-section of a pore. For a compressible

fluid it is __
^-^fe+A^-ft). (28)

In these expressions we can call PL and PG "permeabilities".

As such they will be functions of the membrane thickness,
pore type and size, and of the nature of the gas. By multiplying
each of these permeabilities by Z, the membrane thickness,
one gets "permeability coefficients" PLl and PGl which are
dependent only upon pore type and size, and the nature of
the gas. Finally, multiplying each permeability coefficient by
the viscosity TJ, one obtains the "specific permeability" PLty
or PQITJ, which should depend only on the pore type and size.
In considering the equation of flow of rarefied gases another
set of permeability constants is obtained. The Knudsen
equation, for a capillary system, may be written

(29)
 CONSTANTS, UNITS AND DIMENSIONS 61
an(
= PG^(^1+^2)> ihLasbeencalledbyManegold(i8,i9)
the "molecular permeability". By analogy with the previous
cases {PQIJPL^ o n e n a s a ^ s o a "molecular permeability
coefficient" Pj^l; and finally a "specific molecular per-
meability" PMlyl(M/RT), which is dependent only upon
pore size and type. The equation of molecular effusion through
an orifice gives a "molecular effusion coefficient" PM, and the
equation of orifice flow an " adiabatic effusion coefficient" PA.
These relationships may be summarised as in Table 1.
The various constants defined above have, it will be seen,
different dimensions. It is important therefore to specify
accurately what constants are being employed. When, as is
usually done, the permeability constant is formally defined
as the number of unit volumes passing in unit time through
unit cube having unit pressure difference between its faces,
its dimensions are cm.3sec.1 g."1, which are those of the
"permeability coefficient" of Table 1, for Poiseuille flow. The
dimensions of the diffusion constant D defined by Fick's law
dC 32C
-— = D^r-z are cm.2 sec."1, which are those of the molecular
dl dx2
permeability coefficient of Table 1.
The numerical measure of the permeability constant is
dependent upon the units in which it is expressed. The
literature shows a lack of uniformity with respect to units.
It is specially important to state the thickness of the membrane
used. In some systems it may also be an advantage to give
other membrane properties such as porosity, defined by the
ratio
Pore volume
Total volume *

SOME EXPERIMENTAL INVESTIGATIONS OF GAS FLOW

IN CAPILLARIES

Warburg(i) first established the validity of Poiseuille's law

hi the region of high pressures. These measurements were
made before the development of high vacuum technique, but
at the lower pressures it was even then necessary to assume
 TABLE 1

Dimensions
Equation of flow Perraeabilit}7 constants Properties in O.G.S. units

Poiseuille equation for

(i) Liquid V/t = APL(px - p2) Pi (permeability) Dependent upon type of pore system, cm.2 sec.1 g."1
(ii) Gas Vp/t=APG{pl-pt) iiPt +P«) PQ (permeability) on thickness, and on nature of gas
PjJ, (permeability coefficient) Dependent upon pore system, and cm.3 sec.1 g. -1
PQI (permeability coefficient) on nature of gas
Pl,lr) (specific permeability) Dependent upon pore size and type cm.2
Paty (specific permeability) only
Knudsen equation
Vp/t=APM(pl-pt) PM (molecular permeability) Dependent upon pore type and size, cm.1 sec."1
on thickness of membrane and on
nature of gas
PMI (molecular permeability Dependent upon pore size and type, cm.2 sec."1
coefficient) and on nature of gas
PMlyl(M/BT) (specific mole- Dependent upon pore size and type cm.1
cular permeability)
Effusion through an orifice (low P M (molecular effusion co- Independent of orifice dimensions cm.1 sec.-1
pressure) efficient) and shape. Dependent upon the
Vplt = APM(Pl-pt) mol. wt. of the gas and T
Orifice flow, adiabatic, high pressure Pj[ (adiabatic effusion co- Independent of orifice dimensions. cm.1 sec."1
3 efficient) Dependent upon T and Cp
V*/t=APA V{(PxlP*)V -(Pi/P*) ?}
for air, oxygen, nitrogen, or hydrogen
* V is measured at the pressure on the high pressure side of the orifice.
 EXPERIMENTAL INVESTIGATIONS OF GAS FLOW 63
a slip between the wall of the capillary and the adjacent gas
layer due to specular reflection (compare with equation (9)).
Warburg expressed his results by the formula

0 = (30)

where his 0 expressed the quantity of gas flowing per unit of

time (measured as the product of pressure in bars and volume),
n lr4
a is a constant = - - — (r being the capillary radius), p is the
o 71 Ld
mean pressure = %(Pi+p2), and £is the "coefficient of slip".*
G
0-04
0-03
0-02
0-01

Fig. 11. The gas flow as a function of pressure.

Knudsen (2) expressed his results by the empirical formula

where b was called the "coefficient of molecular streaming"

and C\ and C2 were two quantities depending upon slip. At
high pressure the first term in the square bracket alone is
significant, but at the lowest pressure the second term in the
bracket was predominant, and so reduced to b. On the
basis of the kinetic theory, values may be assigned to 6, Cx
and C2, and to £. Fig. 11 (6) gives the quantity of gas flowing
as a function of pressure, curve I referring to the simple
Poiseuille law G — wpiVi—V?)) curve II to the calculated rate
of flow according to equation (31), and curve III a typical
run by Knudsen. The experimental curve shows a minimum
2 -f
* £ = — ~ A, as denned on p. 55.
64 MOLECULAR FLOW IN CAPILLARY SYSTEMS
when G is some 5 % less than G at p = 0. This minimum t
which is characteristic of most of the results obtained(2,6,20),
being true even for mixtures(21) of gases, does not appear to
have been explained. The minimum occurred in this case(6>
when the mean free path was about five times as large as the
capillary radius. Similarly, Fig. 12(6) gives another series of
curves, calculated and experimental, by several authors.
Curve I is for simple Poiseuille flow for hydrogen, curve II
for Poiseuille flow with slip, curve III for hydrogen employing

150 300 450 600 750P

mrn.Hg
Fig. 12. Gas flow as a function of pressure.

Knudsen's empirical equation for Gaede's(3) apparatus and

results (the calculations being made by Klose (6)), while curve IV
gives the experimental measurements of Gaede. The point
marked " £ra" on the axis of G was calculated by Klose from
Smoluchowski's formula for a tube of rectangular cross-
section. Fig. 12 shows the degree of divergence in theory and
experiment of the various authorities. Clausing's(7) experi-
ments, which are not cited in Figs. 11 or 12, are probably more
accurate than any series. In the region of molecular streaming
or Knudsen flow he found Knudsen's formula satisfactory, as
did Klose, who, however, still noted small deviations which
he ascribed to an experimental error.
 EXPERIMENTAL INVESTIGATIONS OF GAS FLOW 65
One may summarise the situation as follows:
(i) For high pressures theory and experiment both lead
to the Poiseuille law wherein slip may be neglected.
(ii) For the lowest pressures theory and experiment are
agreed on the type of molecular streaming proposed by
Knudsen.
(iii) In the region of intermediate pressures theory and
experiment do not yet satisfactorily agree, nor is it certain
to what extent the deviations are due to experimental errors
and to what extent to inadequacies in the theory.
Adzumi(20) has recently tested the Knudsen equation for
the mixtures H2-C2H2, H2-C3H6, finding a flow G = b(p1-p2)
which was that given by the ideal mixture law b = n^ + n2b2,
the n'& being mol. fractions. In the mixtures at the lowest
pressures the constituents diffused by molecular streaming
independently, but when Poiseuille flow became predominant
at high pressure the separation of constituents by flow became
negligible.

FLOW OF GASES THROUGH POROUS PLATES

Since a porous plate consists of a medley of pores of different
sizes, flow through such a plate need not be of one type.
Instead, one would anticipate that flow Would be a mixture
of the various kinds, reducing to Knudsen flow at the lowest
pressures. Effusion through an orifice or molecular streaming
down capillaries should occur for all gases at rates inversely
proportional to the square root of the molecular weight, the
condition for effusion or molecular streaming being a suitably
low pressure. It is interesting to see under what conditions
the law is fulfilled. Ramsaj?- and Collie(22), by diffusing helium
through unglazed clay and comparing the data with the corre-
sponding diffusion of hydrogen, found the atomic weight of
helium to be rather low (3-74 compared with 4-00); oxygen,
argon and acetylene however obeyed the Knudsen flow formula.
Donnan(23) measured the effusion of argon, carbon dioxide,
carbon monoxide and oxygen through an orifice, finding small
66 MOLECULAR FLOW IN CAPILLARY SYSTEMS
deviations from the lj*JM law, which they attributed in part
to orifice flow as described by Buckingham (5) and others. These
measurements were made however at pressures approximating
to atmospheric.
Recently, Sameshima (24) has measured the rates of flow of
various simple gases through a compact unglazed earthenware
plate. The rates of flow definitely did not obey the Knudsen
formula t = k JM, where t denotes the time required for the
effusion at constant pressure of a volume V of a gas of molecular
weight M, and where k is a constant. On the other hand, the
law t = k^M was accurately obeyed when the gases effused
through a platinum plate with a single orifice. For the
earthenware plate Sameshima found a formula t = krinM^1~ny>
to apply. If the wall was very thin n approached zero, and
the simple behaviour of the perforated platinum plate was
found. If the wall was thick n approached unity and the
equation became t = kq (rj denotes viscosity).
Adzumi(25) has elaborated a semi-empirical theory for flow
processes through porous plates, based on the assumption
that the plate is perforated by numerous fine holes, the
diameters of which can vary down their lengths, so that
effectively each capillary is a number of capillaries of very
numerous diameters arranged in series; and each such com-
posite capillary is in parallel with all the other capillaries. At
high pressures the Poiseuille law Gv = ap(p1—p2) applies
(p. 55), and at low pressures the law Gm = b(px—p2) (p. 54).
For medium pressures, or for a porous plate, the two laws
may be considered to overlap:
G = Gv + yGm = K(Pl-p2), (32)
where y is a term due to the coefficient of slip, and where
K = A yP + yBy. If one has n capillaries in parallel,
0 = K'(p1-2h)>
K' = K1 + Ki + ...

(33)
 FLOW OF GASES THROUGH POROUS PLATES 67
When one has two capillaries of radii rx and r2 in series,

IT Aripo-p2
1 + 1-—j~ A,-
where px and p2 are the pressures at the inlet and outlet sides
of the capillary, and p0 the pressure a t the junction of its two
halves of radii rx and r2.
The last equation takes the form

where Ax proves on inserting numerical values for a typical

plate to be less than 1 % of ^77. This gives

(36)

which may in turn be written as

<37)
^
where A2 again proves to be less than 1 % of K". One obtains
therefore as the final constant K" for m capillaries in series

while for the series-parallel arrangement of capillaries, pre-

sumed to comprise the whole pore system of the plate,

(39)
Fig. 13 gives the quantities of gas flowing through a porous
plate (expressed as c.c. x atm.) as a function of pressure for
3-8
68 MOLECULAR FLOW IN CAPILLARY SYSTEMS

a number of gases. It will be seen that, as the theory requires,

the rate of flow is proportional to the pressure, for a constant
pressure difference, but with an intercept (= yBF) on the
iC-axis.

(HO

0-08

| 0-06

0-04

0-02

OB l'C 1-5
Ptktm.

Fig. 13. Gas flow through a porous plate, where K is defined by equation (39).

The treatment of Adzumi allows one to estimate a mean

pore size, and the number of pores per unit volume. Should
all the n pores be of radius r,

E = n-j\ F = nj (I = thickness).

If N = the number of pores/unit volume,

and one may readily compute EQ and Fo from E and F and so

find N and r. This is done for a few permeable plates in Table 2.
 FLOW OF GASES THROUGH POROUS PLATES 69

For further examples of flow through porous plates one may-

study the permeability of various refractories, the properties
of which, in relation to the diffusion problem, are given in
the next section.

TABLE 2. The mean radii of pores, and the

number of pores per unit volume
Thick- 5
Porous plate ness Area Eo x 1011 Fo x 107 rxlO N
cm. cm.2 cm.

Unglazed earthen- 015 0-28 0-97 1-83 5-3 1-2 xl0«

ware
Compact porous pot 0-27 2-22 0054 0195 2-77 2-4 x l 0 s
Rough porous pot 0-41 2-22 0134 0-318 4*84 2-8 xlO 5
Mantle of Daniell 0-2 0-27 0-80 4-98 1-61 9-6 xlO 7
cell, I
Mantle of Daniell 0-2 0-27 109 6-26 1-74 1-19 x 108
cell, II

PERMEABILITY OF REFRACTORIES

Diffusion through refractories is of technical importance.

According to the fineness of the pore structure and the
pressure the flow should conform predominantly to the
equations for Poiseuille flow or Knudsen flow. In general
the numerical value of the permeability depends on a variety
of factors(26) which include:
(1) The material of which the refractory is made (silica,
fireclay, alumina, chrome, or magnesite).
(2) The physical constitution of the refractory, the grain
size of the grog, the percentage of grog to bond, and the
method of manufacture.
(3) The nature of the diffusing gas.
(4) The testing temperature, and the thermal expansibility
of the solid.
(5) The pressure difference maintained between the ingoing
and outgoing surfaces.
(6) The thickness of the specimen being tested.
70 MOLECULAR FLOW IN CAPILLARY SYSTEMS
As a rule the physical constitution produces greater changes
in the permeability than the chemical nature of the refractory.
Thus one finds a very great range in permeability in different
samples of the same chemical nature, as is illustrated in the
following data (Table 3) due to Kanz(27). In agreement with
TABLE 3. The range of permeability in
some refractories
Permeability range
Type of substance (c.c./sec./cm.2/cm./sec./cm. of
water difference in pressure)
Fireclay product 0-00278-0-476
Silica product 00158 - 0 1 3 2
Magnesite product 0-0275 -0-265
Chromite product 0147 -1-42
Insulating product 0-00597-0-539

/J / /

/
/
/
/

A 'A 30UCT i
»UCT M / y
/

f 0 /
/

v/
/

/
Reciprocal of thickness
Fig. 14. Gas flow through refractories as a function of thickness.

both Poiseuille's and Knudsen's equations the velocity of

transfusion is inversely proportional to L, the thickness of the
specimen. Fig. 14, originally due to Clews and Green (28), shows
 PERMEABILITY OF REFRACTORIES 71
this relationship for two refractories. The product I shows a
gradual change in texture with thickness, so that the curve
does not pass through the origin. Usually it appears that the
velocity of flow is proportional to the pressure difference
causing it, provided in the case of Poiseuille flow one measures
the permeation velocity in c.c./unit time at the mean pressure
2(^2+^1) obtaining inside the material. This proportionality,
required by the equation already obtained,
dn dhr 1 p
(8)
- ^ ^
was observed by Bansen(29) for a mortar joint, by Kanz(27)
for a specimen refractory, and by Clews and Green (28).

Temperature and permeability in refractories

One may look for interesting results by considering the
influence of temperature upon the rate of permeation of
refractories. The effects of temperature help to sort out the
different types of flow, Poiseuille, Knudsen, and activated
diffusion. In a refractory should occur large pores and tubes,
small pores, and pores of molecular or sub-molecular dimen-
sions. In the large pores one would anticipate Poiseuille flow;
the permeability PG is proportional to 1/rf (T/ = viscosity);
and the viscosity increases as temperature increases. Thus
Poiseuille flow in capillary systems is marked by a diminution
in flow rate with rising temperature. When Knudsen flow is
occurring the permeability PM increases according to a *JT
relationship. Finally, in a very compact solid where activated
diffusion predominates the permeability constant P contains
a term e~EIRT
For systems in which Poiseuille flow is important Preston (26)
has pointed out that for a given refractory, at all temperatures,
the quantity PGlrj is constant. Then if PGITJ be the specific
permeability (p. 62) at a temperature T° C, and P°ol be the
permeability coefficient at a standard temperature, the ratio
PGV/PG should also be a constant. The data of Clews and
Green (28) are summarised by Preston (26) in Table 4, where it
72 MOLECULAR FLOW IN CAPILLARY SYSTEMS
is seen that the product V(PG/PQ) shows small trends with
temperature, but tends to be constant.
The role played by alternative processes of Poiseuille,
Knudsen, or activated flow is perhaps to be seen in the data

TABLE 4. The influence of temperature upon

permeability constants
Fireclay product Silica product
Viscosity
Temp. ofN 2
°C. (poises V(PG/PG°) PQ/PG° V(PG/PG°)
xlO4)

10 1-71 0-997 1-71 0-992 1-70

100 213 0-841 1-79 0-860 ]L-81
150 2-33 0-785 1-83 0-828 ]L-93
200 2-52 0-740 1-86 0-759 ]L-91
250 2-70 0-710 1-92 0-670 ]L-81
300 2-87 0-687 1-97 0-599 ]L-72
350 303 0-662 201 0-546 ]L-65
400 319 0-639 204 0-517 ]L-65
450 3-34 0-613 205 0-493 1-65
500 3-48 0-592 206 0-473 1-65

TABLE 5. A comparison of theoretical and experimental

permeation rate ratios

Temp. Permeability CO2 Permeability H2 Permeability SO2

°C. Permeability air Permeability air Permeability air
Theoretical ratio
0-81 3-79 0-67
17 0-80 3-79 0-65
100 0-80 3-82 0-98
145 — — 1-10
190 0-95
200 0-79 3-55 0-96
300 0-79 307 1-11
400 0-80 2-92 113
500 0-83 2-98
600 0-84
700 0-84 — —

collected by Bremond(30) on the diffusion of gases througli

unglazed porcelains between 17 and 700° C. This author
discovered that for air, carbon dioxide, hydrogen, and sulphur
dioxide the permeabilities at first decreased and then increased
 PERMEABILITY OF REFRACTORIES 73
with temperature. The temperatures of minimum permeation
velocity were:
Gas Air C0 2 H2 S0 2
Temp. ° C. 250 240 340 250
The theory of Knudsen flow requires that the ratios of the
rates of flow should be in the inverse ratios of the square roots
of the molecular weights. Here again the ratios diverge from
the requirements of the theory at high temperatures (Table 5).
These ratios, as well as the increase in diffusion velocity, may
imply a certain amount of activated diffusion at the highest
temperatures.

FLOW IN CONSOLIDATED AND UNCONSOLIDATED SANDS

A number of experiments have been carried out upon fluid flow
through sands, sandstones, or columns of beads. The experi-
menters (31) have sought to correlate the phenomena of gas or
liquid flow with permeation rates of petroleum and its vapours
through oil-bearing sands, and with the possibility of displacing
oils from these sands by another fluid. Thus their measure-
ments, whose technical application may become considerable,
were carried out at high pressures, and as one would expect
cover the regions of Poiseuille flow and of turbulent flow. The
quantitative application of Poiseuille's law to each of the pores
in sand or a sandstone must have corrections for features such
as the following (32,33,34^:
(1) Deviations of the cross-section of an average pore from
circular shape.
(2) The increased length of a sinuous path through the
medium, as compared with its apparent length, and so the
greater pressure gradient across the specimen needed to
maintain a flow rate equal to that for a straight path.
(3) The energy consumption due to alternate expansions
and contractions of cross-section.
There is little that can be done to allow for (1). Schlichter(35)
calculated that in an assembly of spheres the actual path was
74 MOLECULAR FLOW IN CAPILLARY SYSTEMS

1 • 2-1 • 5 times the apparent path; while Chilton and Colbourn (36)
claimed that 80% of resistance to flow should be due to
alternate contractions and expansions of the cross-section,
and only 20 % due to the viscosity. However, equations
having the form of Poiseuille's law must and do apply, and
one has as examples Adzumi's data for flow at moderately
small pressures through porous plates (p. 66), and the data
cited in the previous pages (69-73) for the permeability of
refractories. In papers dealing with the permeabilities of un-
consolidated and consolidated sands it has become customary
to plot the logarithm of the so-called "friction factor", / ,
against the logarithm of Reynold's number [R] = dWpj4trj(\.Q.).
The curve is rectilinear at low pressures, but undergoes a
smooth continuous transition to another rectilinear portion of
smaller slope at the onset of turbulent flow (37,38) (Fig. 15, p. 75).
The friction factor is usually defined by Fanning's (39) equation
for a capillary
7 , , ,
f_gdPl
J _gdPl-p2 l \
L (
~ \W*r
where g = the acceleration due to gravity, d = the diameter
'
of a capillary, p = the density of the gas at the temperature
of the experiment, and the mean pressure J(Pi+p 2 ) °^ fl°w>
W = the velocity of flow in c.c./sec. Fanning's equation is
derivable from an equation of turbulent flow

\-Pl), (41)

similar to that described on p. 57. A denotes the area of

cross-section of the capillary; and v the volume passing
through per second measured at pressure p. Stream-line flow
obeys a law wherein (pv) appears only to the power unity
(p. 58), and so one sees how the curve log (/) versus log [R]
is a straight line of slope — 1 while stream -line, viscous or
Poiseuille flow is occurring; and —0-5 when turbulent flow
has set in. In Fig. 15, taken from the work of Fancher and
Lewis(3i), log (/) versus log [R] plots have been given for a
great variety of porous solids. The pressure at which the
experiments were carried out varied from 10 atm. per square
 NOMEMCLATUF E
=F ••77] f - « CTON
fT
FACTOR.N0 UNfTS
1 SAMPLE NO. SAND POROSITY
1-0 *'t-'; FEET PER SEC.PER SEC.
rr CONSOLIDATED
1
METER OF AVERAGE GRASN.FI %
_——
^ t.
ESSURE OROP LBS PER SQ. T j BRADFORD 12-5
T L=LE OTH OF C O B E . f E E T T.
BRADFORD
JID DENSITY LBS. PER CUB 2 12-3
3 3 « ° VENANGO 18-9
J L| U"*P AREMT VELOCITY C FT.
4 CERAMIC A 370

r V | j CROSS SECTIONAL AREA

. PkVVrT
PER SEC 5
6
ROBINSON
CERAMIC B
2O-3
37-8

s | SYMBC )LS
7 WOOOeiNE
WILCOX
19-7

ii
> =4=t 6 15-9

m z°£ "Aas 9
O
3RD VENANGO
ROBINSON
11-9
19-5
W OIL uo J ROBINSON 18-4
wr.ci_c
—_J. Ui.BUf
OF M
2
3
3ROVENANGO
WILCOX
22-3
16-3
1 WARREN
" t 4 192
\ l! •f
5
6
3R0 VENANGO
ROBINSON
21-4
206
7 CERAMIC C 332

[i 8 3RDVENANGO 21-9
—:?: ^
1-0 xlO 7 1 9 WOODBINE 238
zz:
=
-Hi ?O 26-9
—
-f ?1 27-7
^jo 2 221
7
"Clj I i 3 28-8

V i-i-' I 2*
25
FLINT
OTTA WA
38-5
30 9
l*-
\
}
\ #
N 11 , 1 !i j
4—•-> 4 T =
26
27
Z0-3C OTTAWA
SHOT
34 •5
•?

!
+u |
j
— ill
ill' L 1
1-0.1C5
•v.
T*l

fiti
UNCONSOLIOATtp^ "*
5AMD
'. \ •Si

1O«1O4
!
^, ifNl
p—-1[— ... V
i L "^
sar \

1-CxlO3

10X102 %
h-H—«
1
1
t.
1

10 1
0-001 001 0-1 swp 10 100 100 1000

Fig. 15. The curves log (friction factor) versus log (Reynold's number) in the flow of
various simple fluids through porous materials.
76 M O L E C U L A R FLOW I N C A P I L L A R Y SYSTEMS

inch downwards. Fancher and Lewis's notation and the sub-

stances used are explained in the figure. It is interesting that
the unconsolidated sands all give curves which fall near a
single mean curve, so that they may be said to exhibit a
typical behaviour.
Several types of the apparatus(40,41) devised had arrange-
ments of pressure gauges at intervals down the length of the
column of sand through which permeation was occurring. These
workers then were able to measure the pressure gradient along
the length of the column. It was found that dp2/dx was
constant, and thus equal to (p\ — pl)/L. The curves pv (OTpv),
against (pi—pi), obtained by Muskat and Botset(40) for un-
consolidated and consolidated sands, emphasise still further
that the law of flow is of the form
(pi—pi) = k(pv)n, (42)
where 1 <n<2 (see pp. 57 and 58) according to the amount
of turbulent or stream-line flow occurring. The chance of
turbulent flow occurring decreased with decrease in grain
size, whilst of course the permeability decreased as turbulent
flow set in, i.e. as n increased from 1 towards 2.
Some typical permeability constants for the stream-line
flow of gases and liquids through media such as have just been
considered are presented in Tables 6-8. The data recorded in
TABLE 6. Data of Green and Ampt{&) for flaw through
unconsolidated media
Specific permeability
c.c./sec./cm.2/mm. ^hick/cm.
Hg/unit viscosity, when,
JMean ineq. (28), 2p/(Pl+p2)=l
radius Porosity
cm. % Specific per- Specific per-
1 A—2
X 1U meability by meability by
gas flow liquid flow
xlO~ 6 xlO- 6
Glass spheres 4-69 36-4 372 357
3-55 37-3 250 252
2-49 361 113 115
1-59 36-3 47-6 46-5
1-25 36-6 23-6 28-3
Quartz sand 413 34-7 157-5 154
1-45 34-7 22-3 22-3
0-93 37-7 101 ~9-75
TABLE 7. Data of Muskat, Botset and co-workers (41); and Fancher
and Letvis(Si). Stream-line flow through unconsolidated media

Mean Per- Specific

radius of Porosity meability permeability
Substance particles % co-
.cm. efficient PGlV PLIV
xlO" 2 PGl xl0-« xlO"6
Glass beads 33-8 1350 248
40-45 mesh sand 1025 188 188
60-65 mesh sand 440 233 42-6
80-100 mesh sand 183 33-5 29-6
Heterogeneous sand 420 271 49-6
Lead shot 5 34-5 3190 584
Ottawa sand, unsieved
Ottawa sand, sieved
Flint sand, Pennsyl-
36-7
35-7
18-75
30-9
34-5
38-5
1175
1117
308
215
204
56-3
z
vania

TABLE 8. Data of Fancher and Lewis (Si). Stream-line flow

through consolidated media. The specific permeabilities
are measured with air, water, and crude petroleum;
"2p/(p1+p2)^l. The permeability coefficients are those
calculated from PGlrj for air
Specific
Mean Per- per-
particle Porosity meability meability
Sandstone radius coefficient
%
cm. /o{ PGlV
xlO~ 3 tor air xlO- 6
Woodbine, Texas 811 221 24-2 4-62
8-35 26-9 18-2 3-33
6-39 28-8 17-3 316
Venango, Pennsylvania 12-53 11-9 8-79 1-61
Woodbine, Texas 4-51 22-7 6-35 1-16
Wilcox, Oklahoma 6-99 16-3 4-08 0-75
Venango, Pennsylvania 4-54 21-9 3-32 0-607
Wilcox, Oklahoma 6-97 15-9 2-57 0-47
Venango, Pennsylvania 419 21-4 1-78 0-326
Woodbine, Texas 313 23-8 1-73 0-317
Warren, Pennsylvania 3-26 19-2 103 0189
Robinson, Illinois 2-74 20-6 0-925 0170
>» >> 2-74 18-4 0-498 0-091
>> >» 2-71 19-5 0-466 00855
Venango, Pennsylvania 4-90 16-9 0-347 00636
Ceramic C (5 % binding 1-43 33-2 0-273 00502
agent)
Ceramic B (10% binding 1-43 37-8 0-065 00119
agent)
Ceramic A (20% binding 1-43 370 00380 0-0070
agent)
Bradford, Pennsylvania 2-79 12-3 0-0209 000384
» >» 2-82 12-5 00194 000356
78 MOLECULAR FLOW IN CAPILLARY SYSTEMS
these tables show how particle size modifies permeability, at
nearly constant porosity (Table 6) and how the nature of the
particle (e.g. its smoothness) also controls the gas flow. In
general, particle size controls the flow rate more than porosity;
and the rate of flow is less through consolidated than through
unconsolidated media. The tables also show a good agreement
between specific permeabilities determined with liquids and
with gases (Tables 7 and 6).

FLOW THROUGH MISCELLANEOUS POROUS SYSTEMS

In the literature will be found measurements on liquid or gas
flow through a great variety of substances. The permeability of
papers, leathers, textiles, and fibreboards is treated elsewhere
(Chap. IX). The permeability of various kinds of wood(43),
of a silicic acid gel (44), of building and heat-insulating sub-
stances (45,46), and of various soils (42) illustrate the systems
studied. Manegold(i9) summarised a number of measurements
on these systems, employing the permeabilities or specific
permeabilities of Table 1.

SEPARATION OF GAS MIXTURES AND ISOTOPES

Phenomena connected with effusion and molecular streaming
through porous plates, and in tubes, have had some interesting
applications in separating the components of gas mixtures
and, in particular, isotopes. The original method for the
separation of isotopes developed by Hertz (47) required a
battery of mercury-in-glass diffusion pumps connected in
series. Each pump circulates the gas mixture through a clay
tube; a greater part of the lighter than of the heavier gas
diffuses through the wall in transit down the tube, and enters
a counter-current rich in the light fraction on the other side.
These lighter fractions tend to enter another higher separating
unit, and the heavier fraction a lower unit. The method has
served to effect a complete separation of the isotopes of
hydrogen (H2 and D2)(48), and a partial separation of the
isotopes of neon, carbon and nitrogen (49,50). A schematic
 SEPARATION OF GAS MIXTURES AND ISOTOPES 79
representation of a Hertz porous wall separation unit is
shown in Fig. 16(49).
Recently, however, the type of diffusion unit with a porous
wall has been abandoned by Hertz and his collaborators (51),
and instead batteries of diffusion pumps are joined together
directly, separation occurring by effusion together with
mercury vapour through jets. The theory of the separation of

1st Separation i 2nd Separatio?i j 3rd Separation i 4th Separation i

unit I unit J unit ! unit !

Fig. 16. A Hertz porous wall diffusion apparatus.

isotopes by this method has been given by Barwich(52). For

example, one might be employing a battery of pumps to
separate the isotopes of carbon, C12 and C13, in gaseous form
as methane (53). Then according to Barwich one set of conditions
gives
v r/1 i \ n.i
(43)

where Dl9 D2 denote the diffusion constants of the isotopes,

V is the speed of the mercury vapour jet, B, s, and x0 are
geometrical factors of the apparatus, and q is the separation
coefficient defined by
C13 (exit) C12 (entrance)
(44)
C12(exit) C13 (entrance)'
80 MOLECULAR FLOW IN CAPILLARY SYSTEMS
This equation holds only when the pressure of diffusing gas
is high. When the pressure and temperature are such that the
densities of the diffusing gases are small compared with the
density of the mercury vapour, one has D = \wA (w = the
molecular velocity; J. = the mean free path), and the expression
for the separation of the isotopes becomes
In q = const. pHgV[(r2 + r Hg ) 2 Jm2 - (rx + r Hg ) 2 ^ r a j , (45)
where rv r2, r Hg are the molecular radii of the two isotopes
and of mercury atoms respectively (r1c^.r2), and p H g is the
density of the mercury vapour.
A diagram of the diffusion unit (54) used to separate isotopes
by this method is given in Fig. 17. It is not necessary, according
to a recent communication (53), to connect the pumps by a
capillary tube. With a battery of fifty-one pumps it was found
that, a 32 % separation of C12 and C13 could be effected from
300 c.c. of methane at 1*8 mm. pressure. The advantage of this
newer arrangement of Hertz's original method is that there
is no accumulation of impurities within a porous wall, and that
higher separation factors are possible.
Another method of separating liquid and gaseous mixtures,
which was first developed by Clusius and Dickel(55), is based
upon Chapman's early study of thermo-diffusion in gases(56).
This method promises to be very efficient in separating
isotopic mixtures, and is simple to operate. If a hot wire passes
axially down a tube containing a mixture of gases, the gases
under the combined influence of radial and axial diffusion and
of convection undergo a partial separation into the two
components (57,58). Let the separation unit consist of a hot and
a cold surface, plane and parallel, between which is confined the
solution being disproportionated. The temperature gradient
exists across this solution, in the ^-direction, and convection
occurs in the ^-direction. Then the separation velocity is
governed by the differential equation (57)

(46)
sW*»--ar'
where v(x) denotes the convection velocity as a function of x,
 SEPARATION OF GAS MIXTURES AND ISOTOPES 81
and T the temperature. For a separating unit of length I
there is a characteristic time

which gives the order of magnitude of the time required for

the separation equilibrium to be approached. In a normal

B"
Fig. 17. Mercury diffusion pumps. " A " is the Hertz pump and " B " is a unit
with a modified mercury jet. The pumps are connected in series as shown.
(Scherr (54).)

apparatus this would be 1000 days for liquids, and < 1 day
for gases. The equilibrium state is defined by the equation
G
^Cv^=°- (47)
82 MOLECULAR FLOW IN CAPILLARY SYSTEMS
It was found in a liquid mixture D20(32°/o) + H2O(68o/o) that
a separation of up to 4-8 % was established in two days—
about j 1 ^ of the equilibrium separation in the stationary state.
The method was employed successfully to disproportionate
the isotopes Zn64, Zn66, Zn68, as aqueous solutions of zinc sul-
phate (57). The isotopes were estimated spectroscopically. The
method has also been used (59) in the separation of isotopes
of Cl and of Hg. The method is of quite general applicability
in disproportionate or sedimenting such mixtures as
ZnCl2, ZnSO4 in H2O(57), w-Hexane in CC14(57),
NaCl, Na2SO4 in H2O(57), C6H6 in C6H5C1(57),
Chlorophyll in CC14(57).
Table 9 illustrates its efficiency in separating gas mixtures (55).
TABLE 9. The Clusius-Dickel method of separating
gas mixtures
Total pressure approximately 1 atm.
Length of Composition at
Original gas separating Temp,
mixture column diff. "Heavy" "Light"
% cm. °C. end end
25 Br2 75 He 65 -300 100% Br2 (liq.) 100% He
40CO 2 60 H 2 100 -600 100% CO2
20 O2 80 N 2 (air) (a) 100 -600 42% O2
(b) 290 -600 85% O2
Normal Ne 260 -600 At. wt. 20-68
(at. wt. =20-18)
23H 87 C1 77H 35 C1 290 -600 40% HS7C1
60% H35C1

NON-STATIONARY STATES OF CAPILLARY GAS FLOW'

Stationary streaming proceeds at the same rate whether the

molecules have a long life in the adsorbed phase or not,
according at any instant to the equation
dn IB 1 (Pl p2\(2-f)
(2)
dt 2L<J(2nMR)\y/T1 JT2) f
already discussed. In the stationary state every molecule
leaving the wall is replaced by one striking the wall. Clausing (7),
 STATES OF CAPILLARY GAS FLOW 83
however, has shown how non-stationary streaming may be
used to determine the lifetime of molecules in the adsorbed
phase, and heats of adsorption. His experiments and theoretical
treatment are of interest because they obviate any necessity
for direct measurement of the very minute quantities ad-
sorbed on certain surfaces such as glass and crystals. The
essential apparatus consists of two large flasks connected
by a fine capillary. In one flask, for all times t > 0 the
pressure is p±; while in the capillary and in the second flask
the pressure p2 ~ 0. At time t = 0 the gas enters the capillary.
A certain time must elapse before the first molecule enters
the second flask, and this time is a function of the lifetime
of the molecules in the adsorbed state, being greater the
greater this lifetime.
In obtaining a solution for the diffusion into the capillary,
Clausing assumed that surface diffusion need not be considered
in comparison with bulk diffusion. The diffusion equation is
then once more* 2 -„
D
^x* = ~dl' (48)

The initial condition is, inside the capillary,

for t = 0, G = 0.
The boundary conditions are
for x = 0, C = Gx for all t,
for x = L, C ~ 0 for all t.
The solution of the problem then is

( ^ ) } ] (49)
2 2
if a = L /n D.
When x = L

) i^
^1 2(e
~" a *"*"* +c ~ 9 " a) * (5o)

* In equations (48) to (58) C denotes the number of molecules per unit length
of the capillary, including those adsorbed on the wall. It is given by equation (55).
6-2
84 MOLECULAR FLOW IN CAPILLARY SYSTEMS
Thus in time t there flows into the second vessel a number of
molecules Nt given by

so that

4 ;
£ ^9V ' (51)

The value of Z>, Clausing was able to deduce from kinetic

theory and from his own consideration of the mean life, r e , of
an adsorbed molecule. This value of D is

D— L 9 (52)

so that a=-

and T=
' -3"U) a -¥- (54)

It was also shown from simple considerations that

(55)

In equations (52)-(55) r denotes the pore radius, w the root

mean square velocity of a molecule, and N± the number of
molecules in the first vessel per c.c; Cx is of course the same
for stationary as for non-stationary flow. By treating
stationary flow as a diffusion, one obtains
dn 8nr* ,„ _ _ dC n

Thence, by combining (51), (52), (55) and (56), one obtains

(57)
 STATES OF CAPILLARY GAS FLOW 85
in which

For large values of t

) ( ^ y P , (58)
That is, the diffusion rate tends for longer times to become
that for stationary streaming, as would be anticipated. The
function ft(t/oc) then approaches the asymptote ir(t/oc) = 1,
corresponding to stationary flow.
In Clausing's apparatus the flow through two capillaries
with diameters in the ratio 1:10 was measured. The pressure
on the ingoing side was virtually constant, and on the outgoing
side was measured by two ionisation gauges. The current,
which was then amplified, was proportional to the pressure,
and was registered by a linear galvanometer with a vibration
period of ^ sec. Thus the deflection was found as a function
of time, and so, from the calibration curve for the instruments,
the value of p and therefore of Nt = Pcc\Jr(tj(x) was deduced.
The record of the diffusion process was obtained photo-
graphically on a moving photographic plate. Employing the
foregoing theory to interpret his photographs, Clausing
obtained the lifetimes of molecules of argon, nitrogen, and
neon on glass. He also obtained the lifetime of argon, r e , as
a function of temperature, his results being expressible as
linear logr e versus I/I 7 curves (T in °K.) (Fig. 18). According
to the theory of Frenkel(60),
Te~ToeAHlRT, (59)

where T0 denotes the vibration period of the adsorbed mole-

cule .on the solid, and AH is the heat of adsorption. Thus
the slope of the curve logr e versus l/T gives the heats of
sorption, which are collected for argon on glass in Table 10.
The heats of sorption obtained by Kalberer and Mark(6i
86 MOLECULAR FLOW IN CAPILLARY SYSTEMS

for argon on dehydrated silica gel are fairly constant at

2500 cal./atom, a figure somewhat lower than the above value
for glass. Clausing remarks that the results are not always
easy to reproduce. For argon on glass between 78° and 90° K.,
the approximate value for re is
7 c ^l-7 x io-14e38O°/*T sec. (59a)

-S5
WO 1300

Fig. 18. The linear relationship between log re and l/T (T in ° K.). ®, cleaned
with chromic acid; • , A> cleaned with hydrofluoric acid.

At 78° K., r c ^ 7 5 x 10~5 sec; at 90°K. r 6 ^3-l X 10"5 sec. For

nitrogen re was of the same order of magnitude, whilst for
neon re is less than 2 x 10~7 sec. These lifetimes accord well
with our knowledge of the relative adsorbability of these

The establishment of sorption equilibria in porous solids is

another very important case of non-stationary flow into
capillaries. These capillaries may range in size from molecular
dimensions, when activated diffusion occurs, to capillaries
 STATES OF CAPILLARY GAS FLOW 87
of macroscopic dimensions. Any theoretical treatment is
rendered more difficult by the occurrence of an unknown
distribution of pore sizes and channel sizes. It is true of a
limited number of adsorbents, however, that the pores are
nearly all of one size, for example in dehydrated chabasite
or analcite crystals (62), from which heat and evacuation
removes water without destroying or fundamentally changing
the crystal skeleton. A formal treatment based on Fick's law
without a kinetic picture of what is happening has been applied

TABLE 10. The adsorption heat for argon on glass

Cleaned with hydrofluoric acid Cleaned with chromic acid
AH AH
Remarks Remarks
(cal./atom) (cal./atom)
2450 The first three results 3320 The first two results are
2430 not quite trustworthy, 3430 not so trustworthy, but
3300 the last result free 4050 the last four are free
3930 from objection 3440 from objection
3810
3880
Mean of
last four
results:
3800

to heulandite and analcite systems by Tiselius(63>; and a

kinetic theory interpretation of the diffusion has been outlined
by Hey (64) (Chap. III). Other attempts have been made, based
upon the kinetic theory of gases, to derive formulae for the
rate of flow of gases into capillary solids (65,66). Owing to
the complexity of the problem such attempts are not very
successful. Often sorption kinetics, after some time has elapsed,
tend towards a pseudo-unimolecular law. When oxygen,
nitrogen, and hydrogen were sorbed by charcoal (67) at low
temperatures, the "unimolecular" velocity constants then
stood in the ratio O 2 :N 2 :H 2 = 1:1-1:3-2. Molecular flow
would require O 2 : N 2 : H 2 = 1:1-07 : 4-0. Finite sorption velo-
cities may in some cases have their origin in slow dissipation
of the heat of sorption, or in the slow displacement of surface
88 MOLECULAR FLOW IN CAPILLARY SYSTEMS
impurity. In the experiments described, however, the first
possibility was avoided by using minimal amounts of gas, and
the second by the most thorough heat-evacuation treatment.
Other authors (68,69,70) have observed finite sorption rates
in charcoals, and a number of the equations used to express
the velocity of sorption in various adsorbents have been
summarised by Swan and Urquhart(7i).

REFERENCES
(1) Warburg, E. Ann. Phys., Lpz.9 159, 399 (1876).
(2) Knudsen,M. Ann. Phys., Lpz., 28, 75 (1909); 35, 389 (1911).
(3) Gaede, W. Ann. Phys., Lpz., 41, 289 (1913).
(4) v. Smoluchowski, M. Ann. Phys., Lpz., 33, 1559 (1910).
(5) Buckingham, E. Technol. Pap. U.S. Bur. Stand. No. 183 (1921).
(6) Klose, W. Phys. Z. 31, 503 (1930); Ann. Phys., Lpz., 11, 73
(1931).
(7) Clausing, P. Ann. Phys., Lpz., 1, 489, 569 (1930).
(8) Herzfeld, K. and Smallwood, M. In Taylor's Treatise on Physical
Chemistry, 1, 169 (1931).
(9) Treatise on Physical Chemistry, 1, 171 (1931).
(10) Brillouin, M. Lecons sur la Viscosite, Paris (1907).
(11) Hagenbach, E. Ann. Phys., Lpz., 109, 835 (I860).
(12) v. Smoluchowski, M. Bull. Acad. Krakau, 143 (1903).
(13) Reynolds, O. Philos. Trans. 174, 935 (1883).
(14) Ackerel, T. Handb. Phys. 7, 304, Berlin (1927).
(15) Stodola, A. and Lowenstein, L."* Steam and Gas Turbines, 1, 55-
67, New York (1927).
(16) Eason, A. B. Flow and Measurement of Air and Gases, London
(1919).
(17) Blasins, H. ForschArb. dtsch. Ing. 131 (1913).
(18) Manegold, E. Kolloidzschr. 81, 164 (1937).
(19) Kolloidzschr. 81, 269 (1937).
(20) Adzumi, H. Bull. chem. Soc. Japan, 12, 285 (1937).
(21) Bull. chem. Soc. Japan, 12, 292 (1937).
(22) Ramsay, W. and Collie, N. Proc. Roy. Soc. 60A, 106 (1897).
(23) Donnan, F. Phil. Mag. 49, 423 (1900).
(24) Sameshima, J . Bull. chem. Soc. Japan, 1, 5 (1926).
(25) Adzumi, H. Bull. chem. Soc. Japan, 12, 304 (1937).
(26) Preston, E. J. Soc. Glass. Tech. 18, 336 (1934).
(27) Kanz, A. Arch. Eisenhiittenw. 2, 843 (1929).
(28) Clews, F. and Green, A. Trans, ceram. Soc. 32, 295, 472 (1933).
(29) Bansen, H. Arch. Eisenhiittenw. 1, 687 (1927-8); Stahl u. Eisen,
48, 973(1928).
(30) Bremond, P. C.R. Acad. Sci., Paris, 196, 1651 (1933).
 REFERENCES 89
(31) Fancher, G. and Lewis, J. Industr. Engng. Chem. 25, 1139
(1933).
(32) Bartell, F. J. phys. Chem, 15, 659 (1911); 16, 318 (1912).
Bartell, F. and Carpenter, D. C. J. phys. Chem. 27, 252 (1923).
(33) Bartell, F. and Miller, F. Industr. Engng Chem. 20, 738 (1928).
(34) Bartell, F. and Osterhof, H. J. phys-. Chem. 32, 1553 (1928).
(35) Schlichter, C. Rep. U.S. geol. Surv. 2, 305 (1897-8).
(36) Chilton, J. and Colbourn, A. Industr. Engng Chem. 23, 913
(1931).
(37) Chalmers, J., Taliaferro, D. and Rawlins, E. Trans. Amer. Inst.
min. (metall.) Engrs, 98, 375 (1932).
(38) Wilde, H. and Moore, T. Oil Weekly, 67, 34 (1932).
(39) Cf. Fancher, G. and Lewis, J. Industr. Engng Chem. 25, 1139,
1143 (1933).
(40) Muskat, M. and Botset, H. Physics, 1, 27 (1931).
(41) Wyckoff, R., Potset, EL, Muskat, M. and Reed, D. Bull. A?ner.
Ass. Petrol. Geol. 18, 161 (1934).
(42) Green, H. and Ampt, G. J.Agric. Sci. 4, 1 (1911-12); 5, 1 (1912-
13).
(43) Narayanamurti, O. Z. Ver. dtsch. Ing. Bleiheft, Nr. 2, 13 (1936).
(44) Manegold, E. (with Hennenhofer). Kolloidzschr. 81, 278 (1937).
(45) Raisch, E. and Steger, H. Gesundheitsing, 57, 553 (1934).
(46) Raisch, E. Z.. Ver. dtsch. Ing. 80, 1257 (1936).
(47) Hertz, G. Z. Phys. 79, 108 (1932).
(48) Farkas, A. Light and Heavy Hydrogen, p. 120, Cambridge (1935).
(49) Harmsen, H. Z. Phys. 82, 589 (1933).
Harmsen, H., Hertz, G. and Schutze, W. Z. Phys.90,703 (1934).
(50) Wooldridge, D. and Smythe, W. Phys. Rev. 50, 233 (1936).
(51) Hertz, G. Z. Phys, 91, 810 (1934).
(52) Barwieh, H. Z. Phys. 100, 166 (1936).
(53) Capron, P., Delfosse, J., Hemptinne, M. de and Taylor, H. S.
J. chem. Phys. 6, 656 (1938).
Hemptinne, M. de and Capron, P. J. Phys. Radium, 10, 171
(1939).
(54) Scherr, R. J. chem. Phys. 6, 252 (1938).
(55) Clusius, K. and Dickel, G. Naturwissenschaften, 33, 546 (1938).
(56) Chapman, S. Phil. Mag. 7, 1 (1929).
(57) Korsching, H. and Wirtz, G. Naturwissenschaften, 20/21, 367
(1939).
(58) Waldmann, L. Naturwissenschaften, 14, 230 (1939).
Grinten, W. Naturwissenschaften, 19, 317 (1939).
Wirtz, K. Naturwissenschaften, 20/21, 369 (1939).
(59) Clusius, K. and Dickel, G. Naturwissenschaften, 28, 487, 148
(1939).
Groth, W. and Harteck, P. Naturwissenschaften, 34, 584 (1939).
(60) Frenkel, J. Z. Phys. 26, 117 (1924).
(61) Kalberer, W. and Mark,.H. Z. phys. Chem. 139A, 151 (1929).
90 MOLECULAR FLOW IN CAPILLARY SYSTEMS
(62) Barrer, R. Proc. Roy. Soc. 167A, 392 (1938).
(63) Tiselius, A. Z. phys. Chem. 169A, 425 (1934); 174A, 401 (1935);
Nature, Lond., 133, 212 (1934).
(64) Hey, M. Miner. Mag. 2\\ 99 (1935).
(65) Damkohler, G. Z. phys. Chem. 174A, 222 (1935).
(66) Wicke, E. Z. Elektrochem. 44, 587 (1938).
(67) Barrer, R. and Rideal, E. K. Proc. Roy. Soc. 149A, 231 (1935).
(68) Blythswood, Lord and Allen, H. S. Phil. Mag. 10, 497 (1905).
(69) Herbst, H. Kolloidchem. Beih. 21, 1 (1925).
(70) Dietl, A. Kolloidchem. Beih. 6, 127 (1914).
(71) Swan, E. and Urquhart, A. J. phys. Chem. 31, 251 (1927).
 C H A P T E R III

GAS FLOW IN AND T H R O U G H CRYSTALS

AND GLASSES
STRUCTURES OF SOME SILICATES AND GLASSES
In this chapter are described the phenomena of gas flow in
glasses and in crystals. While the structure of alkali halides
and similar more simple crystals is well known, it is not so
with the large families of derivatives of silica. Nevertheless,
considerable advance has been made towards an under-
standing of structural relationships in some at least of these
families of compounds. Since many silicates are of importance
to the present chapter, a few of the known structural relation-
ships will be given for substances such as silica, silicate glasses,
zeolites, mica, clays, feldspars and ultramarines, the first four
of which are of great importance in discussing the permeability
data.
Certain units occur throughout all these substances, for
instance, a silicon atom (or S i + + + + ion) surrounded by four
oxygen atoms (or 0 — ions). Sometimes one finds an analogous
aluminium tetrahedron, whose resultant negative charge is
balanced by cations, and which, as in the aluminosilicates,
may replace the silicon tetrahedron. These tetrahedra are
then linked together to form chains and closed rings. Four
tetrahedra can form a 4-ring; one can also find analogous
6-rings, and in certain irregular or acrystalline silicates (such
as glass) also 5- and 7-rings.
If the structure is composed entirely, or nearly entirely, of
linked silicon-oxygen tetrahedra, one can obtain the three
crystalline silica structures, showing a- and ^-modifications.
These are quartz, cristobalite and tridymite. Also the irregular
netting of the tetrahedra leads to fused silica glass. Of the
crystalline forms, quartz is the most dense; and /?-cristobalite,
next to fused silica, is the most open. In /?-cristobalite, zigzag
92 GAS FLOW IN CRYSTALS AND GLASSES
chains are cross-linked forming wide 6-rings, about 6 A. across,
but in quartz one has spiral chains also cross-linked to give a
ring structure, the widest aperture in the lattice being about
4 A. (from ion centre to ion centre).
As already observed, aluminium tetrahedra may replace a
part of the silica tetrahedra in the three-dimensional network,
which becomes anionic, since Si + + + + is replaced by Al +++ .
Cations now enter the lattice to restore electrostatic balance,
with a corresponding distortion of the parent structure. In
this way nepheline is built up from tridymite. Sometimes more
cations are incorporated than are needed for electrostatic
balance, and the nett positive charge is then in its turn
balanced by the incorporation of anions SO4", Cl', S". This
type of structure is typical of the ultramarines, of which
lazurite, whose simplest empirical formula is Na8Al6Si6024
(S", SO4"), is an example.
These substances in which extra cations or anions are
incorporated in the network are called interstitial compounds.
Another group of interstitial compounds, of the greatest
interest from the viewpoint of gas flow, is the zeolites, in which
the very open anionic frameworks, electrostatically balanced
by cations, contain also the neutral molecule water. The
water may sometimes be replaced with varying lattice changes
by the substances NH3, H2S, N2O, H2, N2, Ar, He, I 2 , Hg, or
CH3OH, so giving rise to very diverse types of diffusion
system, and heterogeneous equilibrium. In the zeolites, the
ultramarines and the clays, interstitial cations may undergo
replacement and diffusion, although these properties are most
readily observed in zeolites.
W. H. Taylor (l) has classified the zeolites, according to their
structural properties, into
(1) rigid frameworks,
(2) semi-rigid frameworks,
(3) platy frameworks.
In the first class are chabasite and analcite, in which the
water-containing interstices do not collapse markedly upon
 GAS FLOW IN CRYSTALS AND GLASSES 93
dehydration at moderate temperatures (i, 2). In these structures
the frameworks are thus to a large extent independent of the
neutral molecules occupying interstices inside the frame-
work. In the second class are the fibrous zeolites natrolite,
scolecite, mesolite and edingtonite, all of which are formed
by cross-linking chains of (Si-Al)-O tetrahedra in different
ways (Fig. 19). The fibrous zeolites undergo reorganisation
around the water-bearing interstices on dehydration, without
any appreciable shrinkage of the chain length. The third class
of zeolites is exemplified by heulandite, of which the great
shrinkage in one dimension on dehydration and the X-ray
diffraction patterns are compatible with a platy structure of
two-dimensional lattice layers, separated by water molecules
and cations. These cations serve to bind the anionic laminae
together. In Fig, 20 a possible laminated lattice is shown (i).
Mica is another case of a laminated silicate crystal, while
laminar crystals of silica itself have recently been reported (3).
Clays are aluminosilicate structures often having close simi-
larities to the zeolites. Montmorillonite, for example, may be
regarded as the counterpart of the zeolites heulandite and
stilbite. Like heulandite it swells greatly upon hydration(4),
as is indicated by Fig. 21. The clays also show base-fexchange
properties like the zeolites. In general they contain larger and
less definite channels.
Glasses are anhydrous silicates,* in which A l + + + or B + + +
may replace some S i + + + + , and are based upon the irregular
network of silica-glass. Electrostatic balance is restored by
the various basic oxides of types M2O, MO, M2O3 which they
contain. By varying the proportions of constituents the most
diverse properties may be conferred.
Thus one finds in silicate chemistry many examples of
complex structures resolved by the classical chemical and
X-ray methods. While diffusion phenomena have been little
studied in many of the structures (feldspars, ultramarines, and
* Experiments upon the diffusion of deuterium through silica glass showed,
however, that some hydrogen, probably as hydroxyl groups, still remained in the
silica, since the deuterium content of the diffusing gas decreased (4a).
 Thomson/ft

n't

0 1 2 3 • 54

Fig. 19. The structures of some fibrous zeolites. The structure of one tetrahedron
chain is shown in (d'), in which the large circles represent silicon or
aluminium atoms, the small circles oxygen atoms, and the heights of the
atoms are given in A. The same chain is represented diagrammaticallv
in (d), where the numbers show the heights of silicon and aluminium
atoms as multiples of c/8 (the c-axis is 6-6 A., so that 3c/8 = 2-5 A..
5c/8 = 4-1 A.). The linked chains are shown in (a), (b), (c) as arranged in
edingtonite, thomsonite and the natrolite group, respectively, and in each
the unit cell is indicated by dotted lines.
Fig. 20. A lattice of two-dimensional laminae. Large circles represent oxygen
atoms, small circles silicon or aluminium atoms. The upper diagram repre-
sents a portion of an infinite sheet of tetrahedra in which all vertices are
supposed to point upward. If these vertices fie on a reflection plane, a
second similar tetrahedron sheet, in which all vertices point down, is linked
to the first. The lower diagram represents the appearance of the linked
sheets, which form a tetrahedron framework of finite thickness, when viewed
in the direction indicated by the arrow in the upper diagram.

MOLECULES OF WATER PER UNIT CELL

2 U 6 8 10 12 15 20 25 30

0 10 20 3Q HO 50 60 70 80 90 100 1fO
WATER ~ % OF IGNITED MATERIAL
Fig. 21. The one-dimensional swelling of montmorillonite on hydration.
96 GAS FLOW IN CRYSTALS AND GLASSES
clays), in others (glasses and zeolites) considerable data have
been collected. It must be in the light of the main structural
features outlined that these data are examined. The mobility
of ions has been observed in ultramarines, clays, zeolites and
glasses—in the three former groups by base exchange experi-
ments, and in the latter because of its function as a hydrogen
electrode, or even as a sodium electrode. The mobility of
sodium (and potassium) ions in glass may be demonstrated in
an experiment used to prepare pure sodium in vacuo. The cell
NaNO 3 Na Conducting Glowing
Cu (molten) glass evacuated space tungsten filament
is constructed by dipping an evacuated soda glass bulb into
molten sodium nitrate, into which dips a copper electrode.
Inside the evacuated bulb is a hot tungsten filament, which,
by thermionic emission, renders the evacuated space con-
ducting. On making the tungsten negative with respect to
the copper, metallic sodium appears inside the glass bulb.

THE DIFFUSION OF WATER AND AMMONIA IN ZEOLITES

Introductory
Having briefly reviewed zeolitic structure and properties,
one is in a position to consider diffusion within these
crystals. The most complete and satisfactory studies are
those of A. Tiselius (5,6), whose methods and data will now
be considered. In crystals such as NaCl, lattice diffusion is
possible only at elevated temperatures because no channels
are available for the diffusion. Very often, as when thorium
diffuses in tungsten, diffusion occurs down the only available
channels, the grain boundaries and faults in the tungsten
crystal. In interstitial compounds such as zeolites there are,
when the crystal is dehydrated, numerous channels through
the lattice left by the evacuated water, and so diffusion by
spreading through interstices, as opposed to high temperature
processes of diffusion by place exchange, may occur.
 DIFFUSION OF WATER AND AMMONIA 97

The experimental and theoretical interpretation

of diffusion data
Tiselius(5,6) found that the double refraction of light by
these crystals was dependent upon the extent of hydration
of the lattice. The crystals were cut for this purpose in the
form of thin plates, and the rays were examined in a polarisa-
tion microscope. As a first approximation, Tiselius then
considered Fick's law to hold. To interpret his data he used
the solution of the diffusion equation —- = Z) ^-^ for the
^ dt dx2
diffusion of a substance along the co-ordinate x into a medium
of infinite length through a surface normal to its length. The
solution is (Chap. I):
c_p-a7T — "Tz: I
where /? = xj2^J(Dt); Cp = the concentration at plane x = 0,
assumed constant; Cx = the concentration at plane x == x
at time t\ Co = the concentration, constant for all x, at the
beginning of diffusion.
For this solution of Fick's equation the boundary conditions
are:
C = Op for x = 0 and all times t,
G = Co for x > 0 and t = 0,
C = Co for x = oo and all t.
The equation provides three methods of obtaining D, the
diffusion constant, which are:
(i) Measure a; as a function of t for a constant known (7;
(ii) Measure G as a function of t for a given x;
(iii) Measure C as a function of x for a given i
All three methods were employed successfully in Tiselius's
experiments.
The laminated lattice of heulandite has the laminae parallel
to the 010 plane. The diffusion of water normal to this plane
was found to be D olo < 7 x lO-^cm^sec." 1 at 20° C; but
diffusion of water parallel to the 010 plane was rapid, though
98 GAS FLOW IN CRYSTALS AND GLASSES
proceeding at different rates across 201 planes and 001 planes.
Thus one is introduced to the phenomenon of diffusion
anisotropy, exhibited also in "platz-wechsel", or place-
change, diffusion of ions in salts (4).

The diffusion of water in the 010 plane, normal

to the 201 plane
The first of the three methods of determining the diffusion
constant, based on Fick's law, depends upon the linear
relation between x2 and t, at constant C. Fig. 22 gives the
x2 — t curve at a constant Cx value of 13-87 % of water in the

2000
Seconds
Fig. 22. x'2 as a function of t at constant C.

lattice. Here Co was 13-45 % and Cp was 19-67 %, the experi-

ment being performed at 20° C. Thus
7~> O.A v l f | - 7 p m 2OAn —1
J-^OQI — O T: ^s 1 v U1JL1* o C l i ,

When repeated at different values of Cx, there was evidence

that D varied according to the water content of the lattice.
When C was plotted as a function of t at a constant x} the
curve of Fig. 23 was found; and when C was plotted as
a function of x at constant t, the curve of Fig. 24 was
obtained.
In this latter figure the dotted curve gives the value of a
C — x curve computed using D = 3-9 x 10- 7 cm. 2 sec.-\ and
 DIFFUSION OF WATER AND AMMONIA 99
assuming D to be independent of C. In Tables 11 and 12 are
collected the values of D according to these last two methods,
as functions of G, using the simple Fick law — = D -^-^ in
their evaluation.

500 WOO 1500 2000

t in seconds

Fig. 23. C as a function of t at constant r.

0 50 100 150
t in seconds
Fig. 24. C as a function of x at constant t.

When the difference between Cp and Co was small the values

of D were nearly constant, and the simple equation-^- = D-^
^
more closely obeyed, as is indicated by Table 13.
dO
Hitherto the simplest
r equation -z- = has been used.
dt dx
7-2
100 GAS FLOW IN CRYSTALS AND GLASSES
TABLE 11. Diffusion constants from the relation between C
as a function of t at constant x
Cv = 19-67%, Co = 12-30%, x = 9-832 x 10~3 cm., T = 20° C.

t C9-Cx .0 £>xlO7
% sec. cP-c0 p cm.2 sec."1
130 $1 0-905 1181 2-4
13-5 93 0-837 0-986 30
140 112 0-770 0-849 3-4
14-5 140 0-701 0-734 3-6
15-0 169 0-634 0-639 3-9
15-5 212 0-566 0-553 4-2
160 283 0-498 0-475 4-2
16-5 377 0-430 0-402 4-4
170 577 0-363 0-334 4-4
17-5 1060 0-294 0-267 3-6
180 1755 0-227 0-204 3-7

TABLE 12. Diffusion constants from O as a function of x

at constant t
Cp = 19-67%, C'o = 8-3%, t = 1200 sec, T == 20°C.

x in units Cp-Cx DxlO7

% 2-809 x lO"3 cm. cP-c0 cm.2 sec."1
9 108 0-938 11
10 107 0-850 1-8
11 105 0-762 2-6
12 99-7 0-675 3-4
13 930 0-587 4-2
14 830 0-499 50
15 70-8 0-411 5-6
16 56-5 0-323 60
17 380 0-235 53
18 210 0147 4-2
19 7-2 0058 3-3

TABLE 13. Diffusion constants when the crystal ivas nearly

saturated with water
Cp = 19-67%, Co = 16-20%, T = 20 c C , x = 11-353 x 10"3 cm.

t c,-cx DxlO7
% sec. P cm.2 sec. *
op-c0
16-5 60 0-915 1-218 40
170 119 0-770 0-894 40
17-5 218 0-625 0-627 41
180 398 0-482 0-457 4-2
18-5 984 0-342 0313 3-8
190 1520 0-266 0-240 41
 DIFFUSION OF WATER AND AMMONIA 101

However, there is considerable evidence that D may vary

with C(T), in which case the equation to be solved is

dt ~ dx\
for which solutions(5,6) have been given (Chap. I). Of use in
evaluating D is the expression
~ 1 dxc
xdC,

when C is measured as a function of x for a given time interval.

The integral can be evaluated graphically, and the derived
value of D refers to that value of C at x = xv Employing this
method, Tiselius succeeded in obtaining a number of diffusion
constants for various values of C, which, as Table 14 shows, do
indeed depend noticeably on C.

TABLE 14. Diffusion constants evaluated from -=- = ^ - 1 D - ^ I

Temp. 20° C, £„ = 19-67 %.

DxlO 7 DxW DxlO 7

c (cm.2 sec."1) (cm.2 sec.*1) (cm.* sec.-1)
(exp. with (exp. with (exp. with
% tfo = *3%) Co = 13-20%) Co = 16-20%)
10 004*
11 0-2f — —
12 0-7 —
13 1-3
14 20 21 —
15 2-7 2-6 —
16 30 3-5
17 4-0 4-2 40
18 40 41 4-2
19 3-3 3-5 41

* Very uncertain. t Uncertain.

These results are numerically somewhat different from

those obtained earlier with the simple Fick treatment. They
show in each case a maximum which might be explained as
follows. The heat of sorption is greatest at small Co; therefore
the water molecule is most strongly anchored and is not very
mobile. When the lattice has its full complement of water,
mobility must also be slight, since few water molecules will
102 GAS FLOW IN CRYSTALS AND GLASSES
have vacant lattice cells to migrate into, and therefore for
intermediate values of Co the greatest mobility is possible.

Diffusion anisotropy in heulandite

The permeability perpendicular to the plane of laminae in
heulandite is negligible; parallel to this plane it is > 105 times
as rapid, at room temperature. But different directions in the
010 plane have different diffusion constants. This is shown by
the photomicrographs in Fig. 25 in which the dark band gives

Fig. 25. Progress of diffusion in heulandite, showing diffusion anisotropy.

a measure of the rates of advance of diffusion normal to the

201 and 001 faces, the most rapid advance being found normal
to the 201 face. The relative rates of advance of the dark
bands are given in Table 5.
TABLE 15. Diffusion anisotropy

c% •^201 • a '001

13-21 3-6
14-43 3-4
15-50 3-5
16-25 3-2
17-36 3-4

The ratio = 11*6, and is constant at a series of

#001
values of Co. When the diffusion velocities normal to the faces
201 and 201 were measured, it was found that they were equal.
Thus for the particular sample of heulandite studied
DOoi'D2O1:D2ol 1:11-6:11-6.
 DIFFUSION OF WATER AND AMMONIA 103
The diffusion anisotropy phenomenon was studied in a
number of other heulandites, and proved substantially the
same in all of them. Always the minimum diffusion rate
parallel to the 010 plane was normal to the 001 plane, and the
following anisotropy ratios were found:
i) 2 0 1 at 20° C. Aoi
Heulandite from and C = 15-0% Z>001
Dalur, Faroerne 2-7 x 10- 7 13-6
Teigarhorn, Iceland 3-7 15 to 20
Rodefiord, Iceland 3-5 15 to 20
Sulitelraa, Norway 2-3 15

The energy of activation for the diffusion of water

in heulandite
Tiselius measured the diffusion constants at a number of
temperatures, finding the values given below:

Temp. D201 x 10' AmxlO7

°C. cm. 2 sec." 1 cm.2 sec."1
20-0 2-7 0-23
33-8 41 0-45
461 4-8 0-66
600 7-6 1-45
750 111 2-8

From these figures he computed the energy of activation

for diffusion normal to the 201 and 001 faces respectively to
be 5400 and 9140cal./mol. The temperature dependence of
the diffusion constants he found did not depend appreciably
upon the amount of water in the lattice, although we have
seen that their absolute magnitudes do.

Diffusion processes in analcite-ammonia

These experiments were made upon a most complex zeolite,
heulandite, and were later extended to analcite, which has
only one type of lattice hole and a cubic or pseudo-cubic
symmetry(7). .Once more the optical experiments were per-
formed upon narrow plates cut parallel to cube or diagonal
faces. The material was thoroughly outgassed and the sorption
processes with ammonia were observed. In Table 16, which
104 GAS FLOW IN CRYSTALS AND GLASSES

gives the absolute diffusion constants, A TMenotes the difference

of the refractive indices of the ordinary and extraordinary
rays, which was used as a measure of the amount sorbed.

TABLE 16. Diffusion of NH 3 through 0-92 mm. plate

cut parallel to cubic surface
Temp. = 302° C , pressure = 761-2 mm., volume of NH 3 sorbed = 27-5 c.c./g.

Z)xlO 8
t
sec. v< zLTxlO7 cm.2 sec."1
0 0 0 1-4
900 30 40 10
3,600 60 60 1-4
8,100 90 119 1-2
14,400 120 151 1-2
22,500 150 183 —

The diffusion rate was the same for a plate cut parallel
to the diagonal surface, so that no anisotropy occurs. The
temperature dependence of the diffusion velocity was found
using powdered analcite, and the equation

AQoc(Cp-(
where AQ denotes the amount sorbed. This equation supposes
that D does not depend upon AQ. If one writes
AQ =0=f{Dt)
Cp — Co
only, the curve 6 versus the time, t, or Jt, should be the same
at all pressures. I t was shown that this was so for two
rate curves at 302° C, one with analcite outgassed and
p = 758-6 mm., the other with the analcite saturated at
p0 = 578-6, a further dose of ammonia being admitted at
1385-4 mm. Then since 6 =f(D,t) only,
Dm In
Vrp
1
1
Dn
and a simple method is available to determine the temperature
coefficients and activation energies. The temperature de-
pendence of the diffusion coefficient gave the activation energy
for diffusion as 11,480 cal./mol. It was also noted that the
 OF WATER AND AMMONIA 105
ratio of the diffusion constants at two different temperatures
was not dependent upon the amounts of ammonia sorbed.

Other observations on zeolite systems

In a study (2) of sorption by various zeolites, platy, fibrous,
and rigid three-dimensional networks, several points of interest
were established. Ammonia sorption was most rapid in the
lattices of the three dimensional networks (chabasite, analcite).
It was least rapid in suitably outgassed fibrous and platy
zeolites (natrolite, scolecite, and heulandite). Heulandite, a
420

^—~~~

J ! 3J0
autooa. alytic >' 08
absorption >v/
/
220 «

^-adsorption
120 00 /
1
t (in days) *in min )
Fig. 26. Sorption velocities. NH3 on natrolite, showing tendency to autocatalytic
sorption rate curves. (NH3 sorbed before commencing expt. = 12-11 c.c.
at N.T.P. Inset shows NH3 on heulandite, sorption rate following the
parabolic diffusion law.)
laminated crystal like mica, outgassed at low temperatures
(130° C.) to prevent lattice collapse, sorbed ammonia a t first
rapidly and then more and more slowly, suggesting a rising
activation energy with charge. When heated to 330° C , how-
ever, heulandite sorbed ammonia much more slowly. Some
profound change in the bonding between laminae must have
been effected by the high temperatures, associated with the
more complete removal of water. For heulandite crystals it
will also be seen (Fig. 26) that the sorption rate obeyed the
parabolic diffusion law initially, although this law quickly
broke down. I n natrolite, a fibrous zeolite, sorption in its early
106 GAS FLOW IN CRYSTALS AND GLASSES

stages was autocatalytic (Fig. 26) and a more or less well-

defined ammoniate resulted. The experiments bring out an
important point: activated diffusion in zeolites may be a
diffusion of interfaces in certain cases (natrolite) and not
a diffusion of molecules down a concentration gradient (chaba-
site, analcite, heulandite).
Another point of interest was the difference in the velocity
of ammonia sorption by analcite observed by Barrer(2) and
Tiselius(6). The former found sorption rapid at temperatures
as low as 200° C ; the latter found that sorption equilibrium
could not be established at 270° C , so slow was the ammonia
uptake. These experiments all serve to show the complexity
of behaviour of this very interesting series of diffusion systems.

A kinetic theory of diffusion in zeolites

These systems provide a very complete investigation of
the diffusion of vapours in zeolites, and the types of
phenomena likely to be encountered. They also provide an
interesting application of the methods outlined in Chap. I.
Kinetic theory may also yield equations of acceptable form
for velocities of sorption and for equilibrium data. Such
equations were derived by M. Hey (7,8), who expressed
Tiselius's data in terms of them and concluded that they were
satisfactorily able to represent Tiselius's experiments. Hey
derived his equations on the assumptions that water or other
vapours may occupy all or part of definite lattice positions,
and that the water if given an energy of activation E may
become mobile. A plate of zeolite is mounted with a free
surface of unit area exposed to the vapour, which diffuses
normal to this surface. Water molecules impinging on the
surface with an energy Ex normal to the surface may, if they
strike within certain areas, enter that lattice hole nearest the
surface. Conversely, molecules in the first lattice hole by
acquiring an energy E2 may re-evaporate, or by acquiring an
energy E may diffuse into the lattice. E2 is usually regarded
as being greater than E. In the diffusion problem two extreme
cases arise:
 DIFFUSION OF WATER AND AMMONIA 107
(1) Hydration of the zeolite is governed by the sorption
and desorption rates at the surface.
(2) Hydration is governed by the diffusion rate into the
crystal.
In the latter case __ d^(2E) _EjRT

where d denotes the distance between two successive points

of equilibrium of a water molecule in the lattice, ^ is a
correction term for anharmonicity of vibration of water in
the crystal, and the other symbols have their usual significance.
When E is a function of x, the vacant fraction of water posi-
tions, one may write E = EO{1 -+/(#)}. The equations involve
simplifying assumptions, but were applied to Tiselius's data for
water-heulandite diffusion systems. Hey found for this system
that for diffusion across the 201 face, from Tiselius's data,
E201 = 4-79(1 + 0-13#) x 103 cal./mol. from which d\x = 3-66 A.,
a reasonable value since x is approximately unity. The diffusion
data across the 001 face gave a value of
EW1 = 7-3(1 + 0-08^) x 103 cal./mol. and d/x = 19 A.,
by no means so likely a value. One difficulty is that all the
water in heulandite is not held with the same energy; Hey
considered three groups of water lattice positions to occur. The
data above apply only to the most volatile group (13-11 %
hydration). Table 17 gives the calculated and observed diffusion
constants computed using the data given above.
Similar agreement was found for diffusion across the 001
plane. It may be concluded that an equation of the type
deduced by Hey can represent the experimental results when
the constants are suitably chosen.
Tiselius's data for ammonia-analcite systems however
allow the simplified form

to be applied, since E does not appear to vary with x. Hey

calculated Eo = (1-35 + 0-05) x 104 cal./mol. of ammonia from
Tiselius's data, while from X-ray examination d — 5-93 A
108 GAS FLOW IN CRYSTALS AND GLASSES
Thus D*02°c = (3-5 ± 2) x 10~8 cm.2 sec."1, which constant
Hey (8) showed for various simple cases was one-third the
diffusion constant along a particular set of channels, as
measured by Tiselius. Thus from X-ray data and the calcu-
lated value of Eo the diffusion constant was (1*2 ± 0-7) x 10~8
cm.2 sec."1 compared with the directly measured value
1-2 x 10-8 cm.2 sec."1, both at 302° C.
TABLE 17. Diffusion data for heulandite, calculated
and observed
% H20 19 18 17 16 15 14
X 010 0-255 0-41 0-56 0-71 0-865
3-3, 4-0, 4-0, 3-5, 2-6, 21,
cm.2 sec."1 (obs.) 3-5, 41, 4-0, 3-6 2-7 20
41 4-2 4-2
107 x2Dim0'1 4-47 4-27 3-80 3-32 2-70 214
cm. sec." (calc.)
Temp. °C. ... 200 33-8 461 60 75 -—
7
10 x Z>201, for 2-7 41 4-8 7-6 111
x=0-71, in cm.2
sec.- 1 (obs.)
107 x D 201 , for 2 2-70 410 5-60 7-90 110 —
x = 0-71, in cm.
sec."1 (calc.)
% H2O 18-5 17-5 16-5 15-5 14-5 13-5
X 018 0-33 0-48 0-64 0-79 0-94
D^c- (obs. 1-7 20 1-8 1-8 1-8 1-7
•#201 °' ^calc- 1-93 1-93 1-94 1-98 203 209

DIFFUSION IN ALKALI HALIDE CRYSTALS

Introductory
Experiments designed to clarify photochemical processes in
alkali metal and silver halides have led to some interesting
studies on the diffusion of metal vapours (9, io), halogen
vapours (li) and hydrogen (12) in alkali halides. It is possible
to prepare transparent single crystals of the halides of the
greatest chemical purity. It was found that the halides can
act as solid solvents for a number of substances, and that
the properties of these mixtures or solutions permit one to
study photochemical processes and diffusion. Thus, potassium
 DIFFUSION IN ALKALI HALIDE CRYSTALS 109

vapour dissolves in potassium halides giving a number of

'' colour centres'' or' 'Farbzentren'' which give rise to a definite
absorption band in the crystal (Fig. 27) (13). In a similar way,
a solution of potassium hydride in potassium bromide gives
rise to another typical absorption band (Fig. 27). Exposing
a crystal containing colour centres to hydrogen produced
a solid solution of potassium hydride in potassium bromide,
whereas exposing the hydride-bromide solution to light re-
formed the potassium colour centres. What is true for one
Wove-length (m/t)
300 +00

2 eVoftf
Fig. 27. Optical absorption bands of KH-KBr solid in which about half
the KH has been decomposed photochemically giving "Farbzentren".

halide may in general be taken as true for others (RbBr(i3),

KI(i3), KCl(i3)). The studies on'halides containing potassium
and potassium hydride were extended to cover solubility and
diffusion of halogens and hydrogen, and it is with these
systems as with the mobility of Farbzentren that the data now
to be given are concerned.
The number of colour centres per unit volume of halide
may be measured optically, after impregnating the crystal
with alkali vapour. It can be shown that
N = constant x Km x H
= constant x area of the band " F " of Fig. 27,
where Km is the absorption constant at the maximum in
the band, and H is the breadth of the band at K = \Km.
110 GAS FLOW IN CRYSTALS AND GLASSES
The concentration of colour centres may also be determined
from the conductivity, since each potassium atom can give
rise to an electron (e.g. K in KI).
The corresponding investigations upon solutions of the
halogens in halide crystals were also attempted using the
conductivity of the crystal as a means of estimating the
halogen excess (ii). For potassium bromide and iodide a rise
in conductivity followed their exposure to Br2 and I 2 vapour
respectively, but in C12-KC1 systems no rise in conductivity
was observed. The diffusion of halogens in halide crystals was
also followed by saturating the crystal with thallium vapour,
which imparted to it a brown tint. Then, as the halogen sub-
sequently diffused inwards, the brown tint disappeared as the
boundary advanced, due to the formation of colourless or
white thallium halide. The progress of the colour boundary
served to measure the diffusion rate.
In H2-KBr systems the concentration of hydrogen was very
slight, and was best determined by subsequent heating of the
crystal containing hydrogen in potassium vapour. In this way
potassium hydride was formed, the concentration of which in
the mixed crystal KBr -f KH can be determined optically, the
concentration in atoms per c.c. being

(Km = absorption constant of optical absorption maximum).

The diffusion of hydrogen was followed by the same method
as was employed for the halogens, using as indicator metal
potassium and following the movement of the boundary
K-KH (blue to colourless) through the crystal.

The experimental results on solubility and diffusion

Potassium dissolved in both potassium bromide and potassium
chloride exothermally, the heats of the processes being
given by
vap.->KinKBr) = - 5-8 k.cal./atom of K,
vaP.->K in Kd) = - 2*3 k.cal./atom of K.
 DIFFUSION IN ALKALI HALIDE CRYSTALS 111

At high temperatures the potassium was dispersed in atomic

form, but on slow cooling to lower temperatures it could be
condensed into colloidal aggregates. The halogens and
hydrogen dissolved endothermically in halide lattices, the
heats of solution being
= 18-5k.cal./g.mol. I 2 ,
-= 27-6 k.cal./g.mol. Br2,
= 1 6 k.Cal./g.mol. H 2 .
The main feature of interest in the results is that the solubility
in each case is proportional to the external gas pressure, so
that the halogens and hydrogen dissolve in molecular form.
The results on diffusion show too that thermal diffusion occurs
mainly in molecular form, there being little dissociation, save
perhaps for the halogens at the lowest pressures when simul-
taneous atomic and molecular diffusion takes place (ii).
It is doubtful, when solution occurs as molecules (e.g. H 2 ,
Br2, Cl2, in KBr, KC1), whether the molecules are homo-
geneously dispersed. The "solute" is most likely to be
dispersed along faults and glide planes in the crystal. The mere
introduction of large foreign molecules into a perfect lattice
would distort the lattice locally, and create a fault. Hilsch
and Pohl(i3), however, pointed out that when KH is formed
in KBr the system KH-KBr is a true solution, since the lattice
constant of potassium bromide is a linear function of the
potassium hydride content.
When both hydrogen and alkali metal vapour were in
contact simultaneously with the crystal, which contained a
constant amount of potassium, the amount of potassium
hydride at the surface was fixed by the equations

* = ( (KM)
ST^° (Mass Action).
It was established through the optical absorption of successive
thin layers of the crystal that the potassium hydride gradient
decreased inwards nearly linearly with distance, and it was
112 GAS FLOW IN CRYSTALS AND GLASSES
then found that two equations for the depth of penetration, x,
of hydride could be derived (12) according as the hydrogen
diffused as molecules (1) or as atoms (2):

(1)

(2)
In the relation (1), k is the Henry's law solubility coefficient
for molecular hydrogen in the crystal. Experiment showed (12)
that equation (1) was correct, so that hydrogen is both dis-
solved as molecules and diffuses as molecules. The derived
value for D was 2-3 x lO^cm^sec." 1 at 680° C.

Fig. 28. Diagrammatic representation of penetration by hydrogen of KBr

containing an initial uniform excess of K.

If a crystal containing a given uniform concentration, N9

of alkali metal is exposed to hydrogen, after the lapse of
time t, an approximately linear hydrogen concentration
gradient may be assumed to extend into the crystal (Fig. 28)
a distance x. During the time interval dt the gradient will
extend a further distance dx. If dnjdt denote the hydrogen
stream (as molecules),
dn Ndx
~ll~di~ X
2
Nx 1
whence D = if k is defined by (K2)Xs=0 = kp0. Thus
Nx2jt should be a linear function of p0, a prediction which the
 DIFFUSION IN ALKALI HALIDE CRYSTALS 113

diagrams of Fig. 29 show is substantially correct. The diffusion

constants computed from this relationship were
•^6oo°c. = ^'^ x 10~ cm
* sec.~ ,

The measurements on the movement of bromine and iodine

were made along similar lines, the halogen diffusing into the
salt containing thallium as indicator. The concentration of

-to' •10'
25
X /

*/ 20

A{• / o

15
/*
>
X
V
/
5 20° 6 00°

10 20 30 <W 10 20 30 VO
H2-pressure in Atm.
Concentration N Concentration N Concentration N
• 4,G.1O 6 | • 1,5.10" I • 4,4-lO17>|
x l,6.1017[K/cm? x 4,4.10 !7 }K/cm? x 14,-1018 JK/cm?
O911<H O2,2.l0l8j
G 5 -10 1 7 Br/cm? D 3,2-10" Br/cm? D 3,2-10" Br/cm?
Fig. 29. The linear relation between Nx2/t and p0.

the thallium within the crystal was large compared with the
concentration of absorbed halogen (Fig. 30). If n is the
quantity of halogen (expressed in atoms) diffusing through
unit cross-section in time t, one may write
dn
- DdC - n{C)*
at dt — U —z — JLJ
ax gradient
since an approximately linear halogen x was established.
Thus 2
114 GAS FLOW IN CRYSTALS AND GLASSES
and so according as diffusion occurs as atoms (or ions), or as
x2
molecules, one would have — N a linear function of y/p or p
respectively. When the experiments were made it was shown
that the diffusion occurred mainly as molecules (ii). Table 18
then gives the results for D for
Boundary
several temperatures and at several of Crystal
halogen pressures. In Table 18 are Ha tide Excess Thallium
N
included some diffusion constants
for potassium vapour, determined
by an electrical method (ii). It is Halogen
Vapour
possible that some diffusion of
bromine occurs as a diffusion of ions
or atoms, but parallel measurements
of DBf by the electrical conductivity
method showed that at high pressures
(above 1 atm.) transport of bromine Pig . 30. Schematic representa-
as Br~ was much less important tion of the permeation of
than its transport as Br2. Below ^alKum-containing halide
r
* by halogen.
1 atm. a greater percentage of
dissociation occurred, with consequent enhancement of the
percentage of transfer as ions.
Since the diffusion of hydrogen and of halogen is mainly as
molecules, and since they are dissolved within the halide in the
molecular state, one may include them in the same family of
diffusion systems as gas-silica (p. 117) and gas-rubber systems,
i.e. non-specific activated diffusions. They differ from the
latter systems in one respect—the solution processes in the
halides are highly endothermic (p. I l l ) compared with the
very slight endothermicity of hydrogen solubility in silica
(p. 140) and the slight exothermicity of hydrogen solubility in
rubbers (p. 418, Chap. IX). The temperature coefficients of
bromine diffusion in potassium bromide are small (Table 18).
and correspond to energies from 3k.cal./g.mol. at 1 atm.
pressure to 8k.cal./g.mol. at 16 atm. pressure. For hydrogen
the energy is about 8 k.cal./g.mol. in potassium bromide, while
for potassium it is about 16 k.cal./g. atom in the same crystal.
 DIFFUSION IN ALKALI HALIDE CRYSTALS 115

In addition to these studies of solution and diffusion in

halide crystals, so important when one considers photochemical
and conductivity phenomena in crystals, there has been a
number of investigations upon the sorption of dipole gases by
alkali halide crystals. That water may penetrate a rock salt
crystal has been shown from the infra-red absorption spectrum
of the crystal after dipping it in water. In the interior of the

TABLE 18. Diffusion constants of Br2 and of K in KBr

Temp. ° C. 500 600 700
Br2 pressure (atm.) 1 4 16 1 4 16 1 4 16
DUt%=N 2 1(0)^*1* 2-33 1-33 0-9 2-64 2-28 1-65 3-33 30 2-7
in cm. 4 sec."1
( x 10- )
DK from an electrical 0-8 3-0 8-0
conductivity
method ( x 10~4)

crystal the infra-red spectrum of water may be observed.

No study has been made of the velocity of sorption of water
by alkali halide crystals, but a number of workers have
observed and measured slow sorption processes for the
following systems:
S0 2 inNaCl(i3a),
HClinKCl(i4),
NH 3 in
In the case of the two latter systems the data obtained showed
that a slow activated diffusion occurred into the crystal
dC d2C
substance, which followed the Fick diffusion law -^- = D^-^,
Ol OX*
and gave as the temperature coefficient of D
~7000cal./mol.
^HCI-KCI
^NH,-Naci~6300cal./mol.
These experiments were performed upon polycrystalline
masses, and it is probable that diffusion occurred down grain
boundaries. No experiments upon single crystals designed to
8-2
116 GAS FLOW IN CRYSTALS AND GLASSES
test this point are, however, available. Little experimental
work has as yet been carried out upon the diffusion of water
in hydxated crystals. The progress of such a diffusion may be
followed by using heavy water as diffusing liquid in a crystal
with light water of crystallisation. Preliminary measurements
of this kind have been made by Kraft (13d) upon the diffusion
of D 2 0 into alum. Kraft obtained data at 75, 65 and 55° C.
which conform to the expression
D = 0-56 x 10-7 e-e

DIFFUSION OF HELIUM THROUGH SINGLE CRYSTALS

OF IONIC TYPE
The passage of helium gas. through a perfect ionic crystal has
not so far been detected with any degree of certainty. On
account of the laminar structure of mica, and the crystallo-
graphic perfection of the laminae, it would be of great interest
to establish helium diffusion across the laminae. Rayleigh
attempted to measure the helium permeability both at
TABLE 19. The helium permeabilities of ionic crystals
Permeability at ° C.
Substance in c.c./hr./cm.2/nim./
thickness/atm. pressure
Quartz (cut J_ to optic axis) <0-05xl0- 8
Mica (cleavage plate) <0-06xl0- 9
Calcite (cleavage plate) <005 x 10-88
Fluorite <0-2 xlO- 6
Rocksalt <0-2 xlO- 9
Selenite (cleavage plate) <0-7 xlO-
Beryl (cut J_ to optic axis) < 0 1 xlO"78
Beryl (cut j| to optic axis) <015xl0"

20°C.(io) and at 415° (3. (16). At the latter temperature the

permeability was still below 7 x 10-8c.c./day/cm.2/mm. thick-
ness/atm. pressure, more than 10~4 times as small as the helium
permeability of silica glass at 20° C. The figure given was
about the limit of sensitivity of the apparatus.
No better success has attended the efforts of Urry (17), or of
Rayleigh (15) to measure the helium permeability of quartz,
the densest of the crystalline forms of silica, and of beryl. The
 DIFFUSION OF HELIUM 117
latter substance is of special interest since channels run parallel
to the optic axis, of a diameter slightly greater than that of
a helium atom. Table 19 gives the results of Rayleigh's
attempts to measure helium permeabilities of a number of
crystals. The permeabilities are less than the sensitivity limits
of the apparatus.

THE PERMEABILITY OF GLASSES TO GASES

That silica glass is permeable at high temperatures has been
known for a considerable time. Reference to this property was
made by Watson (18), who used it for the interesting object of
purifying helium for an atomic weight determination by means
of its selective diffusion through silica glass. Some earlier
and some later references have been made by Villard(i9),
Berthelot (20) and others (21,22,23). Among the earliest quanti-
tative measurements were those of Wustner(24), who studied
both permeation rates and absorption coefficients up to
800 atm. Alty(25) recognised the phenomena as examples
of activated diffusion and applied to them a treatment akin
to that evolved by Ward (26) and Lennard-Jones (27) for the
activated sorption of hydrogen by copper. The experimental
and theoretical aspects were extended by Barrer(28) who
succeeded in identifying the type of interaction involved in
the migration process.

Methods of measurement of gas flow through

glasses and crystals
It is usual in the study of gas flow through glasses to use an
apparatus which is in essence a chamber containing the gas
separated from an evacuated chamber by the membrane whose
permeability is being studied. The pressure in the gas-filled
chamber is sensibly constant, since the rate of passage of gases
through a glass is small, and under ordinary conditions will
be varied from one experiment to another from a few centi-
metres to an atmosphere. The rate of growth of pressure on
the high vacuum side is followed by pressure-measuring
devices such as the McLeod gauge, Pirani gauge or ionisation
118 GAS FLOW IN CRYSTALS AND GLASSES

gauge. The problem of mounting the membrane offers little

difficulty where soda glass, pyrex,borosilicate glass, silica glass
and other glasses are concerned, for these glasses may be blown
into double-walled vessels of the types illustrated in Fig. 31.
When the glass is available in the form of plates, the latter
may be mounted only with difficulty (of. Fig. 32). Urry(33)
mounted flat disks of quartz, basalt, and other rocks upon
the ground-glass flanges of a glass tube which led to a gas-
analysis apparatus. The specimen was made to adhere to the
To HELIUM
RfSEflVOIft
"i- Tc Me L too
GALCC

-Fusto SIUCA

«- SorrGLASS

V V
Fig. 31. Types of double-walled diffusion cell.

ground-glass plate by means of a little tap grease, and was

sealed off from leakage between it and the ground-glass
flanges by filling the apparatus with mercury up to the top
edge of the specimen.
The experiments of Wiistner(24) were made at very high
pressures and temperatures, so that special requirements had
to be met. The diffusion cells were contained in a platinum
oven which in turn was in the interior of a steel bomb. The
pressure was transmitted to the bomb from the pump (with
oil as transmitting fluid), through the separator which was a
U-tube containing mercury, to water, which filled the bomb.
The requisite temperature within the bomb was reached by
the use of the platinum oven. The silica diffusion cells inside
 PERMEABILITY OF GLASSES TO GASES 119
the oven oontained the hydrogen imprisoned by mercury seals.
This hydrogen was compressed when the pressure was raised,
its pressure being given by the
gauge, and its temperature by
thermo elements leading from
the oven. The pressure of
hydrogen on one side of the
silica was thus high (700-1000
atm.); on the other side it was
negligible, and diffusion con-
sequently occurred through the
small thin-walled silica bulb
into which the hydrogen was
compressed. The ingoing and
outgoing sides of the silica wall
were both open to the pressure-
transmitting liquid, so that
there was no nett pressure
gradient across the wall, and
so no danger of mechanical
rupture. The volume of hydro-
gen which diffused through the
silica was estimated from the
movement of the mercury seal
in the silica diffusion cell.
Rayleigh (16) succeeded in the
technically difficult task of
mounting a mica plate for dif-
fusion experiments at tempera-
tures up to 415° C. The principle
of his method was to keep an
outer annulus of the plate cold
and thermally insulated as far
as possible from the inner part of the plate which was kept
hot (Fig. 32). The mica plate was held between steel disks,
which had a hot inner section and an outside ring water-cooled.
The outside ring was separated from the inner disk by narrow
120 GAS FLOW IN CRYSTALS AND GLASSES
steel ribs which gave rigidity to the whole, and reduced heat
conduction from the hot central disk to the outer annulus.
The mica disk was sealed with red wax to the steel annuli on
either side of it. In its central part, through which diffusion
was to occur, the mica bore sufficiently lightly upon the central
steel disks for the helium to pass easily between mica and steel
and so to collect in the low-pressure chamber. The mounted
membrane was fixed across a steel tube, the two halves so
formed serving as chambers to supply the helium at constant
pressure, and to collect it. Not only the annuli but also the
steel tube was water-cooled. The heating was carried out by
means of nichrome wire bent back and forth and passing
through silica tubing, there being one heating coil on each side
of the mica disk. The diffused gas was removed for analysis by
a Toepler pump.

The principal characteristics of the permeation process

The main features observed when one studies the passage of
gases through membranes are:
(1) The Fick diffusion law P = D— (P denotes the per-
ox
meability and D the diffusion constant) is true in the stationary
State (33,24,32).
(2) Stationary flow is established in a period of minutes, the
actual time depending on the temperatures.
(3) The permeation rates are usually proportional to the
pressure and inversely proportional to the thickness of the
membrane (29).
(4) The velocity of diffusion is only slightly altered by
roughening the outgoing surface (29).
(5) The passage of an electric discharge through either the
glass wall or through the gas (hydrogen or helium) has no effect
upon the diffusion rate(i7).
(6) The process of permeation through these glass mem-
branes is highly selective, and markedly temperature de-
 PERMEABILITY OF GLASSES TO GASES 121
pendent, so that by proper choice of temperature zones one
could in theory very effectively separate certain gases (O2 and
He; air and He; argon and He; H 2 and argon, etc.).

Dependence upon pressure of the rate of permeation

A linear pressure dependence of the permeation rate is indi-
cated by the results of various observers (24,30,28,31,32). It was
immaterial whether the gas was monatomic (He) or diatomic
(H2, O2, N2), the same linear relation was obtained. There is
one contrary result (33), data on the velocity of permeation of
helium through pyrex, soda, lead, and Jena 16 m glasses being
reported as conforming to an equation dp/dt = ocpn. The
values of n were given as follows:
Pyrex glass n = 0-88,
Lead glass n = 0*56,
Soda glass n = 0-64,
Jena 16 m glass n = 0-66.
It was also stated that the value of n was nearly proportional
to the amount of SiO2 + B2O3. These results need further
testing, since helium cannot be dissociated in diffusing (which
would give a law dp/dt = ocp*), so that the only possibilities
remaining are that helium entered the glass from an adsorbed
phase which does not obey a distribution law x = kp (where x
denotes the amount adsorbed), or that irreversible changes
were occurring in the structure of the glass. At room tem-
perature on glass helium more than all gases should be
adsorbed according to Henry's law (x = kp).

The nature of the permeation process through silica glass

The observation that the diffusion rate (save for the exceptional
data of Urry) is proportional to the pressure is typical of a very
wide class of diffusion membranes of silica, glass, basalt, rubber;
porcelain, cellulose, collodion, and various polymeric products
from styrene, vinyl acetate and similar substances. It may be
characteristic of activated diffusion (p. 125) or of various types
of flow down tubes (Chap. II). On the other hand, there are
122 GAS F L O W I N C R Y S T A L S A N D G L A S S E S

systems where permeation velocities are proportional to ^p

(p denoting pressure), exemplified by the diffusion of hydrogen
through nickel, iron, copper, palladium, platinum and other
metals. In this type of system diffusion appears to be a specific
property of the system considered. No trace of helium will
diffuse through palladium (34) or copper (35), though the helium
atom is much smaller than the hydrogen molecule. In ex-
planation of all the facts, one may regard specific activated
diffusions as occurring when the membrane can dissociate the
diffusing molecule into atoms or ions. An insight into the
nature of the gas-solid interaction in the case of the non-
specific type of activated diffusion is provided by the experi-
ments of Barrer (28) on gas-silica glass systems. It was observed
that the temperature dependence of the permeation rate in
calories per mole, though much greater than the sublimation
energies, follow in the same order (Table 20).

TABLE 20. Heats of sublimation of gases, and the temperature

dependence (in caLjmol.) for flow through silica glass
Temperature
Sublimation coefficient
Gas heat from P=P0 e~EIR T
for flow
He 126* 5,600
Ne 590 9,500
H2 529* 10,300
oN22 2150 31,200
1860 26,000
A 2030 32,100

* Calculated (io) from the force law E= -oc/W+fi/R12, when the constants
a and ft are derived from the virial coefficients of the equation of state of the gas.

Table 20 suggests that van der Waals and repulsive forces

contribute to the temperature coefficient of flow (in cal./mol.)
just as they do to the latent heat of sublimation; in the latter
instance the attractive forces, which are small, are pre-
dominant, but in the former the much larger repulsive forces
are predominant. This provides a basis for the classification
of activated diffusion processes into specific types (H2-Pd) and
non-specific types (He-SiO2), the basis being the nature of the
 PERMEABILITY OF GLASSES TO GASES 123
interaction between solid and gas. Table 21 attempts a
classification of some diffusions having exponential tem-
perature coefficients into specific and non-specific activated
diffusions. In addition to these "natural" diffusions there are
numerous "forced" diffusions wherein ions may move under
an impressed potential. These will be considered separately
(Chap. VI).
TABLE 21. Diffusion processes

Nature of Nature of interaction

system Specific Non-specific
Gas-solid H2-Pd, Ni, Pt, Cu H2-KBr
O2-Ag, Cu-O, FeO H 2 -Si0 2 , pyrex, rubber,
N2-W, Mo, Fe cellulose
0 2 -Si0 2
N 2 -Si0 2
AT, Ne, He-SiO2, B2O3,
borates, silicates, Jubber,
cellulose
Vapour-solid H2O-zeolites? NH3-pure rubber
NH3-zeolites? H2O-pure rubber, cellulose?
HC1, NH3-NaCl? CO2-pure rubber
Solid-solid S-FeS
Ca++, K+, Na+-glass,
zeolites, ultramarines,
clays
Metal-metal systems?
Liquid-liquid NH3-H2O? C6H6OH-CH3OH
D2O-H2O? C6H6OH-C6H6
s-C2H2Br4-s-C2H2CL
D2O-H2O?
Gas-liquid C0 2 -Na0H aq. H2-H2O
O2-H2O, NaOH aq.

The influence of temperature on the permeation rate

Three formulae have been employed to express the permeation
rate as a function of temperature. The formulae are

(i)

(2)

(3)
dt
124 GAS FLOW IN CRYSTALS AND GLASSES

in which A, ra; B, a; and C and EjR, are characteristic

constants. The most satisfactory agreement over the high

4-5 \
Log (Permeation velocity
jti

1-4 78
1000/T (Tin °K.)
Fig. 33. Passage of helium through silica.

\
4-5
\

s \

5-5 \

10 1-5
7000/7 (T in °K.)
Fig. 34. Passage of hydrogen through silica.

temperature range is undoubtedly given by equation (3)

(Figs. 33, 34), which has also the advantage of having a
theoretical basis. Thus when an energy of activation is required
 PERMEABILITY OF GLASSES TO GASES 125
to make the molecule enter the pore and to move it along the
energetically periodic pore length, one would by analogy with
the Arrhenius theory of chemical reactions expect the number
entering the pore to be given by
N =
where Nx denotes the number of molecules available to enter
the pore and A denotes the chance that a molecule having a
sufficient energy will actually enter the pore. While a kinetic
theory of permeation of a more elaborate nature can be
derived, the applicability and significance of equation (3)
will now be assumed, and from the experimental data the
temperature coefficients in cal./mol. for the permeation process
are collected in Table 22. These coefficients include the tem-
perature variation of Nx (which may refer to adsorbed
molecules, or even dissolved molecules). Since the heat either
of sorption or of solution (p. 140) is small, these temperature
coefficients may be approximately identified with activation
energies for diffusion within the solid.

Irreversible effects
Several interesting points arose from Barrer's(28) study of the
influence of temperature upon the permeation rate. First, the
permeability towards the heavier gases is affected by prolonged
heating to high temperatures. This is shown by the permeation
rates of air through a silica tube, given as a function of tem-
perature for various periods of heat treatment of the silica
(Fig. 35). The flowing curves are drawn through points
determined in sequence, and the permeability decreases as the
heating is prolonged. That the effect was solely a surface one
was shown by treatment with hydrofluoric acid, which restored
the permeation rate to its original value. This superficial
change was probably connected with a visible clouding of the
surface, and it was thought that it might be due to the
formation of tiny crystals, possibly platy (3) or else of /?-cristo-
balite type. With an ensuing decrease in permeability there was
an increase in the activation energy which is clearly indicated
 TABLE 22. The temperature coefficients of permeability constants in glasses in cal./mol.
Gas Glass Energy Author Remarks
He Fused silica 5,600 T'sai and Hogness(30> The energy of activation increases as
Fused silica 5,700 Barrer (28) the percentage of SiO2 decreases
Fused silica 5,390 Braaten and Clark (29)
Pyrex 8,700 van Voorhi8(36)
Thuringian 11,300 Piutti and Boggiolera(37)
Lead — Urryd7) Permeation observed but influence of
Soda — Urryd7) temperature not studied
Jena 1 6 m 8,720 Urryd7)
Ne Fused silica 9,500 T'sai and Hogness(30)
H2 Fused silica I 9,300 Williams and Ferguson(32)
Fused silica I I 10,000 Williams and Ferguson (32)
Fused silica III 10,000 Williams and Ferguson (32)
Fused silica IV 10,800 Williams and Ferguson (32)
Fused silica I 10,900 Barrer(28)
Fused silida II 10,800 Barrer(28) After prolonged heating, causing super-
ficial crystallisation
Fused silica 8,500 Mayer (38)
Fused silica 9,200 Johnson and Burt(39)
Fused silica 12,000 Wustner(24) At up to 800 atm.
Pyrex — Urryu7) Permeation observable
Pyrex — Williams and Ferguson(32) Permeation not observed
Jena — Williams and Ferguson (32) Permeation not observed
N2 Fused silica 26,000 Johnson and Burt(39)
Fused silica 22,000 Barrer(28)
Fused silica 29,900 Barrer(28) After prolonged heating with surface
change
02 Fused silica 31,200 Barrer(28) After prolonged heating and cleaning
with hydrofluoric acid
Air Fused silica [18,500] Barrer<28) Apparent energy only
Fused silica, after pro- [22,000] Barrer(28) A different sample of silica from that
longed heating used for author's other data
Argon Fused silica, after pro- [48,000] Barrer(28) A very slow rate of permeation indeed
longed heating
Fused silica 32,100 Barrer(28> A fresh surface obtained by treatment
with HF, and much greater perme-
ability
 PERMEABILITY OF GLASSES TO GASES 127
in Table 22. These effects were most marked for the heavy gases
(N2, Ar, and air), whereas for hydrogen and helium the influence
of heating upon the permeability was slight. This fact, and the
close agreement of the temperature coefficients of the per-
meability when expressed in cal./mol.,* determined by various

1-25

£> 100 t
l 0'75
J 0-50
$

V.
^ 0-25

800 woo noo

Temperature, °K.
Fig. 35. The rate of flow of air through silica, as affected by
temperature and time of heating of the silica.

workers on different samples of silica led to the postulate (28)

that there are two types of activated diffusion in silica
glass:
(1) A structure sensitive diffusion down faults, and cracks
of molecular dimensions, predominant when argon, nitrogen
and air diffuse.
(2) A structure insensitive diffusion through the anionic
network of the glass itself, and corresponding to a solution
process. This diffusion is predominant when helium, neon and
hydrogen pass through silica glass.
* The permeability constant P, diffusion constant D and solubility k are
related by P = Dk Ap jl, where Ap denotes the pressure difference across, and I the
thickness of the specimen. When as for H2 and He (p. 140) the temperature
coefficient of k is small, the temperature coefficient of P in cal./mol. is that for />,
i.e. the activation energy for diffusion.
128 GAS FLOW IN CRYSTALS AND GLASSES
Even for helium, neon and hydrogen, grain boundary
diffusion must occur simultaneously with ''lattice" diffusion,,
and the predominance of one or the other is conditioned by
the temperature. As the temperature is lowered, the more

-4-5

8
1§ -6-5 \

-7 5 \

-85
10 • •5 2O

Fig. 36. The structure sensitive and structure insensitive

regions in the diffusion of helium in silica glass.

temperature sensitive "lattice" diffusion is increasingly sub-

merged in the grain boundary diffusion. So one should find a
flattening of the curve log (permeation rate) against 1/T
(T in ° K.) for small values of T. Figs. 34 and 36 show this effect
for hj^drogen and helium.
 PERMEABILITY OF GLASSES TO GASES 129
On the other hand, a specimen of silica glass which had
undergone prolonged heating gave a log (permeation rate)
against l/T curve linear even at room temperature, suggesting
that the heating had diminished (by the surface change noted
(p. 125)) the structure sensitive part of the diffusion.
Table 23 gives mean values of the temperature coefficient
of the permeation rate for diffusion processes occurring in
various zones of temperature. Table 23 also illustrates in what
TABLE 23. Temperature coefficients of permeability in
cal./mol. at low temperatures*
Nature of Energy
Gas Glass Worker predominating cal./ Temp.
diffusion mol. ° C.

H2 Fused silica Barrer<28) Lattice 10,800 >400

Fused silica Barrer(28> Grain boundary 4,300 193
He Fused silica T'sai and Hog- Lattice 5,700 >300
ness (30)
Burton, Braaten Grain boundary 4,190 110 toO
and Wil- 3,040 0 to - 41
helmoi)* 2,310 - 41 to - 78
Pyrex van Voorhis(36) Lattice 8,700 >300
Urryd?) Grain boundary 5,840 283 to 172
Grain boundary 4,540 172 to 81
Jena 161" Urry(i7) Uncertain 8,720 283 to 134
Grain boundary 6,900 134 to 22

* In a later communication(29) two of the authors state that-no further

diminution in E could be found below - 20° C. Their new values of E were:
from 180 to 562° C, 5390 cal.; from 180 to - 78° C, 4800 cal.

way the silica content affects the activation energy for the
process. The smaller the silica content, the larger is the
activation energy and the smaller the permeation rate (cf.
p. 137). The passage of helium through a number of glasses of
known composition (36) also revealed that acidic oxides such
as B2O3 or SiO2 increased the permeability, while basic oxides
such as K 2 O, Na2O, or BaO decreased it approximately in
proportion to their amount. Feebly basic or amphoteric oxides
such as PbO or A12O3(36) were stated to have little effect
upon the permeability. It is to be noted that A12O3 may
replace SiO2 in the anionic network of the silica membrane.
* See footnote, p. 127.
BD 9
130 GAS FLOW IN CRYSTALS AND GLASSES
Roeser(40) studied the permeability of various samples of
unglazed and glazed porcelain in air. He found many samples
with channels so large that stream-line or Knudsen flow
occurred in them; but in other more perfect specimens only
a strongly temperature sensitive diffusion was observed. The
experiments, which were made up to 1300° C, gave varied
agreement with the law
Permeability constant* = Poe~EIRT,
tne products of some manufacturers giving linear curves
of log (P) versus l/T, while other manufacturers' products
behaved more or less capriciously. The slope of the curve
log (P) versus l/T even at the highest temperatures varied
from specimen to specimen. The slopes observed at high
temperatures for one set of porcelain tubes varied from 29,000
to 63,000 cal. All these results suggest that grain-boundary
diffusion is a more usual process than lattice diffusion, in
conformity with Barrer's(28) findings for the migration of the
heavier gases through silica glass. In this connection it may
be mentioned that Roeser's results on silica glass and on glazed
and unglazed porcelain give permeability constants of similar
magnitude. Porcelain may be considered to consist of crystals
of mullite (3Al2O3.2SiO2) embedded in a glass magna, and
often with undissolved quartz or clay in the structure. It is
clear that such a chemical will be far from homogeneous, and
to this chemical inhomogeneity as well as to its physical
inhomogeneity may be ascribed its capricious behaviour.
Further studies on irreversible phenomena associated with
helium diffusion through pyrex glass were made by Taylor and
Rast(4i). The effects noted by them were not confined to the
surface of the glass, as were those observed by Barrer (28). They
found that the permeability constant rose by 10 % at about
550e C , after annealing at that temperature, and that there-
after the new permeability-temperature curve lay above the
* Even with a single energy of activation for each gas, this law could not be
rigidly obeyed when air diffuses, since air is a mixture of N2 and O2 and the law
should be P =P1e-E°JRT+P2e-E*JRT.
 PERMEABILITY OF GLASSES TO GASES 131
old one. The effect was considered to be the result of strain
removal in the original glass, the curve log (permeability)
against 1/7 being now continuous into the region where the
glass could be regarded as a viscous liquid. The diminution in
slope at lower temperatures which have been interpreted (28)
as increasingly important contributions of grain-boundary
diffusion were also noted, the slopes giving values of
^440-eoooc. = 7150cal./atom,
^35o-44O°c. = 6480 cal./atom,
which may be compared with the values of E in Table 23.
However, the authors regarded this diminution as due to a loss
of rotational or vibrational freedom in the silicate complex,
occurring at a critical temperature. As a further change in the
slope of the log (P) versus \\T curve had occurred at 225° C,
one must on this theory suppose a second loss of rotational or
vibrational freedom.

On the relationships between some types

of molecular flow
It will be interesting to point out at this stage how several
important types of mechanism for gas transference through
solids are related. In the normal stream-line flow of fluids the
diameter of the pore and the pressure are such that the number
of collisions on the pore wall in unit time per unit area is com-
pletely outweighed by the number of collisions in unit time
and in unit volume of the gas phase. When the pressure and
pore diameter are such that collisions with the wall completely
outweigh collisions in the gas phase, one finds not stream-
line flow but molecular streaming. Also as one continually
restricts the pore diameter, the gas molecules must spend
a greater and greater fraction of time in the surface field
of the solid, that is adsorbed on the solid. Ultimately, as a
logical conclusion to this process of restricting the pore
diameter, the diffusing molecules are always within the surface
fields of the solid. Owing to the periodic or crystalline nature
9-2
132 GAS FLOW IN CRYSTALS AND GLASSES

of the solid, the energy distribution along the surface is also

periodic, and a molecule diffusing along the surface finds itself
moving to successive energy hollows, separated by energy
barriers.
When the periodic fields of the opposite sides of the pore
overlap, energy will be needed to make the molecule enter the
pore. In illustration of this, Fig. 37 shows the energy needed to

Curye I. Diam.square,5-4A:
- £. .. .. 5-6A
n m. " - S-OA.
- J7. * » 6-4A

5 4 3 2 / 0 7 . 2 3 4 5
Distance, (A).
Fig. 37. Energy needed to make an argon atom pass through
a square of argon atoms.

make an argon atom pass through a square of argon atoms,

when the diameter of the square is varied to make the atomic
force fields overlap to varying extents.
Experiments upon the transition region from molecular
streaming to activated diffusion are few. Rayleigh(is) has
recently made interesting experiments upon the passage of air
and helium through very narrow artificial channels. When
optically plane glass plates were placed in contact and heated
to remove adsorbed gases, but not heated sufficiently to destroy
the planeness of the glass, it was found that contact was so
intimate that no flow of helium between the two plates could
be measured, although the rate of flow of helium through silica
glass can be measured at room temperature. In another
experiment the plane glass plates were placed in contact,
 PERMEABILITY OF GLASSES TO GASES 133
but not heated. Their distance apart was of the order of
10 A., and the permeation rates for helium and for air were
9-5 x 10~ 2 cu.mm./dayandl-4x 10~2cu.mm./day respectively.
The ratio of these velocities is
He
Air
instead of the ratio
He 2-7 28-8
Air
which Knudsen's law of molecular streaming would require.
That is, Rayleigh succeeded in passing beyond the limits of
Knudsen flow into what must be at least a transition region to
activated diffusion. In this region the ratio of the permeation
rates would be governed principally by a ratio of exponentials:
He

Numerical values of the permeabilities

The volumes in c.c. at N.T.P. of gas diffusing per sec./cm.2
through a glass wall 1 mm. thick, when a pressure difference of
1 cm. is maintained across the wall, is given in Tables 24-34.

TABLE 24. He-SiO9

Permeability constant x 109
Temp. (e.c\ at N.T.p./sec./cm.2/nain./cm. Hg)
o p
A B C D
150 0-73 0-78 ,
200 1-39 1-52 019
300 315 413 0-48 0-46
400 615 8-25 0-99 0-92
500 10-4 13-8 1-72 206
600 16-4 19-3 300 4-62
700 21-9 i 4-25
800 28-5 — 5-50 —
900 36-2 6-72 —
1000 45-4 — 8-42 —

Authors: A = T'sai and Hogness; B = Braaten and Clark; C = Barrer;

D = Williams and Ferguson.
134 GAS FLOW IN CRYSTALS AND GLASSES

TABLE 25. He-SiO2 from -200 to 150° C.

Permeability constant x 109
Temp. (c.c. at N.T.p./sec./cm.2/mm./cm. Hg)
° C.
A A A B
-200 0-0028
-180 00035
-160 — — 00038
-140 00044
-120 00053
-100 0-0066
-80 000070 0-0084
-70 000176 00101
-60 000315 00121
-50 00052 00145
-40 0-0077 0-0179
-30 0-0109 00224
-20 00174 0-028
-10 0022 0037
0 0035 0029 0028 0050
10 0051 0-046 0040 0073
20 0070 0062 0055 0104
30 0090 0080 0073
40 0114 0-106 0095 —
50 0135 0132 0-119
70 0-205 0-198 0176 —
90 0-304 0-29 0-264 0-274
110 0-44 0-42 0-39 0-45
130 0-62 0-59 0-52 0-55
150 0-79 1 —

Authors: A = Braaten and Clark; B = Burton, Braaten and Wilhelm.

TABLE 26. Ne-SiO9

Permeability9
Temp. constant x 10
°C. (c.c./sec./cm.2/mm./cm. Hg)
500 0-139
600 0-282
700 0-50
800 0-81
900 118
1000 1-63

Authors: T'sai and Hogness.

 PERMEABILITY OF GLASSES TO GASES 135

TABLE 27. Argon-SiO2

Permeability constant x 109

Temp. (c.c./sec./cm.2/nini./cm. Hg)
°C.
A A* B
850 00161
900 0-58
950 0062, 0031 000022
1000 000050

Authors: A = Barrer; B = Johnson and Burt.

Decreased permeability due to long heating of silica glass specimen.

TABLE 28. H 2 -Si0 2

Permeability constant x 109

Temp. (c.c./sec./cm.2/nim./cm. Hg)
°C
A B B B B C D

200 0022 __ _ _ __
300 0099 0051
400 0-366 0-48 0-44 0-50 0-275
500 0-70 0-92 0-84 — 106 — 0-58
600 1-43 1-75 1-54 216 200 0-81
700 2-52 31 2-70 2-45 3-9 2-76 1-70
800 4-25 4-8 4-4 40 60 4-5 2-53
900 6-4 70 5-9 3-6
1000 100 — — — — 51

Authors: A = Barrer; B = Williams and Ferguson; C=Wustner; D = Johnson

and Burt (mean of three samples).

TABLE 29. N 2 -Si0 2

Permeability constant x 109

Temp. (c.c./ s e c -/ c m - 2 / m m -/ c m - Hg)
°C.
A B B*
650 0065 0066
700 0132 0146
750 0-268 0-271 0161
800 0-43 0-39
850 0-80 0-64
900 119 0-95
950 1-44 0-65

Authors: A=Johnson and Burt; B = Barrer.

* Decreased permeability due to long heating of silica glass specimen.
136 GAS FLOW IN CRYSTALS AND GLASSES

TABL£ 30. He-pyrex glass

Temp. Permeability constant x 109
°C. (c.c./sec./cm.2/inm./cm. Hg)
A B
0 00037
20 0-0064 —
50 00128
100 00264 —
150 0-058 —
200 0124 0-069
250 0-229
300 0-38 0-243
400 0-70
500 — 1-57
Authors: A=Urry; B = van Voorhis.

TABLE 31. He-Jena 16 m

Temp. Permeability constant x 109
°C. (c.c./sec./cm.2/nini./cm. Hg)
20 00000095
50 00000471
100 0-000071
150 0-000183
200 0-00077
250 000176
300 000362
Author: Urry.

TABLE 32. He-miscellaneous glasses

Temp. Permeability constant x 109
°C. Glass (c.c./sec./cm.2/mm./cm. Hg)
283 Lead 00037*
283 Soda 0-0098*
610 Pyrex ! l-9f
Authors: *=Urry; f = Williams and Ferguson.

TABLE 33. He-Thuringian glass

Temp. Permeability constant x 109
°C. (c.c./sec./cm. 2 /mm./cm. Hg)
100 0-00000106
200 0000117
300 0-00084
400 0-0045
f>00 00132
Authors: Piutti and Boggiolera.
 PERMEABILITY OF GLASSES TO GASES 137

TABLE 34. Air-porcelain

Permeability constant x 109
Temp. (c.c./sec./cm.2/nim./cm. Hg)
°C.
Sample I Sample II
25 000106
400 0-00106
600 000212
800 00032
1000 00161 0-0012
1200 0077 0022
1300 0-32 0117

Author: Roeser.

In deriving these permeability data graphs of the experimental

measurements were used to obtain figures at comparable
temperatures. Several features of this group of tables are
interesting. First one may consider the selectivity of the
permeability of silica glass towards a number of gases, illus-
trated by the following series:
For He-SiO2(30) at 900° C, P = 36-2 x 10~9,
H2-Si02(28) at 900° C, P= 6-4 x 10-9,
Ne-SiO2(30)at 900° C., P= 1-18 xlO" 9 ,
N2-Si02(28)at 900° C, P = 0-95 x 1O"9,
Ar-SiO2(39)at 900° C., P = 0-58 x 10-9.
It must, however, be remembered that permeabilities towards
certain of the gases vary from specimen to specimen of glass;
thus much smaller permeabilities both to argon and to helium
have been reported (28). In yet another series may be given
the permeability of a number of glasses to helium:
For He-SiO2(30) at 300° C, P = 3-15 xlO- 9 ,
He-pyrex(i7) at 300° C, P = 0-38 x 10"9,
He-soda glass (17) at 283° C, P = 0-0098 x 10-9,
He-lead glass(17) at 283° C, P = 00037 x 10-9,
He-Jena 16i"(i7) at 300° C, P = 00036 x 10-9,
He-Thuringian glass at 300° C, P = 0-00084 x 10 - 9
138 GAS FLOW IN CRYSTALS AND GLASSES
and from the series one observes the high sensitivity of
permeability to chemical composition (cf. p. 129). Further
data along these lines are provided by Rayleigh's studies on
the helium permeability of membranes, including silicate
glasses, silica glass, and boron trioxide melts, as well as a
number of metallic and organic membranes. The following
Tables (35 and 36) give the helium permeabilities of a number
of glasses; and also allow a comparison of the permeabilities

TABLE 35. The helium permeability of glasses at room

temperature in c.c./sec./cm.2lmm. thick/cm. Hg pressure
Terence

Substance Permeability Substance Permeability

Silica 0-058 x 10- 99 Fused B2O3 0055 x 10- 9
Optical silica 0043 x lO" Fused B2O3 0056 x lO"9
Silica sheet 0-025 x 10- 99 Fused borax <0-000103 x 10- 11
Thin silica tube 0053 x lO" 9 glass
Pyrex 00040 x 10~
Corex glass (mainly < 0-000040 x 10- 9
Ca3(PO4)2)
Extra white sheet <0-000025 x lO"10
glass
Micro cover glass <0-000020 x lO"10
Flint glass <0000030xl0-n
Soda glass .<0-000043 x lO"13

TABLE 36. The helium and air permeabilities of various kinds

of membrane at room temperature in cc/sec/cmf/mm.
thick Icm. Hg pressure difference

Helium Air Ratio of

Substance permeability permeabilities
permeability
Cellophane 0023 x 10- 99 0049 x 10"11 48
Silica 0052 x 10-9 <OO52xlO~ 13 > 10,000
Gelatin 014 x lO" 0-077 x 10"911 184
Celluloid 6 1 x lO"9 0-305 xlO"9 20
Rubber 12 x 10- 9 4-16 x 10- 2-9

of certain organic membranes with the inorganic glasses.

All the measurements reported were at room temperature,
 PERMEABILITY OF GLASSES TO GASES 139
2
and the constants have been converted to c.c./sec,/cm. /mm.
thickness/cm, of mercury.
Examination of Table 35 shows an extreme variation in the
helium permeability of the silica membranes of only 2-3-fold
at room temperature, while Table 24 indicates an extreme
variation of 6-4-fold at 500° C. The sensitivity of the permea-
bility to the composition of the glass is again apparent, and
it may be that the variations in Table 24 are due to small
amounts of the alkali metal oxides in the less permeable of
the silica glasses as well as to a greater amount of grain-
boundary diffusion in the more permeable glasses. It is
interesting that fused B2O3 compares with fused SiO2 in
permeability, and also that the organic membranes cellophane
and gelatin have permeabilities of the same order as the
inorganic oxides. The ratio of helium to air permeability
clearly bears no relationship to molecular masses, being
mainly governed by the ratio of exponential terms of the
type e~EIRT.

THE SOLUBILITY OF GASES IN SILICA AND DIFFUSION

CONSTANTS WITHIN IT

When discussing the temperature coefficients of the per-

meabilities, we considered these temperature coefficients to
be approximately those for the activated diffusion process
within the silica. This viewpoint was justified because, as
the data now to be given show, the solubility of hydrogen
and helium in silica varies only to a very minor extent with
temperature, and the permeability constant (P), diffusion
constant (D) and solubility k are related by

where Ap/l is the pressure gradient across the membrane. If

the values of P and of k are known one may compute the more
fundamental quantity D. The values of &(24,42) are given in
Table 37. It may be noted that the solubilities are quite
140 GAS FLOW IN CRYSTALS AND GLASSES

comparable with the solubilities of gases in rubber mem-

branes (p. 418), in liquids and in crystals (p. 111). Further,
the differences in permeability which one encounters for gases
in silica and pyrex are, one is now led to believe, governed
mainly by the differences in D, which are in their turn (p. 125)
governed partly by an exponential term.*

TABLE 37. The solubilities of hydrogen and helium in silica

Solubilities (average values in c.c.
Temp. at N.T.p./c.c./atm.)
System °C.
Wiistner(24) Williams and Ferguson(42)
H 2 -Si0 2 1000 00103
900 00102 —
800 00109
700 00099
600 00082
400 00057 0-0095 (extrapolated)
300 00055 00099
He-SiO2 500 — 00101
450 00103
He-pyrex 500 00084

When the permeability constant is expressed as c.c./sec./

cm.2/mm. thick/atm. pressure, and the solubility as c.c./c.c. of
silica/atm. pressure, the use of the equation
^ , Av
P

leads to a value of D expressed as cm.2 sec."1 These values of

D are of interest for purposes of comparison with corresponding
values of D obtained for liquid-liquid or gas-liquid systems;
for gas-metal, and gas-rubber systems; and for ion-ionic lattice
diffusion systems. This comparison will be made elsewhere
(p. 426). In Table 38 the values of D are computed from the
solubilities of Table 37 and the permeabilities of Tables
24-34.
* But where grain-boundary diffusion predominates the number of internal
surfaces, which is determined by the history of the specimen, is also
important.
 PERMEABILITY OF GLASSES TO GASES 141

TABLE 38. The diffusion constants inside

and glass in cm.2 sec.-1
Solubility
Solid Gas Temp. taken D Authors
O /I
C. c.c./cc. solid cm.2 sec."1
SiO2 He 20 001 0024-0055 x 10"8 Rayleigh
SiO2 He 500 001 0017-014 x 10- 6 Authors of
(Williams and Table 24
Ferguson)
SiO2 H2 500 0-01 OOOfr-0011 xlO~ 6 Authors of
(Williams and Table 28
Ferguson)
SiO2 H2 500 00055 0012-0021 x 10- 6 Authors of
(Wiistner) Table 28
8
SiO2 H2 200 00055 005-0-08 x lO" Barrer
Pyrex He 20 00084 00045 x lO"8 Rayleigh
Pyrex He 500 00084 002 x 10~6 van Voorhis
(Williams and
Ferguson)

The diffusion constants of Table 38 may be written as

D = Doe-E'RT,
and then D takes the following values:
= (7*9 ~ 3 ' 5 ) 1 0 ~ 6 e- 560 °/* T cm. 2 sec.- 1 ,
= (5-2 - 0-64) 10~6 e- 5600 ^^ cm. 2 sec.- 1 ,
= ( 8 ' 3 - 1 4 * 5 ) 1 0 " 6 e-101Q°iRT cm. 2 sec.- 1 ,
= (13-7 - 35) 10- 6 e-ioioo/tfr c m 2 s e c -i ?
= 1.3 x l O - ^ - s ^ / ^ c m ^ s e c . - 1 ,
= 5-5 x 10~6 e" 8700 /^ cm.2 sec."1.
The result for the He-pyrex system at 20° C. studied by
Rayleigh suggests that grain-boundary diffusion (see Table 23)
has been taking place, with a lower energy of activation than
the 8700 cal. assumed, which is the activation energy* for
"lattice" diffusion. With this exception the values of D o are
all of the same order of magnitude, and do not alter markedly
with temperature, but do show random fluctuations which
would conform well with the theory (p. 127) of mixed grain-
boundary and lattice diffusion.
* See p. 139.
142 GAS FLOW IN CRYSTALS AND GLASSES

Theories of the diffusion process

Several authors (33,27,25) have attempted theories of the
diffusion process. Urry (33) who considered flow to be a process
of molecular streaming, as observed when rarefied gases pass
through capillaries, is obviously incorrect for silicate glasses.
The other theories (27,25) are based upon more acceptable
premises, but do not lead to any notable advance in the study
of the subject. They need not therefore be discussed here.

REFERENCES
(1) Taylor, W. H. Proc. Roy. Soc. 145A, 80 (1934).
(2) Barrer, R. M. Proc. Roy. Soc. 167A, 392 (1938).
(3) Shishacow, N. A. Phil. Mag. 24, 687 (1937).
(4) Nagelschmidt, G. Z. Kristallogr. 93, 481 (1936).
(4 a) Farkas, A. Private communication.
(5) Tiselius, A. Z. phys. Chem. 169 A, 425 (1934).
(6) Z. phys. Chem. 174 A, 401 (1935).
(7) Hey, M. Miner. Mag. 24, 99 (1935).
(8) Phil. Mag. 22, 492 (1936).
(9) Mollwo, E. Z.Phys. 85, 56 (1933).
(10) Rogener, H. Ann. Phys., Lpz., 29, 387 (1937).
(11) Mollwo, E. Ann. Phys., Lpz.,29, 394 (1937).
(12) Hilsch, R. Ann. Phys., Lpz., 29, 407 (1937).
(13) E.g. see Faraday Society Discussion, "Chemical Reactions in-
volving Solids", pp. 883 et seq. (1938).
(13a) Durau, F. and Schratz, V. Z. phys. Chem. 159A, 115 (1932).
(136) Herbert, J. Trans. Faraday Soc. 26, 118 (1930).
(13 c) Tompkins, F. C. Trans. Faraday Soc. 34, 1469 (1938).
(13d) Kraft, H. Z. Phys. 110, 303 (1938).
(14) Bradley, R. Trans. Faraday Soc. 30, 587 (1934).
(15) Rayleigh, Lord. Proc. Roy. Soc. 156 A, 350 (1936).
(16) Proc. Roy. Soc. 163 A, 377 (1937).
(17) Urry, W. J. Amer. chem. Soc. 55, 3242 (1933).
(18) Watson, W. J. chem. Soc. 97, 810 (1910).
(19) Villard, P. C.R. Acad. Sci., Paris, 130, 1752 (1900).
(20) Berthelot, M. C.R. Acad. Sci.t Paris, 140, 821 (1905).
(21) Jaquerod, A. and Perrot, F. C.R. Acad. Sci., Paris, 139, 789
(1904).
(22) Richardson, O. and Richardson, R. C. Phil. Mag. 22, 704 (1911).
(23) Bodenstein, M. and Kranendieck, F. Nernst Festschrift, p. 99
(1912).
(24) Wustner, H. Ann. Phys., Lpz., 46, 1095 (1915).
(25) Alty, T. Phil. Mag. 15, 1035 (1933).
 REFERENCES 143
(26) Ward, A. F. Proc. Roy. Soc. 133 A, 506, 522 (1931).
(27) Lennard-Jones, J. E. Trans. Faraday Soc. 28, 333 (1932).
(28) Barrer, R. M. J. chem. Soc. 378 (1934).
(29) Braaten, E.O. and Clark, G. J. Amer. chem. Soc. 57, 2714 (1935).
(30) T'sai, L. S. and Hogness, T. J. phys. Chem. 36, 2595 (1932T).
(31) Burton, E., Braaten, E. O. and Wilhelm, J. O. Canad. J. Res.
21, 497 (1933).
(32) Williams, G. A. and Ferguson, J. B. J. Amer. chem. Soc. 44,
2160 (1922).
(33) Urry, W. J. Amer. chem. Soc. 54, 3887 (1932).
(34) Paneth, F. and Peters, K. Z. phys. Chem. I B , 253 (1928).
(35) Smithells,C.andRansley,C.E. Proc. Roy. Soc. 150 A, 172(1935).
(36) van Voorhis, C. C. Phys. Rev. 23, 557 (1924).
(37) Piutti, A. and Boggiolera, E. R.C. Accad. Lincei (5), 14 (1923).
Also R.C. Accad. Sci. Napoli (3) 29, 111 (1923).
(38) Mayer, E. Phys. Rev. 6, 283 (1915).
(39) Johnson, J. and Burt, R. J. opt. Soc. Amer. 6, 734 (1922).
(40) Roeser, W. Bur. Stand. J. Res., Wash., 7, 485 (1931).
(41) Taylor, 1ST. W. and Rast, W. J. chem. Phys. 6, 612 (1938).
(42) Williams, G. A. and Ferguson, J. B. J. Amer. chem. Soc. 46, 635
(1924).
 C H A P T E R IV

GAS FLOW T H R O U G H METALS

INTRODUCTION

It is difficult to prepare a pure metal, for not only do metals

contain traces of other metals, carbon, sulphur, or phosphorus,
but they also contain combined or occluded gases—hydrogen,
oxygen, nitrogen, and sulphur dioxide. These impurities exert
in many instances a profound effect upon the properties of the
metal. The present discussion concerns the behaviour of such
gas-metal systems. Pioneer researches on hydrogen-palladium
systems were made by T. Graham (i) in 1866; even earlier
observations on the system hydrogen-iron were made by
Cailletet(2)? who in 1864 found that some of the hydrogen
evolved when an iron vessel was immersed in dilute sulphuric
acid was absorbed in the iron. Deville and Troost(3) first
showed that hydrogen diffused through platinum, and the
interesting permeability of silver towards oxygen was observed
by Troost(4) in 1884. These workers have been followed by
others (5,6,7,8), amongst whom must be mentioned Richardson,
Nicol and Parnell(9), who developed an equation for the flow
of gas through a metal which is to-day the basis of interpre-
tations of diffusion processes:

where P denotes the permeability constant, k and b are

constants, I denotes the thickness, and p and T are respectively
pressure and temperature.
It has also been found (2,5,7,10,11,12) that hydrogen gas in
nascent form can penetrate metals such as palladium, iron, or
nickel at room temperature, if the metal is made the cathode
during electrolysis or if the hydrogen is generated by chemical
reaction at the surface. Pickling a metal in hydrogen in this
 GAS FLOW THROUGH METALS 145

way may alter its mechanical properties to a marked degree (13),

and the process of absorption is also very sensitive to traces
of poisons (14). The earliest observations on this type of
diffusion we owe to Bellati and Lussana(T), Cailletet(2), and
Nernst and Lessing(i2), but the field is one which has not
been extensively studied, although the results should give
information concerning interface reactions, and diffusion
within the metal.
Before discussing the experimental data upon the per-
meability of metals to gases it will be of advantage to consider
the cognate subject of gas solubility in metals. The differences
in behaviour met with there may be reflected in the per-
meabilities (P), since the latter are defined by

P- D™
dx
where D is the diffusion constant and dC/dx is the con-
centration gradient, defined in certain circumstances by the
solubility of the gas in the metal.

THE SOLUBILITY OF GASES IN METALS

For the study of gas-metal systems one has all the apparatus
and technique developed for measuring the sorption of gases
by solids (15). The powerful X-ray method then gives the crystal
habit of the products (16), or shows the effect the absorption has
upon lattice constants (17), so that one may construct the phase
diagrams for the system. Absorption equilibria have been
studied for over a decade by Sieverts and his co-workers (18)
and by many others, and as a result these often remarkable
systems are being increasingly understood. Summaries of
findings on gas-metal equilibria will be found in the books of
McBain(is) and of Smithells(i9), and it is not intended to give
more than a resume of the data here.
The metals which absorb common gases are summarised in
Table 39. Oxides are not mentioned in the table; but apart
from oxygen, hydrogen interacts most freely with metals, not
as a rule to give hydrides but rather alloy systems, or solid
146 GAS FLOW THROUGH METALS

solutions. Solution of gases such as hydrogen, oxygen, or

nitrogen occurs with dissociation, as is shown by a pro-
portionality between the solubility (at small concentrations)
and the square root of the jpressure. Compound molecules,
such as sulphur dioxide, ammonia, carbon dioxide, or carbon

TABLE 39. Summary of the reactivity of metals

toivards gases
Metals which Metals which do not
Gas Group dissolve gas Group dissolve gas
H2 IA Hydrogen gives salt- IB Au
like hydrides IIB Zn, Cd
IB Cu, Ag (slight) IIlB In, Tl
IIA Hydrogen gives salt- IVB Ge, Sn, Pb (but give
like hydrides covalent hydrides)
IIIA Al. Rare earths Ce, VB As, Sb, Bi (but give
La, Nd, Pr covalent hydrides)
IV A Ti, Zr, Hf, Th VIB Se, Te give covalent
VA V, Nb, Ta hydrides
VIA Cr, Mo, W VIII Rh
VIIA (Mn)
VIII Fe, Co, Ni, Pt, Pd
O2 IB Cu,Ag
IVA Zr
VIII Fe, Co, (Ni)
N, IIIA Al (molten) IB Cu, Ag, Au
IVA Zr II B Cd
VA Ta (nitrides) IIIB Tl
VI A Mo (nitrides only), IVB Sn, Pb
W (nitride) VB Sb, Bi
VIIA Mn (various nitrides) VIII Rh
VIII Fe (various nitrides)
CO VIII Ni,Fe (above 1000° C.) IB Cu
so2 IB Cu (liquid), Au (liquid) VIII Pt
He Do not dissolve in any
Ne metal so far studied
Ar either when liquid
Kr or solid
Xe
CO2 VIII Fe VIII Rh, Pt

monoxide, must also dissociate before they can penetrate into

the body of a metal. Temperature alters the solubility according
to an exponential law, so that curves of log (solubility) versus
\jT are often linear, save where two alloy phases co-exist
(H2-Pd, H2-Th, H2-Ti), where a limiting composition is
approached (H2-Pd, -Th, -Ti, -Zr, -V). or where various
 SOLUBILITY OF GASES IN METALS 147
allotropic forms of the metal exist, and show different
capacities to dissolve the gas (H2-Fe).
The greatest diversity of interaction is shown by hydrogen,
which reacts with metals to give three different types of
products:
(a) covalent hydrides: B2H6, SbH3, SiH4, AsH3,
(6) salt-like hydrides: [Na+][H~], [Ca++][2H-],
(c) alloys: PdH0.55.
Those systems with which experiments on hydrogen per-
meability have to do are alloy systems, at least over an
appreciable range of compositions. The hydrogen exists in the
metallic lattice most probably in an ionised or partially ionised
condition.
Metals
A metallic crystal is to be regarded as a giant molecule,
composed of positive nuclei and electrons, so that electrostatic
forces ensure the cohesion of the whole. The electrons occupy
definite energy levels, there being two electrons in each level
5 7 9 13...
6 8 10 12

Fig. 38. Representation of electron distribution in a metal at

0 and 300° K. (after Emeleus and Andersons).

at 0° K. At room temperature a few of the electrons by virtue

of thermal energy are promoted to higher energy levels. In
Fig. 38(20) one sees the number of electrons per level plotted
against the number of levels, the full curve representing the
 148 GAS FLOW THROUGH METALS
distribution at 0° K., and the dotted curve that at 300° K. In
a metal with the electron distribution obtaining at 0° K. no
electrical conductivity can occur because, since all levels are
filled, the Pauli exclusion principle indicates that no nett flow
of electrons may take place under any external potential, for
no level can receive further electrons. There can be con-
ductivity with the 300° K. distribution because there are some

UJ

UJ

Fig. 39. Graphical representation of electron distributions in metals

and semi-conductors (after Emeleus and Anderson(20)).

levels with only one electron, or no electrons. These levels may

receive electrons and there may be a nett flow of electrons
when a potential is applied.
The distribution curve of Fig. 38 is highly idealised. In the
actual lattice the positive nuclei cause a periodic variation in
the potential encountered by an electron moving through the
lattice. The solution of the wave equation for such a potential
distribution leads to the result that the electrons cannot
assume any energies from zero to a maximum, but that there
are bands or zones of permitted energies alternating with
 SOLUBILITY OF GASES IN METALS 149

bands of forbidden energies. If there are fewer electrons than

levels in a given band, or Brillouin zone, the condition of
partially filled levels and so of metallic conduction is fulfilled.
Correspondingly, if two Brillouin zones overlap (Fig. 39 A) (20),
and there are not enough electrons to fill both zones, metallic
conduction is again observed. If the zones do not overlap
but are adjacent, and one zone is full and the other empty
(Fig. 39B), the substance is a semi-conductor, for clearly the
input of a small activation energy will promote electrons
from the first zone to the second, and so leave incompletely
filled levels in both zones, with consequent electrical con-
duction (Fig. 39c). In an insulator the completely occupied
zone and the unoccupied zone are so far apart that with
normal activation energies no electrons are promoted, and so
electrical conductivity is absent.
Hydrogen-metal systems
The nature of alloy systems of hydrogen with metals is now
easily understood. The alloying hydrogen atom may provide
electrons which fill the empty levels in a band, while the metal
loses its para-magnetic properties (20). An H2-Pd alloy ceases
to be para-magnetic at the composition H0.55Pd and then, since
no more electrons can be supplied, very little more hydrogen
will dissolve. Hydrogen and deuterium give the well-known
isobaric curves (21) of Fig. 40, in which there is an apparent
invariant region, and a hysteresis effect. This is due to the
formation first of an a-alloy, and then of a /?-alloy, the two
alloys co-existing along the vertical lines of Fig. 40. The
occurrence of two phases may be explained by supposing that,
as the concentration of hydrogen atoms in the palladium
lattice rises, the atoms interact with each other as well as with
the lattice. When a critical interaction energy is reached, some
of the atoms gather together in closer association in the
palladium lattice, and so a new phase appears which is in
equilibrium with the original more dilute phase. As the
concentration of hydrogen in the lattice increases, the dilute
phase diminishes and the concentrated phase increases in
150 GAS FLOW THROUGH METALS

amount, and one travels along the invariant part of the curve.
When the initial dilute or a-phase is all consumed the system
becomes once more univariant, and also the lattice is nearly
saturated. The limiting composition is not certain but is of
70
7U0 mm. Isobar
per gm. Pd

60
\

50
5
5

V*
20

10 -04

£0 60 80 100 120 W 160 1800~ 200

Fig. 40. The solubility of hydrogen and deuterium in

palladium at atmospheric pressure.
the order PdH0.55 to PdH0.59. Lacher(22) analysed the solu-
bility-pressure-temperature data for the hydrogen-palladium
system and concluded that the heat of absorption of hydrogen
in palladium could be expressed as

where the second term expresses the interaction energy

between dissolved atoms, or protons, nK is the number of
 SOLUBILITY OF GASES IN METALS 151
gram-atoms of hydrogen dissolved and ns is the number of
gram-atoms of potential energy "holes" in the palladium*.
Various other hydrogen-metal systems have properties ana-
logous to the hydrogen-palladium system, except that the
interaction energy between dissolved hydrogen atoms or
protons is not usually sufficient to cause the separation of two
phases, at some critical composition. In Fig. 41 are given
absorption isobars for metals which dissolve hydrogen (23).
It is evident from the slopes of such isobars that the solution
process is strongly exothermic for metals such as V, Th, Zr,

0 200 400 600 800 jooo mo

Temperature, CC.
Fig. 41. Some isobars for metal hydrogen systems.

Ti, Ta, but endothermic for Cu, Fe, Co, and Ni. From the
slopes of log (solubility)- IJT curves (24) (Fig. 42) the heats of
solution given in Table 40 have been calculated. In some of
the metals, notably those in which hydrogen dissolves endo-
thermically, there is no appreciable alteration in the lattice
constants on solution of the gas in the metal. On the other
hand, when hydrogen dissolves in palladium one may have
at saturation 10 % expansion of the lattice. In the metals
Ti, V, Zr, Th, and Ta, where great quantities of hydrogen are
* The dissolved hydrogen is supposed to occupy interstitial positions in the
palladium lattice. These positions of minimum energy are referred to as potential
°nergy "holes" in the lattice.
152 GAS FLOW THROUGH METALS
absorbed, the systems approach a limiting composition and
have a new lattice structure. The densities and limiting com-
positions in the expanded states are illustrated by Table 41.

Tl
Zr __ - — —*
~~—*•

/
V-
7
- — ^ — I —
Th
V

4 /

/
—^^ Pd
f.
-———
°-

CO

g
?

1 •^—*.
• * - —

0 V s

-1 \ s e

-2 \
-3
\
Cu

-4
10 - 12 18 20 22
10.000/T
Fig. 42. Observed solubilities s of hydrogen in various metals subjected to one
atmosphere pressure of H2, shown by plotting log^s against 104/JT. The
solubility s is the volume of H2 gas (reckoned in c.c. at N.T.P.) absorbed by
100 g. of metal.

The methods of statistical mechanics (24,22) have provided

an approach to the problem of gas-metal solubility. The
metal can be regarded as containing a series of holes of low
 SOLUBILITY OF GASES IN METALS 153
potential energy distributed periodically according to the
lattice structure of the metal. The hydrogen molecules in the
gas phase dissociate and are absorbed into the lattice where as
atoms or protons they vibrate in the holes, among which they
TABLE 40. Heats of solution of hydrogen in metals
Exothermic Endothermic
Heat Heat
Metal (caL/mol. H2) Metal (cal./mol. H2)
Ti 10,000 Cu 14,100
Zr 17,500 Co 7,300
Th 22,500 Fe 7,000
V 7,700 Ni 5,600
Pd 2,040 Al 45,500
Pt 35,400
Mo 3,500
Ag 11,600

TABLE 41. Densities and limiting compositions

of some alloy systems
Limiting Density
Metal Density composition Density ratio
Ti 4-523 TiHa 3-91 0-864
Zr 6-53 ZrH2 5-67 0-867
Ta 16-62 TaH 1510 0-906
V 611 VH 5-30 0-867

are distributed at random. Diffusion occurs by jumps from

one hole to another, when a sufficient activation energy has
been acquired. The partition function is then constructed for
the systems
(i) H 2 molecules in the gas phase,
(ii) H atoms in the gas phase,
(iii) H atoms (or protons) in the metal lattice,
and so an expression for the equilibrium between gas molecules
and absorbed atoms is obtained:
{2nmkTf__

kT
c
154 GAS FLOW THROUGH METALS
where v8 = concentration of dissolved atoms, m
k = Boltzmann constant,
h — Planck's constant,
m = the mass of the hydrogen atom,
I = the moment of inertia of the "hydrogen molecule,
w2 = the weight of the normal electronic state of the
hydrogen molecule,
Xa = the heat of solution of a hydrogen atom in the
metal,
Xd = the heat of dissociation of a hydrogen molecule.
Smithells and Fowler (24) showed that if one defines s, the
solubility, as the number of c.c. of molecular hydrogen at
N.T.P., dissolved in 100 g. of metal, the formula above
reduced to ^

where pm = the density of the metal. These formulae are

applicable to solutions of hydrogen in metals such as Fe, Co,
Cu and Ni. Those metals such as Ta, V, Ti, Zr, or Th, where
the amount of hydrogen absorbed reaches a limiting value
and is thereafter constant, can also be treated. The same
methods led to the equation

-s \JcTJ r
L A3 2h2 J
where s0 denotes the saturation solubility. Inserting standard
numerical values gives

Lacher(22) was able to use the statistical mechanical

approach in the same way to explain the occurrence and form
of the peculiar isothermals of hydrogen-palladium systems
(Fig. 40). All that it was necessary to add to the previous
treatment was the assumption that as the concentration of
 SOLUBILITY OF GASES IN METALS 155
hydrogen atoms increased they interacted with one another
until at a critical concentration they formed clusters in the
lattice of the palladium instead of being distributed uniformly.

Oxygen-metal systems
When definite chemical compounds are not formed it should
be possible to apply statistical considerations similar to those
above to other gas-metal systems. Often, however, as in the
system O2-Ag, the nature of the solution process is not

,- PRB88 - tO CMS f
a 40 •<
1
3 .< 90 it I
4 10 .»

I 1

s /I
V' // /

V
N O~^
0
y

// / i

o
o .

Fig. 43. The solubility of oxygen in silver.

clear (25,26,27). The oxygen molecule is dissociated and must

be associated with the silver in at least two ways. First, an
unstable oxide Ag2O is formed, but this oxide decomposes as
the temperature rises, until, above 400° C, it should under
ordinary pressures disappear. However, an endothermic
solubility has now set in, and so the isobaric solubility-
temperature curves (Fig. 43) (26) first have a negative slope
(dissociation of the oxide) and then a positive slope (endo-
thermic solution). Other oxygen-metal systems exist where
there is a solubility of oxides in each other or in the metal,
or of oxygen in oxides. Here one has really to consider
156 GAS FLOW THROUGH METALS
compounds which do not conform exactly to the law of fixed
proportions, and where there is a lattice excess of one or other
component. This may occur when some lattice spaces of one
component are vacant, or when the other component can
occupy interstitial as well as lattice positions. Compounds

r.
7700 Liquid (Fet0)
+
/ r
Liquid(feOtFe)

7500
S-t-Liquid Fe

7300

y+FeO

7700
Solid Solution
Oxygen in Iron
$00
ct+FeO

700

0-5 7-0 20 25
Oxygen Atoms. %
Fig. 44. Temperature-composition phase diagram for O2-Fe.

having a variable composition are usually referred to as

Berthollide compounds, and the phenomena is encountered
with the following typical substances:

Oxides of Fe, Co, Ni ) ( t h e ^ °xideS a n d sulPhides

Sulphides of Fe, Co, Ni o™J •**"«*& solutions with
J x v gen and sulphur).
Tungsten bronzes) v a n a b l h t r a n e
Spinels J( y S considerable).
Zinc oxide (can contain excess metal).
 SOLUBILITY OF GASES IN METALS 157
Nickel oxide, for example, can vary in composition between
MOi-ooo and M O - ^ .
The solubility of oxygen in metals is complicated by the
formation of oxide phases, so that complex phase diagrams
result. As an example, the system Fe-O2 may be taken
(Fig. 44) (28,29). The temperature-composition diagram given
shows that oxygen is probably more soluble in y- than in
a- or 5-iron. The solution of oxygen in molten iron occurs
endothermically, since the solubility increases with rising
temperature (it is 147c.c./100g. at the melting point and
387c.c./100g. at 1734° C). The diagram illustrates the
great variety of solid solutions possible, and thus indicates
how iron-oxygen systems may deviate from the law of fixed
proportions.
The oxygen-copper system (30) has received considerable
attention, and also shows a complex temperature-composition
phase diagram (31). The solution of oxygen in metallic copper
occurs endothermically, as the solubility data of Rhines and
Matthewson(30) show:
Temp. ° C. 600 800 950 1050
Solubility (c.c./lOOg. metal) 5-0 6-6 7-0 10-9
Other systems which have been studied are O2-Co(32),
O2-Ni(33), and O2-Zr(34). One interesting feature of oxygen-
metal systems is the capacity of liquid metals to dissolve
large quantities of oxygen or oxides, which on cooling are
frozen out, as oxide or as bubbles of oxygen.

Nitrogen-metal systems
In order for nitrogen to be absorbed by a metal it is necessary
that the metal should be capable of forming a nitride. Thus,
while nitrogen is not absorbed by Cu, Co, Ag, and Au, it is
taken up by the metals Fe, Mo, W, Mn, Al, and Zr. A great
variety of phases is observed in some of these systems.
Nitrogen-molybdenum(35,36) gives the following:
a (Mo): body-centred cubic. No solid solution with nitrogen.
 158 GAS FLOW THROUGH METALS
/? (M03N): face-centred tetragonal. Stable only above
600° C. Contains 25 % atomic nitrogen.
y (Mo2N): face-centred cubic. Stable at all temperatures,
and contains 33 % atomic nitrogen.
S (MoN): hexagonal. Contains 50 % atomic nitrogen.
Because of the use of iron catalysts for ammonia synthesis,
the nitrogen-iron phase diagram is especially interesting (37,38)
(Fig. 45). Other systems, such as nitrogen-manganese (39),
nitrogen-aluminium (40), and nitrogen-zirconium (34), need not

5 70 75 20 25
Nitrogen Atoms, %
Fig. 45. Temperature-composition phase diagram of N2-Fe.
a = solid solution N 2 in Fe, y = solid solution N"2 in Fe, y' - Fe 4 N,
c = Fe3N.

be considered here, since the behaviour already described for

nitrogen-molybdenum and nitrogen-iron is typical. It is
interesting that nitrogen dissolves in molten aluminium to
give an approximately linear plot of log (solubility) against
l/T. This is characteristic also of solutions of oxygen in
silver at high temperatures, and of solutions of sulphur
dioxide in liquid copper (41,42). In the latter case, as in the
two former, the solubility is proportional to the square root of
the pressure, indicating a dissociation of the sulphur dioxide
molecule in solution.

THE SOLUBILITY OF GASES IN ALLOYS

The study of gas-alloy systems has been confined principally
to hydrogen. One might recognise two possible classes of alloy,
one compounded of two hydrogen-dissolving metals, and the
 SOLUBILITY OF GASES IN ALLOYS 159
other composed of one hydrogen-dissolving metal and a second
metal inert to hydrogen. The behaviour in the second of these
systems is somewhat unexpected, for as the summarising data
below show, instead of the hydrogen solubility diminishing
steadily to a zero value as the percentage of the inert metal
increases, one often finds a solubility first rising to a maximum
and then falling towards zero as the percentage of the inert
metal increases. When hydrogen dissolves in both metals A
and B of the alloy, however, one usually finds a continuous
change in solubility from 100 % A to 100 % B. The system
H2-Mo-Fe appears to be an exception:
(a) Alloys in which the gas solubility changes continuously:
Fe-V(H2)(43),
Pt-Pd(H2)(±4),
Al-Cu(H2)(44),
Ag-Au (O2)(45).
(b) Alloys in which the gas solubility exhibits a maximum:
B-Pd(H2)(46),
Au-Pd(H2)(47),
Ag-Pd(H2)(47),
Fe-Mo (H2)(48).
In the following figures are given data illustrating these
types of behaviour. The maxima are not always sharp as in
the H2-B-Pd system, nor are they always so well defined. The
curve of log (solubility) against 1/T in Fe-Mo alloys, as Fig. 42
shows for many hydrogen-metal systems, is linear (49).
It would be out of place to give further details of gas-metal
systems here. The purpose of this summary has been to
illustrate the very interesting types of equilibria observed, so
that one may have some idea of the state of combination of
the gas diffusing in the solid. We have seen that hydrogen will
diffuse in the form of atoms or protons, that oxygen or nitrogen
diffuses after dissociation, and that oxygen-metal and nitrogen-
1(30 GAS FLOW THROUGH METALS
metal reactions lead much more frequently to the formation
of new phases—oxides or nitrides—than do hydrogen-metal
reactions. Thus we must think of the diffusion of oxygen or
nitrogen as a handing on of the
W8°
dissolved gas atom by successive
decompositions of unstable oxides
or nitrides. There is little evidence 30
to show what is the condition of the
diffusing particle in its transition
state, but it would be anticipated
that oxygen and nitrogen, being
much more electro-negative than
hydrogen, may move as negative
ions while hydrogen may very well
be considered to diffuse as protons, r
I t has been
been seen t h a t even the OPd 10 20 30
Fi
solution of sulphur dioxide follows 8- 46- T h e solubilit
y of H * i n
,. . ,. 5 ,, , , , Pt-Pd alloys.
dissociation of the molecule, al-
though the nature of the fragments remains unknown. Gases
such as carbon monoxide, carbon dioxide, or ammonia must

10 15 20
Atoms % of Boron
Fig. 47. Solubility of H2 in B-Pd alloys.

in the same way undergo dissociation into their components

before they can penetrate a metal. The specificity of some of
 SOLUBILITY OF GASES IN ALLOYS 161
the systems is one of their most fascinating features, and a
great deal of further work must be undertaken before the
reactions occurring can be properly understood.
THE MEASUREMENT OF PERMEATION VELOCITIES
The usual high vacuum technique employed in studying
sorption equilibria and kinetics may be used in obtaining the
permeability of metals (is, 19). In addition a special problem
arises, the mounting of metal membranes in a manner which
will be vacuum-tight and which will permit of heating the
specimen to high temperatures.
Ham (50) in an early apparatus mounted a sheet of platinum
between two heavy steel tubes with flanged ends. The flanges
were ground flat, and the joint put under great pressure by
means of bolts passing through heavy steel rings on either side
of the flange. Unless the surfaces are very smooth, or the
pressure so great as to cause flow of the metal, this arrange-
ment is not entirely vacuum-tight. In later arrangements (51)
the membrane was welded across a tube by atomic hydrogen.
This method is very satisfactory, but it may not always bfe
possible to do the welding. Such systems have the advantage
that eflEects of temperature variations along the furnace are
obviated.
In other arrangements tubes of the metal may be heated
in a furnace. One end of the tube is closed and the other is
open to a manometer. The whole tube may be surrounded
by the diffusing gas. This very simple and useful method was
used by Borelius and Lindblom(52).
It is best to have the tube in the form of a bulb of large
area in the centre of the furnace, with a narrow-necked tube
leading from it out of the furnace. The narrow tube may then
be fitted to glass by a ground joint outside the furnace.
The arrangement avoids the effects of temperature inhomo-
geneities near the ends of the furnace. Another way of
doing this requires a short tube of the permeable metal
(e.g. palladium) sealed at one end by gold solder, and at the
other end welded by gold solder to a less permeable metal
162 GAS FLOW THROUGH METALS
such as platinum or nickel. If the metal is welded to platinum,
the platinum may be sealed through soft glass, and the system
taken up to 370° C. under vacuum for the diffusion measure-
ments. Another variation of this arrangement is to solder
(with gold or platinum solder) a plug of palladium across the
mouth of a platinum tube (53), which is again sealed through
soft glass. In this connection it should be remembered that
copper may be brazed to soft glass, and tungsten sealed
through pyrex, provided the arrange-
ment is such that these seals need
not withstand high temperatures.
The pressure on the high-vacuum
side may be given by mercury mano-
meters, McLeod gauges, or Pirani
gauges according to the rate at which
the pressure rises on this side. At Fig. 48. Apparatus for mea-
high pressures robust methods of ^2 %&££?
mounting membranes across tha tubes
are necessary. A suitable arrangement is that in Fig. 48(54),
used to measure permeabilities up to 112 atm. The mem-
brane^ was brazed between the faced ends of two stout-walled
tubes B, the joint being heated by a small external furnace.
The apparatus required to measure the diffusion of nascent
hydrogen is simpler, since permeation occurs at low tem-
peratures. It is usual to use a hollow tube of the metal being
studied, as cathode, and by a ground joint, an inseal, or by
welding to another metal which is then insealed, to connect
it with a manometer system. A typical apparatus of this
kind is that of Borelius and Lindblom(52). If the hydrogen is
generated by chemical means, the apparatus of Edwards (55)
may be employed (Fig. 49), where the hydrogen diffused is
measured by displacement of mercury. It would also be possible
to adapt Borelius and Lindblom's apparatus to this case.

THE INFLUENCE OF TEMPERATURE UPON PERMEABILITY

The velocity of diffusion through a metal increases very
rapidly as the temperature rises, as Fig. 50 illustrates for
 Fig. 49. Apparatus for measuring the rate of passage of
nascent hydrogen through iron.

500 000 700 800 000 1000 1100 1200

Fig. 50. The effect of temperature upon the permeation velocity of

hydrogen through platinum at various pressures.
164 GAS FLOW THROUGH METALS
the passage of hydrogen through platinum. One notes the
similarity of these curves to the exponential rise of vapour
pressure with temperature, or of chemical reaction velocities
with temperature. As is well known, this is due to an
exponential term erAH!RT, e~ElRT in the vapour-pressure
equation, or expression for the velocity constant respectively;
and a similar term arises in the permeability constant P:
P = Poe-EiR1\
However, some of the earliest studies of permeation velocities
expressed these velocities as
Rate = AT>\
Rate = AebT,
Rate = Ae-w.
For example, Winkelmann (8) gave for the passage of hydrogen
through iron n~5; while Johnson and Larose(56) for the
diffusion of oxygen through silver gave a value of n — 14-6.
The second of these equations has been applied to hydrogen-
nickel (57) and hydrogen-palladium(58) (for the permeability);
and to carbon-iron, and nitrogen-iron (for the diffusion
constants) (59) with some success. There is no theoretical
interpretation for these first two equations however, and the
agreement is in general better when the expression P = Po e~ElRT
is used, and log P plotted against 1/T. This has been done by
Smithells and Ransley (60) for most of the gas-metal systems
whose permeability has been studied, and some of their
diagrams are reproduced here (Figs, 51-53). They found
the equation to be very satisfactory indeed, although ex-
ceptions may occur where allotropic modifications of metal
exist in the temperature range considered (e.g. iron (58)), or
if the diffusing gas can form two or more types of alloy or
phase with the metal (H2-Pd, N2-Fe).
From the slopes of the curves logP against 1/JP. one may
calculate the temperature coefficient in cal./atom of gas
transferred. The first attempt to give a meaning to this
temperature coefficient is due to Richardson, Nicol and
Parnell(9), and no subsequent theories have improved greatly
 >

«
\ .

H^drogen-Iron
Lindbloi n)

\ \
V
\ \>

Nitrogen-Steel
itrogen-St

\
monoxide-Ste
(Ryderf

(Johr Son &i I.arose)

>encer)

0
Oxyg en- \

\
Hydrogen - Platinum
Silver ( W d N !
S i \

SO
omba^p

\
V- <
\

(Den ingj o*
o
\

N Hyd ogen-F Ifckel

(Lombard *Eichner)

0 10 12 14
10000

Figs. 51 and 52. Effect of temperature on the rate of flow of gases through metals.
166 GAS FLOW THROUGH METALS

upon their treatment. They solved Fick's law for the special
case of a gas dissociating in the medium and diffusing as atoms.
In the stationary state, when d2C/dx2 = 0, they found

Rate of permeation = y — I — I p-,

1
P - 95-4 mm.

X
N
N
b/mm.
s.
\

X\ \
\

s_
'1
p - )-08mn

fs \ N
ion Ve

°\
\ \
to
s
I \ H> drogef

o
N trogen
\ \

)rnrn. ^

\
\
10000

Fig. 53. Effect of temperature on the diffusion of hydrogen and

nitrogen through molybdenum.

where D = the diffusion constant,

kL = the dissociation constant of hydrogen in the metal,
s0 = the solubility of molecular hydrogen in the metal,
p = pressure.
From the experiments they considered that
 INFLUENCE OF TEMPERATURE 167
7
where C = constant and GD/2s0 = 8-59 x 10~ ; and so for the
hydrogen-platinum system

Q = 6-60 x io

where Q denotes the mass of gas diffusing per sec/cm. 2 of

platinum of thickness Z, from a pressure p into a vacuum. We
now know of course that the heat of dissociation of hydrogen
in the metal is not 36,500, because the diffusion coefficient of
hydrogen atoms in a metal is itself exponentially dependent
on temperature:D = Doe~Eo/RT. Richardson's equation should
then become
Q = 6-60 x 1
I

and until Eo is known one cannot find the heat of dissociation

of hydrogen in platinum. The term in T* can as a rule be
neglected in a formula such as that above, and it is usual to
write P = Poe~EfRT, where P denotes the permeability constant
(c.c. at N.T.P. diffusing/sec./cm.2/mm. thick/cm., or atmo-
sphere, of mercury). The values of E and Po for a number of
systems are given in Table 42. As Smithells and Ransley have
pointed out, the variation in the term PQ (103-fold) is small
considering the variety of systems listed. Since permeability
constants depend upon so many variables, this consistency
is notable, and so far susceptible to no interpretation. In
systems like H2-Ni the temperature coefficient is nearly the
same in all the specimens considered; but in H2-Pd a remark-
able variation in the temperature coefficient occurs, and this
coefficient appears to depend upon the previous history of
the specimen. In its active most permeable state, the tem-
perature coefficient is low; but in impermeable palladium
it is high. When hydrogen passes through composite metal
sheets of Cu-Pd and Ni-Pd, the temperature coefficient
approximates to the values for the least permeable metal
(copper or nickel), while for Ni-Pt and Pt-Ni the temperature
coefficient is that of the metal at the outgoing face. I t should
168 GAS FLOW THROUGH METALS

TABLE 42. Permeability data for gas-metal

systems (P = Poe-EfRT)

(c.c./sec./cm.2/ E
System mm. thick/ (cal./g. atom) Author
atm. press.)
H2—Ni 14,600, 13,100 Post and Ham(ei)
(below Curie
point)
— 13,100, 12,040 Post and Ham(ci)
(above Curie
point)
1-3 xlO" 2 15,420 Lombard (57)
0-85 x 10- 2 13,860 Deming and Hendricks(62)
1-4 xlO- 2 13,800 Borelius and Lindblom<3-2)
105 x lO- 2 13,400 H a m (50)
1-44 x lO"2 13,260 Smithells and Ransley(60)
— 13,400 Post and Ham(5i)
H2—Pt-Ni — 13,400 H a m (50)
H2—Pt 1-41 x 10- 2 19,600 Richardson, Nicol and
2
1-18 xlO~ 18,000 H a m (3-j)
2-6 xlO" 1 19,800 Jouan(63)
H2—Ni-Pt — 18,000 H a m (50)
H2—Mo 0-93 x lO- 2 20,200 Smithells and Ransley(60)
H2—Pd 5.000 Melville and Rideal(64)
— 17,800 Melville and Rideal(64)
2-3 xlO" 1 4,620 Lombard, Eichner and
Albert (53)
3 0 xlO~ 2 10,500 Barrer<53)
H2—Ni-Pd — 14,300 Melville and Rideal(64)
H2—Cu 2-3 xlO- 3 16,600 Smithells and Ransley(60)
1-5 xlO- 3 18,700 Braaten and Clark (65)
H2—Cu-Pd 13,700 Melville and Rideal<64)
— 11,400 Melville and Rideal(64)
H2—Fe 1-63x10-3 9,600 Smithells and Ransleycco)
1-60 x lO- 3 9.400 Borelius and Lindblom(52>
2-40 x lO- 3 11,000 Ryder (60)
— 8,700 Post and Ham(3i)
(below 900cC.)
— 18,860 Post and Ham(5i)
(above 900° C.)
H2—Al 3-3-4-2 30,800 Smithells and Ransley(67)
O2-Ag 3-75 x 10- 2 22,600 Spencer (OS)
206 x lO- 2 22,600 Johnson and Larose(56)
N2—Mo 8-3 xlO-- 45,000 Smithells and Ransley(eo)
N2—Fe 4-5 xlO-* 23,800 R y d e r (66)
CO—Fe 1-3 x l O 3 18.600 Ryder <s«>
 INFLUENCE OF PRESSURE 169

be remembered that in every case the diffusing molecule

dissociates; when carbon monoxide passes through iron, for
example, the carbon and oxygen diffuse separately.

THE INFLUENCE OF PRESSURE ON THE PERMEABILITY

OF METALS TO GASES

In this field some very thorough studies have been made

for hydrogen-metal systems, especially by Lombard and
Eichner(69), Smithells and Ransley (60) and Post and Ham(6i).
Even for hydrogen-metal systems, however, the interpretation
of the permeation rate-pressure isotherms is not clear. Studies
of glass, crystal, and organic membrane permeability have
shown that two possible permeation rate-pressure relationships
emerge:
(i) Activated diffusion without dissociation:
dp/dt = kpe-VT.
(Gas diffusing in SiO2, KBr, zeolites and organic polymers
such as rubber, bakelite, ebonite, cellulose esters.)
(ii) Activated diffusion with dissociation:

(H2, O2, N2, SO2, CO diffusing through metals.)

It is with the second type of diffusion system we have now
to do. The first type requires an open crystal structure with
large interstices (e.g. zeolites); in metals the crystal form is
never sufficiently open, and so one finds that inert gases
cannot either dissolve in or pass through a metal. This is true
of any gas which cannot in some way react specifically with
the metal under consideration. One criterion of the specific
interaction is the *Jp law in the expression dpjdt — kp* e~b/T.
Richardson (9) deduced the first specific expression for the
permeation velocity through a metal, and his expression
(p. 167), dp/dt = Ap^T* e~bdT, can be taken as the basis of the
present discussion of isotherms. However, not all experiments
gave an exact relationship dpjdt = Ax *Jp. Thus one finds, on
170 GAS FLOW THROUGH METALS
writing dpjdt = Axpn, the following values of n given for
hydrogen and palladium:
Schmidt (1904) (70) n = \,
Holt (1915) (71) n= 1,
Winkelmann (1901) (8) n = 0-7,
Lombard and Eichner( 1932) (72) n = 0-8, n = 0-62,
Lombard and Eichner (1933) (69) (i) n = 0-58, 0-59,
(ii) n = 0'56 (mean of a
number of results).
Since Schmidt's and Holt's results are not very accurate,
and could also be expressed approximately by a yjp law, one
is justified in saying that the exponent is less than one, and
indeed the later values show it to be very nearly one-half.
Finally, Ham and Sauter(73,74) working with very pure
palladium obtained many values of the exponent n between
0*535 and 0-50. They found a maximum deviation from the
xlp law at 248° C. when n = 0*585. One may summarise the
position by saying that there seem to be small and somewhat
variable deviations from an exact y/p law, but that this law
is very nearly fulfilled.
One explanation of a ^ law is, as previously indicated, that
diffusion occurs as atoms. One then has the following relations:

(i) J^ ^ = kl9 for small concentrations (Nernst dis-

tribution law).
C2 (gas)
(ii) J1 . — - = Jc2 (law of mass action).
C(gas)
(iii) Rate of permeation, P = -Dd°n|SoM) (Fick's law).
ax
From (i) and (ii)
(iv) CH(solid) = ^y/k^y/lCHg(gas)}.
dC f CH —l CH*~
Then if one writes - r ~ = ° and substitutes from (iv)
for CH (solid), one finds

* IAJ denotes the concentration of H-atoms just inside the ingoing surface
(x = 0) and jCH the concentration at the outgoing surface (x = /).
 INFLUENCE OF PRESSURE 171
which gives the observed relationship if VG^HJ^AAO^H.)* a s
is the case when one side of the metal is held at a near-vacuum.
This preliminary treatment is, however, very much too simple.
It was first noted by Borelius and Lindblom (52) that even when
one side of the metal was held under a vacuum, the diffusion
isotherms appeared to obey a relationship (Fig. 54)

They therefore suggested that a threshold pressure must be

reached before permeation commences—an explanation
difficult to base upon theory.

30
0
702°C. y
*70 s /
20

532X.
10
******
360*1
0
10 15 20 25 30
, in mms.
Fig. 54. Permeation rate-pressure isotherms of Borelius and Lindblom (52).

Then Smithells and Ransley(60) noted that the diffusion

isotherms bent round at low pressures and thus did pass
through the origin. They suggested that at low pressures the
rate of permeation was proportional to the fraction of the
surface covered by an adsorbed layer, 6, as well as to the square
root of the pressure. Thus, at low pressures the permeability P
is given by p_

and so steadily increases until 0 ~ 1. Thereafter the relationship

is P = k*Jp. Smithells and Ransley (60) plotted many of their
own and other workers' permeability-pressure isotherms to
illustrate their suggestion. Some of their curves are reproduced
in Figs. 55 and 56. The figures show, however, as do the
authors' calculations, that 6 approaches unity at lower
pressures the greater the temperature. That is, at high tern-
172 GAS FLOW THROUGH METALS

peratures the y/p law is obeyed at lower pressures. It would

be necessary to suppose that adsorption was endothermic for
this to be true, whereas it is well known that for these systems

/ 760°K

7
z
7

0 5 10 15 20 25
Fig. 5J5. Permeation rate-pressure isotherms for H2-Pd.

y 723 9 K.

y
1 yX y

0 2 4 6 ft 10 VP

Fig. 56. Permeation rate-pressure isotherms for H2-Cu.

adsorption occurs in atomic form with evolution of heat. The

results of Borelius and Lindblom on the diffusion of hydrogen
through iron at 702° C. (Fig. 54) and at 100° C. (Fig. 57) will
 INFLUENCE OF PRESSURE 173

illustrate the seriousness of this objection. Fig. 54 shows that

at 702° C. if the rate is expressed as P = k(^p — ^p(), Pt has a
value of about 4 mm., while Fig. 57 shows that at 100° C.
pt has a value of nearly 8 atm. These, according to Smithell's
theory, must give approximately the pressure needed to
saturate the surface—1500 times as great a pressure at 100°
as at 702° C , although adsorption occurs exothermally.

I
0 1 2 3 4 5
Vp(Atm/2)
Fig. 57. H2-Fe, the permeation velocity as a function of pressure, at 100° C,

In other ways Smithells and Ransley's theory explains the

observed facts. For if
P = kOyJp and 6 = -

(Langmuir isotherm for sorption with dissociation), one has

P = k kzp at low pressures,
P = k^Jp at high pressures,
expressions which cover the observed facts.
Another viewpoint has been developed by Ham(6i) and his
co-workers. These authors find that the slopes of permeation
rate-pressure isotherms for hydrogen-nickel systems are not
exactly 0-5, as Fig. 58 shows. The deviations from a <Jp law
are at a maximum near the Curie point, and are related to
the purity of the specimen of nickel employed. The isotherms
do not have a slope of 0-5 near the Curie point unless the
nickel is thoroughly decarburised. Similarly, the isotherms
for carbonyl iron(6i) at 500° C. remained with an exponent
well above 0-5 until the carbon was removed. The authors
174 GAS FLOW THROUGH METALS

considered that all deviations from the exponent of pressure

of 0-5 could be attributed to volume effects—some of the
diffusing substance was associated with the impurities (often
carbon, and possibly nitrogen or oxygen) in atom pairs, or as
ions H2+. They pointed out that Smithells and Ransley's
materials had not been decarburised, and attributed the
deviations from a ^jp law to this.

0-6

\
/ \

0-5 200
500 400
Temperature *C.

0-6
1

/ r \
\
>/
V
05
500 4O0
1^ 300 200
Temperature °C.
Fig. 58. Exponents n in expression permeation rate = A;pw(H2-Ni)
for two different samples of Ni.

However, it is likely that these authors were dealing with

two different phenomena. A satisfactory explanation of
Figs. 54-57 may be given in terms of phase-boundary pro-
cesses both at the ingoing and outgoing surfaces (53). Adopting
the nomenclature of p. 170, one may suppose that the trans-
ference of an atom from both
(i) adsorbed layer to solid at the ingoing surface,
(ii) solid to adsorbed layer at the outgoing surface,
is comparable in speed with diffusion in the solid. Then much
of the material transferred across the ingoing interface is
removed by diffusion, and
Actual ( 0 CH) < Equilibrium (0CH),
 INFLUENCE OF PRESSURE 175

as defined by equation (i) p. 170. At the outgoing interface,

the slow rate of transfer across the interface, but considerable
diffusion velocity in the solid, leads to an accumulation just
inside the outgoing surface:
Actual (/CH) > Equilibrium (/CH).
Thus the actual concentration gradient is smaller than the
concentration gradient assumed in equation (v), p. 170. The
dependence of this gradient upon the pressure is complex, but
the curve of the permeation rate plotted against the pressure p
must pass through the origin, because when p = 0 both 0 C H
and jOH are zero, and therefore the permeation rate is zero.
At intermediate pressures, the permeation rate depends upon
a complex function of pressure (see the feet of the curves of
Figs. 54-57), but as the pressure rises, or at high temperatures,
experiment shows that
CAI-AI = *V* (Figs. 54-57).
The universality of this relation must have important im-
plication^ for the actual phase-boundary processes. This point
will be discussed in more detail on pp. 178 et seq.

PERMEATION VELOCITIES AT HIGH PRESSURES

Wustner's(75) studies upon the hydrogen permeability of silica
glass (Chap. I l l ) showed that even up to 800 atm. the per-
meation rate-pressure isotherm was linear:

Rate = i f e~blT.

A few comparable studies have been made upon the per-

meabilities of metals to oxygen and hydrogen. Smithells
and Ransley(76) observed that at pressures of 112 atm. the
permeation rate-pressure isotherm obeyed accurately the
Richardson equation:

Their data are shown in Fig. 59. A similar result was obtained
by Lombard and Eichner(09) at pressures of 26kg./cm.2 for
176 GAS FLOW THROUGH METALS

the H2-Pd system, and by Borelius and Lindblom for H 2 -Fe

at 28 atm. The system O2-Ni, on the other hand, gave a
limiting permeation velocity at high pressures (76) (Fig. 60).
In this case a visible film of oxide forms on the surface of the
metal, and it is likely that in the presence of the solid oxide

>
n
f

e
y
o

5 !0
VP (Atmospheres)
Fig. 59. Diffusion of hydrogen through nickel at 248° C.

as1 o

If
I I
0L _ ] o

VP (Mm.)
Fig. 60. Diffusion of oxygen through nickel at 900° C.

the concentration of dissolved oxide (or oxygen) in the nickel

in contact with the oxide film had reached a saturation value,
for this would lead to a limiting permeation rate, according
to Fick's law
C —C
Rate of permeation = —D-~—',

where Ca is the saturation oxygen concentration at x = 0, and

Ct its value at the outgoing surface x — I. The contrary must be
 PERMEATION VELOCITIES AT HIGH PRESSURES 177
true of the systems H 2 -Si0 2 , H2-Ni, H 2 -Pd. The fact that many
interface processes may occur in the diffusion can, as is shown
subsequently (p. 182), lead to the conclusion that in the
stationary state of flow concentrations within the solid may
be very much less than their values in equilibrium systems.
Among the possible phase-boundary processes one might
include the following: An adsorbed gas molecule or atom is
driven into the solid by molecular bombardment by an
activated gas molecule. Smithells and Ransley tested this
possibility by introducing argon at 100 atm. into a diffusion
system H2-Ni where the hydrogen pressure on the ingoing side

2) Molecule (J\__/J\ Molecule

3) Adsorbed atom
atom
Metal Metal

(T)-(?) Molecule
00 /f777777y77Z (T) Adsorbed atom

Metal * Metal ^-'Dissolved atom

Fig. 61. Models of possible phase-boundary processes.

was 4 atm. No difference in the permeation velocity was

observed, so that at least certain types of penetration process
by molecular bombardment do not occur. The experiment does
not eliminate the two types of penetration by bombardment
represented by the diagrams above (Fig. 61). The gas molecules
are supposed to be hydrogen, which are adsorbed in atomic
form. To distinguish the atoms they have been numbered.
Other diffusion systems in which conditions might corre-
spond to very high pressures indeed are met with in the
passage of nascent hydrogen through metals. For example,
Barrer(53) found that hydrogen supplied at a current density
of 0-44 amp./cm.2, by electrolysing sulphuric acid at 20° C,
passed through a hollow palladium cathode 103-fold as rapidly
as hydrogen diffused from molecular hydrogen gas through
 178 GAS FLOW THROUGH METALS

a sample of palladium in a similar state of activity. If the

phenomena encountered are essentially the same, then
* elect.
t. / ff elect. .
al
* thermal V ^thermal
^
and as ^thermal = 1cm-> Select. ~ 104 atm. While in some ways
the analogy may be false, there is no doubt that the pressures
or concentrations of hydrogen atoms at the surface of the
metal were extremely great. Similarly, Borelius and Lind-
blom (52) found that if hydrogen were generated electrolytically
at the surface of an iron tube some of it passed through the
tube and the rate of permeation P was given by

and was completely analogous to the expression for thermal

diffusion p = kiyp _ ^ e_E/RT^
even the temperature coefficients E being the same. Assuming
the complete analogy, Borelius and Lindblom found from
their data _ 1 7 o n oT
^(atmos.) — 1 / > U U U i (amp./cm 2 .)-
Their own experiments were conducted at current densities
of up to 0-043 amp./cm.2 By carrying out these experiments
at large current densities, it should be possible to reach a
current density at which the surface layers of metal were
completely saturated. The permeation velocity should then
no longer increase as the current density increases. This point
has been tested (53), and has important implications in the
discussion on mechanisms of diffusion in the next section.
It was found that only in very active palladium tubes could
a limiting permeation velocity be reached (see Fig. 62).

SOME MECHANISMS FOR THE PROCESS OF FLOW

In 1935, Ham (77) proposed a specific equation for the rate
of flow of gas through a metal based upon kinetic theory.
Ham's equation approximates to
 DETAILED MECHANISMS OF PROCESS OF FLOW 179
where A, y, a and b are constants, and y~\. It will be seen
that this equation is similar in form to the Richardson
equation, but since it involves a number of assumptions it
need not be considered further.
Riemann(78) has suggested another variant of Richardson's
equation based upon thermodynamic considerations and
Fick's law. His equation took the form

% O f V ) T<
In this equation a = Gp/B (Gp denotes the atomic heat of
dissolved gas atoms), pl9p2 = pressures of gas at the ingoing
and outgoing surfaces respectively of a plate of thickness Z,
and AH0 is defined by AH = AH0 + (|—2a) RT (AH denotes
the heat of desorption of two atoms of gas from the metal to
the gas, as a molecule). But since we are dealing with an
activated diffusion, D can be further written as D = Doe~~EIRTy
where E denotes the activation energy for diffusion.
Riemann's equation, Richardson's equation (p. 167) and
equation (v), p. 170 all suffer from one grave defect. It has
been assumed that the concentrations just inside the solid are
equilibrium concentrations, and this is only true where there
are no rate-controlling phase-boundary processes. There do
in fact seem to be no slow phase-boundary processes for
diffusion through rubbers (79), but this is not so for diffusion
through metals. Melville and Rideal (C4) made the first attempt
to include possible phase-boundary processes in the equation
of flow. Other possible phase-boundary processes were given
by Smithells and Ransley(76), Wang(80), and Barrer(8i). They
regarded the following as possible:

(i) An adsorbed atom passes into the metal:

Rate = M J 1 - ^ )
(kx is the velocity constant of the reaction, dx the fraction of
the surface covered by adsorbed atoms, C H is the concentration
180 GAS FLOW THROUGH METALS
of hydrogen gas just inside the metal, and Cs is the concen-
tration of hydrogen in the metal when saturated with the gas)
(ii) A dissolved atom re-enters the surface:
Rate = &2^H(^ — ^i)#
(iii) A molecule strikes the surface and is adsorbed as atoms:

(iv) Two adsorbed atoms evaporate as a molecule:

Rate = &40f.
(v) A molecule strikes the surface, one atom being absorbed
and the other adsorbed:

(vi) An absorbed atom combines with an adsorbed atom,

and evaporates as a molecule:
Rate = k^C^d^
Other processes may be conceived, but they are not probable
ones, and in any case the six listed above are adequate for
this discussion. Smithells and Ransley, considering the first
four of the preceding rate equations, deduced as a general
expression for the velocity of permeation

where A and K are terms involving d, and considered that a

simple form of this equation

agreed with their experimental findings that at low pressures

dpjdt approximates to Bp9 where B is a constant; and at high
pressures to C Jp, where C is also a constant.
Wang (80) and Barrer(8i) gave equations defining the per-
meation velocity which included all six of the processes above.
 DETAILED MECHANISMS OF PROCESS OF FLOW 181
Wang's treatment led him to conclude that the permeation
velocity would increase indefinitely with pressure, according
to a ^p law when p is large. This would explain Smithells
and Ransley's observation that even at 100 atm. the rate
of permeation of hydrogen through nickel was still propor-
tional to the square root of the pressure. At these pressures
it had previously been supposed that the surface layer would
be saturated with adsorbed gas, and so the processes (i) to (iv)
would give a limiting value for the permeability. Wang
considered that the inclusion of (v) and (vi), however, made
this no longer necessary, and so his equations indicated
that some such processes as (v) and (vi) must have occurred.
However, Barrer(8i) pointed out that Wang reached these
conclusions because the rate equations given by him and by
Smithells and Ransley took no account of the approach
towards saturation in the metal itself. The rate equations (i)
and (v) were in fact written by them as
(i) Rate = k±6v
(v)
The correct equations for the permeability showed that it
always reached a limiting value at infinite pressure. Barrer (82)
also showed that the phase-boundary processes could result
in very great concentration discontinuities at the surface of
a metal, and that under these conditions, even at high pres-
sures, the concentration just within the metal could be a
fraction only of the equilibrium value found in the absence of
phase-boundary processes. He advanced the view that in the
expression
G —C
Permeation rate = D-^-:—-,
v
the concentration C± just inside the ingoing surface was even
at 100 atm., for the H2-Ni system, nowhere near its saturation
value, so that Wang's deduction of a <Jp law is valid at these
pressures. The manner in which concentration discontinuities
arise is indicated on pp. 174-5.
 182 GAS FLOW THROUGH METALS

Later (53), it was shown that for certain H2-Pd systems

(C\ — C2) was indeed much less than its equilibrium value. This
was indicated by measuring both the permeability constant and
the diffusion constant (Chap. V). When gas flow occurred from
a finite pressure through palladium into a vacuum the values
of (Cx — G2) were those in Table 43, for palladium samples of
low permeability. On the other hand, when a palladium
sample was alternately oxidised and reduced the permeability
was high, and for some measurements of Lombard and his co-
workers (69) the values of {Cx — C2) approached the equilibrium
ones (Table 43). Here the phase-boundary processes have

TABLE 43

Pi (Ci-Ca) P 2 (extrap.) (Ci-C,) (Ci-Ca)

D 1 (Barrer) (Barrer) (Lombard) (Lombard) (equilibrium)
T°C. om.'sec." c.e./sec./cm.2/ at 1 atm. c.c./sec./cm.2/ c.c./c.c.Pd/ c.c./c.c.Pd/
xlO- 5 mm./cm.Hg pressure mm./cm.Hg atm. atm.
xlO- 5 c.c./c.c.Pd xlO- 3
350 6-8 0-65 0-73 0-84 11-8 120
334 5-4 0-52 0-73 0-78 12-6 12-5
310 3-7 0-366 0-75 0-70 16-5 13-6
272 2-0 0193 0-74

accelerated so much that diffusion and equilibrium solubility

at the interfaces control the permeation velocity. Finally,
in support of this view Barrer (53) found that the permeation
rate through a very active palladium sample was independent
of the current density, but that as the activity decreased it
became dependent upon current density, approaching more
and more nearly to the relation
Rate = Jc^JI
found for inactive samples (Fig. 62).
Earlier attempts were made tofixthe nature of the rate-
controlling process by noting the type of kinetic expression
followed (83,84,85). Wagner (83) claimed to have found in an
H2-Pd system that various laws were valid under different
conditions:
 DETAILED MECHANISMS OF PROCESS OF FLOW 183

(ii) — — hf 2 — &! C (penetration of atoms into the

metal).
(iii) -x- "p H 2 —&2Cf2(sorption of hydrogen as molecules).
Mechanisms based upon kinetic expressions are to be
accepted with some.reserve.

© Ser
•**

6*0
1

4-0

•8 ft
Series I I 0 ^ - —
2*0

Series III

0-465 0*93 1-395

^// (/ = current in amps.)
Fig. 62. Influence of current density upon diffusion of H 2 through Pd in varying
stages of activity. Series I, most active; Series II, less active; Series III,
still less active.

THE BEHAVIOUR OF HYDROGEN ISOTOPES IN DIFFUSION

AND SOLUTION IN METALS
There are several reasons why hydrogen and deuterium should
react at different velocities:
(1) They have different masses. Kinetic theory leads one
to expect that on this account the reaction velocity constants
k
should be in the ratio 77^= V^-
D2
184 GAS FLOW THROUGH METALS

(2) They have different zero point energies. Hydrogen has

the greater zero-point energy, and so should not need so large
an additional activation energy before passing over a given
energy barrier as would deuterium. The effect of these
differences may be expressed by an exponential term e~AEIR1\
where AE is a small energy increment.
(3) According to quantum theory there exists a finite
probability that the atoms H or D may pass through an energy
barrier without having sufficient energy to surmount it. The
mass appears in a negative exponential term, and so may
result in very great velocity differences, hydrogen always
reacting more rapidly. It may here be said that hitherto no
velocity difference has been great enough to suggest that this
quantum mechanical leakage occurs (80).
In activated diffusions as in chemical reactions a component
of the system passes over an energy barrier, or succession of
energy barriers, in passing from its initial to its final state.
Thus the same phenomena which govern the one process
govern the other also. No difference in the diffusion velocities
of hydrogen and deuterium is large enough for it to be necessary
to assume a quantum mechanical leakage. The processes
involving (1) and (2) remain to be considered. In their earliest
publication on the subject, A. Farkas and L. Farkas(ST)
reported a permeation velocity ratio for hydrogen and
deuterium which could be expressed as

Subsequently, however, Jost and Widmann(88) measured the

diffusion velocity of hydrogen and deuterium inside the
palladium lattice, and found
£>302-5°

which was nearly the value ^/2 indicated by simple kinetic

theory. Further, Jouan(89) stated that the permeation rate
ratio of hydrogen and deuterium through platinum was
 BEHAVIOUR OF HYDROGEN ISOTOPES 185

constant at approximately ^2:1 = P H a : P Do from 550 to

950° C. However, a re-examination of these authors' results
shows that this is by no means so, the mean ratio at a series
of temperatures being:
T°C. 550 650 750 850 950
P P
HJ DZ 1-55 1-50 1-35 1-36 1-27

These ratios conform to the expression

PH
p —e
with some accuracy. A. Farkas(90) reinvestigated the per-
meability of palladium to hydrogen isotopes, using the half-
life of diffusion as a measure of the permeation velocity. He
also used the half-life of the process of conversion at the
surface of para-hydrogen to equilibrium hydrogen to measure
the velocity of sorption or desorption of hydrogen at the
surface. It must be remembered, however, that the solution
of the diffusion equation is in the form of an infinite series
of exponentials, and thus the "half-life period" does not
accurately measure the velocity of diffusion* and is, moreover,
a function of the thickness. Also, the conversion to the
equilibrium mixture of para-hydrogen need not occur by
sorption and desorption of an atomic layer of hydrogen, but
may result from an exchange reaction between a molecule
and an atom in the hydride layer. However, the ratios of
the half-life periods for diffusion of hydrogen and deuterium
depend on temperature, and are often too large for ex-
planations involving only a factor ^/2. These velocity ratios
depend at low temperatures in a somewhat capricious manner
upon the history of the palladium specimen.
A convincing proof that an exponential factor e~AElRT
governs the permeation rate ratio is provided by the measure-
* As it does for a first order reaction, where, since C = Coe~kt, one has

In -^ = Jet, and kt j = In 2

if t± denotes the half-life period of the reaction.

186 GAS FLOW THROUGH METALS
ments of Melville and Rideal (64) on the diffusion of the isotopes
through palladium, using tubes and disks of palladium. Their
data are presented in the following tables:

TABLE 44. Area P d tube 1-51 cm.2, thickness 0*1 mm.

Volume of system 41 c.c.

Quarter- Half-life Three-

life of of per- quarter-
T°C. Gas lifeof Ratio AE
permeation meation permeation H/D (kg.cal.)
(min.) (min.) (min.)
228 H 16-7 42-2
D 1-80 0-76
.30 79
281 H 2-55 8-0 18-4
D 12-7 290 1-58 0-68
4-40
322 H 1-40 40 8-95 1-73 0-86
D 2-52 6-6 151
362* H 0-65 1-76 3-95 1-20 0-4
D 0-80 2-30 4-95

TABLE 45. Area P d disk 0-78 cm.2, thickness 0-075 mm.

Volume of system 44-1 c.c.

Half-life of per- AE
T°C. Gas meation process Ratio
(min.) H/D (kg.cal.)

154 H 3-30 2-42 104

D 8-0
167 H 2-31 1-90 0-90
D 4-41
189 H 215 1-75
D 3-75 0-83

Farkas (90), in another set of data, gave the ratios as

T°C. 186 131 106 20
P P 1 2 4 1#36 1#4 1>84
HI D <>
which conform satisfactorily to the relationship

Data have also been obtained (04) for composite membranes

in which copper or nickel were deposited electrolytically upon
palladium. The deposited films were very thin, about 10~3 cm.
 BEHAVIOUR OF HYDROGEN ISOTOPES 187
thick, and the possibility of there bekig holes in the copper
membrane was checked by measuring the apparent activation
energy, which is different for copper and for palladium. As
a measure of the permeation velocity the time required for a
given pressure drop was used—a procedure which may give
only an approximate measure of the actual permeation rate
process. The deposition of copper films reduced the per-
meation rate by a large factor; the mean AE for the process
was, for a series of measurements involving membranes
Cu-Pd-Cu and Pd-Cu-Pd, 770 cal. Analogous measurements
for Ni-Pd membranes gave a AE of 600 cal. It is interesting
that the deposition of a copper layer 8-4 x 10~5 cm. thick, on
one side of the palladium, gave a velocity twice that observed
when»a copper film 8-4 x 10~5 cm. thick was also deposited on
the other side of the disk. Therefore the palladium did not
appreciably affect the characteristics of the copper membranes,
which alone governed the velocity of gas permeation.
Since the permeability constant depends on the diffusion
constant, upon phase boundary processes and upon the solu-
bility i P = — D —J, it is not possible to interpret these
results in any exact way. The difference in cal./mol. of the
temperature coefficients for hydrogen and deuterium may
depend on the following factors:
(i) The activation energy for adsorption and for desorption.
(ii) The activation energies for penetration from the ad-
sorbed layer into the solid, and the converse process.
(iii) The activation energy for diffusion in the metal lattice,
(iv) The heat of solution of the gas in the lattice.
Even when all phase-boundary processes occur much more
rapidly than any other processes, the temperature coefficients
of the permeability constants are governed by the terms (iii)
and (iv) above. But one notes the general correspondence
between the differences AE in the temperature coefficients of
permeability of metals to hydrogen and deuterium, and
188 GAS FLOW THROUGH METALS
between differences in the temperature coefficients of a variety
of heterogeneous reaction velocity constants (Table 46).
There is also an agreement in the order of magnitude of
the AE observed in diffusion and in heterogeneous reactions,
and the AE calculated by Sherman(95) as the difference in
zero-point energy of a large number of oscillators such as

TABLE 46. Some differences in temperature coefficients for

heterogeneous processes involving H and D

•^(apparent)
Reaction Author
cal./mol.

H 2 + 2C(8oiid)->2CH(8urfaCe) ^ 15,700 Barrer(9i)

I>2 + 2C(-olid)-^2CD(BUrface)
Tungsten
2NH3 —^ 2 -1- 3H2 42,400 800, 790, Barrer (92)
filament
Tungsten 890, 000
filament
Tungsten
2PH3 >2P(fl0lld) +3H 2
filament 32,200 510, 550 Barrer(92)
Tungsten
2PD3 ^2P(8Oiid) -f 3 H 2
filament
H 2 0 + A1 4 C 3 -»4A1(OH) 3 + 3CH 4 14,200 750 Barrer(86)
D 2 O + A1 4 C 3 ->4A1(OD) 3 + 3CD 4
C H 4 + 3C(8Oiid)->4CH(8Urface) ^26,700 780 BarrerOi)
CD4-^3C(8Olid) +4CD (8ur£a ce)
H 2 +CuO->Cu+H 2 O — 400 Melville and
D 2 + CuO->Cu+D2O Rideal(64>
Hydrogenation of styrol — 540 Cremer and
(Pd-BaSO4 catalyst) Polanyi(93)
H2, D 2 + ^O2(Ni)->H2O, D 2 0 — 750 Melville (94)
H2, D 2 + N2O(Ni)->H2O, D2O + N2 720 Melville (94)
H2, D 2 + C2H4(Ni)->C2H6, C2H4D2 — 700 Melville (94)

Ni-H, Ni-D (AE = 0-7 k.caL) or Pt-H, Pt-D (AE = 0-5k.cal.).

The calculations are only approximate; they assume that
the metal atom behaves as though it were independent
of all other lattice atoms, save only the hydrogen atom;
and that the same is true of its associated hydrogen atom.
They also assume that the oscillation is a simple harmonic
motion.
 BEHAVipUR OF HYDROGEN ISOTOPES 189

The solubility of the isotopes in palladium has been made

the subject of an experimental study by Sieverts and his co-
workers (21), and shows that in an equilibrium system there
are considerable solubility differences. Sieverts and Danz's (21)
results are shown in Fig. 40. The higher temperature results,
where Nernst's distribution law, S = k^jp, may be expected
to hold (S denotes the solubility and & is a constant), lead to
the following solubility ratios at one atmosphere pressure:
T°C. 200 220 240 260 280 300 320 340 350 400
SD/SH 0-60 0-63 0-64 0-67 0-685 0-68 0-675 0-71 0-71 0-74

The heat of solution per g.mol. of hydrogen dissolved is, in

the range 220-350° C , 2200 cal., while the heat of solution of
deuterium is 1760cal./mol., using the ratio #D/>SiH = eu°/RT
calculated from the above data. Melville and Rideal in an
analogous temperature range obtained
Ia = 2500cal./mol.,
-AHBt = 740 cal./mol.
It is then clear that interpretation of relative permeation
velocities may have to take account of this large difference in
solubility. In another study (37) the solubilities of hydrogen
and deuterium were measured in iron, giving results of which
the following are typical:
T°C. 600 700 800 950 1000 1200 1350 1450
£H 1-8 2-4 (3-2) 5-9 6-6 9 0 (10-8) 12-6
(c.c/100 g. Fe)
SD2 1-5 2-2 2-9 5-5 6 1 8-5 100 11-7
(c.c/100 g. Fe)
h2 1 0 9 1#1 107 1 08 1 0 6 1 0 8 1#08
#H 2 /£D 2 ° ° '
A theoretical study of t h e solubility of t h e isotopes in
palladium (22) led t o t h e expression

AHn - AHB =

where AH denotes the heat of absorption of gaseous hydrogen

in palladium, Xo *st n e n e a t °^ solution of a hydrogen atom
in the lattice, Xa ^ n e n e a ^ °^ dissociation of a hydrogen
190 GAS FLOW THROUGH METALS
molecule into atoms, and V(T) is the partition function of a
dissolved atom. The subscripts H and D refer to hydrogen and
deuterium respectively. The difference in the zero-point
energies of the isotopes in solution may be calculated if the
vibration frequency v is known. Assuming
vn = 8-75 x lO-^sec.- 1 ,

and that — = ITT^ , one finds

"H V ^ D '

and
ET2 ^logy^-^ = +260 cal. (T = 1200° K.).

An additional assumption in the latter calculation is that

Finally one knows, from an argument analogous to that for

computing (xf-Xo)>tnat UXd-Xd) = 900cal., and so
zl# H -ZLff D = - 4 7 0 + 900 + 260 = 690 cal.
Lacher found he could express Sieverts and Zapf 's(2i) data on
the solubility of the isotopes in palladium at high tempera-
tures as

410
t = log 0D - -^ + log 104

(p in atmospheres; 6 = fraction of saturation hydrogen

content). There is then a good agreement between the experi-
mental and calculated heat difference. Although fitting
equations of the type given above to experimental data is
prone to small errors, these are greater in the temperature-
independent than in the temperature-dependent term.
One can thus see that many factors control the relative
behaviour of the isotopes in solution, difiFusion, and permeation
in metals. The solubility ratio combines the effects indicated
above; the diffusion constant ratio involves differences in the
zero-point energies of dissolved hydrogen and deuterium
 BEHAVIOUR OF HYDROGEN ISOTOPES 191
atoms in initial and transition states respectively; and the
permeability constant ratio combines the effects of differing
diffusion constants, and differing solubilities, or, if phase-
boundary processes are also important, differences in the
relative rates of reaction such as (i) to (vi) on pp. 179 and 180.
One may express these different possibilities in a single
potential-energy distance diagram, in which a molecule is
supposed to approach a membrane, be adsorbed, penetrate
into the solid, diffuse through it, emerge into the adsorption
layer on the other side, and be desorbed into the gas phase
from this layer. The heights of the various energy barriers
may vary greatly from case to case.

THE INFLUENCE OF PHASE CHANGES UPON PERMEABILITY

It has been found that in the region of concentrations where

two hydrogen-palladium alloys coexist (the perpendicular
sections of the curves of Fig. 40) it is not possible to trace out
the same isobaric curve on sorption and desorption. There is
a hysteresis effect illustrated by Fig. 40. It is interesting to
find in the study by Lombard, Eichner and Albert (96) that the
permeability-temperature curve follows in an inverse manner
the absorption isobar, there being a great increase in the
permeability in the region 180-200° C. It may be that this
rapid alteration in permeability marks the change from
yff-phase to a-phase alloy.
The same type of hysteresis loop which was noted in the
absorption of hydrogen in palladium, where a- and /?-phases
co-exist, is observed also in the absorption of hydrogen by
iron (31), where now the two phases are produced by allotropy
in the metal. The permeability-temperature curve shows a
break at this point (Fig. 63) (51). Indeed, wherever a phase
change occurs one may look for a variation in the permeability,
so that the property of permeability may be used to determine
transition points. The change in permeability of nickel towards
hydrogen has similarly been used to characterise the Curie
point in nickel (62).
192 GAS FLOW THROUGH METALS
When the phase changes are brought about by alloying two
metals, one would anticipate that in the same way the per-
meability would vary with the composition of the alloy, and
discontinuously so wherever a new phase is formed. Evidence
upon these points is scanty, but the data of Baukloh and
Kayser(97) indicate that discontinuities do exist in the
permeability-composition curve of alloys of nickel with copper
and with iron (Fig. 64).
75
*w~5
/

10
800*C.
j
1
/
/
>
sso'c.

35 36 37 38 39 20 40 60 60
Temperature (in millivolts) Per cent Nickel
Fig. 63. Fig. 64.
Fig. 63. The permeability curve for iron.
Fig. 64. Effect of composition upon rate of H2-permeation through Ni-Cu alloy.

THE INFLUENCE OF PRE-TREATMENT UPON PERMEABILITY

A metallic crystal, like a salt crystal, normally consists of a

mosaic of small crystallites—dendritic, block-like, or columnar.
Impurities exist in part at surfaces of separation of the
crystallite components, and in part in true solution. Per-
meability is, like diiFusion, a structure sensitive property and
one is accordingly likely to find permeation anisotropy
in aleotropic single crystals; diminution in permeability with
growth of crystallites following annealing, if grain boundary
diffusion occurs; effects due to impurities in the metal; and
sensitivity to the state of the metal surface, its roughness,
degree of oxidation, or the extent of foreign metallic films.
 INFLUENCE OF PRE-TREATMENT 193
The experimental difficulty of preparing and mounting
metallic single crystals has so far prevented the discovery
of any diffusion anisotropy in metallic systems, with few
exceptions such as bismuth (Chap. VI). Effects which may be
due to any of the other possibilities are more frequently
encountered, and in a few instances have been systematically
studied. Some of this evidence may be discussed.

V
A
\
\
A\ V
\V / > s
... •M ^« AA-
'h 1 S 10 1S 20
r
Time in hours
Fig. 65. H2-permeability of nickel as a function of time of heating.

Baukloh and Kayser(97) showed that while the hydrogen

permeability of nickel was not affected by prolonged heating
at temperatures of 600° C. or even higher, at temperatures of
950-1050° C. a decrease in permeability followed. The influence
of heating upon the metal is not the same, at a given tem-
perature, for all samples, as their figures show. One such
diagram is given in Fig. 65 and illustrates a rather extreme
case, where a diminution can be observed even at 680° C. These
effects could be attributed either to movements of impurities,
or to recrystallisation with a diminution in the amount of
grain boundary diffusion. Ham and Sauter (73,74) also made the
observation that the permeability of both palladium and iron
was very much affected by the heating given the metal.
BD 13
194 GAS FLOW THROUGH METALS
One of the most complete investigations upon the influence
of heating upon permeability has been carried out by Lombard
and his co-workers (96). Their observations upon the per-
meability of hydrogen to palladium led them to the conclusion
that it was possible to reduce the permeability of palladium
100-fold or more by heating the metal. This loss in the per-
meability of pure palladium towards hydrogen, which occurred
progressively and the more rapidly the higher the temperature,
was irreversible, and could amount almost to a total loss. By
heating palladium to 500-520° C. for specified periods and
then cooling it to temperatures below 450-500° C, any steady
state of permeability could be attained. Heating the metal
in air at 500° C, and then cooling it in air and subsequently
reducing it in hydrogen at 150° C. partly if not wholly restored
the permeability; but oxidising the metal at 500° C. and
reducing it again at this temperature failed to increase its
permeability. The authors found that a given sheet of palladium
could be regenerated a number of times by oxidation and
reduction, and pointed out that this behaviour recalls the
preparation of metallic catalysts for hydrogenation, and
focuses attention upon the state of the surface of the mem-
brane. The loss of permeability was ascribed to a process of
agglomeration of fine particles at the palladium surface, with
a consequent reduction of the surface, and decrease in ease of
access of hydrogen to the lattice of the palladium. This
temperature of agglomeration was considered to be in the
vicinity of 500° C. for pure palladium, and to be retarded by
some impurities and accelerated by others. One membrane
lost its diffusing power at 315° C. Experiments by Barrer(53)
also revealed a great diminution in the permeability of
palladium due to long heating at temperatures between
270 and 360° C., the palladium at these temperatures tending to
approach a final steady state. Thus two states of permeability
may be possible for palladium, one when the metal surface is
activated by suitable oxidation and reduction, and one when
the metal has undergone the maximum crystallisation at the
surface. In the former state surface processes have their
 INFLUENCE OF PRE-TREATMENT 195
minimum effect on the diffusion velocity; in the latter they
have their maximum.
In Fig. 66 is illustrated the effect of prolonged heating upon
the permeability of palladium for a number of heating and
cooling cycles, with regeneration of the palladium between
the cycles. In the first cycle the metal was maintained at a
constant temperature, and its permeability fell steadily; in
other cycles the temperature was raised or lowered and the
J I M M M M i l I 1i M i |_

/
50
/1
\ :
:
•A / {
t
:
/
1

; }
20
I / I

w • I• —
/ . /

0
100 200 300 400 500 600 700
Temperature in °C.
Fig. 66. Effects of heat-treatment upon the permeability of palladium (96>.

permeabilities were measured simultaneously. The different

permeabilities which palladium may possess have usually
different temperature coefficients. This may be illustrated by
Table 47.
Smithells and Ransley(C7) made a study of the effects of
oxidation, reduction, polishing, and etching upon the per-
meability of nickel and iron. Polished nickel membranes were
less permeable than oxidised and reduced nickel. Etching
increased the permeability of iron more than oxidation and
reduction, but oxidation without adequate reduction of the
iron poisoned it, and rendered it impermeable to hydrogen.
Table 47a gives the data obtained. Another metal, the per-
meability of which is sensitive to surface treatments, is
aluminium, also studied by Smithells and Ransley. The
apparatus was so arranged that the aluminium could be
13-2
196 GAS FLOW THROUGH METALS
scratched with a steel brush without exposing it to air. Results
of a number of the experiments are shown in Fig. 67. The
effects are not simple; oxidation reduced the permeability;
scratching the surface appeared to reduce the permeability
when only the outside was abraded; but increased the per-
meability when the inside was abraded as well. The most

TABLE 47. The permeabilities for a number of

palladium membranes
Permeability
Sample c.c./cm.2/hr./atm. Condition of sample
QiRT
1 201ZTI e-*°* High permeability (69)
Average for a 8-3 x 10 2 c- 4 6 2 0 ^ 2 ') High permeability (96)
RT
number of samples 18T* e-™°' j
2 600/i 2
3 10-7 x 10 e-' * ' Inactive (by heating)06)
4 9-5 x 10 2 e-
10500lRT
Inactive (by heating) (53)

TABLE 47a. Effect of surface treatment upon permeability

Permeation rate
Temp. Pressure at this pressure
Metal Treatment °K. mm. c.c./sec./cm.2/mm.
thick
Ni Polished 1023 0042 l-39xlO-«
Oxidised and reduced 1023 0042 2-70 x 10-«
Ni Polished 1023 0091 2-91 x 10-«
Oxidised and reduced 1023 0091 4-23 x 10"«
Fe Polished 673 0-77 0-47 x 10-'
Etched 673 0-77 4-4 xlO" 7
Fe Polished 863 0073 1-28 x 10-'
Oxidised and reduced 863 0073 0-76 x 10- 7
at 600° C.
Oxidised and reduced 863 0073 1-54 xlO- 7
at 800° C.

important effect was the steady diminution in permeability

with the time of heating, whatever the surface condition. This
phenomenon may compare with the surface changes in
palladium (96), or the effect of heat treatment of nickel (97,98,99).
Sometimes the chemical treatment of a metal with a gas may
increase its permeability to a second gas; for example, Ham
 INFLUENCE OF PRE-TREATMENT 197
and Sauter noted that heating iron in nitrogen increased the
velocity of permeation of hydrogen by 10 to 15-fold, although
baking out reduced the permeability to its former value.

I Z 3 i 5 6 7
Hours
Fig. 67. Effect of surface treatment on the permeability of
aluminium to H2 at 580° G.
Run 1. Al exposed to air.
Run 2. Outer surface scratched.
Run 3. Outer surface scratched again.
Run 4. Outer surface scratched again.
Run 5. Inner and outer surfaces scratched.
Run 6. Outer surface anodically oxidised.

GRAIN-BOUNDARY AND LATTICE PERMEATION

Photo-micrographs may illustrate both the mosaic structure

of metals, and a concentration of impurity between crystal
blocks. If diffusion occurs mainly down these boundaries, the
permeability of a metal would be governed by their number
and nature. The mere fact that oxide is found concentrated
in these boundaries does not, however, necessarily imply that
diffusion of oxygen occurs along them, since the oxide may
be formed within the lattice and then be thrown out of solution
in the zones between crystallites. That grain-boundary
diffusion plays an important part in diffusion processes such
198 GAS FLOW THROUGH METALS

as that of thorium in tungsten, of ions in microcrystalline

sodium chloride, or of gases through silica glass is undoubted,
and these and other examples have been considered elsewhere
(Chaps. VII andlll). It is not easy, however, to establish similar
cases when gases diffuse through metals. I t is true that great
variations in permeability may be encountered with different
samples of metal (e.g. H 2 -Pd systems), but these differences
may also be attributed to the variable influence of phase-
boundary processes. The evidence available casts doubt upon
the special importance of grain boundary diffusion for
hydrogen-nickel (97) and hydrogen-iron systems (ioo) at high
temperatures. In both cases it was shown that crystal size had
no marked effect upon the permeability. Edwards(55) found
the same permeability towards hydrogen in a single crystal
plate of iron before heating it, and after heating it to refine
the grain. Smithells and Ransley(ic) observed that a single
crystal iron tube had the same permeability as a similar tube
with 100 grains/mm.2 (Table 48). Grain boundary diffusion is

TABLE 48. Permeability to hydrogen of iron of

different grain sizes
Permeation rate
c.c./sec./cm.2/nim.
Temp. Pressure
°C. rnm. Fine grain Single crystal
xlO~ 6 x 10-«
245 140 2-4 1-2
413 140 17-6 171
621 140 92-8 89-5
779 140 2030 2050

more easily observed the lower the temperature, and thus

experiments of this type should be conducted at the lowest
possible temperature.
The permeability of hydrogen-iron systems has received
considerable attention at low as well as at high tempera-
tures. Ham and Rast(ioo) studied the permeability from 55 to
920° C. In addition to discontinuities in permeability at every
 GRAIN-BOUNDARY AND LATTICE PERMEATION 199

thermal critical point, they observed a hysteresis loop in the

permeability-temperature curve illustrated in Fig. 68. At
room temperatures, when the diffusing atoms were supplied by
electrolysis, Barrer(53) found that a bright steel tube was at
first impermeable, but quickly became more and more per-
meable to hydrogen in successive diffusion experiments.
Poulter and Uffelman(ioi) found that hydrogen diffused

-2-0

Fig. 68. The low-temperature permeability of hydrogen to iron(ioi),

showing a typical hysteresis loop.

through iron at room temperature, but not until a pressure of

4000 atm. was applied. When the diffusion had thus been
started the steel was permeable in a second run at only 100 atm.
Some if not all of these observations point to an opening up of
grain boundaries as diffusion proceeds. The grain structure
can develop due to desorption between grains, of hydrogen
which has diffused in the lattice. Thus the system may
commence as one where lattice diffusion predominates (p. 198
and Table 48), and end as one where grain boundary diffusion
o minates.
200 GAS FLOW THROUGH METALS
Not only may diffusion occur through the lattice, and
probably ib. certain conditions down grain boundaries, but
also in each crystallite preferred directions of diffusion may
be encountered. Smith and Derge( 102,103) prepared palladium
specimens of various grain sizes and showed that grain size
did not affect the sorptive capacity. They considered, from
crystallographic evidence, and from the marked influence of
deformation upon the velocity of uptake, that slip planes in
each crystal grain are of special importance in absorption.
Etching rolled foils previously charged with hydrogen revealed
that fissures had developed inclined at 45° to the direction of
rolling. It was thought that the fissures were caused by
preferential penetration of hydrogen along slip planes which
were dilated as hydrogen diffused into the lattice.
Much more evidence of this kind is necessary before any
generalisation concerning the relative importance of grain-
boundary, slip-plane, or isotropic lattice diffusions may
be made.

FLOW OF NASCENT HYDROGEN THROUGH METALS

Atoms of hydrogen may be generated by electrolysis or by

chemical reaction at the metal surface. Some of the atoms of
hydrogen thus liberated may leave the surface as molecules
and others may enter the metal and diffuse through it! The
possible sequence of processes is

Except for the processes which liberate the atomic hydrogen,

the reactions are the same as those occurring during thermal
diffusion. Morris (104) showed in agreement with the scheme
above that some hydrogen gas, generated by the action of
citric acid on iron, diffused through it, and some was evolved
(Fig. 69). The fraction diffusing was greater the more rapid
the rate of corrosion of the steel (the acid strength being
constant).
 NASCENT HYDROGEN THROUGH METALS 201
The possible variables in this type of diffusion include
current density, temperature, the dimensions of the mem-
branes, concentration of acid, time, and the concentration of
added salts. Bodenstein(ii) considered that his results on the
diffusion of hydrogen through iron obeyed the relation
Permeation velocity =

1*
r

0? 02
VT (i in amps per cm.2)
Fig. 69. Fig. 70.
Fig. 69. Volumes of hydrogen simultaneously evolved and diffused by the
action of citric acid on fast corroding steel.
Fig. 70. The electrolytic diffusion of hydrogen through iron as a function of
current density. Fe, 0-18% C. x = untreated. O = annealed.

Borelius and Lindblom(52), however, showed that over a

considerable range of current densities the relationship was
Permeation velocity = k(^I — *JI(),
where ^jli is a constant, called the threshold current density
(Fig. 70). The relation is analogous to their equation for the
thermal permeation rate (p. 171):
Permeation velocity = k^p — y/pt).
The explanation that the rate of permeation can be expressed
as P = kd*JI and that the surface is not fully saturated at
low-current densities is not more tenable than Smithells and
Ransley's earlier explanation of the same facts for thermal
diffusion (p. 171). The most satisfactory explanation is that
advanced by Barrer (p. 174).
202 GAS FLOW THROUGH METALS
The influence of temperature upon the permeation velocity
of hydrogen through iron and through palladium (53) shows
that the rate of flow rises exponentially with the temperature.
The slope of the curves log (permeability) against 1/T(T = °K.)
for the H2-Fe system (Fig. 71) is the same at all current
densities from 0• 007 5 to 0• 045 amp./cm.2, and is very nearly that
found for the thermal permeability (EelecU = 9400 cal./atom,

100°

Fig. 71. Electrolytic diffusion of hydrogen through iron as a function of tem-

perature. Curves for current densities: 0-0075 arap./cm.2, 0*015 amp./cm.2,
0030 amp./cm.2 and 0045 amp./cm.2.

compared with -^thermal = 9100 cal./atom). This relationship

and the similar dependence of the permeability upon current
density and gas pressure (p. 178) led Borelius and Lindblom (52)
to assume that all rate-controlling processes were the same for
thermal diffusion and diffusion of nascent hydrogen.
In Edwards' (55) experiments, the gas was generated by the
action of hydrochloric and sulphuric acids upon the iron
membrane. In the steady state the permeation velocities
 NASCENT HYDROGEN THROUGH METALS 203
gave a linear logP — ijT curve whose slope gives E = 7300
cal./atom for the data using iVH2SO4. Barrer's(53) data for
the H2-Pd system also gave linear logP— \jT curves, and
the slopes corresponded to E — 8500 and 9200 cal./atom.
As Edwards' and Barrer's data show, the steady state of
flow through the membrane is established slowly (Fig. 72), and
the interval required depends strongly upon the temperature.
During this interval the metal is absorbing a quantity of gas
and setting up its steady state concentration gradient. It is

/ / /

r /i°c.
/7r°c.

1-0
/
/
/

^2^-0—
300 600

Time in minutes
Fig. 72. Time lag in establishing steady state of flow of H 2 through
a steel cathode (53).

connected therefore with the velocity of diffusion of the gas

within the bulk metal, and the slower this diffusion the more
slowly is the steady state approached.
Other studies of the diffusion of nascent hydrogen have given
empirical relations between the rate of passage through the
metal and the concentration of the acid generating the gas at
the surface of the metal (55); and for the influence of various
capillary active substances (105) upon the velocity of per-
meation. For example, the addition of mercuric chloride
caused an acceleration in the rate of permeation. This accelera-
204 GAS FLOW THROUGH METALS
tion in the passage of hydrogen results in a more rapid decrease
in the mechanical strength of the iron(i3) which is a normal
consequence of "pickling 15 the metal in acid.

REFERENCES
(1) Graham, T. Phil. Mag. (4), 32, 401 (1866); Ann. Phys., Lpz.,
129, 549 (1866); Chem. Zbl. 38, 129 (1867).
(2) Cailletet, L. CM. Acad. Sci., Paris, 58, 327 and 1057 (1864).
(3) Deville,H. andTroost, L. C J ^ c a d . £c*\, Pans, 56, 977 (1863).
(4) Troost, L. CM. Acad. Sci., Paris, 98, 1427 (1884).
(5) Reynolds, 0. Proc. Manchr lit. phil. Soc. 13, 93 (1874).
(6) Ramsay, W. Phil. Mag. 38, 206 (1894).
(7) Bellati and Lussana. Atti. 1st. veneto, 1, 1173 (1890).
(8) Winkelmann, A. Ann. Phys., Lpz., 6, 104 (1901); 17, 591
(1905); 19, 1045 (1906).
(9) Richardson, O. Phil Mag. (6), 7, 266 (1904).
Richardson, O., Nicol, J. and Parnell, T. Phil. Mag. (6), 8, 1
(1904).
(10) Charpy, G. and Bonnerot, S. CM. Acad. Sci., Paris, 154, 592
(1912).
(11) Bodenstein, M. Z. Elektrochem. 28, 517 (1922).
(12) Nernst, W. and Leasing, A. Nachr. Ges. Wiss. Gottingen, 146
(1902).
(13) Alexejew, D. and Polukarew, O. Z. Elektrochem. 32, 248 (1926).
(14) Aten, A. and Zieren, M. Rec. Trav. chim. Pays-Bas, 49, 641
(1930).
(15) SeeMcBain, J. W. Sorptionof Gases by Solids,Routledge (1932).
(16) E.g. Zr-H 2 , Hagg, G. Z. phys. Chem. 11, 433 (1931).
(17) E.g. Pd-H 2 , Linde, J. and Borelius, G. Ann. Phys., Lpz., 84, 747
(1927).
Krtiger, F. and Gehm, G. Ann. Phys., Lpz., 16, 174 (1933).
(18) A full list of references to work of Sieverts and his co-workers
up to 1930 is given in J. W. McBain, Sorption of Gases by
Solids (1932). Other later references will be given subsequently.
(19) Smithells, C. Gases and Metals, Chapman and Hall (1937).
(20) Emeleus, H. and Anderson, J. Modern Aspects of Inorganic
Chemistry, Chap. 13, Routledge (1938).
(21) Sieverts, A. and Zapf, G. Z. phys. Chem. 174A, 359 (1935).
Sieverts, A. and Danz, W. Z. phys. Chem. 34 B, 158 (1936).
(22) Lacher, J. Proc. Roy. Soc. 161 A, 525 (1937).
(23) Smithells, C. Gases and Metals, pp. 153 et seq. (1937).
(24)- Smithells, C. and Fowler, R. H. Proc. Roy. Soc. 160 A, 38 (1937).
(25) Sieverts, A. and Hagenacker, J. Z.phys. Chem. 68, 115 (1910).
(26) Steacie, E. and Johnson, F. Proc. Roy. Soc. 112A, 542 (1926).
(27) Simons, J. H. J. phys. Chem. 36, 652 (1933).
 REFERENCES 205
(28) Mathewson, C, Spire, E. and Milligan, W. J. Amer. Steel Treat.
Soc. 19, 66 (1931).
(29) Smithells, C. Gases and Metals, Fig. 124.
(30) Rhines, F. and Mathewson, C. Trans. Amer. Inst. min. (metall.)
Engrs, 111, 337 (1934).
(31) Smithells, C. Gases and Metals, Fig. 123.
(32) Seybold, A. and Mathewson, C. Tech. Publ. Amer. Inst. Min.
Engrs, 642 (1935).
(33) Merica, P. and Waltenburg, R. Bur. Stand. Sci. Pap. 281 (1925).
(34) de Boer, J. H. and Fast, J. Rec. Trav. chim. Pays-Bas, 55, 459
(1936).
(35) Sieverts, A. and Briining, K. Arch. Eisenhiittenw. 7, 641 (1933).
(36) Hagg, G. Z. phys. Chem. 7B, 339 (1930).
(37) Sieverts, A., Zapf, G. and Moritz, H. Z. phys. Chem. 183 A, 33
(1938).
Earlier references are given in Smithells, C, Gases and Metals,
p. 171.
(38) Smithells, C. Gases and Metals, Fig. 119.
(39) Hagg, G. Z.phys. Chem. 4B, 346 (1929).
(40) Rontgen, P. and Braun, H. Metallwirtschaft, 11, 459 (1932).
(41) Sieverts, A. and Krurnbharr, W. Z. phys. Chem. 74, 295 (1910).
(42) Smithells, C. Gases and Metals, Fig. 107.
(43) Kirschfeld, L. and Sieverts, A. Z. Elektrochem. 36, 123 (1930).
(44) Sieverts, A. Z. Metallkunde, 21, 37 (1929).
(45) Toole, F. and Johnson, F. J. phys. Chem. 37, 331 (1933).
(46) Sieverts, A. and Briining, K. Z. phys. Chem. 168, 411 (1934).
(47) Sieverts, A. and Hagen, H. Z. phys. Chem. 174, 247 (1935).
(48) Cf. ref. (35). See also Takei, T. and Murakami, T. Sci. Rep.
Tohoku Univ. 18, 135 (1929).
(49) Smithells, C. Gases and Metals, p. 186, Fig. 132.
(50) Ham, W. J. chem. Phys. 1, 476 (1933).
(51) Post, C. and Ham, W. J. chem. Phys. 5, 915 (1937).
(52) Borelius, G. and Lindblom, S. Ann. Phys., Lpz., 82, 201 (1927).
(53) Barrer, R. Trans. Faraday Soc. 36, 1235 (1940).
(54) Smithells, C. Gases and Metals, p. 84 (1937).
(55) Edwards, C. J. Iron and Steel Inst. 110, 9 (1924).
(56) Johnson, F. and Larose, P. J. Amer. Chem. Soc. 46, 1377 (1924).
(57) Lombard, V. C.R. Acad. Sci., Paris, 177, 116 (1923).
(58) Lombard, V., Eichner, C. and Albert, M. Bull. Soc. chim. Paris,
4, 1276 (1937).
(59) Bramley, A. Carnegie Schol. Mem., Iron and Steel Inst., 15, 155
(1926).
(60) Smithells, C. andRansley,C. E. Proc.Roy.Soc. 150A, 172 (1935).
(61) Post, C , and Ham, W. J. chem. Phys. 6, 599 (1938).
(62) Deming, H. and Hendricks, B. J. Amer. chem. Soc. 45, 2857
(1923).
(63) Jouan, R. J . Phys. Radium, 7, 101 (1936).
206 GAS FLOW THROUGH METALS
(64) Melville, H. and Rideal, E. Proc. Roy. Soc. 153A, 89 (1936).
(65) Braaten, E. and Clark, G. Proc. Roy. Soc. 153 A, 504 (1936).
(66) Ryder. Elect. {Cl.) J. 17, 161 (1920)
(67) Smithells, C. and Ransley, C. E. Proc. Roy. Soc. 152A (1935).
(68) Spencer, L. J. chem. Soc. 123, 2124 (1923).
(69) Lombard, V. and Eichner, C. Bull. Soc. chim. Fr. 53, 1176
(1933).
(70) Schmidt, G. Ann. Phys., Lpz., 13, 747 (1904).
(71) Holt, A. Proc. Roy. Soc. 91 A, 148 (1915).
(72) Lombard, V. and Eichner, C. Bull. Soc. chim. Fr. 51, 1462
(1932).
(73) Ham, W. and Sauter, J . D. Phys. Rev. 47, 337 (1935).
(74) Phys. Rev. 47, 645 (1935).
(75) Wustner, H. Ann. Phys., Lpz., 46, 1095 (1915).
(76) Smithells, C. and Ransley, C.E. Proc. Roy.Soc. 157 A, 292(1936).
(77) Ham, W. Phys. Rev. 47, 645 (1935).
(78) Smithells, C. Gases and Metals, p . 90 (1937).
(79) Barrer, R. M. Trans. Faraday Soc. 35, 628, 644 (1939).
(80) Wang, J . S. Proc. Camb. phil. Soc. 32, 657 (1936).
{81) Barrer, R. M. Phil. Mag. 28, 353 (1939).
(82) Phil. Mag. 28, 148 (1939).
(83) Wagner, C. Z. phys. Chem. 159 A, 459 (1932).
(84) Engelhardt, G. and Wagner, C. Z. phys. Chem. 18 B, 369 (1932).
(85) Doehlemann, E. Z. Elektrochem. 42, 561 (1936).
(86) Barrer, R. M. Trans. Faraday Soc. 32, 486 (1936).
(87) Farkas, A. and Farkas, L. Proc. Roy. Soc. 144A, 467 (1934).
(88) Jost, W. and Widmann, A. Z. phys. Chem. 2 9 B , 247 (1935).
(89) Jouan, R. J. Phys. Radium, 7, 101 (1936). See Fig. 4.
(90) Farkas, A. Trans. Faraday Soc. 32, 1667 (1936).
(91) Barrer, R. M. Trans. Faraday Soc. 32, 482 (1936).
<92) Trans. Faraday Soc. 32, 490 (1936).
(93) Cremer, E. and Polanyi, M. Z. phys. Chem. 19 B, 443 (1932).
(94) Melville, H. J. chem. Soc. 1243 (1934).
(95) Eyring, H. and Sherman, J. J. chem. Phys. 1, 345 (1933).
(96) Lombard, V., Eichner, C. and Albert, M, Bull. Soc. chim. Fr.
3, 2203 (1936).
(97). Baukloh, W. and Kayser, H. Z. Metallic. 26,157 (1934); 27, 281
(1935).
(98) Lewkonja, G. and Baukloh, W. Z. Metallic. 25, 309 (1933).
(99) Baukioh, W. and Guthmann, H. Z. Metallic. 28, 34 (1936).
(100) Ham, W. and Rast, W. Trans. Amer. Soc. Metals,26,885(1938).
(101) Poulter, T. and Uffelman, L. Physics, 3, 147 (1932).
(102) Smith, D. and Derge, G. Trans. Amer. electrochem. Soc. 66, 253
(1934).
(103) J. Amer. chem. Soc. 56, 2513 (1934).
(104) Morris, T. J. Soc. Chem. Ind. 54, 7 (1935).
{105) Aten, A. and Zieren,M. Rec. Trav.chim.Pays~Bas,49,641 (1930).
 CHAPTER V

D I F F U S I O N OF GASES AND NON-METALS

I N METALS

INTRODUCTION
In the previous chapter were considered the main phenomena
of gas flow through metals. It was shown there that the gas
flow (P) is governed by the equation

P--Dd-°
where D is the diffusion constant, but so far we have had little
to say concerning the important constant D. It has, however,
been stated (Chap. IV) that the diffusion constant obeys an
exponential law D = D e~EIRT
analogous to the expression
p = p0e-*y*21
for the permeability constant. It is not difficult to see why
an energy of activation should be involved in diffusion pro-
cesses in solids. The solid lattice contains atoms distributed
in a periodic manner, to give the regular crystalline array.
Just as when the solubility of gases in metals was discussed
(Chap. IV, p. 153), one may regard the interstitial positions in
the lattice as positions of minimum potential energy—potential
"holes"—separated by energy barriers. The diffusing atom
requires an activation energy before it may pass from one
minimum of energy to another, exactly as does an atom under-
going a chemical reaction. In this way an exponential term
e-EiRT i s introduced. While the value of E in a diffusion process
is a relatively easily interpreted quantity, the Ex in the
expression for the permeability is a complex quantity, which
may be made up of E for diffusion processes, and phase-
boundary processes, and the heat of solution (AH) of the gas
in the metal. Thus the energy of activation for diffusion can
208 DIFFUSION OF GASES AND NON-METALS

be related to the crystal structure and force fields, and the

dimensions of the diffusing particle, but the temperature
coefficient for the permeability is not easily related to these
quantities.
When one examines the literature one finds that relatively
few measurements have been made of the diffusion constants
of gases in solids. One has to be sure that phase-boundary
processes do not control the velocities of absorption, but this
is not always easy to establish. Again, the mathematical
analysis of the data often presents difficulties, and very
frequently a number of phases co-exist which render inter-
pretation even more uncertain. One has sometimes to deal
with the case when the diffusion constant is a function of the
concentration, necessitating the use of Fick's law in the form

dt dx\" dxj
However, the mathematical handling of the data in this instance
is now possible (Chap. I), and the formal treatments of Fick's
law for the great proportion of the possibilities which arise
have been completed.

THE MEASUREMENT OF DIFFUSION CONSTANTS IN METALS

To measure the concentration gradient in a metal, one may

conveniently employ the method of Bramley (1,2,3*4,5) and his
co-workers. They heated iron in a suitable gas atmosphere of
which the following are examples.
CO and CO2 (diffusion of oxygen or carbon),
CO and CH3CN (diffusion of nitrogen or carbon),
CO and \ _ _ / N (diffusion of nitrogen or carbon),
CO and < )>CH3
<^ y (diffusion of carbon),

CO and paraffins (diffusion of carbon),

 MEASUREMENT OF DIFFUSION CONSTANTS 209

CO and NH 3 (diffusion of carbon or nitrogen),

N2 and NH 3 (diffusion of nitrogen),
H 2 and PH 3 (diffusion of phosphorus),
CS2 and \^ / C H 3 (diffusion of carbon and sulphur).
After the metal in the form of bars had been heated to a
definite temperature for a suitable period it was removed, and
successive thin layers taken off in the lathe and analysed.
In this way the concentration gradient was determined and
the solution of Fick's law in the form

was used to evaluate D. Here Go is the concentration at x = 0,

supposed constant and conditioned by the gas atmosphere.
Bramley and Lord (4) showed that bars of the same steel heated
for the same time and at the same temperature but in very
different atmospheres gave the same distribution of carbon
in all cases.
The concentration-distance curves into the metal are,
however, not always of a shape to which a simple solution of
the Fick law is applicable. Figs. 73 and 74 give different types
of concentration-distance curve, for which the solution given
above is manifestly not correct. In the former case a saturated
carbon layer has established itself at the surface, and in the
latter the surface has been to some extent decarburised by
hydrogen present in the carburising atmosphere. An attempt
has been made to treat diffusion problems of the type illus-
trated by Fig. 73(6), but a complete treatment of Fig. 74 is
not available. However, Bramley and his co-workers neglected
the initial parts of these unusual forms of concentration
distance curve, and applied the simple solution of Fick's law
only to the tail of the curves. The error involved in evaluating
D is therefore smaller.
In addition to this analytical method of determining the
concentration gradient it should in some cases be possible to
BD 14
210 DIFFUSION OF GASES AND NON-METALS
use an X-ray method. This would reveal the presence of
embedded crystals of carbide, nitride, or other compounds
of the diffusing element and the metal, and the intensity of the
typical pattern may be used as a measure of concentration.
The method might be used where the lattice is expanded by
its content of dissolved gas, for the expansion is often a
1-2

1-0

0-8 \

\
V
v
V
DEPTH IN MM.

Fig. 73. Types of concentration gradient.

function of the charge of gas. The X-ray method has been

employed for following the inter-diffusion of metals; it has
also been used in studies (7,8,9,10) of the hydrogen-palladium
system in which the lattice constants both for the a- and
/?-phases increase with the concentration of hydrogen in the
lattice (Fig. 75). However, while the method is frequently
available in investigating equilibrium (ii, 12,13) systems, it is
not so often employed in measuring concentration gradients.
When one does not require the concentration gradient, but
simply the total quantity absorbed or desorbed, one may use
MEASUREMENT OF DIFFUSION (CONSTANTS 211

10 v
0-9
\
0-8 \V
2°' 7
K0-6
I.
\
oO-5
X

O0'4
V <

0-3

0-2

0*1
\ • * ^ ^

0 0-5 1-0 ,1-5 2-0 2-5 30 3-5 4-0 45 50

DEPTH M M.

Fig. 74. Types of concentration gradient.

3-92

3-87

Fig. 75. Lattice constants in H2-Pd alloys as a function of

14-2
212 D I F F U S I O N OF G A S E S A N D N O N - M E T A L S

the metal in the form of a wire, and measure its electrical

resistance. This method has been used for carbon-oxygen
(carbon at very high temperatures is believed to dissolve
oxygen(i8)), and for hydrogen-palladium(14,15). The assumption
is usually made that the change of resistance is a linear function
of the charge of hydrogen. Similarly Coehn and Jurgens(i6)
followed the absorption of hydrogen by a large number of
palladium-silver alloys by measuring their electrical resistance.
It is probable that the absorption of gases by any other metal
or oxide which reacts with or dissolves the gas appreciably
could be followed by electrical resistance changes. As suitable
systems one may give
H2-Ta* 02-Cu20(i7)
H2-V O2-NiO(i8)
H2-Ti O2-Zr(i9)
H2-Ce, La, and certain rare earths N2-Zr(i9)
H2-Th
N2-Fe
N2-Mo
N2-Mn
The E.M.F. produced by a palladium wire charged with
hydrogen and immersed in a solution containing hydrogen
ions is different from the E.M.F. given by uncharged palladium,
and Coehn and Specht(2i) measured the diffusion of hydrogen
along a palladium wire whose centre was electrolytically
charged with hydrogen, by measuring the potential at other
points along the wire as a function of time. The speed of
movement of a point of constant potential was related to the
time by an expression

and from curves such as thosQ in Fig. 76 the authors were

able to measure D. The numbers of the curves relate to the
point on the wire at which the potential was measured.
* Sieverts and Briining(20) have shown that the resistivity rises in direct
proportion to the. quantity of hydrogen absorbed, being 30 % greater when the
metal is saturated.
 MEASUREMENT OF DIFFUSION CONSTANTS 213

Another method for following the diffusion of hydrogen

along filaments was devised by Coehn and Sperling (22). The
centre of a palladium wire was charged with hydrogen gas,
and the wire clamped under a photographic plate. The
hydrogen liberated at the surface of the wire blackened the
plate. As before, the movement of the black strip, by means

-02
¥0 SO tzo 160 200 2*0 280 320
Time in hours
Fig. 76. Flow of hydrogen in a palladium wire followed by measuring
the change of potential with time.

of the relationship x2 = 4tDt (x denotes a point of a given

darkness on the plate), was used to calculate D.
The thermo-electric properties of metals are also altered
when they absorb gases. If a thermocouple is made between
a pure metal and a hydrogen-saturated metal, the pure metal
is electronegative, and the thermal E.M.F. per 1° C. is(23)
Fe 0-705 xlO- 7 V.,
Pt 0-223 xlO~ 7 V.,
Ni 1-08 xlO- 7 V.,
Pd 174-5 xlO"7V.
As the capacity of the metal for dissolving gas increases, so
does the thermal E.M.F. The method ought therefore to be
214 DIFFUSION OF GASES AND NON-METALS

applicable to the solution of hydrogen in tantalum, titanium,

cerium, thorium and other metals which dissolve hydrogen in
large quantities.
Euringer (24), who measured the evolution of hydrogen from
nickel wires, gave a method for measuring both the solubility
and the diffusion constant. The method, which should be of
general applicability, was based upon the following solution pf
Fick's law (Chap. I):

where a n is the nth root of the Bessel function of zero order,

and Co is the initial uniform concentration in a wire of radius
r0. The rate of evolution, P, of gas from the wire is

since

The rate of evolution of gas therefore depends upon a constant

2C0D/r and an infinite series of exponentials. The shape of the
curve of log P against log t does not therefore change through
alterations in Co or D, save to undergo bodily displacement.
Thus an experimental curve can be compared, for a given Co
and D with a calculated one, upon a log-log scale of axes.
When the curves are correctly ascertained, the theoretical
curve may be made to coincide with the experimental one,
by suitable horizontal and vertical displacements. D and Co
are then observed as follows:
Use the suffix "e" to denote experimental, and "c" to
denote values assumed in calculation. Take a point (tc, Pc) on
the log-log curve which after displacement coincides with a
point (te,Pe). For this point the products Dt must be equal,
00
and so the series 2 erD(X^1 are equal. Thus Dctc = D€te, and
n=l
 MEASUREMENT OF .DIFFUSION CONSTANTS 215

De is calculated. Similarly, if the exponential series are equal,

one must have

or
C0De

and so one may calculate COe, the experimental value of Co.

One might also find the slope and intercept of the asymptotic
curve
• , 2C0D Da\t
l o g £ ^

t*

04

0-2
0-11

0-5 r 113 IS 2
Fig. 77.

valid for large values of t, but Euringer pointed out that then
changes in P may not be easy to evaluate, since they have
become small. Hevtherefore advocated the method already
indicated.
Van Liempt(25) outlined a method of obtaining D approxi-
mately from the evolution of gases from wires and plates. The
Fick law for diffusion into a semi-infinite sheet is
n 2 f*/2VtfH>
77-=1—r- e~w2dw.

If one plots C/Co against x/2 <J(Dt), the typical curve of Fig. 77
is obtained. The curve encloses the same area CJ^Jn as the
216 DIFFUSION OF GASES AND NON-METALS

triangle, of which the hypotenuse cuts the x/2<J(Dt) axis at

the point 2/^/TT = 1-13, where CjC0 = 0-11. This point of inter-
section was called the "apparent penetration depth", a, or
"apparent outgassing depth" if one is considering desorption.
When the real curve is replaced by the straight line in this way,
one can find simple formulae for the diffusion constant D in
terms of plate thickness, d, and breadth B(B$>d) (Fig. 78).
If Q = gas evolved/cm, length of the plate being degassed,

a
\

Fig. 78.

one has so long as a^ \d or QJQQ^Q- : Q = 2[|aC 0 ]. The

original quantity absorbed was
Q0 = dC0B,

so that ~FT — ~3 —

Other cases were given which are listed below:

(i) Degassing a plate, where a > \d or QIQ0^ 0-50:
77 \d
 MEASUREMENT OF DIFFUSION CONSTANTS 217
(ii) Degassing a wire of radius r0, where QjQ0< 0-67:

D=
S2t 1-
(iii) Degassing a wire of radius r0, where Q/Qo > 0*67:

It must be remembered that the values of D calculated by

these formulae were not exact, although all the values of D
were of the correct magnitudes. Table 49 illustrates the data
obtained by outgassing a nickel sheet 3 cm. wide and 0* 15 mm.
thick. The composition of the evolved gases was not given.

TABLE 49. Diffusion constants by van Liempt's method

Temp. Time D Mean value of D
°C. min. QIQ* cm.2 sec."1 cm.2 sec." 1
700 2 0-33 4 x 10-8
5 0-43 1-8 x 10-8 2-5 x lO- 8
10 0-50 1-8 x 10-8
800 1 0-41 12 xlO-8
2 0-52 9-9 x 10-8 10 x lO- 8
10 0-70 9-4 x 10-8
20 0-85 8-5x10-8
900 2 0-67 2 1 x lO" 7 2-5 x 10- 7
4 0-8C 2-9 xlO" 7

Another method, which depends upon the time-lag in setting

up the stationary state of flow through a membrane, is
applicable to gas flow in metals. The first treatment was given
by Daynes(26), and later extended by Barrer(27,28,29). The
initial concentration in the plate is O0, and at the ingoing and
outgoing surfaces, C± and C2 respectively. Then if one plots
the pressure on the high-vacuum side against the time, the
curve takes the course shown in Fig. 79, and tends to the steady
state asymptotically. In the simplest case the asymptote cuts
the time-axis at the point

L = Z: 3 2_T
 218 DIFFUSION OF GASES AND NON-METALS

where d is the membrane thickness. If C2 = 0, Co = 0,

and D may readily be calculated. This formula is applicable

only if interface reactions are fast compared with the diffusion
process (28.29). When the rates of the processes

0-020

.5
0-010

25 50 75
Time (min.)

Fig. 79. Time-lag in setting up the steady state of flow of hydrogen

through a palladium tube (29).

(i) passage of adsorbed atom or molecule into the metal

(velocity constant kx)9
(ii) emergence of dissolved atom or molecule into the surface
(velocity constant Jc2)
are much smaller than D, one can see that the values of L
are no longer governed by D. The value of L is a complex
function of kx and k2 where the process of permeation is
controlled solely by these two phase-boundary reactions.
 MEASUREMENT OF DIFFUSION CONSTANTS 219

When kl9 k2 are much greater than Z>, the expression for L
reduces to

Thus the expression L = d2/6D gives a value of D which is

more nearly equal to the real value the more the phase-
boundary processes are accelerated. The method can be
applied equally to the diffusion of gases through hot metals,
or to the diffusion of nascent gases liberated at room tem-
peratures by electrolysis or by chemical reaction at the metal
surface.
This review of the possible ways of measuring diffusion
constants in metals indicates the lines of approach which have
been developed. Future work should aim at the establishment
of the relative importance of phase-boundary reactions and
volume diffusion; the estimation of the influence of concen-
tration upon the diffusion "constant"; the understanding
of the role of impurities and mechanical treatment; and
measurements of the influence of temperature and other
variables upon the diffusion constants. In this way the in-
vestigations will ultimately contribute to the problems of
mobility and reaction in solids. Unfortunately there is so
far a great scarcity of data.

A COMPARISON OF METHODS OF MEASURING

DIFFUSION CONSTANTS

It is perhaps unfortunate that the only system for which the

diffusion constant has been measured by a great many different
methods is the hydrogen-palladium system, for palladium
membranes have shown very capricious permeabilities
(ChapIV, p. 194), and this variability seems to extend also to
diffusion constants. The permeability to hydrogen of nickel
membranes, on the other hand, has shown itself to be more
constant, as the data of Table 42, p. 168, indicate, so that
hydrogen-nickel systems might prove a better testing ground.
220 DIFFUSION OF GASES AND NON-METALS

The published data for the diffusion constants, in the fora,

of log D against l/T curves, are shown in Fig. 80. These data
are by no means concordant, and their divergences from one
another may be explained in part by the difficulty of the
technique employed. In part, however, differences in the
properties of the palladium may be responsible, for, as Tam-
mann and Schneider showed, the rate of absorption of hydrogen
depends on the state of the metal, e.g. whether it is soft or
tempered. Also one may anticipate effects due to impurities,
and strongly developed grain boundaries. Duhm's(32) result
(9-5 x 10~5 cm.2/sec. at room temperature) is so far away from
all other published data that it has not been given in Fig. 80.
The equations of the different curves in Fig. 80 are given in
Table 50. The most unlikely data are in square brackets.*
By drawing a straight line through the various points in
Fig. 80 one may obtain a mean diffusion equation for all the
different palladium samples and methods.
It is known that when a potential difference is maintained
between the ends of a palladium wire charged with hydrogen,
the hydrogen moves towards the negative end (21,32). I t
therefore carries a positive charge, and may move as protons.
The protonic nature of the hydrogen is, however, by no means
certain, for the effective charge appears to be only a very small
fraction of that of a free proton. Coehn and his co-workers, and
also Duhm, measured the mobility of the hydrogen under an
applied potential, by measuring the E.M.F. or resistance of
sections of the wire after intervals of passing a current. The tem-
perature coefficient of the mobility, U, obeyed the expression
U = Uoe~E'RT,

where E was 6*3 and 4-9 k.cal. respectively by the two methods.
* These figures are considered doubtful not because of the value of E but of
Do. The value of Do is such that a different mechanism of diffusion would have to
be postulated in the transition state, involving a large disturbance in the
surrounding palladium lattice, and thus a large entropy of activation, or la'rge
Do. If the diffusing particle is a proton this disturbance is most unlikely.
 MEASUREMENT OF DIFFUSION CONSTANTS 221

d -5-0

.s
X.
\ V
s

-7-0

0'0016 0-0021 0-0026

(Tin°K.)
0-003T 0-0036
\
Fig. 80. Summary of data on diffusion constants of hydrogen in palladium.
0 =Time lag (Barrer).
x =Rate of sorption (Jost and Widmann).
• = Variations in E.M.F. (Coehn and Specht).
A = Action on photographic plate (Coehn and Sperling).
* = Rate of sorption (Tammann and Schneider).
% =Rate of sorption (Tammann and Schneider).

TABLE 50. The diffusion constants, D, in palladium, defined

by D = D0<
Method of Diffusion constant
measurement cm.2/sec. Author

Time lag 2-5 x 10"13 e-l0100lRT

RT
Barrer (29)
Rate of absorption 5-4 x lO"2 e-™°l Jost and Widmann oo)
740 B2V
Rate of absorption 2-6 x 10- e- °/ Tammann and Schneider (3D
[Rate of absorption 1 0 7 x 10 3 e-
l290
°lRT1[ Tammann and Schneider (3D]
E.M.F. of wire 7-4 x 10" e-™»iRT
1
Coehn and Specht <2i>
[Effect on photo- 2-5 x 10 8 e-
2070
°lRT Coehn and Sperling (22)]
graphic plate
Mean curve from Fig. 80 1-5 x l O - 2 e - M o o A R r

* Soft. | Tempered.
222

THE DIFFUSION CONSTANTS OF VARIOUS

ELEMENTS IN METALS
Euringer(24) employed the method described on p. 214 to
interpret his results on the desorption of hydrogen from
nickel. He obtained the values for D = Do e~EiRT given in
Table 51. There are no data available with which to compare
Euringer's measurements, since van Liempt's data (Table 49)
on the outgassing of commercial nickel sheets and wires refer
to a mixture of gases (H2, CO, CO2).
TABLE 51. Diffusion constants D = Doe-EIRT of
hydrogen in nickel
Solubility of hydrogen
Temp. D Doe-EIRT in nickel in c.c. at
°C. cm.8 sec.- 1 cm.2 sec."1 N.T.P./C.C. metal at
760 mm. pressure
165 10-5 x lO"88 2-04 x 10-»c- 870
°/ Br
0192
125 3-4 xlO- 8 0194
85 116 xlO~ 0-202

Edwards'(33) measurements on the diffusion of nascent

hydrogen may be interpreted by the time-lag method (p. 217)
to give limiting values to the diffusion constants of hydrogen
in iron. The curves of Fig. 81 show a considerable time lag in
setting up the steady state permeation velocity, and by writing
D = Z2/6JS, where L is the intercept on the time axis and I the
thickness of the sheet, the data of Table 52 were obtained.
Hydrogen gas may diffuse in a metallic lattice as protons,
which will not greatly disturb the lattice in the act of diffusion,
since the dimensions of the proton are small. When oxygen,
nitrogen, phosphorus, sulphur, or carbon diffuse in solids, there
will be a tendency for the diffusing particle to be negatively
charged or to diffuse as an atom; its size will not therefore be
less than that of the free atom, and considerable distortion
may occur in solution in the lattice and in the act of diffusing.*
There is therefore a tendency for compounds of these sub-
* T , . , . , ., • ,. radius of diffusing atom . ..
* In metals such as iron, however, the ratio -r. 7-. ^ is small
radius of iron atom
enough for the atoms of nitrogen or carbon to exist interstitially in the iron
lattice.
 DIFFUSION CONSTANTS 223
stances with the metal to separate out of the lattice as crystals
embedded in the solid metal, or aggregated at grain boun-
daries (34). When elements such as sulphur diffuse in iron, the
process may be regarded as the passing on of the diffusing
atom by alternate dissociation and formation of sulphides,

100 150 200

Time (min.)
Fig. 81. Diffusion of nascent hydrogen through iron sheet (thickness 0-005 in.).

rather than as the inter-penetration of alloying metals, by

place change, or as the zeolitic type of diffusion occurring
when hydrogen passes into a metal as atoms or protons.

TABLE 52. Diffusion constants of hydrogen through iron

liberated by HC1)

Temp. Intercept L D=l*/6L

°C. min. cm.2 sec.- 1 D=Doe~ElRT

10 270 1-66 x 10"» 1-65 x 10-*e-9*00lRT

20 173 2-59 x 10"»
40 67 6-68 x 10- 9 i

50 39-4 114 x 10~« !

75 15-6 2-87 x 10- 8
100 3-6 1-24 x lO"7
224 DIFFUSION OF GASES AND NON-METALS

TABLE 53. Diffusion constants of carbon in iron

Values of D
Temp. Carburising DxlO8 (cm.2 sec.-*) at 1000° C ,
°C. mixture cm.2 sec."1 as obtained by
various authors
950 CO+xylene 70 19-3 xlO- 7 (6)
CO + toluene 9-6 4-2,11-2 xlO" 7 (35)
CO + benzene 10-5
CO + petrol 120 10-4,12-4 xlO- 7 (36)
1000 CO + xylene 120
CO + toluene 150
CO + petrol 21-5
1050 CO + toluene 240
CO + petrol 400
1100 CO + toluene 420
CO + petrol 480

The values of the diffusion constants for a number of the

electro-negative elements (C, O, N, P, S) in iron are given in
Tables 53 and 54. Many authors (6,35,36) have given figures for
the diffusion constant of carbon in iron, and the diffusion has
been followed by heating the metal in a variety of gaseous
atmospheres(1,2,3,4,5) (p. 208, and Table 53). The diffusion
constants are similar also if nitrogenous hydrocarbons replace
the hydrocarbons of Table 53. In Table 54 are given the

TABLE 54. Diffusion of various elements into steel

(D in cm.2 sec-1)
Temp. °C. 800 850 900 950 1000 1050 1100 1150
Diffusion system
D x 108 for C-Feu) 1-7 3-8 8-7 200 —
D x 10* for C-Fed) 1-5 3-8 7-5 11-7 200 28 45
Dxl08fotN-Fe(i> 1-2 30
e-o 10-8 13-5 ,250 40
10
2)xl0 9 forS-Fe(5) — • —
30 5-5 7-0 10 13
Z>xl0 forP-Fe<5) — — —
0-72 1-31 2-5* __
Dxl010forO-Fe(5) —. — —

— 10 — —
* At 1040° C.

diffusion constants obtained by Bramley and his co-workers

for different elements in steel at various temperatures. In
considering these data it must be remembered that the diffusion
constants are often very different in the presence of other
 D I F F U S I O N CONSTANTS 225
impurities in the lattice; for example, the diffusion of sulphur
and phosphorus is retarded by carbon (p. 236), while the
diffusion of nitrogen is accelerated by oxygen but retarded (3fo^
by carbon (Table 54a).

TABLE 54a. Diffusion of N2 in a-Fe at 550° C., as a

function of the carbon content
%CinFe 001 006 0-54 0-82 1-40
8 2 1
D x 10 cm. sec." 214 116 0-33 011 005

The data in Table 54 may be presented in the form

D = Doe~ElRT,
and the following formulae are applicable:
C-Fe :Z> = 5-5 x l o ^ e " ); and 3-5-4-7e-3i4o/*T
N-Fe:D = 1-07 x 1 0 "
S-Fe: D = 4-8 x 10"6 e- (5),

P-Fe: D = 4-5 x 10" (5).

EIRT
The expressions D = Doe- are derived from data con-
siderably less accurate for S-Fe and P-Fe than for C-Fe,
N-Fe systems.
A very recent series of measurements by Wells and Mehl(36&)
applied Matano's method of analysis (Chap. I, p# 47) to
determine D for carbon in iron over a range of carbon
concentrations of 0 1 to 1-0 % C. by weight. They found
that the diffusion, which was unaffected by the grain size
of the samples of iron used, followed the expression

e= (o.o7 + 006x %C.) exp. cm.2sec.-1.

Fig. 81a summarises all data on the diffusion of C in steel.

Tt may be noted that the most satisfactory analysis of the
experimental data is that of Wells and'Mehl.
 226 DIFFUSION OF GASES AND NON-METALS

DEGASSING OF METALS

The evolution of gases from metals is a process of some

technical importance. The commonest gases to be evolved are
hydrogen, oxides of carbon, and nitrogen (Table 55). When
-5-0

-6-0

5
-7-0 -VOwt.%C.-
0-1wt.%C:

-8-0
10 8

Fig. 81a. Summary of data on diffusion of carbon in iron and steel.

Tammann and Schonert(35)
Wells and Mehlow)
Runge<6>
0 . . Paschke and Hauttmann(36C)
Bramley and Allen OM)

the metalloid, graphite, for example, is degassed (37) the com-

position of the evolved gas alters with the period of heating,
and the evolution of gas is more rapid the smaller and less
perfect the component crystallites, so that charcoals more
readily evolve gases than graphite. It is very difficult indeed
to remove the last traces of gas from carbon, the general
behaviour being that at a given temperature a state is reached
where no appreciable gas evolution occurs, but as soon as the
temperature is raised again a fresh burst of gas is obtained.
 TABLE 55. Oases evolved in heating metals
Temp. Gases
Metal evolved Remarks

Mo (38) >1000 Mainly H 2 Amounts of gas slight by

1000 CO, CO2, and H t 1760° C.
>1200 Oxides of carbon
andN 2
W(38) <2430 H 2 ,CO,CO 2 ,N 2 No further gas evolved after
outgassing at 2430° C.
Amounts of gas relatively
small by 1760° C.
C(38) 1300 H2 (52%) Graphite (tungar anode)
CO (44%) could be freed of gas com-
pletely by prolonged heat-
1600 H a 2 (30%) ing to 2150° C.
CO (48%)
CO2(6%)
Na (16%)
1900 Ha (11%)
CO (13%)
CO2(5%)
Na (71%)
2110 CO (9%)
CO2(4%) *
N2 (87%)
Fe(39) H2, oxides of Guillet and Roux(39) found
carbon, nitro- 30 c.c/100 g. metal
gen
Zn(4o> Almost pure H 2
from electrode-
deposited Zn
N i (38,41) CO,H 2 ,andCO 2 As much as 100 c.c/100 g.
Mainly CO for Mond nickel,
and mainly H 2 for electrode
deposited nickel
750 CO2(12%) Total 0-67 c.c. in 20 min.
CO (54%) (from 100 g. metal)
H2 (34%)
850 Total 0-39 c.c. in 20 min.
COa(7-7%)
H2 (7-7%?
950 CO2(5%) Total 0-60 c.c. in 20 min.
CO (90%)
Ha (5%)
1050 CO2(9-5%) Total 0-53 c.c. in 20 min.

H2 (5-7%?
1150 00,(11%) Total 0-27 c.c. in 20 min.
CO (48%)
Ha (41%)
Al(42.43) CO. H2, CH4 Quantities from 2 to 30 c.c./
100 g. reported. Solubility
of gases in aluminium small,
and therefore gases evolved
and retained in pinholes.
Hydrogen occluded as a
result of the action of water
upon aluminium
CU (43) SO2 (60%) Total gas extracted 2 c.c./
CO (20%) 100 g.
H 2 (14%)
228 DIFFUSION OF GASES AND NON-METALS

These characteristics of graphite filaments are common to all

metal filaments, and it is also usual to find that if the filament
is degassed at a high temperature and cooled in vacuum to
room temperature it remains free of gas at this temperature
for a long period. This is the result of the high temperature
coefficient of the diffusion constant. The data in Table 55 give
some typical results on the degassing of metals. The com-
position of the evolved gases, besides depending upon the time
of heating and the temperature, depends also upon the nature
of the metallurgical process by which the metal has been
obtained. Therefore the data in this table are by no means
comprehensive, and are intended only to give some typical
experiments on degassing. In addition to considering the
effects of time of heating and temperature upon the gas
evolution, one may also have to consider the place of origin
of the gas. Eltzin and Jewlew (37) considered the gases evolved
from graphite to be different in composition according as they
were evolved from the "surface" of the filament, or from the
"interior" of the filament,* their analysis giving:

Surface CO2 3 4 % Interior CO 6 8 %

CO 5 8 % N2 3 2 %
N2 8%
It is interesting to find nitrogen among the products evolved
from graphite, for its presence as a stable compound in the
lattice may mean that it replaces carbon atoms in the edges
of graphite laminae, or is firmly attached to peripheral carbon
atoms in the laminae.
Kinetic studies of the evolution of gases from metals have
been made which indicate that the processes involved are
often true diffusions. The evolution of carbon monoxide from
nickel wires (41) shows that the curve log (gas evolved) against
time has the typical shape associated with a diffusion process
(Chap. I, Fig. 7), and from this curve the diffusion constants
* Their concept of "surface" and "interior" is however rather vague. They
considered the first gases evolved to come from the surface, and subsequent
gases to come from the interior.
 DEGASSING OF METALS 229

may be calculated. The method which van Liempt(25) applied

(p. 215) to measure the mean diffusion constant of gases from
commercial nickel should give comparable constants, since the
main gas evolved is carbon monoxide. Some of these diffusion
constants are given in Table 56, from which it is evident
that the two methods agree in order of magnitude only.
Van Liempt also calculated "diffusion" constants for the
outgassing of molybdenum wire, obtaining a mean value
of 7-6 x 10-9 cm.a'sec.-1 at 900° C.

TABLE 56. Diffusion constants of CO in nickel

Diffusion
Author Form of constant Temp.
nickel cm.2 sec."1 °C.

Smithells and Ransley(4i) Wire 4 0 x 10- 8 950

140 x 10- 8 1050
van Liempt (25) Thin sheet 2-5 x 10-*8 700
9-9 x 10- 7 800
2-1 x 10- 900
(Using Smithells and Wire Final 14 x 10"77 1050
Ransley's data) Initial 91 x 10~8
Mean 6 x 10~ 950

Filaments glowed in gas atmospheres will frequently absorb

the gas, by reactions which are in part chemical and in part
processes of diffusion. Tantalum (44) absorbs nitrogen slowly
at 1300° C. and rapidly at 1800° C; while oxygen is taken up
at 730° C. and rapidly at 1500° C. Tantalum will also decom-
pose hydrocarbon vapours at high temperatures, the carbon
diffusing into the filament (from 1700 to 2500° C.) with the
formation of carbides, while evacuation at 2200° C. causes
the carbon to evaporate again from the metal until only pure
tantalum remains. Similarly, nitrogen (38) can be absorbed
by molybdenum wire, setting up an equilibrium whose tem-
perature variation implies a heat of 38,500 cal./moL, while the
rate of establishment of the equilibrium involves an apparent
activation energy of 26,600 cal. The analogous tungsten
nitride could also be formed. Another type of filament-gas
reaction may also be found in which the metal filament
230 DIFFUSION OF GASES AND NON-METALS
evaporates and the condensing metal combines with or adsorbs
an otherwise inert gas. In this way metal filaments may be
used in the clean-up of residual gas (45).

DIFFUSION AND ABSORPTION OF GASES IN

FINELY DIVIDED METALS

When a gas is allowed to come into contact with a thoroughly

outgassed finely divided metal, there may take place an
instantaneous adsorption and one or more slow processes.

TABLE 57. Some data on the sorption of gases by

finely divided metals
System Behaviour
H 2 , D2-CU(46) - 78° C. rapid initial and slow subsequent sorption.
0° C. amount taken up by slow sorption increased
H2-CU(47) Initial rapid sorption and slow subsequent sorption
at all temperatures studied
H2-Cu<48) Slow process observed
H 2 , D2-Ni(49> Two processes, fast and slow. H 2 more rapidly sorbed
than D 2
H2-Ni(w> One and possibly two slow processes observed
H 2 -Ni(5i) Slow process observed
H2-Fe<52) One rapid sorption process and two slow ones
suggested
Ha-Fe(«3) -190°C., van der Waals' sorption. Above 0° C.
two slow processes suggested
N 2 -Fe, A12O8(54) Slow uptake of nitrogen by iron, as well as initial
N2-Fe, Fe-Al2O,, rapid sorption
K2O
O2-Ag(55) Slow* uptake of oxygen by silver, as well as initial
rapid sorption

The nature of the slow uptake of gas has been the cause of
considerable discussion. The general similarity of the pheno-
mena observed for a number of sorption systems is illustrated
in the summarising data of Table 57. In all the cases given, at
least two processes have been established. The initial rapid
process is correctly associated with the van der Waals' ad-
sorption on the surface, and it might be anticipated that
the slow process was the solution of the gas in the metallic
lattice, either forming an alloy (H2) or a compound (N2)
with-the metal. However, the behaviour is not necessarily
 DIFFUSION OF GASES IN METALS 231
so easily explained, for not only do some authors describe
two slow processes, but in a large number of studies of
sorption very similar slow processes have been observed

0 200° 400°
Fig. 82. Sorption of hydrogen by Mo-Si catalysts.
where solution is not probable. Among these systems are
the following:
H 2 , D2-Cr203(57,58),
H 2 -Cr 2 0 3 , ZnO(56),
0 2 -CuCr 2 0 4 , ZnCr2O4, CoCr2O4, NiCr2O4, BeCr2O4(59),
H2-MoO3, SiO2(60),
H 2 -C (Charcoal (61), Graphite (62), and Diamond (63)),
CH4-C(6i),
C2H4-Ni(64).
232 DIFFUSION OF GASES AND NON-METALS

All these systems give the characteristic high-temperature

isobar illustrated in Fig. 82, in which at first the amount
sorbed increases with rising temperature, and then decreases.
It is considered that the rising part of the isobar denotes a
non-equilibrium condition, but that the falling part is rever-
sible for such systems as H2-C, H 2 -Zn0. Cr2O3. In other cases,
however, no part of the curve denotes a reversible equilibrium
(CH4-C; C2H4-Ni). I t is also a characteristic of these reversible
and irreversible chemical sorptions that the velocity of
sorption increases exponentially with temperature, and from
this increase in sorption velocity an apparent energy of
activation may be calculated, whose magnitude is that of
ordinary chemical reactions or of activated diffusions. The
apparent energy of activation increases with the amount of gas
sorbed, and so the sorption velocity decreases strongly as the
charge of gas is increased. This means that at low temperatures
one will not get saturation of the available surfaoes in any
finite time.
Now one may compare these properties with those noted in
the processes of slow sorption of gases by finely divided metals.
Once more one finds high temperature isobars in which the
amount sorbed at first increases with rising temperature and
then decreases (Fig. 83). The decreasing part of the isobar is
reversible, and the increasing part is irreversible. The velocity
of sorption increases strongly as the temperature rises, and an
apparent energy of activation is observed. Once again the
irreversible part of the isobar may be interpreted as due to
variable apparent activation energies which increase with gas
charge, so that at low temperatures the velocity of sorption
is so diminished that the available sorption volume is not
saturated in any finite time.
It seems therefore that activated diffusion into the bulk of
a metal and reversible chemical adsorption (called by H. S.
Taylor (65) activated adsorption) may be similar in their
observable properties. Indeed there has been considerable
argument as to whether reversible chemical adsorptions may
not be activated diffusions, and vice versa. The situation is
 DIFFUSION OF GASES IN METALS 233
clarified by the proof that activated diffusion processes into
solids (e.g. H2-Pd(30)) can occur to give, eventually, homo-
geneous solutions; but that reversible chemical adsorptions
which do not involve inter- or intra-lattice diffusion also occur
(e.g. H2-C (61,62,63)). There is no need to strain either the
hypothesis of activated diffusion or of activated adsorption to
include all the features observed in considering gas-solid

-200 -150 -100 - 5 0 0 50 100

Temperature, °C.
Fig. 83. Sorption isobars of hydrogen on nickel, showing low and high tem-
perature sorption. Curve 1, 2*5 cm. pressure; curve 2, 20 cm. pressure;
curve 3, 60 cm. pressure.

systems. In those cases where two slow processes of sorption

of hydrogen by metals have been postulated, it may be that
a slow chemical adsorption process occurs at the surface,
followed by a slow absorption process by the metal. The
adsorption process can be either on the metal surface itself, or
as seems likely under some experimental conditions, it may be a
chemical adsorption of the hydrogen by an oxide monolayer.
One of the few quantitative analyses of sorption kinetics in
gas metal systems was made by Ward (47), who after heating
and evacuating finely divided copper in hydrogen a number of
times, concluded that he was measuring an activated diffusion
234 DIFFUSION OF GASES AND NON-METALS

process, and analysed the data on the rate of sorption by means

of Fick's law. In accordance with this law, for small values of
thje time, t> Ward found that Q, the quantity of hydrogen
absorbed, was proportional to *]t. If n0 denotes the concen-
tration in the adsorbed film of gas from which solution in the
metal occurs, one should have, from Lennard-Jones'(66)
analysis of this problem,

and so by plotting log j against = one may measure the

activation energy for the solution process. The figure obtained
was
E = 14,100 cal./g. atom,
which may be compared with the temperature coefficients foi
the permeability of copper to hydrogen of 18,700 and 16,600 cal.
(Chap. IV, p. 168).

THE INFLUENCE OF IMPURITY UPON DIFFUSION CONSTANTS

It has been observed that the presence of carbon in nickel

decreases the permeability of nickel towards hydrogen below
700° C , but above this temperature increases the perme-
ability (67,68). Also in steel it is very important to know the
effects which various possible impurities (O, N, S, P, C, Si)
have upon the permeability and diffusion velocity of other
elements in the steel. While few experiments have been made
on the permeability, the diffusion velocity within the material
has been very thoroughly studied by Bramley and his co-
workers (1,2,3,4,5). Ham and Sauter(69) showed that nitriding
the surface of steel increased its hydrogen permeability
10-15 times, and that the original permeability was restored
after out-gassing the metal. Bramley and his co-workers
found that the nitriding of a steel rod was accelerated by
the presence of small amounts of oxygen in the metal, as the
following data indicate (Table 58). The nitriding process occurs
 INFLUENCE OF IMPURITY UPON DIFFUSION 235

on heating in an atmosphere of ammonia, and the processes

taking place may be visualised as
4Fe + 2NH3 -> 2Fe2N + 3H2,
2Fe2N -> 2Fe2N (dissolved),
F e ^ ^ 2Fe + N (dissolved).
The role of the oxygen or oxide may be to fix the nitrogen
atoms as oxides of nitrogen, which prevents the formation and
subsequent evolution of molecular nitrogen. These two

TABLE 58. Nitriding after various treatments, involving

solution or removal of oxygen
Diffusion
Condition of Fe before nitriding constant
for Na x 108
Swedish Fe, heated 200 hr. at 1050° C. in dry H 2 1-8
Swedish Fe, in original state 21
Swedish Fe, in original state oxidised in CO-CO. (75- 2-5
25%) for 30 hr.
Armco Fe, in original state 2-6
Swedish Fe, oxidised for 100 hr. in malleabilising 2-9
furnace at 1000° C.
Swedish Fe, oxidised for 200 hr. in malleabilising 3-5
furnace at 1000° C.

systems (H2 diffusing in nitrided Fe; and N 2 diffusing in

oxidised Fe) are the only ones where acceleration occurs. More
usually one finds a retardation, as for instance when sulphur
diffuses through steel of increasing carbon content. Fig. 84
shows how the diffusion constant decreases with the percentage
of carbon and is nearly inhibited by a large quantity of this
element. Analogous experiments were made by heating a
mixture of toluene and carbon bi-sulphide in contact with
steel. The shape of the curves was attributed to a de-sulphuri-
sation brought about by the hydrogen liberated from the
toluene. However, the analysis of the deepest part of the
curves led to the conclusion that the diffusion constant of the
carbon was decreased by increasing the sulphur content, while
the diffusion constant for sulphur was decreased by increasing
the carbon content.
236 DIFFUSION OF GASES AND NON-METALS
The last of this series of experiments was a study of the
phosphorisation of steels. Phosphorisation occurred on heating
the steel sample in a hydrogen-phosphine mixture, and con-
centration gradients of phosphorus into the metal were
established. It was found on analysing the cpncentration-
distance curves that carbon retarded the diffusion of phos-
phorus very strongly, and that phosphorus entering the metal
swept the carbon in front of it.

v\
O 0-1 ©4 O* O6 K>
CARBON PERCENT.
Fig. 84. The decrease in the diffusion constant of sulphur
in steels as the carbon content increases.

REFERENCES
(1) Bramley, A. and Jinkings, A. Carnegie Schol. Mem., Iron aiuj Steel
Institute, 15, 17 (1926).
Bramley, A. and Beeby, G. Carnegie Schol. Mem., Iron and
Steel Institute, 15, 71 (1926).
Bramley, A. and Jinkings, A. Carnegie Schol. Mem., Iron and Steel
Institute, 15, 127 (1926).
Bramley, A. Carnegie Schol. Mem., Iron and Steel Institute, 15,
155 (1926).
(2) Bramley, A. and Lawton, G. Carnegie Schol. Mem., Iron and
Steel Institute, 15, 35 (1927).
(3) Bramley, A. and Turner, G. Carnegie Schol. Mem., Iron and
Steel Institute, 17, 23 (1928).
(4) Bramley, A. and Lord, H. Carnegie Schol. Mem., Iron and Steel
Institute, 18, 1 (1929).
(5) Bramley, A., Heywood, F., Cooper, A. and Watts, J. Trans.
Faraday Soc. 31, 707 (1935).
(6) Runge, B. Z. anorg. Chem. 115, 293 (1921).
 REFERENCES 237
(7) Hanawalt, J. D. Phys. Rev. 33, 444 (1929).
(8) Linde, J. and Borelius, G. Ann. Phys., Lpz., 84, 747 (1927).
(9) Kruger, F. and Gehm, G. Ann. Phys., Lpz., 16, 174 (1933).
(10) Owen, E. A. and Jones, J. Proc. phys. Soc. 49, 587 (1937).
(11) Hagg, G. Z.phys. Chem. 7B, 339(1930).
(12) Z. phys. Chem. 4B, 3.46 (1929).
(13) Meyer, L. Z. phys. Chem. 17B, 385 (1932).
(14) Wagner, C. Z. phys. Chem. 159 A, 459 (1932).
(15) Sieverts, A. and Hagen, H. Z. phys. Chem. 174A, 247 (1935).
(16) Coehn, A. and Jiirgens, H. Z. Phys. 71, 179 (1931).
(17) Dunwald, H. and Wagner, C. Z. phys. Chem. 22B, 212 (1933).
(18) v. Baumbach, H. and Wagner, C. Z. phys. Chem. 24 B, 59 (1934).
(19) de Boer, J. H. and Fast, J. Rec. Trav. chim. Pays-Bos, 55, 459
(1936).
(20) Sieverts, A. and Briining, K. Z. phys. Chem. 174A, 365 (1935).
(21) Coehn, A. and Specht, W. Z. Phys. 62, 1 (1930).
(22) Coehn, A. and Sperling, K. Z. Phys. 83, 291 (1933).
(23) Franzini, T. R.C. 1st. Lombardo, [2] 66, 105 (1933).
(24) Euringer, G. Z. Phys. 96, 37 (1935).
(25) van Liempt, J. Rec. Trav. chim. Pays-Bos, 57, 871 (1938).
(26) Daynes, H. Proc. Roy. Soc. 91 A, 286 (1920).
(27) Barter, R. M. Trans. Faraday Soc. 35, 628 (1939).
(28) Phil. Mag. 28, 148 (1939).
(29) To be published.
(30) Jost, W. and Widmann, A. Z. phys. Chem. 29B, 247 (1935).
(31) Tammann, G. and Schneider, J. Z. anorg. Chem. 172, 43 (1928)-
(32) Duhm, B. Z. Phys. 94, 34 (1935).
(33) Edwards, C. A. J. Iron and Steel Inst. 60, 9 (1924).
(34) Smithells, C. Oases and Metals, p. 173 (1937), Figs. 120, 121.
(35) Tammann, G. and Schonert, K. Stahl u. Eisen, Diisseldorf, 42,
654 (1922).
(36) Calculated by Runge<6) from observations of Giolotti and
co-workers.
(36a) Eilender, W. and Meyer, O. Arch. Eisenhiittenw. 4, 343 (1931).
(366) Wells, C. and Mehl, R. Metals Technol. A.I.M.E. 1940, Tech.
Publ. No. 1180.
(36c) Paschke,M.andHauttmann,A. Arch.EisenhiUt.9,305(1935-6).
(36d) Bramley, A. and Allen, K. Engineering, 11 March 1932.
(37) E.g. Eltzin, I. and Jewlew, A. Phys. Z. Sowjet. 5, 687 (1934).
(38) Norton, A: and Marshall, F. Trans. Amer. Inst. min. (metaU.)
Engrs, Feb. 1932.
(39) Guillet, L. and Roux, A. Rev. MetaU. 26, 1 (1929).
(40) Rontgen, P. and Moller, H. Metallwirtschaft, 11, 685 (1932).
Burmeister,W. and Schloetter, M. MetaUwirtschaft, 13,115 (1934).
(41) Smithells, C. and Ransley, C. J. Proc. Roy. Soc. 155 A, 195 (1936).
(42) Villachon, A. and Chaudron, G. C.R. Acad. Sci., Paris, 189, 324
(1929).
(43) Hessenbruch, W. Z. Metallic. 21, 46 (1929).
238 DIFFUSION OF GASES AND NON-METALS
(44) Andrews, M. J. Amer. chem. Soc. 54, 1845 (1932).
(45) E.g. Bryce, G. J. chem. Soc. p. 1513 (1936).
(46) Beebe, R., Low, G., Wildner, E. and Goldwasser, S. J. Amer.
chem. Soc. 57, 2527 (1935).
(47) Ward, A. F. Proc. Ray. Soc. 133 A, 506, 522 (1931).
(48) Leypunsky, O. Ada phys.-chim. U.R.S.S. 2, 737 (1935).
(49) Magnus, A. and Sartori, G. Z. phys. Chem. 175A, 329 (1936).
(50) Iijima, S. Sci. Pap. Inst. phya. chem. Res., Tokyo, 23, 164 (1934).
(51) Benton, A. and White, T. J. Amer. chem. Soc. 52, 2325 -(1930).
(52) Harkness, R. and Emmett, P. J. Amer. chem. Soc. 56, 490 (1934).
(53) Morosov, N. M. Trans. Faraday Soc. 31, 659 (1935).
(54) Hammett, P. and Brunauer, S. J. Amer. chem. Soc. 56, 35 (1934).
(55) E.g. Benton, A. and Elgin, J. J. Amer. chem. Soc. 51, 7 (1929).
(56) Pace, J. and Taylor, H. S. J. chem. Phys. 2, 573 (1934).
(57) Taylor, H.S. and Diamond, H. J. Amer. chem. Soc.56,1821 (1934).
(58) Kohlschutter, H. Z. phys. Chem. 170A, 300 (1934).
(59) Frazer, J. and Heard, L. J. phys. Chem. 42, 855 (1938).
(60) Griffith, R. and Hill, S. Proc. Roy. Soc. 148A, 195 (1935).
Hollings, H., Griffith, R. and Bruce, R. Proc. Roy. Soc. 148 A,
186 (1935).
(61) Barrer, R. Proc. Roy. Soc. 149A, 231 (1935).
(62) Trans. Faraday Soc. 32, 481 (1936).
(63) J. chem. Soc. p. 1256 (1936).
(64) Steacie, E. and Stovel, H. J. chem. Phys. 2, 581 (1934).
(65) Taylor, H. S. J. Amer. chem. Soc. 53, 578 (1931).
Trans. Faraday Soc. 28, 131 (1932).
(66) Lennard-Jbnes, J. Trans. Faraday Soc. 28, 333 (1932).
(67) Lewkonja, G. and Baukloh, W. Z. Metallic. 25, 309 (1933).
(68) Baukloh, W. and Guthmann, H. Z. Metallic. 28, 34 (1936).
(69) Ham, W. and Sauter, J. Phys. Rev. 47, 337 (1935).
 C H A P T E R VI

D I F F U S I O N OF IONS. IN I O N I C CRYSTALS
AND T H E I N T E R D I F F U S I O N OF METALS

INTRODUCTION AND EXPERIMENTAL

The study of the interdiffusion of solids probably begins with

the empirical facts of carburisation of steel, an art many
centuries old. That solid metals will interdiffuse was early
observed (i); but the velocity of this interdiffusion was not
realised until the quantitative measurements of Roberts-
Austen (2) revealed that at 300° C. gold would diffuse througli
lead faster than sodium chloride would diffuse through water
at 18° C. The first alloys were prepared by Faraday and
Stodart(3) in 1820, by heating together mixtures of metal
powders. It is interesting to find that this original method is
employed today (4,5) for the preparation of special alloys (6,7).
One may also trace, from early studies of carburisation of
steel (8,9), and with increasing research on intermetallic
diffusion, the development of nitriding, chromizing, calorizing,
sherardizing, and siliconizing, and the formation of bi-metal
strip and veneer metals (io, li). The processes of homogenisation
of segregated alloys, rates of transformation in metals, and
of precipitation of crystals in solids (e.g. Fe3N in Fe), are all
closely connected with processes of diffusion in solids. One
may see therefore the technical importance of a knowledge of
the laws governing intermetallic diffusion. The subject has
not yet reached a completeness in itself or in relation to cognate
topics such as the diffusion of gases in metals (Chaps. IV and V)
or of ions in ionic lattices. Since the diffusion usually occurs in
the lattices of the solids, it is likely to yield much information
on the physics of crystals, and phenomena such as annealing,
age-hardening, plasticity, recrystallisation, and order-disorder
transformations in alloys.
240 INTEBDIFFXJSION OF SOLIDS

The problem of diffusion in ionic lattices has been studied

principally by the indirect method of measuring the con-
ductivity of the crystal. Some of the earliest measurements
of conductivity in solids are due to Faraday (12). The study of
conductivity in electrolytes has been developed along three
main lines:
(1) The movement of ions in ionic lattices.
(2) The movement of ions in molten ionic liquids (e.g.
molten NaCl).
(3) The movements of ions in solution.
While we are not concerned tvith (2) and (3), it is interesting
to note that the conductivity of silver halides shows no dis-
continuous jump on passing from the solid to the liquid
state (13,14), although it is more usual to find discontinuities
whenever a phase change occurs (KI(U), HgCu2I4(i5),
Hgl 2 (16)). Early measurements upon the conductivity of solid
oxides (17,18,19) showed that the current carriers were ions, and
also led to the development of the Nernst lamp, using a filament
of zirconia, with thoria and rare earths. One of the earliest
observations upon the increase of conductivity of micro-
crystalline salts under pressure was made by Graetz (20), and
upon photoconductivity by Arrhenius(2i), who noted that the
conductivity of silver chloride and bromide was altered by
light. The modern developments of the subject we owe
especially to von Hevesy, Seith, Jost, Wagner, and Tubandt.
Application of the mobility of hydrogen or sodium ions in
glass is made in the glass electrode, or the preparation of pure
sodium by electrolysing sodium ions from molten sodium
nitrate through glass. Processes of base exchange in zeolites
(as in water softening by "permutit"), or even in clays, must
occur in part by diffusion of ions down concentration gradients
in the individual crystallites composing the mass. Zeolites
contain large interstitial channels down which such a diffusion
is possible.
The determination of diffusion constants in metals may be
made by a number of rather special experimental techniques,
 INTRODUCTION AND EXPERIMENTAL 241

only some of which permit the whole concentration gradient

to be measured. Dunn (22) followed the diffusion of zinc from
a-brass by vaporising the zinc in vacuo, and measuring the
loss in weight of the sample. This method involves the
assumption that the concentration of zinc at the outgoing
surface is zero and, like all methods of averaging, does not
easily show whether the diffusion "constant" depends on the
concentration. Another method of averaging has been em-
ployed to find the rate of diffusion of carbon and nitrogen in
iron (23). The diffusing element is removed as a gas (CO, or Na)
as soon as it reaches the surface A similar method (24) was used
to measure the diffusion rate of oxygen in y-iron, the oxygen
being removed at the surface by hydrogen. If the metal is in
the form of wires, the electrical conductivity may be used to
give the average composition of the wire (25,26,27) The method
has proved successful for hydrogen-palladium, carbon-
tantalum, and other systems. The lattice parameter, in a few
cases (28), undergoes a steady change as the concentration
increases, and this change may be employed to find the mean
concentration.
When it is desired to find the actual concentration gradient,
one may use several methods. The first, used by Bramley and
his co-workers (23) to measure diffusion coefficients in iron, is
to heat the metal with the diffusing substance, and to remove
and analyse chemically thin layers at that interface from
which diffusion proceeded. Other methods of analysis are
available besides chemical ones. The shavings may be
examined by means of an X-ray camera, and variations in
lattice parameter, or occurrence of known alloy phases noted,
and used to establish concentration gradients. Similarly,
spectroscopic analysis of the shavings by giving the position
and intensity of spectral lines will allow the concentration
gradients to be measured. This method was used to follow the
diffusion of a number of metals in silver (29). A micrographic
method of establishing concentration gradients in Cu-Al,
Mg-Al systems has also proved successful (30).
A very interesting method (31), which could be used either as
BD 16
 TABLE 59.* Artificial radioactivity (after Hevesy and Paneth&i))
Atomic no. Mass numbers of Half-life periods of
and element isotopes bombarded radioactive products Nature of radiations emitted

4 Be 10 lOy. (-)
9F 17, 18, 20 l-2m., 108m., 9s. (+M+M-)
11 Na 22,24 3y., 15h.
14 Si 27,31 6-6m., 2-5h. (+M-)
15 P 30,32 3m., 14-5d.
16 S 31,35 26m., 80d. (+)»(~)
18 A 41 110m. (~)
19 K 38, 42 7-5m., 12-5h.
20 Ca 39,45 4-5 m., 2-3 h. (+)»(-) en
21 Sc 41, 42, 43, 44, 46, 48 53m., 4-lh., 4-0h., 52h., 90d., 41h. M
23 V 48, 49, 50, 52 16d., 32m., 3-6h., 3-8m. o
25 Mn ?, 56, ?, ? 46m., 2-5h., 21m., 5d., 7mth.
26 Fe 55,59 8-9m., 40d.
27 Co 55, ?, 58, 60 18h., 150d., l l m . ~ l y . ( + ) , ( + and - ) , ( + ) , ( - )
28 Ni 63 120m. ( ~)
29 Cu 61, 62, 64, 66 3-3h., 10-5m., 12-5h., 5m. ( + ) , ( + ) , ( + and - W - ) O
30 Zn 63,65 38m., 60m.
31 Ga 66, 68, 70, 72 9-4h., 60m., 20m., 23h. d
32 Ge 75, ? 20h., 30m. 03
33 As 76,78 26h., 65m.
34 Se 79 or 81, 83 lh., 17m.
35 Br 76, 80, 80, S2, 83 6-3m., 18m., 4-5h., 34h., 2-5h. ( + ) . ( ~ )> ( ~ ) » ( ~ )>(*")
36 Kr v 74m., 4-5h., 18h, (-),(-),(?)
37 Rb 86,88 18m., 18d. ( _^ (_j
38 Sr 89 3h., 55 d.
39 Y 90 70 h.
40 Zr 97 44h. (-)
 42 Mo ? 2*5m., 17m., 36h.
44 Ru ? 40s., 100s., l l h . , 170h. (_)>(_)>(_)>(_)
46 Pd 60h., 15m., 12h., 3m.
47 Ag i06,106,108,110, 111, 24-5m., 8*2d., 2*3m., 22s., 7-5d., 8-2h.
112 *
48 Cd ?, 115, 117 33m., 4-3h., 58h. (+)»(~) (~)
49 In 111,112,114,114,116, 20m., 72s., 4 1 h., 50d., 13s., 54m., (+).(-)•(-).(-)•(-).(-).(-)
lift 117
11O, 1 1 1 2*3 h
51 Sb X20,122, 124 16m., 2*5d., 60d. (+),(-),(-)
52 Te ? llh. (~)
55 Cs 134 l*5h. (~~)
56 Ba ?, 139 2-5m., 80m ^ (?), (?)
57 La 140 l-3d.
59 Pr 140,142 3m., 19h. (+)»(—)
60 to 71 (other — In nearly every case half-life periods —
rare earths) of suitable length
79
i £t "Rf
ni 181 JIJUI

73 Ta 180, 182 8h., 97d. (+M-)

74 W ? Id.
75 Re ?, 188, ?, ? 20m., 18h., 85h., 40h. (jjj ( - ) , (-)
76 0s 40 h.
77 Ir 192, 192 or 194, 189, 2m., 14m. and 19h., 28m., 8-5h. (?),(-),(-),(-),(-)
192 or 195
78 Pt 197, 199 18h. and 3-3d., 50m. ( • " ) » ( " )

79 Au 198, 199 2-7d., 13h., 4-5d. ( - ) , ( - ) , ( - )

80 Hg 205, 206 41 h., 45 m. ( - ) , ( - )

81 Tl 204,206 5 m., 97 m. ( - ) , ( — )

82 Pb 209 3h. ( - )

* In the table, m. = minute, h. =hour, d. =day, mth. = month, y. =year. Also ( - ) denotes electron and ( + ) denotes positron.
244 INTERDIFFUSION OF SOLIDS

an averaging method, or to determine the actual concentration

gradients, from the radioactive emission of successive thin
layers, is the radioactive isotope technique. The method was
originally employed to determine self-diffusion constants in
lead, using radium D or thorium B as indicator. Thorium B
may be condensed on the metal foil or single crystal. The path
of a-particles in lead is only 50/4, so that-when some of the
thorium B atoms have penetrated more deeply than this their
radioactivity can no longer be detected, and the radioactivity
of the lead sheet falls off. Instead of following diffusion by
measuring the ionisation produced by a-rays, the recoil rays
accompanying emission of a-particles were measured in some
cases (32), the improvement effected by this method being that
recoil particles in lead have a range of only 0*5x 10~6 cm., so
that diffusion coefficients as small as l O - ^ c m ^ d a y 1 can be
evaluated. The a-ray and recoil-atom methods have been used
also to follow the self-diffusion of bismuth using thorium C as
indicator. The results obtained by the two methods are in
satisfactory agreement.
The radioactive indicator method has been used in just the
same way to follow the diffusion of lead ions in lead chloride
and lead iodide (32). With the discovery of artificial radio-
activity the method seems capable of very wide application
indeed. Gold, for example, has been rendered radioactive by
neutron bombardment, and then used (33) to measure the self-
diffusion constant of gold in gold; and radioactive copper has
been used to measure self-diffusion in copper (33a). Some of the
substances which may be rendered radioactive by bombard-
ment with neutrons, deuterons, protons, or y-rays, and whose
half-life period seems adequately long for the duration of
possible diffusion experiments, have been collected in Table 59.
In addition to the normal radioactivity of radium, thorium,
polonium, and uranium, radioactivity may be induced in these
elements. The cyclotron has made it possible to obtain high-
energy particles in considerable concentrations, and so it may
be anticipated that artificial radio elements will become
increasingly accessible to research workers. It is for this reason
 INTRODUCTION AND EXPERIMENTAL 245

that the radioactive isotope method is regarded as extremely

important.
The diffusion of elements from the interior to the surface of
a metal will usually change its thermionic emission(35), photo-
electric emission (36), or contact potential/37). These methods
have been used especially in following the grain-boundary
diflFusion of thorium in tungsten, and the surface migration of
barium, caesium, sodium, and potassium over tungsten (Chap.
VIII). Cichocki (38) demonstrated the diflFusion of metals from
salts through copper, silver, and gold foil by making use of
the positive-ion emission. The salt was enclosed in the foil, and
heated, and after an interval positive ions escaped from the
outer surface.
Diffusion constants in ionic lattices are in many instances
calculated from the conductivities. The actual transfer in the
ionic lattices may be demonstrated by the method of
Tubandt(39) who employed cells such as
Pt / AgCl / Ag2S / Ag2S / Ag2S / Ag,
and passed a current through the cell, so that silver dissolved
at the silver electrode, and was precipitated at the platinum
electrode, the quantity transferred being found by weighing.
If sticks of two salts are pressed together and heated they
diffuse into one another, and the composition could be found
by dividing the material into sections, and determining their
density or chemical composition.

TYPES OF DIFFUSION GRADIENT

The diflFusion gradients established on the interdiffusion of two

solids may take a variety of forms. When two slabs of metal,
e.g. Cu and Ni, are heated together, the simplest type of
diflFusion gradient which may be established is that illustrated
in Pig. 85(40). The figure shows that the presence of a third
substance may alter the shape of the concentration gradient.
The form of the concentration gradient on either side of an
interface is not always of this smooth form. Bramley and his
co-workers (23) heated steel bars in various atmospheres from
246 INTERDIFFUSION OF SOLIDS

which the elements C, N, O, S, and P diffused into the steel.

The variety of shapes of concentration gradient which they ob-
tainedare illustrated in Chap. V, Figs. 73 and 74. Under certain
conditions they obtained gradients with a maximum, as well
Boundary %M

Distance from Boundary

Fig. 85. Diffusion of copper into nickel (Grube and Jedele(40>).
PureNi.
Ni containing Mn.
Curve after heating at 1025° C. for 120 hours.

o
*fe 40 1
t 30

i it 20 24 28 X F 40 44~ 48
Distance in inchesx 10'*
Fig. 86. Diffusion of chromium into iron (Hicks(4i>).
Gradient after 96 hours at 1200° C.

as S-shaped. Obviously the analysis of such curves to give

diffusion constants will be difficult. Another complicating
factor arises when two phases co-exist. In this instance con-
centration discontinuities occur (Fig. 86) (41). Systems in which
concentration discontinuities exist may often be analysed
 TYPES OF DIFFUSION GRADIENT 247

readily to give the diffusion constant, D. The discontinuity

usually marks points of fixed concentration, and the rate at
which this discontinuity progresses into the solid is therefore
likely to be governed (for linear diffusion) by the law
x2
jr- = constant,
where x denotes the distance of the discontinuity from the
origin, and t is the time.

THE STRUCTURE OF REAL CRYSTALS

The ideal crystal consists of a perfectly ordered array of atoms,

ions, or molecules in three dimensions. This ideal is difficult
to attain experimentally, although in one instance at least
(NaCl)(42) it appears to have been closely approached. There
cahnot be any ionic conductivity or atomic diffusion iji a
perfect lattice which conserves its ideal order in all circum-
stances, and the observation that conductivity and diffusion
do occur is only one of a number of lines of evidence which
lead to the conclusion that a real crystal is not perfectly
regular under all conditions.
As a result of a large number of researches it has been
established that two types of fault system may exist in crystals
which may be called reversible and irreversible fault systems.
The former have reproducible properties in many respects,
but properties of the latter depend upon the previous history
of the specimen. In particular, irreversible fault systems
give rise to the "structure-sensitive" diffusion and con-
ductivity data described in the following chapter, while
reversible fault systems give rise to reproducible conductivity
and diffusion phenomena. Some of the reproducible and
equilibrium types of fault systems may be briefly discussed
before the conductivity and diffusion data are given.
Besides defects which have been artificially introduced in
crystals, other types of imperfection exist of the greatest sig-
nificance for understanding ionic mobility in crystal lattices.
Lattice imperfections of this kind are:
248 INTERDIFFUSION OF SOLIDS

(1) Ions existing interstitially within the normal lattice.

This leaves holes in the lattice, and electrolytic conduction or
diffusion may proceed by jumps of the ions from one inter-
stitial position to another; or by diffusion of the holes.
(2) Some positions in the normal lattice are vacant, although
there are no interstitial ions. Again the holes may diffuse, by
jumping of ions from an adjacent lattice place into the hole,
leaving a second hole.
The equilibrium between holes, interstitial ions, and the
normal lattice is maintained at temperatures upwards of about
100-200° below the melting-point (43,44,45); but at low tem-
peratures the amount of disorder in the lattice depends upon
the history of the specimen. This non-equilibrium distribution
of points of disorder in the lattice is referred to as "irreversible
gitterauflockerung"* by von Hevesy(46). It may have the
most remarkable influence upon the ionic conductivity, and,
with the subject of grain-boundary diffusion, is discussed later
(Chap.VII). The equilibrium types of disorder existing in afew
lattices are especially interesting when considering models for
the diffusion process at high temperatures, such, for example,
as that given by Frenkel(47).

SOME EQUILIBRIUM TYPES OF DISORDER IN CRYSTALS

In silver chloride the current-carrying ion is the cation, and

therefore only the cations are in disorder (Fig. 87) (48). In
Fig. 87 the arrows indicate the possible processes of ionic
Ag+ CI- Ag+ CI- • CI- Ag+ Cl-
Ag+—* /
CI- Ag+ CI- Ag+/ CI- Ag+ CI- Ag+

Ag+ CI- Ag+ CI- Ag+ CI- Ag+ Cl-

Ag+
CI- • CI- Ag+ CI- Ag+ CI- Ag+
Fig. 87. Disorder in the silver chloride lattice according to Frenkel.

mobility. These are of two kinds corresponding to migrations

of holes and of interstitial ions respectively.There is an energy
* Gitterauflockerufig = lattice loosening.
 STRUCTURE OF REAL CRYSTALS 249

of activation for the types of jump illustrated in Fig. 87, as well

as anendotherrnieheat of formation of the interstitial positions.
OL-silver iodide^), silver mercury iodide^), and similar com-
pounds give the extreme example of the disorder illustrated
in Fig. 87, for they show an almost perfect anionic lattice,
but a nearly random distribution of cations within it. This
accounts for the observation (p. 240) that the ionic con-
ductivity of solid and molten silver iodide are nearly the same
(~lohm~ 1 cm.- 1 ). Diffusion in potassium chloride occurs,
according to Schottky, because of a disorder in both anion and
cation lattices (Fig. 88) (45) which does not produce inter-
stitial ions. The similar size of K + and Cl~ ions renders the
K+ Cl- • Cl- K+ Cl-

Cl- }£+/ Cl- K+ \ y K+

K+ Ci- K+ Cl- K+ Cl-

Fig. 88. Disorder in potassium chloride according to Schottky.

interstitial spaces too small to accommodate K+ or Cl" ions.

Instead, there are equal numbers of cation and anion vacant
spaces formed by movements of ions into the crystal surface
where they tend to build new lattice layers.
It has been found that a number of compounds exist which
do not rigidly obey the law of fixed proportions. In the spinel
group of compounds, of which MgAl2O4 is the type, certain
constituents of the lattice may be lost but the lattice still
retains its structure. Spinel itself may undergo the continuous
transition MgAl2O4 -> Mg2Al8/304 (y-alumina). The excess of
one component is due either to the existence of gaps in the
lattice where that component should be, or to an interstitial
excess of another component. The behaviour is shown by
certain oxides, sulphides and halides.
As an example of an oxide with metal excess one may take
zinc oxide (51). At 600° C. some oxygen has been lost by dis-
sociation, and a solid solution of zinc in zinc oxide remains in
which the metal is dissociated into cations and electrons. The
250 INTERDIFFTJSION OF SOLIDS

higher the oxygen pressure the less the zinc ion excess, and so
the smaller the conductivity. Cadmium oxide (51) behaves in
a similar manner.
Other oxides give systems with an oxygen excess (oxides of
iron, cuprous oxide, and nickel oxide(52)). Cuprous oxide, for
example, takes up an excess of oxygen as 0 " ions, as a result
of which vacant cation sites appear in the lattice. The electrons
for this process are liberated by the reaction Cu+ -> Cu++ + e,
and electronic conductivity occurs because electrons may be
supplied by the electron transfer of the above reaction. The
higher the oxygen pressure the more excess oxygen is dissolved
in the lattice (e.g. - 0 - 1 % at 1000° C. and at 30mm.Hg
pressure), and so the higher the conductivity. The sizes of the
ions (O" = 1*32 A.; Cu+ = 0-96 A.)do not allow one to arrange
the excess oxygen interstitially, and so the type of disorder
given in Fig. 89(48) is postulated. X-ray studies (53) of ferrous
oxide, sulphide, and selenide showed that empty 'cation posi-
tions similar to those in Fig. 89 may indeed exist.

Cu++ Cu+ • _ Cu+ Cu+ Cu+

O" O" O" O" O"
Cu+ Cu++ Cu+ Cu+ • Cu+
Fig. 89. Disorder in cuprous oxide with excess oxygen.

Cuprous iodide (54) and bromide (39) show in part an electronic

conductivity due to the solution of an excess of iodine or
bromine as ions, the electrons for the formation of the ions
being supplied by the reaction
Cu+->Cu++ + e,
Potassium iodide if pure has a small conductivity ionic in
type. The lattice may dissolve excess either of iodine or of
metal, and the conductivity is then altered by the presence
of iodine ions, or potassium ions:

The alkali halides in general (55) can behave as solvents for small
amounts of halogens, alkali metals, and even hydrogen. The
solute is not always in atomic or ionic form however (Chap. Ill).
 251

THE INFLUENCE OF GAS PRESSURE UPON CONDUCTIVITY

Most important from the viewpoint of lattice disorder are the
observations of Wagner and his co-workers upon changes of
conductivity brought about by surrounding the crystal by a
gas atmosphere of its electronegative component (oxygen for
oxides, or halogen for halides). These measurements were the
basis of the classification of lattices, as in the previous section,
into those with anion excess, cation excess, or stoichiometric
cation anion ratio.
As an example of the influence of pressure we may consider
0 2 -Cu 2 0 systems which contain excess of the electronegative
component oxygen. It may be supposed that Cu+ ions
release electrons, giving Cu ++ ions, and an equivalent number
of Cu+ ions diffuse from their lattice sites. These ions and
electrons react at the surface of the crystal with gaseous
oxygen, producing more crystalline Cu2O. One denotes Cu ++
by the symbol (e)h or electron defect site; similarly (Cu+)j is
a vacant lattice site. The corresponding occupied sites are
denoted by (e)g, (Cu+)a. The symbols "I" and "g" mean
respectively "leerstelle " or "vacant place ", and "gitterplatz "
or "lattice site". The reaction is then
O2(gas) + 4(e), + 4(Cu+), ^ 2(Cu2O), + 4(e),+ 4(Cu+),.
Since (C^O)^ is simply an array of (0—) g + 2(Cu+)g, the net
reaction is
O2(gas) + 4 (e), ^ 2 ( 0 ~ ) , + 4(e),+ 4(Cu+),.
Application of the law of mass action then gives

because (e)g and (O—)g are nearly constant. But (Cu+), = (e),,
and so . . rrr -,•

The conductivity, which in cuprous oxide is mainly electronic,

is thus proportional to the eighth root of the pressure of
oxygen. While the theory predicts an eighth root, experiment
252 INTERDIFFUSION OF SOLIDS

shows an approximate seventh root dependence of conduc-

tivity on pressure (Fig. 90), a satisfactory agreement.
Table 60, after Wagner(43), gives in a simple form all the
available information concerning lattice disorder and the
conduction process in certain oxides and halides. The use of
the table may be illustrated with reference to silver chloride:
Column 2. There are almost equal numbers of interstitial
cations and vacant cation sites in the silver chloride lattice.
The number of quasi free electrons or electron-defect sites is
negligible.

fOOO'C.

*900°C.

+02

-01

Fig. 90. The dependence of the conductivity of CuaO on the

oxygen pressure. (Dunwald and Wagner<52>.)

Column 4. Neither silver nor chlorine is in any appreciable

excess over their stoichiometric ratio.
Columns 5, 6 and 7. Increased partial pressure of chlorine
has no influence upon conduction due to cations, but may
alter the electronic conductivity. Electronic conductivity
(columns 2 and 9) occurs however only to a negligible extent.
Columns 8 and 9. Silver chloride is almost exclusively a
cationic conductor.
 TABLE 60. Types of lattice disorder, with examples if known, and characteristics of conduction of
electricity in corresponding lattices (Wagner(43))
Influence of increasing the partial pressure of Transport
Types of lattice* Component negative component on conduction due to numbers of
Number disorder possible Example present
in compound Me X m excess Cations Anions Electrons Cations Anions
la (Me+) = (e) ZnO Me Decreases Decreases <l 0
b (X'h' = (€)', 9 Me — Decreases Decreases 0 <^1
2a (X')z =(e), 9 X Increases Increases 0 <^1
b Cu26, NiO X Increases — Increases 0
3 (e)z = («h ] — — — — —
a ? Me Decreases — None 0
b {X')t Z<(efz \ CuO Me Decreases None 0 <^ 1
c {X)z + <^(e)i 1 '? X Increases None 0 <^ 1
d (Me )f<^(e)j J 9 X Increases — None 0
4
<o
a (e)z \*(Me+)H 9
Me None Decreases Zi 0
+
b (e)i ^(Me )j) AgCl X None — Increases ~i 0
5 (X')g ?a {X')t \ —
a (e)z <(X')t \ BaCla Me — None Decreases 0 '-"I
b (e), <{X')Z ) ? X — None Increases 0 -1

* "2" denotes "zwischengitter" or "interstitial"; "J" (see p. 251) denotes vtleerstelle", "vacant site", or "electron defect site*'.
254 INTERDIFFTJSION OF SOLIDS

THE ENERGY OF DISORDER IN CRYSTALS

The results of the previous sections lead us to consider im-

portant types of equilibrium ionic disorder to be:
Vacant cation sites, with an equal number of interstitial ions;
Vacant anion sites, with an equal number of interstitial ions;
Equal numbers of anions and cations in interstitial positions;
Equal numbers of vacant anion and cation sites.
The first three examples are known as Frenkel disorder,
and the last example as Schottky disorder. Calculations of
the energies of disorder (56) are likely to be of importance if
they give the absolute values of the energy of disorder from
known force laws and crystal parameters; and if they show
which type of disorder is most likely to be met with in
different salts.
Let E denote the lattice energy, that is, the energy needed
to dissociate 1 Mol. of the lattice into gaseous ions. Then E/No
is this energy referred to one ion pair of the lattice; it is con-
nected with the energy E8 needed to produce a Mol. of Schottky
lattice defects, and the corresponding polarisation energy,
Evol, released around the vacant sites, by E = ES + Evol. In
a preliminary calculation of E one may use the force law

in which the first term gives the Coulombic attractive force,

and the second the repulsive force.* Born's expression for
the lattice energy is then applied, and the lattice energy per
ion pair in sodium chloride for example is

JS0 a
where a denotes the lattice parameter, and e is the electronic
charge. In addition, the polarisation energy involvedis approxi-
mately equal to that when an ion is transferred from a. medium
* The repulsive exponent n is usually given a value of 9-13.
 THE ENERGY OF DISORDER IN CRYSTALS 255
of dielectric constant e to a medium of dielectric constant
unity, given for a pair of ions by

No a \ e)
Accordingly, the net energy of disorder per ion pair becomes

j\o a \ n] a
which is very much smaller than the lattice energy. However,
it must be remembered that many approximations are involved
in so simple a calculation as that above. The nature of these
approximations is now indicated:
(1) The force law used in calculating the lattice energy is
inadequate since it neglects contributions of van der Waals'
interactions, which may become appreciable in lattices where
the mass of the ions is large, or where one type of ion exists
largely in interstitial positions. Thus in a-Agl, or a-Ag2S, where
the cations are in almost complete disorder in an anion lattice,
van der Waals' forces are considerable.
(2) Interactions between induced quadrupoles also con-
tribute to the polarisation energy.
(3) When a vacant site is formed, surrounding ions will tend
to rearrange themselves slightly, and the corresponding energy
term must be allowed for.
(4) A better method of expressing the repulsive potential
replaces the term
B
E -

(r+4-r.-r)
by #reP. -be P ,
where r+ and r_ are the radii of a positive and a negative ion,
distant r from each other, and b and p are constants. Even this
equation is inaccurate at very small separations.
256 INTERDIFFTJSION OF SOLIDS

(5) The crystal cannot be regarded as a continuum, as was

assumed in calculating the polarisation energy, E^, for one
is dealing with phenomena on a molecular scale. Thus one is
uncertain what value to give to e and what the effects of
inhomogeneity of the medium upon the polarisation energy
may be.
Other refinements may be suggested (66), but enough has
been said to indicate the nature of the calculation which must
be made, and its difficulties. The energy of disorder may be
determined experimentally, however, from conductivity data.
Koch and Wagner (57) measured the conductivity of solutions
such as PbCl2, CdCl2, in AgCl, and of pure AgCl. The addition
of cadmium chloride will alter the number of vacant lattice
sites in a silver chloride lattice with which it forms a homo-
geneous solution. For in cadmium chloride the cation to anion
ratio is 1: 2, instead of 2:2 for silver chloride, and there is thus
one cation too few for every Cd ++ ion in the lattice, giving one
vacant site for each cadmium ion incorporated. The authors
have established that the electrical conductivity of solutions
of cadmium chloride in silver chloride is in the main determined
by the product of concentration and mobility of the vacant
sites, the concentration of interstitial ions being much smaller.
But the concentration of vacant sites is equal approximately
to the concentration of the cadmium chloride, and so the exact
analysis of the conductivity data permits one to find the
mobility of vacant sites, and by extrapolating to zero con-
centration of cadmium chloride to find the much smaller
concentration of vacant sites in pure silver chloride, as well as
their mobility. The results of this investigation appear in
Table 61 (48), and from the measurements at various tempera-
tures, by application of the van't Hoff isochore ^T — -pi™
the following energies of disorder were found:
AgCl: E ~ 25,000 cal./ion
AgBr: E ~ 20,200 cal./ion.
These energies show that the mathematical treatment of
 ENERGY OF DISORDER IN CRYSTALS 257
energies of disorder is correct in outline, since, as the theory
indicates, the energy of disorder is much less than the lattice
energy.
The question as to which of the two main types of disorder
(that of Frenkel or that of Schottky) will prevail in a given
system can be treated analogously by the mathematical theory.
Simple geometrical considerations are, however, sufficient to
show the conditions which decide the actual types of disorder.

TABLE 61. Concentrations and mobilities of interstitial

ions and vacant positions in silver halides*
Mobility of Fraction of ions
Temp. vacant sites in interstitial
Compound °C. in Ag lattice positions = fraction
cm./sec./v./cm. of vacant spaces
AgCl 350 6-6 x 10-* 1-5 x lO"2
300 4-2 x 10-* 5-5 x 10-*
250 2-3 x 10-* 2-2 x 10-*
210 1-5 x 10-* 8 1 x 10- 5
Agl 300 7-6 x 10-* 4 0 x lO"33
250 3-4 x 10-* 1-8 x 10-
210 2 0 x 10-* 7-6 x 10-*

* The concentrations are calculated assuming that the mobility of holes and
interstitial ions is the same. This assumption may lead to errors in the data
of Table 61.

If the ions displaced from their regular lattice sites are small
compared with the interstices, they may more easily exist in
interstitial positions, and Frenkel's disorder is possible. If the
anions and cations are of comparable radii, the lattice must be
highly distorted for ions to exist interstitially, and Schottky's
disorder becomes probable.

THE INFLUENCE OF TEMPEEATURE ON DIFFUSION IN METALS

AND CONDUCTIVITY IN SALTS

The earliest attempts to express conductivity data for salts

made use of power series in the temperature (T). Foussereau (58)
represented a number of specific resistances by the expression

17
258 INTERDIFFUSION OF SOLIDS

where a> b, and c are constants. Rasch and Hinrichsen(59) and.

Konigsberger(60) independently employed the equation

where K denotes the conductivity. Phipps, Lansing and

Cooke (61) followed Konigsberger's suggestion that the equation
was really of the form

where E is an energy term concerned as we now know with

the movement of the current-carrying ion or diffusing metal
atom from one position of minimum potential energy in the
lattice to another. All results are now expressed in terms of
Konigsberger's formula, although sometimes two or more
exponential terms are necessary when two or more ions
participate in the transport of electricity. The general and
very satisfactory applicability of the exponential law, both
for diffusion in metals and conductivity in salts, is illustrated
in Figs. 91-93. I t can be seen from these figures and from
Figs. 94 and 95 that one may classify diffusion and conductivity
systems into three groups:
(A) Substances which conduct by transport of one ion, or
where the diffusion constant, Z>, obeys a simple exponential
law D = Doe~EIRT. This is the normal and by far the most
numerous group (Figs. 91 and 92).
(B) Substances in which more than one exponential term
is important, according to the temperature. Thus the diffusion
constants of indium and of cadmium in silver (29), and the con-
ductivity of lead iodide (32) involve two exponentials. One may
write for lead iodide:

because both ions carry the current. At low temperatures,

one exponential predominates, and the behaviour is that of
class (A). Another salt which may follow a similar law is
silver chloride, for which Smekal(02) gives
KAgCl =
 Temperature in °C.
650 700 900

Fig. 91. The diffusion constants of various metals in silver

(Seith and Peretti(29>).

Temperature '/> *C >•

WO 200 300 WO 500 600 700 SOU 900

$///
-e

Fig. 92. The conductivities of a number of salts (Seithcm).

17-2
 Temperature tn CC
300 350 W>0

Fig. 93. The diffusion and conductivity data for lead iodide (Seith(39>).
#pbi, = conductivity of Pbl 2 .
Ky = conductivity of I' ions.
J£pb++ = conductivity of Pb" ions.
Dpb++ = self-diffusion constant of Pb" ions.

1
aCuA Melt
0 •ssl
j
-1
""
-2 "/

-3 y
$ /
'2 rCol 'r

/
-6
/
W ZOO* 300' SOFC.
, 94. The conductivity of cuprous bromide as a function of
temperature (Tubandt<39)).
 INFLUENCE OF TEMPERATURE 261
Where one has a pair of mutually soluble salts the behaviour
tends to be that of group (B). For example, in Table 62 (39) are
given the constants for the conductivity of CuBr-AgBr mixed
crystals. The conductivity may be represented by
K = KCu+ + KAg+ = ACn+ e - W + AAg+ e-
10*
Agl-
zo 18 f6

a-Agl

r
u s •4
1

Y
!

? 2i t 2

Fig. 95. Conductivity of silver iodide and silver mercury

iodide (Seith<39>)

TABLE 62. Constanta in the equation

K = ^ C u + e-^w + A^ e~E*iRT (ohm"1 cm."1)
for the conductivity of CuBr-AgBr mixtures

Composition
mol. % AgBr cal./ion Aca+ cal./ion

100 1-5 x 10* 20,600

90 1245 10,040 66 7540
80 850 8,460 7-3 3720
65 235 6,540 17-3 3960

Similar data for CuI-AgI mixed crystals are given in

Table 63(39).
The tables illustrate the loosening of the lattice by the
262 INTERDIFFUSION OF SOLIDS

addition of the component having the least tightly bound

conducting ion. For example, the energy term required for
the migration of silver ions in CuBr-AgBr mixtures falls from
20,600 to 6540 cal. when 35 mol. % of CuBr have been added.

Pbci2+o.oosm

15 2.0 PJ 3
103/T(Tin°K.)
Fig. 96. The influence of potassium chloride on the conductivity
of lead chloride (6yulai(63>).

TABLE 63. Constants in ihe equation

K = ACu+ e-EiiRT + AAg+ e~E*lRT (ohm"1 cmr 1 )
for tlie conductivity of CuI-AgI mixtures
Composition EAgJr
mol. % Agl cal./ion cal./ion
100 24-9 4600 _ _
95 36 5400 3-7 5400
90 26 4960 5-7 4960
70 10 3720 8 3720
50 6-3 3270 10-7 3270
30 2-8 2400 11-2 2400
0 — — 5-5 1190

The same loosening has been effected for lead chloride by

adding small amounts of potassium chloride, as Fig. 96<G3)
shows. It can be seen that the potassium chloride has reduced
 INFLUENCE OF TEMPERATURE 263
the energy needed for rendering mobile the current-carrying
ion, for the slope of the log K-ljT curve is greater for PbCla
than for PbCl2rf 0-005 KCl.
For KCl-NaCl solutions Smekal(62) proposed an equation
with three exponential terms, since all three ions may act as
current carriers:

(C) Some salts do not obey.any simple conductivity-tem-

perature law, as Fig. 94 illustrates in the case of cuprous
bromide, and Fig. 95 for silver iodide and silver mercury
iodide (Ag2HgI4). Each transition point shows a sharp break
in the conductivity-temperature curve, and the conductivity
of a-CuBr and a-Agl can actually be greater in the solid state
than in the molten state. In the case of cuprous bromide one
notes (Fig. 94) that in addition to the basic curve III some
irreversible curves are also shown (I, Ha, 116). This peculiarity
is due to the extreme sensitivity of the conductivity of cuprous
bromide to excess halogen, and to the difficulty of excluding
traces of such impurities, which are sufficient to raise the
conductivity by several powers of ten.
The extremely high mobility of silver ions in a-Agl and
Ag2HgI4 has been the subject of a number of researches. The
solution of their unique behaviour came from the studies of
Strock (49) and Ketelaar(50) on silver iodide and silver mercury
iodide respectively. Strock showed that while the anion lattice
was in perfect order the cation lattice was almost completely
disordered. The work of rendering the silver ions mobile is thus
very small, and their conductivity high. Ketelaar found that
the same property explained the conductivity-temperature
curve of Ag2HgI4.
a-Ag2S, a-Ag2Se and a-Ag2Te have also extremely high
conductivities, and it was some time before the explanation
of the conductivity values was forthcoming (64). Ultimately,
Wagner (65) pointed out that the conductivity of a-Ag2S was
sensitive to the sulphur-vapour pressure in the surrounding
264 INTERDIFFUSION OF SOLIDS

gas phase. Therefore, just as for Cu2O and similar oxides

(p. 251), the conductivity is in part due to electrons. The elec-
trons are supplied by reactions in the lattice such as
S"-> 8 + 2e, Se"-> Se + 2e, Te"-> Te + 2e.
The values for the conductivity K, and the constants A
and E in the formula K — Ae~ElRT for electrolytic conductors
in which the conductivity is predominantly by one ion, are

TABLE 64. Conductivity, and constants in the conductivity

formula K = Ae~EiRT (Seithm)

K (at the
Melting- melting- A E E
Salt point point) ohm"1 cm."1 cal./ion e.V.
1 1
ohm" cm."
LiF 842 0-6 x lO"2 4 xlO 77 51,000 L. 2-20
LiCl 606 1-5 x 10-* 5 xlO 38,000 L. 1-65
NaF 992 1-7 x lO"33 1-5 x 106 52,000 L. 2-25
fcaCl 800 1-3 x 10" 1 xlO«6 44,000 L. 1-90
NaBr 735 1-3 x 10- 33 1 xlO 41,200 L. 1-78
Nal 661 4 0 x lO" 1-5 x 10» 33,000 L. 1-42
KF 846 8 0 x 10-* 3 xlO 78 54,400 L. 2-35
KCl 768 2-0 x 10-* 2 xlO 47,800 L. 206
KBr 728 2 0 x 10-* 1-5 x 10« 45,600 L. 1-97
KI 680 1-5 x 10-* 3 xlO 5 41,000 L. 1-77
RbCl 717 5 0 x lO"55 3 xlO« 49,200 L. 212
RbBr 681 3-5 x 10~ 1-8 xlO 6 47,000 L. 203
TICi 427 5 0 x lO"3 2-5 x 103 18,320 L. 0-79
TlBr 457 5 0 x 10- 3 1-7 x 103 18,560 L. 0-80
AgCl 455 1 0 x 10"1 3 xlO« 22,200 T. 0-96
AgBr 422 6 0 x 10"1 3 xlO« 20,600 T. 0-89
a-Agl 522 25 5-5 1,186 T. 005
Ag2HgI4 4 xlO 2 8,600 K. 0-37
PbCl2 501 5 x 10- 3 6-6 10,960 S. 0-47
Pbl 2 402 3 x 10- 5 1-2 x 105 30,000 S. 1-30
9-8 x 10-* 9,360 S. 0-40
(I")

L. Lehfeldt. T. Tubandt. K. Ketelaar. S. Seith.

summarised in Table 64. Lehfeldt (66) showed that the energy E

in the conductivity equation K = Ae~EfRT depends upon the
radius of the halogen ions in the lattice to a remarkable extent.
His representation of this effect is given in Fig. 97. As the
 INFLUENCE OF TEMPERATURE 265
anion increases in size, the energy needed to render the
current-carrying ion mobile diminishes considerably. The
effect is to be traced partly to the increasing polarisability of
the anions with increasing radius. On the other hand, the
polarising capacity of the cations increases in the order
Rb < K < Na < Li. The polarisation energy liberated around
vacant lattice sites tends to offset the work of forming the
vacant site.

1 *%
K
22

\
2-0

\
s,
\
N I

\
F 0/* \
H
t6• f8
n o W 2-2
Radius of the anion in A.
Fig. 97. Dependence of the energy E in the conductivity equation
K = Ae'EIRT upon the anion radius (Lehfeldt<«6)).

THE IDENTITY OF THE CURRENT-CARRYING IONS

The method of the radioactive indicator affords a simple

means of determining whether the anion or cation is the
conductor in a given salt. One has only to compare the
expressions
K = Ae~EfRT, for the conductivity K,
and D = Doe-E/RT, for the diffusion constant D.
266 INTERDIFFTJSION OF SOLIDS

If the values of E are equal, the current carrier is the ion used
as the indicator. An example of this method has been given
elsewhere for lead iodide and lead chloride (p. 271).
The second method is the measurement of the transport
number, for which the experimental technique was de-
veloped by Tubandt and his co-workers (39). It was early
discovered (67,68) that Faraday's laws were valid for the salts
barium chloride and silver chloride, and thus that the current
carriers are ions. Tubandt and his school carried these in-
vestigations much further by pressing salt cylinders together
between metal electrodes and electrolysing the system. By
weighing the cylinders and electrodes before and after electro-
lysis the amounts of material transported were estimated
directly. However, it was shown that in many instances
threads of metal formed stretching from anode to cathode, so
that conduction soon became metallic. a-Agl did not behave
in this manner, and it was sufficient to coat the electrodes with
a protective layer of this salt to suppress the formation of
metal threads. With a cell arranged as below:
Pt / a-Agl / Ag2S / Ag2S / Ag2S / Ag,
it was possible to measure the total current, the weight of silver
deposited at the cathode, the weight of silver dissolved from
the anode, and the loss or gain of weight of any intermediate
silver sulphide cylinder. In this way it was found that the
transport number of silver was unity. Nevertheless the result
does not mean that Ag2S is a pure cationic conductor, but
only that any electrons in the silver sulphide lattice do not
enter the silver iodide lattice. At the Ag2S/AgI boundary
both Ag+ ions and electrons are liberated and removed, with
formation of excess sulphur :

The Ag+ ions move to the cathode, while the electrons move
to the anode. Therefore in the Agl phase all the current is
carried by Ag+ ions, and the transport number measured by
deposition of silver at the cathode is unity. The excess sulphur
liberated at the phase boundary Ag2S/AgI eventually reacts
 IDENTITY OF CURRENT-CARRYING IONS 267
with silver from the anode. Other examples of Tubandt's
method do, however, speak unequivocally for cationic or
anionic conductivity, especially if used in conjunction with
a supplementary method such as that described in the next
paragraph. The example of a-Ag2S serves to indicate the
type of difficulty encountered.
Finally, should the conductivity be sensitive to the pressure
in a surrounding gas atmosphere of a component of the crystal,
the investigations of Wagner and his school (51,52,43) have
suggested that part of the conduction may be electronic
(p. 251).
By the application of these methods the current-carry-
ing ions have been identified in the instances given below.
Inspection of these examples shows that in salts with ions of

Cationic Ionic and

Cationic Anionic and anionic electronic Elec-
conductors conductors conductors conductors tronic

AgCl PbFa Pbl 2 a- and /?-Ag2S Metals

AgBr PbCl2 Alkali halides a- and /?-Ag2Se Fe,O4
a-Agl PbBr2 near their a- and /ff-AgjTe PbS
AgNO, BaFa melting- a-Cul
a-Ag2HgI+4 (Hg++ BaCl2 points a-ZnO
and Ag ) BaBr2 a-Cu2O
Alkali halides a-NiO
below 500° C. a-FeO
a-FeS
etc.

different valency it is as a rule the ion of smallest valency which

migrates under the applied E.M.F. The extent to which
electronic and ionic conduction occur in the case of salts
conducting by the two mechanisms may depend very much
upon the temperature. Fig. 98(54) shows that in a-Cul the
conductivity becomes 100 % ionic at high temperatures, and
100 % electronic at low temperatures.
In the studies of the conductivity of metallic alloys (H2-Pd,
Au-Pb) it has been found that the alloying constituents may
move in the lattice under the impressed E.M.F., SO that the
conductivity is not really completely electronic. The move-
ments of dissolved hydrogen in palladium under an impressed
268 INTERDIFFUSION OF SOLIDS

E.M.F. have been used to measure the charge on the hydrogen,

and its mobility and diffusion constant (69) (Chap. V). The
transference numbers of hydrogen are of course small. Seith (70)
showed that if carbon is dissolved in iron it will move in an
electric field towards the cathode, so that it is positively
charged. The transport number for the solution Fe +1 % C

too

[60 /

/
J
.sW

20 /

200 260 360

Tempersture in °C.
Fig. 98. The percentage of ionic conduction in a-Cul at various temperatures.

at 1000° C. was ~10- 6 . The measurements were extended

to alloys of gold in lead(7i), and of gold in palladium or
copper (72,73), transport numbers being respectively 10~10 at
200° C. and 10"11 at 900° C. However, although ionic mobility
may be demonstrated, it is of very slight importance compared
with electronic conduction.

THE RELATION BETWEEN CONDUCTIVITY AND

DIFFUSION CONSTANTS

Nernst(74), and later Einstein(75), found a relation between

the diffusion constant in electrolytic solutions and the con-
ductivity. The relation is
D-RTB
 CONDUCTIVITY AND DIFFUSION CONSTANTS 269

where B denotes the steady velocity, or mobility, of the solute

under unit force. Von Hevesy and his co-workers (76,32),
Braune(77), Tubandt, Reinhold and Jost(78) made use of an
analogous expression in dealing with the conductivity of ionic
crystals. The results justified the use of the Nernst equation
at least qualitatively.
Wagner (79) made a quantitative calculation of the relation-
ships involved, using thermodynamic and kinetic properties
of the crystals. His treatment applied to mixed crystals of the
type Cu2S + Ag2S, PbCl2 -I- PbBr2, AgCl + AgBr. If one denotes
by Yt the equivalent fraction of the species i of valence Z i , then
Z U
V i i

where nt gives the number of gram ions of the substance i in

the solution. If the species "i" is cationic, the summation
ZZjfy refers only to the cations. If it is anionic, the summa-
tion refers only to anions. The concentration of the species
i in equivalents/c.c. is
z n
i i
°« = ~y~>
r

if V is the volume of the system. The mean velocity ut of an

ion of transport number Vi is given by

when ut is the velocity under a gradient of 1 V./cm., F is the

Faraday, and K the specific conductivity. Since a field of
1 V./cm. is equivalent to a force of -^ (^e) dynes on the ion
(e denotes the electronic charge), one obtains as an expression
for the mobility Bi9 defined as the stationary velocity attained
by the ion under a force of one dyne:
300^

If in a mixture of CujS + Ag^ one denotes silver, copper,

270 INTERDIFFUSION OF SOLIDS

and sulphur ions as species 1, 2, and 3 respectively, Wagner

showed by considering both forced and natural diffusion that

D __

In these expressions, /^A^,S denotes the chemical potential of

Ag2S, as defined by Gibbs. When the solution of Ag2S in Cu2S
is dilute, \vtS = „ v , and the expressions for Dx and D2 are
012 ^2-* 2

When the mobility of the anion is small compared with the

mobility of the cations*, the terms containing Bz may be
omitted, and the equations reduce to

D _2> *

for dilute solutions of one salt in the other.

The mobility B2 may be obtained from the transport number
of the cations in the solid solution by means of the relationship
300 U2K2
B9 =
Z2e F (Z2e)C2
The relationship gives, in the simplest case, the connection
between D and K,
jRT 300 UK
D

* In a binary salt mixture, with a common non-diffusing anion, elementary-

considerations of electrical neutrality throughout the crystal will also show that,
for natural diffusion, - - 5 - 1 = -~, and Dx -^ = - D2 — 2 , if the cations have the
dx dx * dx * dx
same valency. Thus DX=D2,-
 CONDUCTIVITY AND DIFFUSION CONSTANTS 271

In this connection mention should be made of Frenkel's

relation between K and D (Table 76). This is

In this expression Nx denotes the number of ions per unit volume,

and the factor of proportionality turns out to be nearly unity.
That the factor is actually small is illustrated by the expressions
for tha conductivity of lead iodide (Fig. 93), and for the self-
diffusion constant of lead ions in lead iodide as measured by
the method of the radioactive indicator. Thus*
30,000
K?y + = 1-15 x 105e KT ohm-1 cm."1,
30,000
DPh++ = 3-43 x 105e RT cm.8 day 1 .
For lead chloride, on the other hand, only anionic conduction
occurs, and the equations are
10,960
Kcl- = 6-55e RT ohm- 1 cm."1,

DPh++ = 6-65 x 105e **T cm.2 day 1 .

TABLE 65. Self-diffusion of Ag+ in Agl from conductivity

and diffusion measurement^
Temp. ° C. 454 500 551 594 651 701 744
Acaio cm.22 day-11 214 2-68 3-32 3-86 4-60 5-25 5-85
D(ob..) cm. d a y 1-53 2-00 2-48 2-88 3-46 3-98 419
a 0-71 0-75 0-75 0-75 0-75 0-76 0-72

Wagner's (79) treatment shows that the relationship

^o
»-¥'
is only approximate. Tubandt, Reinhold and Jost (78) used the
relation.

where a is a constant. They calculated the diffusion constant

of silver in silver iodide from the conductivity, using Nernst's
272 INTERDIFFUSION OF SOLIDS

equation, and compared the value so obtained with the sdf-

diffusion constant (see below). Table 65(78) shows the value
of a to'be about 0-74.

DIFFUSION CONSTANTS IN METALS AND IONIC LATTICES

It has been indicated how the diffusion constant in a salt may
be calculated from the ionic mobility of the current carrier
(p. 268), or from the use of a radioactive isotope as an in-
dicator (p. 244). Von Hevesy(3i) considers that it will be
possible to use the latter method to follow the self-diffusion
of numerous metals (Table 59). The method can be extended
in a few instances by using as indicators small quantities of
certain salts in solid solution in a closely related salt(77,78).
For example, (78) small amounts of CuCl or NaCl were dissolved
in AgCL These mixtures are all cationic conductors in the tem-
perature and concentration range investigated. The diffusion
constants 2)Na+ or DCu+ were measured, and also the conduc-
tivity, and transport numbers of each ion. These latter data
served to calculate the ionic conductivities. The self-diffusion
constant, DAg+, was found for pure AgCl by assuming the
correctness of the relation:
Specific Ionic Mobility* of Ag+ in pure AgCl
Specific Ionic Mobility of Na+ in mixed crystal
_ Self-diffusion Constant of Ag+ in pure AgCl
Diffusion Constant of Na + in same mixed crystal'
It was established that, at 238° C.,
£ W = 3-5x lO^cm.May- 1 ,
where the specific mobility ratio *= 25, and
D Cu+ = 2-1 x 10-* cm.2 day 1 ,
where the specific mobility ratio = 0*01.
Then the above relation gives for the self-diffusion constant
9 x 10~5 and 21 x 10~5 cm.2 d a y 1 respectively. The agreement
is satisfactory.
UK
• Defined by K/ = — *- * where Yx and C7X are the mol fraction and transport
number of species 1 in a crystal of conductivity K. For pure AgCl, K'Ag+ = K.
DIFFUSION CONSTANTS IN METALS AND LATTICES 273

The results of a great many investigations are summarised

in the following tables (Tables 66, 67). I t will be seen that the
diffusion constants in salts may roughly be divided into two
groups, in one of which the energy needed to render the ions
mobile is greater than lOk.cal., and in the other it is less.
In the second group of electrolytes fall those substances such
as a-Agl, Ag2HgI4, CuBr, whose conductivity is unusually
high. The Langmuir-Dushman expression (p. 298) for the
diffusion constant is often used as a means of correlating the
diffusion data. This equation gives for D

Noh
where d is the lattice constant, E the activation energy for
diffusion, and h is Planck's constant. The empirical nature of
this equation is stressed elsewhere (p. 299), but its applicability
in many cases is outstanding. In Tables 66 and 67 in the last
column are given the values of E calculated from D and d.
The point to be noted here is that for the salts of group I I of
Table 66 the equation does not hold (Smekal(62>), but that as
a rule when the activation energy is large the equation is
satisfactory.
Those diffusion systems which have been marked with an
asterisk in Table 67 represent metal pairs which form a con-
tinuous series of mixed crystals. As a rule in such systems
the value of Do is small, an important exception being self-
diffusion processes (Pb in Pb:D 0 = 5-1 cm.2 sec."1; Au in
Au:D 0 = 1-26 x 102cm.2sec.~1). The normal range of values
of Do lies between 10"1 and lO^cm^sec." 1 . There are ex-
ceptions, however; for example, the diffusions of silicon and
tin into copper give Do = 1-0 x 104 and 6-7 x 10s cm.8 sec.-1.
Do for the self-diffusion of bismuth, in a direction perpen-
dicular to the c-axis, reaches a value of (1-33—16-3) x 1045. It
is difficult to assess the reproducibility of some of the data,
since some., of the diffusion processes listed are structure
sensitive, but those for bismuth appear reasonably con-
sistent (see Fig. 100). It is to be noted that the higher the
BD 18
274 INTERDIFFUSION OF SOLIDS
activation energy the larger ihe factor Do, although there is
no simple relationship between them, and exceptions also
occur. Some of these exceptions are, however, for structure-
sensitive diffusions (Th-W, Mo-W) where diffusion does not
occur solely through the lattice. Here the number of channels
available for diffusion may*be restricted to grain boundaries,

TABLE 66. Diffusion constants of ions in salts according

to the equation D = Doe~EIRT
E (cal./ion)
calcu-
E lated by
System cm. 8 sec.- 1 cal./ion Langmuir-
Dushman
formula

Group I. Salts witliE> 10,000 cal./ion

Ag+ in AgCl(46) 23,000 19,300
Ag+ in AgBr(46) — 19,000 21,000
Na+ in NaCl(46) — 11,800 35,000
Cl- in NaCl(46) — 47,200 38,600
Pb+*<- in PbCl2(46,8o,32) 7-7 36,800 34,500
Pb++ in PbI2(46.8o,32) 4-9, 10-6 30,000 29,000
Cl- in PbCl2(46,32) — 6 11,000 18,000
Se" in a-Ag2S(62,8i> 67 xlO- 20,040 —
Ag+ in a-Cu2Te(80,78) 2-4 20,860 —

Group II. Salts with E< 10,000 cal./ion

I ' in PbI2(46,32) 9,300 4,300
Ag+ in a-CuJ (62) 4-5 xlO" 3 6,760 —
Li+ in a-AgI(80f62,78) 58-3 xlO" 54 4,570 —
Cu + in a-AgI(80,62,78) 16-3 xlO" 2,260 —
Cu+ in a-Ag2S(62) 46 xlO~ 5 3,180 —
Cu+ in a-Ag2Se(63) 15-5 xlO" 65 2,940 —
Cu+ in a-Ag2Te(62) 3-85 xlO" 2,660 —
Ag+ in a-Cu2S(62.8i) 32-7 xlO" 5 4,570 —
Ag+ ina-AgI(46) — 2,260 —

and so the factor Do becomes correspondingly small. On the

basis of Eyring's(99) theory of diffusion (p. 302) a large value
of Do implies a big entropy increase on passing into the
aotivated state, or a large disturbance in the lattice in this
state. The larger disturbances are usually found in self-
diffusion processes, where the chemical similarity between the
diffusing atom and the solvent is a maximum. One has in the
large and small values of Z>0 an analogy with the "fast"
 TABLE 67. Diffusion constants in metals according
to the equation D = Doe-E/RT
E (cal./
atom) from
System & Langmuir-
cm.2 sec.- 1 cal./atom Dushman
equation
*Pb in Pb(32r46,80) 51 27,900 24,400
/tf-Tl in Pb (46,80,82) 3-7 xlO- 2 21,000 22,400
/?-Sn in Pb (46, so, 82) 3-4 xlO- 1 24,000 23,200
Au in Pb (46,80,71,83) 4-9 xlO" 21 13,000 13,300
Ag ill Pb(80,83) 7-5 xlO- 3 15,200 —
Bi in Pb(80,82) 7-7 xlO- 1 18,600 21,900
Hg in Pb(so,84) 3-6 xlO- 19,000 —
Cd in Pb(8o,84) -21 xlO- 2 -18,000 20,000
Zn(9-58%) in Cu(80,22) 3-2 xlO" 2 42,000 41,000
Zn (29-08%) in Cu(80,22) 5-8 xlO- 3 42,000 38,000
Sn (10%) in Cu (85,86) — 2 40,200 40,000
*Au in Au(33) 1-26 x 10 51,000
*Pd in Au(80,87) 111x10-3 37,400
*Cu in Au(80,87) 5-8 xlO~ 4 27,400 —
*Pt in Au(80,87) 1-24 x 10-3 39,000 —
fTh in W (small grains) (65,88) 7-5 xlO" 1 94,000 96,700
fTh in W (large grains) (85,88) 4 1 xlO- 3 94,400 118,200
Th in W (volume) (35) 10 120,000 —
fU in W(85,89) 10 100,000 100,500
fYt in W(85,89) 0-46 68,000 70,100
fCe in W(85,89) 10 83,000 82,700
fZr in W(«5,89) 10 78,000 77,400
*fMo in W (polycrystal)(46,80,90) 5 x 10-3 80,500 —
*Mo in W (single crystal) (46, so, 90) 6-3 xlO- 4 80,500 —
N 2 in Fe(85,23) 107 x 10"1 34,000 38,100
C in Fe (80,23,9i) 4-9 xlO"1 36,600 36,700
fC in W2C(62) 108,000 —
Z n i n (Cu + 4 % Zn)(92) 1-57x10-3 34,100 —
Alin(Cu + 4 % Al)(92) 6-34 x lO"4 2 40,400 • —
Si in (Cu + 4 % Si)(92) 1-0 xlO 64,200 —
S n i n (Cu + 4 % Sn)(92) 6-7 xlO 2 54,000 —
*Cu in Ni(8o,93) 104 x lO- 3 35,500 —
Cu in Ag(29,80) 5-9 xlO~ 5 24,800 —
Sb in Ag(29.«o) 5-3 xlO" 55 21,700
Sn in Ag(29,80) 7-9 xlO" 21,400
In in Ag(29,80) 7-3 xlO" 5 24,400 —
Cd in Ag(29,80) 4-9 xlO- 54 22,350 —
*AU in Ag (80,94,95) 11 xlO" 26,600 —
5-3 xlO"64 29,800 —
*Pd in Ag(-so,87) 416 x lO" 20,200 —
Au in Ag-Au (92) 5-2 xlO" 64 29,800 —
Ag-Pd (20% Pd) in Ag-Au(92) 7-0 xlO" 20,200 —
Au-Pd in Ag-Au(92) 8-3 xlO~ 43 37,400 —
Cu in Ag-Au (92) 106 x lO- 27,400 —
Au-Pt in Ag-Au(92) 1-28 x lO- 3 39,000 —
fBi in Bi (J_c-axis)(96,80) (1-33-16-3) xlO445 137,000 —
JBi in Bi (|| c-axis)(96,8O) (2-2-6-5) x-10- 30,000 —
to 6 x 10- 4
Cu in Al(97) 2-3 34,900 31,400
Mg in Al(97) 1-5 xlO 21 38,500 29,000
Cu in Cu(33a) 11 xlO 57,200
Al in Cu(98) 1-2 xlO~ 21 37,500 —
Zn in Cu(98> 8 x 10" 38,000 44,000
Sn in Cu(98) 10 45,000 —
Si in Cu(98) 5-2 xlO" 25 39,950 —
Be in Cu(98> 4-5 xlO- 9 27,900 —
Cd in Cu(98) 3-5 xlO- 8,200 —
Z n i n C u + 20%Zn(98) — 31,000 38,500
A l i n C u + 16% Al(98) — 54,000 39,000
Interdiffusirig metals form a continuous series of solid solutions.
Diffusion may show structure sensitivity (see next Chapter).
18-2
276 INTERDIFFUSION OF SOLIDS

and " slow " reactions of chemical kinetics. In the former case
the kinetic theory suggests accumulation of energy through
many degrees of freedom, a Viewpoint corresponding with
a large disturbance of the lattice. A tentative explanation
of the influence of chemical and physical similarity ^)f
solute and solvent upon the diffusion is advanced on
p. 285.
It is also noteworthy that on the whole the activation energy
for diffusion and for conductivity (Table 64) in ionic lattices
is smaller than the activation energy for diffusion in metallic
lattices. It is probable that this difference may be traced to
the greater influence of polarisation forces in the case of salts,
and to the greater density and so closer atomic packing in
the case of many metals.

DIFFUSION ANISOTROPY

When diffusion occurs with different velocities along different

crystallographic axes one has the phenomenon of diffusion
anisotropy. It is of interest to see what examples can be found
of diffusion anisotropy in salts and metals. This property has
already been encountered in the sorption of gases by zeolites,
heulandite(ioo), for example (Chap. Ill, p. 102), giving a very
great anisotropy indeed, and being in one direction (perpen-
dicular to the anionic laminae of this layer crystal) quite
impermeable, and in two other directions easily but differently
permeable. It should be noted that anisotropy in diffusion
has never yet been observed in cubic crystals. Onefindsthat
diffusion of zinc into copper (ioi), of oxygen into copper-silicon
alloys (102), and of carbon (103) or nitrogen (103) into iron, occurs
with equal velocities in all directions.
Seith(82) showed that the conductivity of lead iodide could
be expressed by the relation
K = 9-78 x io-*e-9*°°tRT+ 1-15 x io5e-3O'oo°/*r,
in which expression the first term refers to iodine ion transport
and the second to the lead ion transport. The equation applied
 DIFFUSION ANISOTROPY 277
to a compressed pellet of iodide for conductivity in the direction
of the pressure, which caused the crystallites to orient them-
selves so that their c-axes coincided with the direction of
pressure. The conductivity as shown by the following figure
depends on the direction of diflFusion (96) in the crystal to a very
marked extent. Fig. 99 shows log K(K = conductivity) plotted
against l/T (T = temperature in ° K.), and the curves 2a and
26 are logK-l/T curves of different samples of lead iodide
perpendicular to the c-axis, while 1 shows this conductivity
curve parallel to the c-axis. The former curves have slopes of
about 9000 cal. and the latter has a slope of 30,000 cal. One
226 ° 250 ' 275'• 300° 325*350' 375° -WO'
-4

,^»—°""
-5 s
.—-—' 2*—-
-6
~" 36.-£*
*5 - "
-7
3a yy
-ff
21 20 19 18 17 16 15
t/Tx /0+ (Tin'K.)—**
Fig. 99. Diffusion anisotropy in lead iodide.

concludes that the iodide ion carries most of the current normal
to the c-axis, but the lead ion carries most of the current
parallel to it. The anisotropy is several powers of ten at low
temperatures, but is naturally temperature dependent, and
seems to vary somewhat for different specimens. The curves
3a and 36 are plotted from the two terms of Seith's equation (32)
given in this paragraph For the self-diffusion of lead ions
in lead iodide the following diffusion constants were obtained
(Table 68).
The table shows that diffusion anisotropy for lead ions may
exist, but is slight compared with the anisotropy of the iodide
ion. Like heulandite, lead iodide is a layer lattice, and it is
the layer structure which is responsible for its behaviour in
diffusion.
Both Warburg and Tegetmeier(io4) and Joffe(i05) observed
278 INTEBDIFFTJSION OF SOLIDS

TABLE 68. Self-diffusion of lead ions in lead iodide

(cm.2 day'1) (using ThB as radioactive indicator)
Number Temp. ° C. D (cm.2 d a y 1 ) Remarks
1 323 614 x 10- 66 J_ to c-axis
2 316 5 0 2 x 10~ ±
3 306 205 x 10-« _L
4 278 5-94 x 10- 7 ±
5 262 3-31 x lO"7 _L
1 317 2-50 x 10- 6 II
2 314 2-57 x 10- 6 II
3 300 3-26 x 10- 8 II ,
4 268 2-67 x 10- 7 II ,

anisotropy in the conductivity of quartz. Perpendicular to

the crystal axis the conductivity
225° 260*
is 104-6-fold greater than parallel 1 1
)
to the axis, although the tem-
perature coefficient was the same
in both directions. A third ex-
U1 /

1
ample of diffusion anisotropy is
provided by bismuth (96). Self-
diffusion parallel and perpendi- -7
cular to the c-axis, using ThC as
radioactive indicator, gave an t . .
anisotropy of up to 106-fold
(Fig. 100). The figure shows
-9
1
that parallel to the c-axis the
1
activation energy for diffusion is
30 k.cal., but that perpendicular ll
to it the activation energy is
/
137,000 cal.
Another study of the same
type (106) was made of the dif-
fusion of mercury into single
12
21 20
I" 19
i/Tx to f(Tin °K.) *
18

crystals of cadmium or fcinc. In Fig. 100. Self-diffusion in Bi.

(1) Parallel to c-axis, (2) per-
both cases the diffusion was a pendicular to c-axis (Seith<96)).
maximum parallel to the basal
planes and a minimum perpendicular to them. On the basal
planes mercury drops gave circular diffusion, while all other
 DIFFUSION ANISOTROPY 279
planes gave ellipses with the major axis parallel to the basal
plane. The ellipticity diminished as the temperature was raised.
Similarly, Spiers (107) observed, during the spreading of mercury
on tin, that the mercury diffused into ellipses.

THE INFLUENCE OF CONCENTRATION UPON THE DIFFUSION

CONSTANTS IN ALLOYS

When silver ions move in a silver halide lattice the concen-

tration of ions within the lattice is fixed by the number of
lattice points available to silver, per unit volume. When,
however, gold diffuses into lead, or lead into gold, concen-
tration gradients of one metal into the other are established,
and it is then necessary to find whether the diffusion constant
D varies with the concentration, in order to interpret the data.
This means that one must use not the equation -=- = B-^-z
^ dt dx2
but the equation -=- = ^-l D-^-1 and the method of
dt dx\ dx)
Matano(i08) (Chap. I). The concentration gradients may be
measured by any of the methods outlined for the normal Fick
law, the most usual being by spectroscopic, X-ray, or chemical
analysis of thin layers adjacent to the interface.
The application of the equation ~^r =
copper-nickel system then gives for the variation of D with
the percentage of copper the curve of Fig. 101 (i08). The value
of D does not alter very markedly until the copper constitutes
80 % of the alloy. Then, however, a rapid increase occurs.
Fig. 102 shows similar effects of composition upon the diffusion
constants in Au-Ni, Au-Pd and Au-Pt alloys. As the per-
centage of alloying metal increases the constants at first remain
fairly steady with decreasing gold content, but after a certain
threshold value a much more rapid increase is observed.
Similarly, Table 69 shows the diffusion constants of a number
of metals in copper at 750° C, along with other data which
will be discussed later. It can be seen again that the diffusion
280 INTERDIFFUSION OF SOLIDS

constants alter with changing concentration of the diffusing

metal.
The diffusion of Al, Be, Cd, Si, Sn and Zn in copper was
subjected to the Matano method of analysis by Rhines and
Mehl(98), with the result that in every instance the diffusion
constant was found to vary with concentration. In all cases
the diffusion in nearly pure copper was less rapid than diffusion

Al

60
;z
80 H» fO 40 60 SO 100
CONC. IN ATOMIC PERCENT AU

Fig. 101. Fig. 102.

Fig. 101. The diffusion constant as a function of composition for a nickel-copper
alloy (Matano(108)).
Fig. 102. The diffusion constant at 900° C. for the inter-penetration of Au and
Ni, Pd and Pt as a function of composition of the Au-Ni, Au-Pd, and Au-Pt
alloys (Mehl(92) after Matano(io8>).

in the alloy (Fig. 103) and rose at first slowly and then rapidly
with increasing amounts of the solute. I t is made apparent
from these data that the diffusion constant is generally con-
centration-dependent, and that the application of the normal
Fick equation to metal-metal systems can yield only an average
value of Z>, and may therefore lead to errors in evaluating Do
and E in D = Doe~B/RT. Many more diffusion studies, using the
Matano method, are essential before the relationships between
Do and E can be put upon a quantitative basis.
From measurements of D which have been derived by the
Matano treatment of the law -=- — -—-\D-^~\, trends in the
ct ox\ ox)
 TABLE 69. The diffusion of metals in copper

Zn from base Zn from base Al from base Si from base Sn from base Q
Metal composition 13-6% alloy 29-6% alloy 17-8% alloy 10-85% alloy 5-6% alloy
0% 4 % 10% 0% 4% 0% 4% 16% 0% 4% 8% O
10% 0% 4 % 5%
Zn Zn Zn Zn Zn Zn Al Al Al Si Si Si Sn Sn Sn Q
D x 10"10 cm.2 sec."1 at 750° C. 1-4 1-5 3-2 31 3-6 5-6 1-6 1-7 10-8 20 2-8 8-6 21 22-2 33-5 O
Activation energy, E — 34,100 — — 31,400 — — 40,400 — — 64,200 — — 56,000 —
Crystal habit of solute C.p.h. — — C.p.h. F.c.c. — — Diam. — — Tetrag. — —
Solid solubility at the peritectic 32 32 16 11 75
temperature in atomic %
Atomic radii (Gfoldschmidt) in A. 1-374 1-374 1*40 — 1-582
Melting-point of solute ° C. 419-4 419-4 660 1420 231-9

t9
GO
 60
REVISED D AT 800 °C
ALUMINUM AND COPPER •••
50
BERYLLIUM AND COPPER*
CADMIUM AND COPPER — (EXTRAPOLATED*
SILICON AND COPPER^"
40
TIN AND COPPER

8 10 12 14 16 IS
CONCENTRATION IN ATOM PER CENT

Fig. 103. Selected average diffusion coefficients interpolated to 800° C. for

Cu-Al, Cu-Be, Cu-Cd, Cu-Si, Cu-Sn and Cu-Zn systems (Rhines and
Mehl(98>).

60000

50000

40000

30000

ALUMINUM AND COPPER

BERYLLIUM AND COPPER <
20000
CADMIUM AND COPPER
SILICON AND COPPER -
10000 TIN AND COPPER
ZINC AND COPPER

6 8 10 12 14 18
CONCENTRATION IN ATOM PER CENT
Fig. 104. Relationship between E and the concentration for the Bystems
Cu-Al, Cu-Be, Cu-Si, Cu-Sn, Cu-Zn (Rhines and Mehl 08)).
 INFLUENCE OF CONCENTRATION 283
activation energy for diffusion may also be obtained. These
trends (Fig. 104) in E with concentration are usually initially
slight, but may become large and rapid at higher concen-
trations. The data of Mehl(92) for the system Al-Cu follow an
approximate relation E = EQ[l + const, x (cone, of Al)2]. The
values of the energy of activation for the diffusion process can
be seen in Fig. 104(98) both to increase and to decrease with
increasing concentration of the alloying material. In some
instances, therefore, the solute is more easily loosened in the
lattice by the alloying process, in others the converse is true.

SUMMARY OF FACTORS INFLUENCING DIFFUSION CONSTANTS

Some factors which influence the diffusion velocity and con-
ductivity have already been indicated:
(a) It has been seen that, when certain types of disorder
exist in a crystal lattice, an increasing partial pressure of the
electronegative component may alter the conductivity (O2 in
FeO, Cu2O, NiO, ZnO, CdO; S in FeS; Br2 in CuBr; I in Cul).
\b) The constants D, Do and E in the equation
D = Doe-E'RT
have been shown to depend very much upon the concentration
of solute in the solvent metal.
(c) The importance of polarisability has been observed. The
data of Table 66 showed how a high polarisability of ions- in a
lattice may reduce the term E in the conductivity equation
K = Ae-ElRT. In polarisable salts (Agl, Ag2S, CuBr) the
activation energy E is small (< 10,000).
(d) Lehfeldt's diagram (Fig. 97), showing the connection
between E for a series of alkali halides and the radius of the
halogen ions, indicated that with increasing radius of the halide
ion the energy E diminished; and with increasing radius of
cation E increased. This relation may also be due in part to
polarisation.
It is now interesting to see what other factors can alter
diffusion constants. Tables 70, 71 and 72 give the diffusion
 to

TABLE 70. Diffusion of metals in lead

Pb (self-
Metal Au Ag Cd Bi -Tl -Sn diffusion)
D cm.2 sec."1 at 285° C. 4-6 x 10-« 9 1 x 10- 8 2 x 10-» . 4-4 x 10- 10 3-1 x 10- 10 1-6 x 10- 10 7 x 10"11
Activation energy 13,000 15,200 18,000 18,600 21,000 24,000 28,000
Crystal habit of solute F.c.c. F.cc. C.p.h. Rhombic F.cc. Tetrag. F.cc.
Maximum solid solubility 005 012 17 35 79 29 100
(atomic %)
Atomic radii (A.) after Gold- 1-44 1-44 1-52 1-82 1-71 1-58 1-74
schmidt
Melting-point of solute ° C. 1062 960 321 271 303 232 327

O
TABLE 71. Diffusion of metals in silver
O
Metal Sb Sn In Cd Au Pd
/•*

)LIDS
D cm.2 sec."1 at 760° C. 1-4 x 10- 9
2-3 x 10-* 1-2 x 10"» 9-5 x 10- 10
3-6 x 10- 10 2-4 x 10- 10
Activation energy 21,700 21,400 24,400 22,350 26,600 20,200
Crystal habit of solute Rhombic Tetrag F.c, tetrag. C.p.h. F.c.c F.c.c.
Maximum solubility of 5 12 19 42 100 100
solute in atomic %
Atomic radii (A.) after Gold- 1-614 1-582 1-569 1-521 1-438 1-372
schmidt
Melting-point of solute ° C. 630 231-9 155 320-9 1063 1555
 FACTORS INFLUENCING DIFFUSION CONSTANTS 285

constants of metals in lead, in silver, and in noble metals (92).

Table 69 (p. 281) gives the same data for copper (92). Examina-
tion of these and other data indicates the following properties
of diffusion systems:
(i) The melting-point and atomic radius of the solute show
no direct connection with the diffusion constant in lead and
silver (Tables 70 and 71), since the trend is in opposite directions
for lead and for silver.
(ii) The diffusion constant is smaller the greater the melting-
point of the solvent.
The influence of the melting-point of the solvent upon the
diffusion constant is large, and there exists in a number of
instances a well-defined relationship between the energy E in
the equation D = Doe~EIRT and the melting temperature:

System Cu in Au CuinNi MoinW

Melting-point of solvent temp. ° K. 1356 1728 3743
20-2 20-6 21-3

The influence of the melting-point upon the diffusion constant

was the basis of attempts by Braune (94) and by van Liempt (109)
to incorporate the empirical relationship of Table 76 in
equations for the diffusion constant D = Doe~E/RT by writing
E = 62(Tm/T), where 6 is a constant. These formulae will be
referred to later (p. 301).
(iii) The diffusion constant depends inversely upon the solid
solubility, being least for metals which form a continuous
series of mixed crystals, or for self-diffusion.
The influence of solubility may be explained in the following
way (cf. p. 274). When the atoms of solute and solvent are
identical, the solute occupies a lattice site in the crystal without
distorting the lattice. When the atoms of solute and solvent
are dissimilar, the solute atom distorts the solvent lattice until
in the extreme case it is thrown out of solution. The difference
in degree of disorder between the normal and transition states of
the solute atom before and in the act of diffusing respectively
 to

TABLE 72. Diffusion of metals in Ag (r = 1-642 A., melting-point 960-5° C.)

and Au (r = 1-438, melting;point 1063° C.)
H
Au Au-Ag Ag-Pd Au-Cu Ni Pd
Diffusion (diffusing alloy alloy Au-Pd Pt Au-Pt
system in Ag) alloy 20% Pd 10% Cu (inAu) (inAu) alloy (in Au) alloy

D cm.2 sec."1 M6xlO-9 1-6 x lO"8 11 x 10-» 5 0 x 10~» 0-34 x 10- 10 0-22 x lO"10 1 0 x 10- 10 0-16 x 10- 10 0-77 x lO"10 CO
at 900° C. Nirich Pd rich Ptrich M
6 0 x 10- 10 2-3 x 10- 10 0-61 x lO"10 o
Au rich Au rich Au rich
Activation , 29,800 26,600 20,200 — — — 37,400 — 39,000
energy
Atomic radii 1-438 1-244 1-372 1-385 CO
(A.)afterGold- O
schmidt
Melting-point, 1063 1452 1555 1755 —
°C.
 FACTORS INFLUENCING DIFFUSION CONSTANTS 287

is therefore much less when solute and solvent are dissimilar

than when they are similar. The additional energy needed to
loosen the lattice sufficiently for diffusion to occur is accord-
ingly less for dissimilar than for similar atoms. At the same
time the entropy of activation is greater for the similar than
for the dissimilar atoms. This entropy of activation is defined
by (p. 303)
D = Do = JL,

fid Ag(Au) Cd Jn J/»

Fig. 105. Diffusion constants of metals in silver at 800° C.

(D in cm.* day""1.)

where d denotes the mean free path (identified normally with

the lattice parameter), h is Planck's constant, and k is Boltz-
mann's constant. These two effects act in opposite directions,
but the net result is a diminution in D with increased solid
solubility.
Data giving the diffusion constants of a number of metals
in lead are given in Fig. 106(46,92). Here also one sees the
diminishing diffusion rates with increasing solid solubility. An
interesting method of representing the diffusion data in lead
and in silver has been adopted by Seith and Peretti(29), and
the data given (Fig. 105(66)) illustrate in another way this
dependence of the diffusion rate upon the mutual solubility.
These figures also show definite trends in D with the position
288 INTERDIFFUSION OF SOLIDS

of the solute and solvent in the periodic table. This position

governs in part the mutual solubility of the diffusing elements.
It may be considered as surprising that so far no relationship
between diffusion velocities and the atomic radii of solute and
solvent has emerged. That such relationships may exist has,
however, been shown by Sen (HO),
who demonstrated the following
rule. In a pair of solids M and
N the direction of most rapid
diffusion is from M to N when
the minimum distance of ap-
proach of atoms in N is greater
than the same distance in M. The
rule is illustrated by Table 73. It
is thus to be anticipated that
small atomic size of solvent will
usually favour slow diffusion, PO 19 18 17
unless the solute diffusing is of
even smaller atomic radius. Fig. 106. Diffusion constants of
various metals in lead (D in
Valence and polarisation ef- cm.2 sec.""1).
fects are observed in metals as
in ionic lattices, especially where zinc, cadmium, mercury, and
thallium act as solvent metals. Similarly with copper, silver,

TABLE 73. Effect of atomic radius upon direction

ofdif
Minimum distance of Direction of most
System approach in cm. rapid diffusion
Cu-Pt Cu : 2-54 x 10"88 Pt : 2-7J8 x 10- 8 Copper into platinum
Cu-Zn Cu : 2-54 x 10" Zn:2-67 x 10"88 Copper into zinc
2-92 xlO"
Fe-Ag Fe : 2-54 x 10"88 Ag : 2-876 x 10~88 Iron into silver
Au-Pb Au:2-88xlO- 8 Pb:3-48 xlO" Gold into lead
Fe-C Fe : 2-54 x 10~ C : 1-50 x 10- 8 Carbon into iron

gold, and some B subgroup metals with atomic radii not very
favourable for solid solution, and with which valence effects
occur also to a certain extent, one obtains numerous alloy
phases. The metals of group 8, on the other hand, have atomic
 FACTORS INFLUENCING DIFFUSION CONSTANTS 289

radii favouring solid solution, so that electrovalence effects are

subordinated and intermediate phases not so numerous.
Although in metals ionisation and deformation are hard to
measure, it has already been indicated (p. 267) that Au-Fe,
Au-Pd, and C-Fe systems may be electrolysed so that there
is definitely polarisation of the constituents. FrenkeFs
theory (47,44) (p. 293) of ionic conductivity illustrates the great
importance of polarisation forces in decreasing the energy
needed to render an ion mobile in an ionic lattice. The
polarisation is a maximum when the electron affinity of the
cation is large, and of the anion small. In Fig. 107 we may,

Fig. 107. The relative mobilities of silver ions in various silver compounds,
arranged around a parabola. On the right is the relative electronic con-
tribution to conductivity.

after von Hevesy (40), arrange the mobilities of silver in a series

of different lattices (ionic and metallic) around a parabola.
It is seen that the ionic mobility rises strongly from fluoride
to telluride as the electron affinity of the anion diminishes?
or its polarisability increases. On the right-hand side of the
parabola it is shown that the small self-diffusion rate of silver
in silver is raised by the addition of tin or antimony, while the
electronic conduction is decreased. Similar considerations
apply when the anion is the same and is combined with various
cations.
The increased mobility of silver in a silver-antimony alloy
as compared with a pure silver lattice may be considered as
due to the loosening of the silver lattice by distortion due to
introducing antimony. Von Hevesy (40) suggested that as a
BD IQ
290 INTERDIFFUSION OF SOLIDS

qualitative measure of lattice loosening in ionic compounds

one might employ the ratio
Conductivity above the melting-point
Conductivity below the melting-point'
The ratio is great in many instances, but in a few cases
(a-Agl, a-Cul) it is less than unity (Table 74). As another
method of comparing the relative loosening of two metallic
lattices one may use the ratio
Self-diffusion constant for first metal
Self-diffusion constant for second metal*
This ratio for lead and gold at 326° C. is 26,000, and one sees
that the lead lattice is loosened but the gold lattice is not at
this temperature.

TABLE 74. The ratio of conductivities above and below

the melting-point
Ag+in Ag+in Ag+in Li+in Na+in Na+in
System AgCl AgBr Agl LiCl NaCl NaN0 3
Conductivity ratio 16 2-5 0-5 10 1-5 10
above and below xlO 4 xlO 5 xlO 5
melting-point

One of the most interesting phenomena allied to the problem

of diffusion in metals is the order-disorder transformation in
alloys. The terms order and disorder are being used in a
different sense from that previously used when discussing
vacant sites and interstitial ions in a lattice, as will be seen
when the transformation is described. When a binary alloy
with constituents in the ratio 1:1 or 1:3 is cooled, the melt
solidifies to give a crystal. At these high temperatures all the
atoms, though regularly arranged in space, are interchanged
at random through the lattice. A given lattice point is as
likely to be occupied by an atom of one kind as the other. As
the lattice is cooled a reorganisation occurs which sets in at'
a fairly definite temperature, in which certain of the lattice
sites tend always to be occupied by one type of atom, and
 FACTORS INFLUENCING DIFFUSION CONSTANTS 291
other definite sites by the second type of atom. In the alloy
Cu3Au, for example, at high temperatures Au and Cu atoms
are distributed at random in the face-centred cubic lattice,
while at^low temperatures gold occupies cube corners and
copper face centres. The transitions may be followed by
resistance or X-ray measurements, and as a result of many
studies (noa) it appears that the process of ordering may not
reach its equilibrium state over a long interval of time at
temperatures below the critical temperature of the trans-
formation. This is because small zones of order are set up
through the crystal which are out of phase with each other.
The incorporation of one zone in another to give a homo-
geneous ordered phase is then very slow.
In any one zone undergoing the transformation disorder to
order the time, r, required for a given fraction of the trans-
formation should be
r = AeElRT,
where A is a constant and E is the energy barrier which must
be surmounted before the reorganisation occurs. This energy
barrier must be very similar to that which occurs during
diffusion, for in both cases a momentary mobility must be
imparted to the atom. Owing, however, to the numerous
antiphase nuclei which form through the crystal, the simple
expressionr = AeEIRT is not valid (iio&), for the slow coalescence
of antiphase nuclei is occurring simultaneously, and one cannot
measure the change in the separate nuclei independently.

MODELS FOR CONDUCTIVITY AND DIFFUSION

PROCESSES IN CRYSTALS

From a number of different viewpoints expressions have been

derived for conductivity or diffusion constants in crystals.
A number of types of conduction may be recognised as
theoretically possible, and examples of some of these are well
established. It is now convenient to review some of the
models for diffusion and also equations for the diffusion
constant which have been derived by a number of workers.
19-2
292 INTEBDIFFUSION OF SOLIDS

An important type of ionic diffusion is that which

occurs in zeolitic structures. In zeolites one has an anionic
framework with large interstices in which exist the cations
necessary for electrical neutrality. The anionic framework is
sufficiently open to allow cations to pass through it, along
cation channels, without any appreciable loosening of the
anionic network. Thus one may often exchange one cation for
another in these interstitial compounds. I t is probable that
the diffusion of hydrogen as atoms or ions in the metallic
lattices of palladium and similar metals may also be referred
to this type. This is also the mechanism by which gases flow
into zeolites, or through silicate glasses and organic mem-
branes. Von Hevesy (46) regarded the very rapid diffusion of
gold into lead as zeolitic, and indeed considered that most
examples of rapid diffusion in pure metals occurred by zeolitic
diffusion. The process of amalgamation may perhaps be
regarded as a zeolitic diffusion, accompanied simultaneously
with a disintegration or reorganisation of the solvent lattice
into amalgams. In order, however, for diffusion in metals to
be truly zeolitic the solute atoms should be small enough to
fit interstitially between the solvent atoms.* It is conceivable
that a metallic lattice could be formed in which one kind of
atom was freely mobile, while the other kind of atom formed
a more rigid lattice framework. Such a system would compare
with the anionic disorder in lattices of Agl, or Ag2HgI4 (p. 263).
A picture which is often given of the process of inter-
diffusion of metal pairs supposes place exchange by a thermal
loosening of the lattice sufficient for the molecules to pass
round each other, as distinct from Frenkel's mechanism (p. 293).
It has been illustrated in Fig. 108. This model has not passed
uncriticised. Bernal(iii), for example, considered that the
activation energy needed for the loosening would be too
great but that diffusion might occur by spontaneous small
gliding processes along different planes. As with the Frenkel
mechanism, however, the polarisation energy may consider-
* The atomic radii of gold and lead are respectively 2-88 A. and 3-48 A.,
probably not sufficiently different for interstitial solution.
 CONDUCTIVITY AND DIFFUSION PROCESSES 293
ably diminish the energy needed to create the momentarily
disorganised lattice.
On the quantitative side one may indicate the theory of
volume diffusion and conductivity in ionic crystals due to
Frenkel(47) and extended by Jost(44,56). The theory is based
upon the concept of equilibrium disorder introduced in a lattice

ooooo ooooo
ooooo oo o oo
OOOOO OO ® x OO
ooooo oo * @ oo
ooooo oo o oo
ooooo ooooo
Undisturbed lattice Lattice momentarily loosened to
permit place exchange of
molecules A and B
Fig. 108. A place exchange process as pictured in a crystal lattice.

O
• b •

Fig. 109. Frenkel disorder in a two-dimensional lattice.

by the thermal energy (p. 248). Thus in Fig. 109 it may be

supposed that the atom or ion at the position (a) has gained
sufficient energy for it to migrate to the position (6), while at
any temperature the average number of vacant sites and inter-
stitial atoms or ions is constant and in equilibrium with the
nortnal lattice. Diffusion can ensue by two mechanisms: an atom
may move over a potential energy barrier Ex into the vacant
294 INTERDIFFTJSION OF SOLIDS

site, giving a diffusion of the vacant sites; or an interstitial

ion may move over an energy barrier to another interstitial
position. When
n = the number of vacant sites/unit volume,
N = the number of lattice sites/unit volume,
Eo = the energy needed to create a hole,
Frenkel(47), by kinetic theory, and Jost(44,56), by statistical
mechanics*, showed that n ~ Ne~Eol2RT. Only a fraction of these
holes, or corresponding interstitial atoms, will diffuse how-
ever, because an activation energy is necessary for diffusion.
Thus the number moving is proportional to e~^E^2E^llRT.
An approximate value of Z>, the diffusion coefficient, can be
deduced if it is assumed that there are six ions around each
hole, distant d from its cejitre, and capable of moving with
the mean thermal velocity in any one of six directions. Each
of the six particles may move in a single direction (to the
hole), so the six of them are equivalent to a single particle
free to move in all directions. Therefore the diffusion constant
for a hole is

and the diffusion constant for all the ions is

since 6n/N is the fraction of ions around the holes. Then

2
In this equation \dv = eScm^/day- 1 , when d = 3A., and
v = 5x 104cm./sec. The case of forced diffusion (electrical
conductivity) may be obtained by using the Einstein equation:
D =; BUT (B denotes the mobility)
and also K = N(Ze)*B,
* For Schottky disorder the analogous relationship is n=a.Ne~~Bo!R1. The
exponential does not contain the factor J, and a may have values as high as 10*.
 CONDUCTIVITY AND DIFFUSION PROCESSES 295

where K denotes the conductivity, and N the number of ions

per unit volume of valency Z, and charge Ze. Then

K=

The factor before the exponential has a magnitude similar to

that for Z>, and normally one may write
K = (10-100) e-<E»+*EdlmT.
We have now to consider how accurately this theoretical
expression predicts the experimental findings. The value of
(E0 + 2Ex) might, at first sight, be thought to be comparable
with the lattice energy (100-200 k.cal.); whereas the experi-
mental values are instead from 1 to 50 k.cal. The suggestion
by Smekal(H2) that the mobility occurs only along internal
surfaces and outer surfaces leads to impossible values of the
constant A in the equation

Reasonable values of the quantity (Eo -f 2Et) may, however,

be computed for lattice ion jnaovement if the polarisation
properties are also considered, for it has been found that
polarisability and polarising properties and conductivity are
connected (cf. pp. 254 and 265) (114). One may proceed to
calculate the energy, Eo, involved in the formation of a hole,
and removal of an ion to an interstitial position, in a
manner analogous to that outlined on p. 254. To |2?0 must
then be added Ev the energy required before the hole, or the
interstitial ion, can migrate. Calculations of this kind have been
made by Jost(44). He determined Eo for NaCl by evaluating
the following energy terms:
(i) Coulombic energies in normal and displaced positions,
(ii) Repulsive energy in the normal and displaced position,
(iii) Polarisation energies around the vacant site and the
displaced ion.
The calculation is subject to the limitations outlined on
p. 255. The next step is to find El9 the energy of activation
296 INTERDIFFUSION OF SOLIDS

for diffusion of either a vacant site or an interstitial ion. The

term \E0-\-E1 might then be compared with the experimental
figure for \E0 -f- Ex obtained from conductivity data (44 k.cals.).
The theory indicates a qualitative agreement with experi-
ment.
Certain salts (e.g. a-Ag2S; D = 10 xe - w <v r cm.2 day- 1 ) have
a very high electrolytic conductivity; %(E0 + 2E1) is only
3220 cal. For this salt the polarisation properties reduce
i(E0 + 2E±) to a very low value. Calculation gives:
Eo = energy of "hole" formation = E^^om* + JS?rep# + EvoL
= (2-31 - 0 - 1 1 - 2-15) e2/a
= ~5-7k.eal.
A similar approximate computation of El9 the potential
energy needed for migration from a displaced position,
~" 1> "~ h ~~ h t>° a displaced position + J, + \, +\% gives
Ex ~0-le 2 /a~10-7k.cal.
and thus \(EQ + 2EX) - 13-6 k.cal.,
which may be compared with the experimental value 3220 cal.
Qualitatively, if not quantitatively, the Frenkel theory can
thus account for the exponential term.
The constant A in the equation
K = Ae~<Eo+*ELy2RT,

to which the theory of Frenkel gives a value 2 x 101 to 2 x 102,

actually takes experimental values of 10 > 4 > 106. To explain
this discrepancy three possibilities arise:
(i) By analogy with the theory of thermionic emission it
may be assumed that there is a temperature coefficient to the
quantity $(E0+2E1) ,which has been neglected in the Frenkel
treatment outlined.
(ii) A molecule having once acquired the energy %(E0 + 2EX)
may retain it while describing a mean free path far greater
than corresponds to a single displacement.
(iii) The activation energy may be stored through many
degrees of freedom (p. 300).
 CONDUCTIVITY AND DIFFUSION PROCESSES 297

Braunbek(H5) developed an analogous expression for the

conductivity of an ionic solid. Using sodium chloride as his
model lattice, he assumed a simple linear vibration of sodium
ions, while the chlorine ions were supposed to remain fixed.
Each sodium ion was situated in the mid-point of an octahedron
of chlorine ions, and the sodium ion on acquiring sufficient
energy could move from the centre of its octahedron through
the mid-point of one of its faces, and enter a new vacant
octahedron whose central sodium has diffused away by a
similar process. The probability of occurrence of such a process
during a vibration was calculated, and from this the self-
diffusion constant, and the conductivity K:

where e = electronic charge,

d = 5-63 x 10~8cm.,
r = the vibration period, ~ 2 - l x 10~13sec.
The quantity E was regarded as comparable with the energy
of melting of the lattice. This equation should however also
. ,. . n No. of vacant Na+ ion sites ,.,
contain the ratio ^ = ^ , XT TKTT-' ^ > which
N Total No. of Na+ ion sites
gives the probability that the second octahedron will be
empty to receive the migrating Na + ion. The agreement
with v. Seelen's(116) data is thus probably fortuitous.
Cichocki(H7) derived an expression for the self-diffusion
constant in ionic or metallic lattices. He used a body-centred
cubic lattice as his model, and then endeavoured to calculate
the probability that a given atom would have the requisite
energy and direction of vibration to move to a new position,
at a time at which the atoms surrounding that position have
the energy and direction of vibration to make an adequate
interstice for it. His theory is interesting as an attempt to
allow for correct timing in the diffusion. An analogous treat-
ment by Dorn and Harder (H8), in which the crystal is regarded
as a periodic system of energy wells, is less satisfactory. It is
298 INTEBDIFFUSION OF SOLIDS

not possible to assume constant potential energy walls about

the hole in which the solute atom resides. Sometimes it may
be almost impossible for the atom to escape in a given direction,
and sometimes it may readily be able to do so. The height of
the surrounding energy barriers fluctuates with time according
to the differences in phase of vibrations in surrounding atoms
in the lattice. Cichocki's formula, which allows for this, is
nevertheless open to the same objection as Braunbek's (p. 297).
Langmuir and Dushman(ii9) proposed a semi-empirical
equation for diffusion in cubic lattices, which has proved
a useful guide to the behaviour of diffusion processes in
ionic and metallic lattices (Tables 66 and 67). I t was derived
by considering the lattice as composed of layers of atoms in
planes a distance d apart, where d denotes the interionic or
interatomic distance. I t was assumed that d was also the mean
free path of a diffusing ion or atom. The number of atoms per
unit area is then d2, and the chance that an atom will leave this
area in unit time is kd2. By analogy with an early expression
for the reaction velocity constant (120), Langmuir and Dushman
hvlkT E 2 EfRT
wrote k = ve~ y and D = = - r d e~ , where v is t h e

vibration frequency of the solid, Nohv = E> and h denotes

Planck's constant, and the other terms have their usual
significance. The Langmuir-Dushman equation is to be re-
garded as empirical, the frequency v being fictitious, as the
following correct derivation of the diffusion constant shows.
Suppose two salts which form solid solutions of cubic symmetry
are interdiffusing, salt A passing in the -f x direction. Draw
planes at a;and x 4- d (Fig. 110), where d is the mean free path of
an activated molecule, and also at x - \d and x + \d, normal
to the ^-co-ordinate. All salt A ions in the region x — \d to
x -f \d can if they acquire the necessary activation energy with
suitable direction (4- x) pass the intermediate plane x + \d. The
chance that an activated ion will move in the + x direction is
one-sixth the chance it will move in any direction and this
latter chance is assumed equal to the probability of activation,
v0 e~EIRT, where v0 = the vibration frequency in the lattice = the
 CONDUCTIVITY AND DIFFUSION PROCESSES 299
number of vibrational collisions/second. The total number of
ions of the salt A moving/sec./unit area in the + x direction
across the plane x + \d is thus yKJvoe~EIRT; while the number
C—d-£- I v^e~EIRT. Thus t h e
nett flow across the plane x + \d per second per unit area in
the +x direction is -x- voe~E/RT> which also equals D-^-.

x+ d
Fig. 110.
Thus
D = %d2vQe~EIRT (kinetic theory deduction),
W
D = d2 ^r~h e-E/RT (Langmuir-Dushman empirical ex-
0
pression),
so that if there were any correspondence E/Noh = %v0; but
since v = E/Noh9 and as E can be even greater than
90,000 cal./mol., v0 must if derived from the Langmuir-
Dushman expression correspond in such a case to an energy
of 540,000 cal. Clearly no physical significance can then be
attributed to the term E/Noh in the Langmuir-Dushman
expression. When not one but two degrees of freedom are in-
cluded for the storing of the activation energy, the rate R at
which the ions are activated becomes
E

with D = \c
which is Bradley's expression (121). When n degrees of freedom
300 INTERDIFFUSION OF SOLIDS

are included the appropriate expressions are those of

Wheeler (122)*
V
°\RT) (n-l)l
l
and D^y^f- -L-99,
Langmuir and Dushman's empirical relationship

has been used fairly extensively as a guide to the behaviour

of diffusion systems, perhaps unfortunately, since there has
been a tendency to read a physical significance into the values
of d thus computed. These happen in many cases to be of the
order of magnitude of the inter-crystalline distances. In
general a better agreement is found from the Langmuir-
Dushman equation, between D calc and Dob9, using for d the
crystal parameters, when the systems have a high activation
energy for diffusion, and this may perhaps be regarded as
evidence that activation energies are then stored in only one or
two degrees of freedom. However this may be, many systems
of diffusing metal pairs obey Langmuir and Dushman's
formula with some precision, as Table 67 shows. Agreement
is much poorer for those ionic solids for which low activation
energies are observed. The self-diffusion of lead may be cited as
E
an example of diffusion obeying the equation D = -—— d2e~E/RT
with precision. Fig. 111(46) shows experimental and calcu-
lated curves of log D versus 1/T, in which the experimental
curve (1) gives an E value of 27,900 cal. When d is taken
as the shortest distance between two neighbours, 4*94 A.,
the whole logD-l/T curve is fixed by a single experimental
point and in the figure is given by curve (2). When d is 8-5 A.
* These three formulae apply in their present form to zeolitic diffusion in
dilute solution, or to place exchange diffusion. In concentrated zeolitic solution
,. ,. No. of vacant interstices , . . , . _ ,.-. ^,
the ratio m . . xT ? — . — - . must be included. For diffusion by Frenkel
J
Total No. of interstices
or Schottky mechanisms the ratio — (p. 303, footnote) must be included.
 CONDUCTIVITY AND DIFFUSION PROCESSES 301

(the unit cube diagonal) the agreement is a little better and

the logD-l/T curve is given by (3) in the figure. For this
system the Langmuir-Dushman empirical equation is re-
markably satisfactory, and a single diffusion constant would
suffice, using crystallographic data, to map out the whole
course of the logD-l/T curve.

r
-9

to

26 25 21 23 22 21 20 79 70 77
l/T x 10*.
(1) Experimentally found curve, ,#=27,900 cal./atom.
• Calculated using d—shortest distance between Pb atoms,
(2) and E=25,500 cal./atom.
Calculated using d = -y/3 x shortest distance between Pb
(3) atoms, and #=26,700 cal./atom.
E
Fig. 111. Self-diffusion in Pb calculated from D =-WT d2e~EIRT calculated from
a single experimental diffusion constant and the distance between neigh-
bouring atoms (d).
The formulae of Braune (94) and of van Liempt (109) recognise
the dependence of the velocity of diffusion upon the melting-
point by introducing in the exponential term the equality
E T E
T OT B
wherein Tm is the melting-point. In these equations the
constant b depends upon the atomic size, and polarisation
properties of the lattice. The usual value of 62 is about 2, and
for lattices of very different melting-points such as lead, silver,
302 INTERDIFFUSION OF SOLIDS

and tungsten respectively (Table 75) the values of b2 are as

1:2:1, while the absolute melting-points are as 1:3:10 (see
also p. 285).
Eyring(99) in 1-936 showed how the transition state theory
of reaction velocity could be applied to viscosity, plasticity,
and diffusion. This treatment has been applied especially to
liquids, and to organic polymers, but it should be still more
applicable to the problem of diffusion in crystals. The ion or
atom in a body-centred cubic lattice, for example, may diffuse
through the Centre of any one of six faces, over a potential
energy barrier, to a neighbouring vacant site or interstitial
position. The velocity constant for passing over a potential
energy barrier is ™, T

In this expression
a = the probability that a system having once crossed the
top of the energy barrier will not recross it in the reverse
direction before losing its activation energy.

TABLE 75. The dependence of the self-diffusion constant

upon the melting-point
Melting- #18° c. in
Metal point ° C. cm. a d a y - 1

Pb 327 2-2 x 1 0 - 1 5
961 9-6 x lO" 20
& 3400 4-3 x lO" 59

Fn = thepartitionfunctionfor the normal state of the system.

Eo = the energy of activation for the transition from the
initial to the final state.
F* = the partition function for the activated state, ex-
cluding the partition function for the co-ordinate in which
the transition occurs. This latter partition function gives
the frequency term kT/h (k = the Boltzmann constant;
h = Planck's constant).
If it is now supposed that there is a concentration gradient
dCjdx in the + x direction, and that the distance between two
 CONDUCTIVITY AND DIFFUSION PROCESSES 303

successive minima of potential energy in this direction is d, one

finds the concentration C at one minimum and C + d(dCldx)
at the next. The number of ions or atoms passing in the + x
direction is 'thus Nodk'C, and in the reverse direction it is
The excess flow in the — x direction is thus

dx ° dx
where No is Avogadro's number. Finally*,
F*JcT
D = dW = d2oc^^

In the simplest cases F*/Fn~ (1 -e~hvlkT), and tends to unity

for large v (v is the vibration frequency of the atom or ion in
the lattice). If the equation is to apply to diffusion processes
in solids for which the Langmuir-Dushman equation also
holds, one may write E = RT + E0 and so

w
Noh "~h Fn' Fn BT272-
For the self-diffusion of lead at 230° C, this relationship
becomes IT*
"X *10-
F*
There is so far no analysis of factors determining the ratio -—•
for' diffusion by place exchange or other mechanisms. In
some instances, the considerations of the footnote below may
have to be allowed for.
Eyring and Wynne-Jones (99) have extended Eyring's equa-
tion for the diffusion constant by introducing the entropy
of activation AS*1, and the heat of activation AH*. The
diffusion constant is now given by

* For Frenkel or Schottky disorder a term

n _ No of interstitial ions, or vacant sites
N Total No. of potentially diffusible ions
should be included (p. 294). In dilute zeolitic solution, or in place exchange
diffusion (Fig. 108), no such term arises. The latter may be the mechanism of
interdiffusion of many metal pairs. (See p. 297, and footnote, p. 300.)
304 INTERDIFFTJSION OF SOLIDS

It is then seen that the temperature independent factor Do is

related to" the entropy of activation by the expression

Therefore a large Do implies a large entropy of activation, or

a considerable loss of order in the lattice on passing from the
initial to the transition state. This is particularly noticeable
for self-diffusion constants.
These formulae based upon different conceptions of the
diffusion process are collected in Table 76. It can here be
seen what physical quantities have been correlated with the
diffusion or conduction process. The process of diffusion is the
necessary preliminary to many metallurgical and chemical
processes involving solids. In addition, mechanisms of self-
diffusion and diffusion are of absorbing interest as a study in
themselves. Yet the very variety of the attempts which have
been made to obtain satisfactory models for diffusion, and the
limited application of so many of them, demonstrate how little
progress has been made towards a comprehensive treatment.
This is rather surprising when one considers the present
knowledge of the crystalline state. Three main viewpoints
have been advanced: the theory of equilibrium disorder
in ionic and metallic lattices; the carrying over of kinetic
theory, and of gaseous reaction kinetics to solid phases; and
the application of statistical mechanics to diffusing systems.
On the experimental side much more numerous and more
accurate data are required. It will be necessary to know with
exactitude how D, Do and E in the expression D = Doe~EIRT
vary with composition in alloy systems. Many more self-
diffusion coefficients obtained by the radioactive indicator
method are required. The connections between polarisation,
atomic radius and density, position in the periodic table, alloy
formation, melting-point, and degree of lattice loosening must
be placed upon a more quantitative basis than the present data
permit. When these properties have been correlated among
themselves, and with existing X-ray data on crystal structure,
it should be possible to understand more clearly phenomena of
diffusion in metallic and non-metallic lattices.
 TABLE 76. Expressions for the diffusion constant and the
ionic conductivity in crystals
System Formula Author

Ion-ionic lattice Frenkel(47),

metal atom, 2 o i , JO8t(44)
d o=energy
E =mean free pathformation
of hole
lattice
E1=energy for diffusion of ion or atom
into hole
v =mean thermal velocity of diffusing
atom or ion
N(Ze)2 Frenkel(47),
K
hT D Jo8t(44)
K = conductivity
Z = valence
e = electronic charge
N = number of ions per unit volume
2 e iV 0 —EIRT Braunbekdis)
Ion-ionic lattice 3T(8-3) dE
(sodium E=activation energy for ion transfer
chloride) r = vibration period, ~ 2 x 10~13 sec.
Ionic or metallic D=2-43 x 108 - 1
v /^me-iEi+£:)tRT Cichockidi7)
lattice (self- M1= atomic or molecular weight
diffusion) V = atomic or molecular volume
Tm=melting-point
Ex=energy needed to form interstitial hole
E = energy needed to bring atom to inter-
stitial position
Ionic or metallic D= — d2e-ElRT Langmuir and
lattice Dushman(ii9)
D=~vd2e-^2-^ van LiemptdO9)
v=vibration frequency of lattice
6 is a constant
Ionic or metallic Bradleyo2i)
lattice
D
1 / •" 1 V(JL u T?iTfT Wheeler (122)
-^\RTJx (/-I)!
/ = t h e number of degrees of freedom
involved in the diffusion process
/ = 2 in Bradley's equation
Ionic or metallic D= —(a.J!-±-\d*e-ElRT Eyring(99)
lattice Fa*/Fn—ratio of partition functions tran-
sition and normal states respectively
(excluding for former the partition
function for co-ordinate of diffusion
process)
a=transmission coefficient, i.e. proba-
bility that system having reached
transition state will pass over the
energy barrier to a new state
±
Ionic or metallic D=e^S /R / *?! ) dze-AH±lRT Eyring and
lattice \ h/ Wynne-
f) =eAS lR / _ ) d2
± Jones (99)
AS±, AH± denote respectively entropy
and heat of activation
306

REFERENCES
(1) Spring, W. Z. phys. Chem. 15, 65 (1894).
(2) Roberts-Austen, W. Philos. Trans. 187 A, p. 393 (1896).
(3) Faraday, M. and Stodart. Quart. J. Sci. 9, 319 (1820). Experi-
mental Researches in Chemistry, p. 57 (1859).
(4) Masing, G. Z. anorg. Chem. 62, 265 (1909).
(5) Masing, G. and Overlach, H. Wiss. Veroff. Siemens-Konz. 9,
ii, 331 (1930).
(6) Diergarten, H. Metal Progress, p. 64, Jan. 1936.
(7) Mehl, R. Trans. Amer. Inst. min. (metall.) Engrs, 122, 11
(1936).
(8) Arnold, J. and M'William, A. J. Iron and Steel Inst. 56, 85
(1899).
(9) Giolotti, F. and Tavanti, G. Gazz. chim. ital. 39, H, 386 (1909).
(10) Rawdon, H. Protective Metal Coatings, New York, Chem. Cat.
Co. (1928).
(11) Grimshaw, L. Trans. Amer. Inst. min. (metall.) Engrs, 120, 363
(1936).
(12) Faraday, M. Philos. Trans. 23, 507 (1833).
(13) Wiedmann, E. Ann. Phys., Lpz., 154, 318 (1875).
(14) Kohlrausch, W. Ann. Phys., Lpz., 17, 642 (1882).
(15) Thomson, S. P. Nature, Lond., 24, 469 (1881).
(16) Beetz, W. Ann. Phys., Lpz., 92, 452 (1854).
(17) Reynolds, O. Diss. Gottingen, 1902.
(18) Nernst, W. Z. Elektrochem. 6, 41 (1899).
(19) Bose, E. Ann. Phys., Lpz., 9, 164 (1902).
(20) Graetz, L. Ann. Phys., Lpz., 29, 314 (1886).
(21) Arrhenius, S. S.B. Akad. Wiss. Wien, 96, 831 (1887).
(22) Dunn, J. J. chem. Soc. 129, 2973 (1926); J. Soc. Chem. Ind..
Lond., 46, 109(1927).
(23) Bramley, A. Carnegie Schol. Mem., Iron and Steel Inst., 15,
155 (1926).
Bramley, A. and Beeby, G. Carnegie Schol. Mem., Iron and
Steel Inst., 15, 71 (1926).
Bramley, A. and Jinkings, A. Carnegie Schol. Mem., Iron and
Steel Inst., 15, 17, 127 (1926).
Bramley, A. and Lawton, G. Carnegie Schol. Mem., Iron and
Steel Inst., 16, 35 (1927).
Bramley, A. and Heywood, F. Carnegie Schol. Mem., Iron and
Steel Inst., 17, 67 (1928).
Bramley, A. and Turner, G. Carnegie Schol. Mem., Iron and
Steel Inst., 17, 23 (1928).
Bramley, A. and Lord, H. Carnegie Schol. Mem., Iron and Steel
Inst., 18, 1 (1929).
Bramley, A., Heywood, F., Cooper, A. and Watts, J. Trans.
Faraday Socf 31, 707 (1935).
 REFERENCES 307
(24) Brower, T., Larsen, B. and Shenk, W. Trans. Amer. Inst. min.
(metall.) Engrs, 113, 61 (1934).
(25) Van Arkel, A. Metallwirtschaft, 7, 656 (1928).
(26) Bruni, G. and Meneghini, D. Int. Zeit. Metallog. 2, 26
(1912).
(27) Tanaka, S. and Matano, C. Mem. CoU. Sci. Kyoto, 14A, 59
(1931).
(28) Linde, J. and Borelius, G. Ann. Phys., Lpz., 84, 747 (1927).
Kruger, F. and Gehm, G. Ann. Phys., Lpz., 16, 174 (1933).
29) Seith, W. and Peretti, E. Z. Elektrochem. 42, 570 (1936).
(30) Brick, M. and Philips, A. Trans. Amer. Inst. min. (metall.)
Engrs, 124, 331 (1937).
(31) v. Hevesy, G. Trans. Faraday. Soc. 34, 841 (1938).
v. Hevesy, G. and Paneth, F. Z. anorg. Chem. 82, 323 (1913).
Groh, J. and v. Hevesy, G. Ann. Phys., Lpz., 63, 85 (1920).
v. Hevesy, G. and Obrutscheva, A. Nature, Lond., 115, 674
(1925).
(32) v. Hevesy, G. and Seith, W. Z. Phys. 56, 790 (1929).
Z. Phys. 57, 869 (1929).
v. Hevesy, G., Seith, W. and Keil, A. Z. Phys. 79, 197 (1932).
(33) McKay, H. Trans. Faraday Soc. 34, 845 (1938).
(33a) Steigman, J., Shockley, W. and Nix, F. Phys. Rev. 56, 13
(1939).
(34) v. Hevesy, G. and Paneth, F. Manual of Radioactivity, Oxford,
1938.
(35) Langmuir, I. J. Franklin Inst. 217, 543 (1934).
(36) Bosworth, R. C. Proc. Roy. Soc. 150 A, 58 (1935).
(37) Proc. Gamb. Phil. Soc. 34, 262 (1938).
(38) Cichocki, J. Ann. de Physique, 20, 478 (1933).
(39) Tubandt, C. Z. Elektrochem. 39, 500 (1933).
Tubandt, C, Eggert, S. and Schibbe, G. Z. anorg. Chem. 1, 117
(1921).
Seith, W. Z. Elektrochem. 42, 635 (1936).
Tubandt, C, Reinhold, H. and Liebold, G. Z. anorg. Chem. 197,
225(1931).
(40) Grube, G. and Jedele, A. Z. Elektrochem. 38, 799 (1932).
(41) Hicks, L. Trans. Amer. Inst. min. (metall.) Engrs, 113, 163
(1934).
(42) Renninger, M. Z. Kristallogr. 89, 344 (1934).
(43) Wagner, C. and Schottky, W. Z. phys. Chem. 11B, 163 (1930).
Wagner, C. Z.phys. Chem.p. 177 (1931) (BodensteinFestband).
Z. phys. Chem. 22 B, 181 (1933).
(44) Jost, W. J. chem. Phys. 1, 466 (1933); Z. phys. Chem, 169 A, 129
(1934).
(45) Schottky, W. Z. phys. Chem. 29B, 335 (1935).
(46) v. Hevesy, G. Handbuchder Phys. 13, 286 (1928); Z. Elektro-
chem. 39, 490 (1933).
308 INTERDIFFUSION OF SOLIDS
(47) Frenkel, J. Z. Phys. 35, 652 (1926).
(48) Wagner, C. Trans. Faraday Soc. 34, 851 (1938).
(49) Strock, L. Z. phys. Chem. 25B, 441 (1934); 31B, 132 (1935).
Rahlfs, P. Z.phys. Chem. 31B, 157 (1935).
(50) Ketelaar, J. Z. phys. Chem. 26B, 327 (1934); 30, 53 (1935);
Z. KristaUogr. 87, 436 (1934).
(51) Baumbach, H. and Wagner, C. Z. phys. Chem. 22B, 199
(1933).
(52) Dunwald, H. and Wagner, C. Z. phys. Chem. 22B, 212 (1933).
(53) Jette, G. and Foote, F. J. chem. Phys. 1, 29 (1933).
Hagg, G. and Sucksdorff, J. Z. phys. Chem. 22B, 444 (1933).
Hagg, G. and Kindstrom, A. L. Z. phys. Chem. 22 B, 453
(1933).
(54) Seith, W. Z. Elektrochem. 42, 635 (1936).
(55) Hilsch, R. and Pohl, R. W. Z. Phys. 108, 55 (1937).
Z. Phys. 57, 145 (1929); 59, 812 (1930).
Rogener, H. Ann. Phys., Lpz., 29, 387 (1937).
Mollwo, E. Ann. Phys., Lpz.9 29, 394 (1937).
Hilsch, R. Ann. Phys., Lpz., 29, 407 (1937).
(56) Jost, W. Trans. Faraday Soc. 34, 861 (1938).
Diffusion ii. Chem. Reakt. imfesten Stoffen, Dresden (1937).
Jost, W. and Nehlep, G. Z. phys. Chem. 32 B, 1 (1936).
(57) Koch, E. and Wagner, C. Z. phys. Chem. 38B, 295 (1937).
(58) Foussereau. Ann. Chim. (Phys.), 5, 241, 317 (1885).
(59) Rasch, E. and Hinrichsen, F. Z. Elektrochem. 14, 41 (1908).
(60) Konigsberger, J. Phys. Z. 8, 883 (1907).
(61) Phipps, T., Lansing, W. and Cooke, T. J. Amer. Chem. Soc. 48,
112 (1926).
(62) Smekal, A. Handbuch der Phys. 24, ii, 880 (1933).
(63) Gyulai, Z. Z. Phys. 67, 812 (1931).
Jost, W. Diffusion ii. Chem. Reakt. im festen Stoffen, p. 130,
Dresden (1937).
(64) Tubandt, C, Eggert, S. and Schibbe, G. Z. anorg. Chem. 117,1
(1921).
Klaiber, F. Ann. Phys., Lpz., (5), 3, 229 (1929).
Baedeker, K. Ann. Phys., Lpz., 22, 749 (1907).
Tubandt, C. and Reinhold, H. Z. phys. Chem., Bedenstein.
Festband, p. 874 (1931).
Jost, W. Z. phys. Chem. 16 B, 129 (1932).
(65) Tubandt, C. and Reinhold, H. Z. Elektrochem. 37, 589 (1931).
Wagner, C. Z. phys. Chem. 21B, 42 (1933).
(66) Lehfeldt, W. Z. Phys. 85, 717 (1933).
(67) Haber, F. and Tolloczko, St. Z. anorg. Chem. 41, 407 (1904).
(68) Bruni, G. and Scarpa, G. R.C. Accad. Lincei, 22, 438 (1913).
(69) Coehn, A. and Sperling, K. Z. Phys'. 83, 291 (1933).
Coehn, A. and Specht, W. Z. Phys. 62, 1 (1930).
(70) Seith, W. and Kubaschewski, O. Z. Elektrochem. 41, 551 (1935).
 REFERENCES 309
(71) Seith, W. and Etzold, H. Z. Elektrochem. 40, 829 (1934).
(72) Jost, W. Z. angew. Chem. 45, 544 (1932).
Jost, W. and Linke, R. Z. phys. Chem. 29 B, 127 (1935).
(73) Nehlep,G., Jost, W. and Linke, £ . Z. Elektrochem. 42,150 (1936).
(74) Nernst, W. Z. phys. Chem. 2, 613 (1888).
(75) Einstein, A. Ann. Phys., Lpz., (4), 17, 549 (1905).
(76) v. Hevesy, G. S.B. Akad. Wiss. Wien, 129, 549 (1920); Z.phys.
Chem. 127, 401 (1927).
(77) Braune, H. Z. Elektrochem. 31, 576 (1925).
(78) Tubandt, C, Reinhold, H. and Jost, W. Z. anorg. Chem. 177,253
(1928); Z. phys. Chem. 129 A, 69 (1927).
(79) Wagner, C. Z. phys. Chem. 11B, 139 (1931).
(80) Jost, W. Diffusion u. Chem. Reaht. im festen Stoffen, p. 132,
Dresden (1937).
(81) Braune, H. and Kahn, O. Z. phys. Chem. Ill A, 270 (1924).
(82) v. Hevesy, G. and Seith, W. Z. Elektrochem. 37, 528 (1931).
(83) Seith, W. and Keil, A. Z. phys. Chem. 22B, 350 (1933).
(84) Seith, W., Hofer, E. and Etzold, H. Z. Elektrochem. 40, 322
(1934).
(85) Mehl, R. Trans. Amer. Inst. min. (metall.) Engrs, 122, 21
(1936), collected from various sources.
(86) Matano, C. Jap.J. Phys. 9, 41 (1934).
(87) Jost, W. Z. phys. Chem. 21B, 158 (1933).
(88) Fonda, G., Young, A. and Walker, A. Physics, 4, 1 (1933).
(89) Dushman, S., Dennison, D. and Reynolds, N. Phys. Rev. 29
903 (1927).
(90) van Liempt, J. A. Rec. Trav. chim. Pays-Bas, 51, 114 (1932).
(91) Paschke, M. and Hauttmann, A. Arch. Eisenhiittenw. 9, 305
(1935).
(92) Mehl, R. F. J. appl. Phys. 8, 174 (1937).
(93) Matano, C. Mem. Coll. Sci. Kyoto, 15, 351 (1932).
(94) Braune, H. Z. phys. Chem. 110, 147 (1924) (above 750° C).
(95) Jost,W. Z.p%s.C/&em.9B,73(1930)(between200and600°C.).
(96) Seith, W. Z. Elektrochem. 39, 538 (1933).
(97) Brick, M. and Philips, A. Trans. Amer. Inst. min. (metall.)
Engrs, 124, 331 (1937).
(98) Rhines, F. N. and Mehl, R. Trans. Amer. Inst. min. (metall.)
Engrs, 128, 185 (1938).
(99) Eyring, H. J. chem. Phys. 4, 283 (1936).
Eyring, H. and Wynne-Jones, W. J. chem. Phys. 3, 492 (1935).
(100) Tiselius, A. Z. phys. Chem. 169 A, 425 (1934).
(101) .Elam, Q. J. Inst. Metals, 43, 217 (1930).
(102) Smith, C. Min. and Metall. 13, 481 (1932).
(103) Wells, C. Metals Res. Lab., Carnegie Inst. Tech. unpublished
research.
(104) Warburg, E. and Tegetmeier, F. Ann. Phys., Lpz., 32,442(1887)
(105) Joffe, A. Ann. Phys., Lpz., 72, 495 (1923).
310 INTERDIFFUSION OF SOLIDS
(106) Bugahow, W. and Breschnewa, N. Tech. Phys. U.S.S.R., 2,
435 (1935).
Bugahow, W. and Rybalko, F. Tech. Phya. U.S.S.R: 2, 617
(1935).
(107) Spiers, F. Phil. Mag. 15, 1048 (1933).
(108) Matano, C. Jap. J. Phys. 8, 109 (1933); Proc. Phys. Math. Soc.
Japan, 15, 405 (1933).
(109) van Liempt, J. Z. Phys. 96, 534 (1935).
(110) Sen, B. C.R. Acad. Sci., Paris, 199, 1189 (1934).
(110a) Borelius, G. Proc. Phys. Soc. Extra Part, 49, 77 (1937).
Bragg, W., Sykes, C. and Bradley, A. Proc. Phys. Soc. Extra
Part, 49, 96(1937).
(1106) Sykes, C. and Evans, H. J. Inst. Metals, 58, 255 (1936).
(111) Bernal, J. D. Int. Conf. on Phys. 2, 119 (1934).
(112) Smekal, A. Phys. Z. 26, 707 (1925).
(113) Jost, W. Z. phys. Chem. 6B, 88, 210 (1929); Z.phys. Ghent. 7B,
234 (1930).
Joffe, A. Z. Phys. 62, 730 (1930).
(114) Fajans, K. Fortschr. Phys. Wiss. 5, 294 (1926).
Reis, A. Z. Phys. 44, 353 (1927).
(115) Braunbek, W. Z. Phys. 44, 684 (1927); 38, 549 (1926).
(116) v. Seelen, D. Z. Phys. 29, 125 (1924).
(117) Cichocki, J. J. Phys. Radium, 7, 420 (1936).
(118) Dorn, J. and Harder, O. Trans. Amer. Inst. min. (metatt.) Engrs,
128, 156 (1938).
(119) Langmuir, I. and Dushman, S. Phys. Rev. 20, 113 (1922).
(120) Polanyi, M. and Wigner, E. Z. phys. Chem. 139A, 439,(1928).
(121) Bradley, R. S: Trans. Faraday Soc. 33, 1185 (1937).
(122) Wheeler, C. Trans. Nat. Inst. Sci. India, 1, 333 (1938).
 C H A P T E R VII

STRUCTURE-SENSITIVE DIFFUSION

TYPES OF IRREVERSIBLE ^AULT IN REAL CRYSTALS

It is usual to classify diffusion processes into volume, grain-

boundary, and surface diffusions. While the preceding and
following chapters show that the problems, of volume and
surface diffusion have been attacked in a fairly adequate
manner, the data with which the present chapter has to deal
are much more fragmentary. Jost, Wagner, Frenkel, and
Schottky (l) have developed the theory of disorder in equili-
brium with order in crystals. It transpires from theory and
experiment that the transitions
Order ^ Disorder
occur in both directions with an energy of activation, and that
the change from left to right is an endothermic process. As
with chemical reactions which proceed with an energy of
activation, by suddenly chilling the system the high tem-
perature equilibrium is frozen, and a non-reversible low
temperature system results. The degree of disorder in a crystal
increases with increasing temperature, so that the disordered
crystal chilled suddenly should have more interstitial ions and
vacant lattice sites than it would possess in its equilibrium
state, and should show an increased ionic conductivity.
Annealing, provided cracks and grain boundaries did not
occur, should diminish the conductivity to a basic value.
A second kind <rfstructure-sensitive diffusion is also possible.
The crystal, whether metallic or ionic, consists of a mosaic of
small blocks,* separated by grain boundaries, or submicro-
scopic flaws. It is well known that diffusion on an external
surface (Chap. VIII) proceeds more easily than volume dif-
fusion. By an analogy, which will later be supported by
experimental data, one may suppose that diffusion along
* The evidence for this point of view is presented in the next section.
312 STRUCTURE-SENSITIVE DIFFUSION

these "internal surfaces " will occur more readily than volume
diffusion. The velocity of diffusion is largely governed by the
exponential term in the equation D = Doe~EIRT, so that if
-^surface < ^graln boundary < ^lattice*
the velocity of grain-boundary difiFusion may exceed that of
lattice diffusion even though the number of paths available
for lattice difiFusion is much greater than the number of paths
available for grain-boundary diffusion.
A third type of structure-sensitive difiFusion process is to
be attributed to impurities in the lattice. These may be incor-
porated during the growth of the crystal due to accidental
impurity of the mother liquor, or of the vapour. They may also
be introduced intentionally, as when alkali-halide lattices are
heated in alkali-metal vapours, halogen vapours, or mixed
gas atmospheres of hydrogen and alkali metal. Wagner (2) and
his co-workers introduced excess of one component by heating
the crystal in an atmosphere of its electronegative component
(O2 for CdO, ZnO, NiO, Cu2O; S for FeS; Br2 for CuBr; I 2
for Cul). It is not difficult to recognise the irreversible low-
temperature diffusion processes resulting from this type of
disorder because of the sensitivity of the high-temperature
conductivity to gas pressure. In the high-temperature region
there exists an equilibrium
Gaseous component ^ Component in excess in the lattice,
which Wagner (i) used to study the different kinds of equili-
brium disorder occurring at high temperatures in crystals.
When a crystal with an equilibrium excess of one component
is chilled to low temperatures, irreversible conductivity and
diffusion properties result (cf. Fig. 94, Chap. VI).
When the conductivity cannot be attributed to excess of
one component, i.e. when the conductivity is not sensitive to
changes in partial pressure of a surrounding gas atmosphere
of that component, and yet irreversible conductivity or
diffusion phenomena are observed, the behaviour may be
attributed either to grain-boundary difiFusion, or to lattice
disorder, frozen by rapid cooling of the high-temperature
 IRREVERSIBLE FAULT IN REAL CRYSTALS 313
equilibrium state of disorder. To determine to which type of
fault the structure sensitivity is due requires that one should
know a great deal concerning the properties of the crystal.
The ways in which these properties have led one to envisage
different lattice imperfections may now be briefly reviewed.

NON-EQUILIBRIUM DISORDER IN CRYSTALS

Some properties of crystals depend upon the previous history
of the specimen, while others are insensitive to all treatments.
On this basis a classification (3) of crystal properties may be
made (Table 77). The structure sensitivity of some of these

TABLE 77. Classification of crystal properties

Property- Insensitive Semi-sensitive" Sensitive
Character Additive, contribu- Additive, but con- Selective
tion of anomalous tribution of ano-
parts of smaller malous parts of
order of magnitude same order as that
than that of nor- of normal parts
mal parts
Examples Specific gravity, Electrical conduc- Tensile strength,
specific heat, re- tivity in ionic plasticity, dielec-
fractive index, X- crystals, diffusion, tric strength, mag-
ray interference, X-ray extinction, netisation curve of
elastic properties vibration damping ferromagnetic sub-
stances

properties has been related to various types of imperfection,

distributed over the surface and in the interior in a random
manner.
The most obvious imperfections in crystals are cracks on
and within the crystal surface. It is sometimes possible to
see the cracks with the eye, or under a microscope, especially
after suitable etching. Examination reveals that the surface
consists of blocks, or grains (Fig. 112) (4), whose interfaces
extend into the solid as cracks. Cracks of this size result from
thermal and mechanical treatments of the crystal, and also
according to Smekal(5) as a result of strains due to primary
flaws in the crystal. The primary flaws are considered by
Smekal to be due to occluded impurity, or to too rapid rates
314 STRUCTURE-SENSITIVE DIFFUSION

of growth of the crystal which leave gaps or local variations

in orientation. I t is also possible that they result from ex-
tension of the very shallow surface cracks which Lennard-
Jones and Dent (6) consider to occur as a result of unbalanced
forces at the crystal surface, which cause a lateral contraction
of the surface. The reality of contraction due to surface tension
is demonstrated by the experiments on the electrical con-

Fig. 112. Appearance of bright platinum surfaces ( x 100),

Bhowing crystallites, furrows and gram boundaries.

ductivity of thin films mentioned in the following chapter (7).

Due to the break up of evaporated films on surfaces, under
surface tension forces, the resistance rises after a time interval.
The presence of cracks and grain boundaries is revealed by
a number of additional experiments. Poulter and Wilson(8)
found that water, ether, and alcohol would penetrate into glass
or quartz for considerable distances when a pressure of
15,000 atm. was maintained for a quarter of an hour. If the
pressure was released quickly, the glass was shattered, due to
expansion of the diffused liquid in the grain boundaries. The
measurements suggested also an upper limit to the thickness
 NON-EQUILIBRIUM DISORDER IN CRYSTALS 315

of the crack, since larger molecules such as oil, or glycerine,*

caused no analogous shattering. Similarly, the diffusion of
hydrogen through iron which can be made to occur at pressures
of 9000-4000(9,10) atm., at room temperature, while in part
a true lattice diffusion, also opens up grain boundaries, so
that it is possible to force mercury and oil through the metal
after diffusing hydrogen through it. These cracks in the iron
were not visible to the naked eye.

Fig. 113. Particles on SiO2 glass, under strong grazing illumination ( x 850).

The presence of cracks, especially at the surface, governs the

tensile strength of crystals (ii). If a specimen of rock-salt is
dipped in hot water the tensile strength rises by 20-fold.
Etching glass or silica fibres in hydrofluoric acid increases the
tensile strength 5-fold. Some vapours sorbed on silica fibres
diminish their tensile strength (water, 3-fold; alcohol, 3-fold;
benzene, 2-fold). These effects may be explained as penetration
of cracks by the sorbed liquid, increasing the bonding between
grains or blocks in the solid rock-salt when water is sorbed,
but diminishing it in glass, or silica, possibly due to hydration
and swelling of silica powder down the grain boundaries and
cracks, so thrusting the grains apart. Evidence of surface
* It is to be noted, however, that these larger molecules are not spherical.
316 STRUCTURE-SENSITIVE DIFFUSION

cracks was adduced from experiments on the condensation of

metallic atoms on diamond and silica surfaces (12). The metal
atoms aggregated in lines along the surface (Fig. 113), and
these lines were thought to trace out the course of surface
cracks. Also the tensile strength of glass fibres is oftei* found
to increase as their diameter is diminished. This may be
explained by supposing that cracks exist in the fibre, but that
the probability of finding a sound fibre is greater the less its
diameter, because the surface or volume of the fibre, in which
the cracks exist, is in this way also diminished.
The continuation of cracks throughout the crystal results
in a mosaic structure of the crystals. The existence of these
secondary flaws through crystals of rock-salt may be demon-
strated by heating sodium chloride in sodium vapour (13).
Some vapour is absorbed and the crystal becomes coloured.
Aggregation of the dissolved sodium occurs, under suitable
conditions, along faults in the crystal, and the course of these
faults may be traced by examination using the ultramicro-
scope. Aggregations of sodium of four kinds have been found
(Fig. 114) (14), corresponding to tree-like,* spheroidal, striated,
and nearly homogeneous distributions of the metal.
When a crystal is. placed on an X-ray goniometer and
rotated in the X-ray beam, the intensity of reflection rises to
a maximum and then falls away. The more ideal the crystal
the sharper is this "sweep curve". Calculations have been
made of the angular breadth of the "sweep curve " for an ideal
crystal, which show that only a few seconds of arc will cover
the whole breadth of the curve. Analogous calculations for a
mosaic crystal indicated a much greater angular breadth of
the "sweep curve", which may amount in extreme cases to
several degrees. In Table 78 (15) are illustrated the observed
breadths of sweep curves at their points of half intensity; and
also values of the "integrated reflection" defined as

(Total reflected energy for uniform velocity of rotation)

_ x (velocity of rotation)
Intensity of incident radiation
 NON-EQUILIBRIUM DISORDER IN CRYSTALS 317

It is seen that Renninger's(i5) artificial rock-salt crystal re-

presents as near an approach to the ideal crystal as can be
obtained, as far as the mosaic structure type of fault is
concerned.
There has been discussion as to the size and distribution of
the Smekal blocks in a mosaic crystal. Zwicky (18) suggested
that a lattice is subdivided into a periodic block structure with
definite spacings, basing his theory in part upon the appearance

Fig. 114. Types of crystal fault in rock-salt, revealed by aggregation of

absorbed sodium, and examination in Tyndall light (H).

of regular triangular etch pits on bismuth (19), and on the

persistence of structure in liquids near their melting-points.
He made calculations which purported to show that a crystal
with such a super-lattice would be more stable than a crystal
without the super-lattice. Since the calculations are not
correct (20), and the evidence from the etch pits on bismuth
does not necessarily imply a super-lattice of Zwicky type, his
view may be discarded in favour of that of Smekal. The latter
considered that any real crystal tends to become an a-periodic
mosaic of blocks. The boundaries of the blocks grow from
318 STRUCTURE-SENSITIVE DIFFUSION

primary flaws, or from mechanical and thermal treatment, and

penetrate through the mass. Since diffusion processes down
these systems of faults obey a law
D = Doe-El*T,
where E is greater than the corresponding energy term for a
surface diffusion (p. 312), it may be concluded that the Smekal
cracks can be of molecular dimensions, so that the crystal
force fields on either side of the crack overlap.

TABLE 78

Half-breadth Integrated
of sweep reflection
Crystal curve xlO 8
(NaCl) (200 face)
sec. 200 400 600
Natural crystal, 900 270 45 16
polished cleavage
Natural crystal, 40 to 50 102-5 26-3 9-8 For CuKa
untouched cleavage radiation
Artificial untouched 71 47-8 10-5 4-6 and NaCl
cleavage,
Renninger(JS)
Calculated for ideal
crystal:
Darwin (16) 4-2 450 12-1 6-9
PrinsdT) 4-9 410 9-9 51

It must be remembered that in addition to the subdivision

of the crystal into Smekal blocks, each block may have its
own glide planes and that glide planes may act as regions for
further break-up of the crystal block, or for preferential
diffusion into the block. Palladium, after sorption of hydrogen,
shows not only a block structure, but also a herring-bone
pattern on the surface of each grain or block (21,22) which has
been attributed to preferential penetration of hydrogen down
slip planes.
The manner in which a mosaic crystal may grow from a melt
is shown by Buerger's(23) studies on dendritic crystals. A
needle-like crystal first forms, the crystal branches into other
needles, these yet again into more needles, and so all the space
is quickly occupied by the dendritic mosaic crystal. Typical
 NON-EQUILIBRIUM DISORDER IN CRYSTALS 319

dendritic mosaics of bismuth are illustrated in Fig. 115.

Dendritic mosaics can be grown from solution, where once
again the peculiar lineage structure may be recognised. Also
galena and quartz may show tree-like markings and boundaries,
while haematite crystals may be grown which are a mosaic of
plates, laid down in a direction approximately normal to the
direction of growth. Buerger considers that a complete range
of structures is possible from dendritic mosaics, parallel crystal
intergrowths, crystals with multiple terminations, or block
structure, to crystals which, like certain specimens of gypsum

Fig. 115. A dendritic surface of rapidly cooled bismuth ( x 2).

and calcite, show no imperfections under optical or X-ray

examination. This lineage theory of crystal growth allows one to
visualise readily how mosaics of various kinds may be formed.
The non-equilibrium disorder which is due to interstitial
ions and vacant lattice sites, obtained by heating the crystal
to high temperatures and then chilling it quickly, can be
revealed by the absorption spectrum, if the concentration of
the disordered points is large enough. It is generally necessary
to incorporate impurities, however, to obtain an adequate
concentration of interstitial ions or vacant sites for spectro-
scopic examination. For example, the blue crystal obtained
by heating an alkali halide in alkali metal vapour gives a new
absorption band(i3) (Fig. 116). The breadth of the band, H, at
320 STRUCTURE-SENSITIVE DIFFUSION

its point of half-intensity allows the concentration of colour

centres, or .F-centres, as they are called, to be estimated
(cf. Chap. I l l ) by the equation
N = 1-31 x l O 1 7 -

where Km9iX denotes the absorption coefficient at the peak of

the absorption curve, and /i is the refractive index for light
having the wave-length of the absorption maximum. The
number of colour centres/c.c. (N) is 1015-1017 with the sodium
vapour pressures normally employed. The F-centres are con-
sidered to be due to electrons occupying vacant chlorine-ion
sites in the lattice (24). They may be electrolysed out of the
Wave-length (w/x.)
¥00 600800 ¥00 600800 ¥00 $00800 ¥00 600800 WO.. 600

UCl NaCll CsCl

V 3 2 3 2 3 2 .» 2 3 2
Energy (eV.)
Fig. 116. Absorption bands of colour centres in solution in
alkali halides at 10° C. (Pohlus)).

crystal, and move towards the anode giving a sharp colour

boundary in the crystal (13). Irradiation with blue light causes
the nature of the absorption band to change, and a new type
of colour centre appears, the F'-centre, regarded by Mott(25)
as consisting of two electrons in a vacant lattice site. Irradia-
tion with infra-red light, or heating, causes the reverse change
into .F-centres to occur, so that a photo-equilibrium can
result: , , ,. , ,
blue light
^-centres < ~* jF'-centres
infra-red radiation or heat
Heating a crystal of alkali halide containing dissolved
alkali metal in hydrogen discharges the colour of the crystal
and causes the absorption spectrum to change (13). A new peak
occurs in the ultra-violet. These centres, which are usually
designated as [/-centres, are in a reversible equilibrium with
 NON-EQUILIBRIUM DISORDER IN CRYSTALS 321

jF-centres, into which they may be transformed by heat or by

ultra-violet light, [/-centres consist of alkali-hydride (13) that
has been formed in the lattice, and dissociates on heating or
irradiation as follows:

The electron occupies a vacant chlorine-ion site and con-

stitutes an F-centre.
These centres of disorder may contribute to the conduc-
tivity, under the influence of light and of heat.
STRUCTURE-SENSITIVE CONDUCTIVITY PROCESSES
Conductivity data may be employed to demonstrate the
properties of structure-sensitive dnTusion in crystals. It is
often found that the conductivity-temperature curve of ionic
crystals divides itself into two sections, a reversible high-
temperature curve, obeying a law
K = Ae~ElRT,
and families of low-temperature curves obeying analogous laws

The positions of the low-temperature curves depend upon the

treatment accorded the specimen and are therefore structure
sensitive. The high-temperature curve is obtained for all
specimens of a given crystal, and is structure insensitive (27)
(Fig. 117). When crystals of sodium chloride were heated for
periods of 10 hr. at a series of temperatures, the conductivity
rose as the temperature of heating rose, corresponding to an
increased number of faults and flaws resulting from the pre-
heating (Fig. 118). Smekal (26) expressed the results of Figs. 117
and 118 for the conductivity (in ohm*1 cm.-1) by the expression
K = J.1e-10'30°/T+ 1-4 x l06e-23>00°/r + 3-6 x
Structure sensitive Anion conductivity
Cation conductivity
showing that the slopes of the structure-sensitive conductivity
curves were all nearly the same, but the temperature inde-
pendent factor altered very markedly.
322 STRUCTURE-SENSITIVE DIFFUSION

Single crystals of sodium nitrate (27), because of the smaller

internal surface, show a lower conductivity than the poly-
crystalline mass solidified from the melt. If a powdered salt
is put under pressure, the coherent block, composed of a mosaic

•"¥

^ ^

\
-e
\

\
-$
\
\
-10 i
7/Tx103(Tin«K.)
Fig. 117. The conductivity of various NaCl crystals. (The lowest
curve is for a single crystal.)

of small crystallites, shows a higher conductivity than a single

crystal (28). Quartz sand when mixed with the crystals of
sodium nftrate (62 % SiO2), by increasing the number of grains
and diminishing their size, was observed to double the con-
ductivity (29). Rock-salt crystals, prepared by crystallisation
from aqueous solution, possessed at 90° C. a conductivity
 CONDUCTIVITY PROCESSES 323

100-fold smaller than that of a rock-salt poly crystalline mosaic

prepared from a melt(30). These observations lead to the
conclusion that internal surfaces are of great importance in
conductivity measurements.

1S0 fiffl
t/TxW3(T/n"K.)
Fig. 118. Rock-salt crystals, heated for 10 hr. at a series of temperatures, show
an increasing conductivity. After 10 hr. at 160° C. (the lowest curve), rising
to 200, 300, 400, 500, 600, 700, 780 (the highest curve).

Another kind of experiment which has thrown light upon

the nature of structure-sensitive conductivity requires the
addition of small quantities of impurity to the crystal lattice.
As early as 1897(31) it was noted that the addition of sodium
chloride to lead chloride caused an increase in the conductivity
of the latter. One of the most remarkable examples of this
phenomenon was given by Ketzer(32), who by adding 0'001 %
of rock-salt to lead chloride raised the conductivity of the lead
chloride 50-fold. Gyulai (33) repeated these experiments, adding
324 STRUCTURE-SENSITIVE DIFFUSION

small amounts of potassium chloride, and showing that in the

equation R = A

both A and E altered (Table 79). Lehfeldt(34) reversed the

procedure of Gyulai, and added small amounts of copper and
lead salts to potassium chloride, obtaining as usual an increase
in the conductivity of the solvent salt. Since the slope of the
curves log (conductivity) against \jT (T in °K.) are almost

TABLE79. The effect of adding KC1 to PbCl2 (Gyulai)

on the constants of the equation K = Ae~EIRT
PbCla +0-005 %KC1
PbCl2 melted in Cl, gas melted in Cl2 gas
A E A E
1-41 10,620 4-36 8680
1-65 10,860 6-88 9000
1-29 11,000 404 8720
108 10,840 902 9360
PbCl, + 0-005 %KC1
PbCl2 melted in N a gas melted in N 2 gas
A E A E
512 10,920 4-78 4400
6-55 11,070) SQu ,b limed 1
6-29 10,860f th(35)
1-43 11,700 Cry stallised fSei
frcm solutionJ

the same, the change in the conductivity is due to an in-

crease in the factor A in the equation K = Ae~EjRT. In
general, however, the effects encountered are in part due to
a loosening of the lattice (see Table 79, KC1 in PbCl2), when
the impurity is added, and in part to a decrease in the size
of constituent crystal grains. A loosening of the lattice sug-
gests an increase in the number of interstitial ions and vacant
sites, and this viewpoint is supported by Lehfeldt's(34) obser-
vation that long-continued electrolysis will free some crystals
of impurity (e.g. of PbCl2 in KC1(34)). On the other hand,
Tubandt and Reinhold(36) found no redistribution of solute
and solvent by electrolysis for NaCl or KC1 in PbCl2, so
 CONDUCTIVITY PROCESSES 325
that this experiment supports the viewpoint that the solute
here enhances the conductivity solely by increasing the
number of crystal grains in a given mass of crystal.
When a current is passed through a crystal, the resistance
in the low-temperature structure-sensitive conductivity region
rises rapidly, as a counter-electromotive force is set up. This

Fig. 119. The two conductivities for rock-salt. (1) True conductivity,
(2) conductivity after space-charge redistribution (Beran and Quittner(37>).

counter-electromotive force is not always due, however, to

polarisation at the electrodes, but to a redistribution of charges
in the body of the crystal (28,37). The experiments therefore
suggest that interstitial ions exist in the crystal which are not
in reversible equilibrium with the lattice. Both the initial, or
true, conductivity of the crystal, and the conductivity after
the redistribution of space charge in the crystal, conform to
the well-known exponential equation K = Ae~E/RT (Fig. 119),
the values of E being respectively 7510 and 9600 cal./ion.
Mechanical deformation will create centres of disarray in
a crystal (38) which result in a momentary increase in con-
326 STRUCTURE-SENSITIVE DIFFUSION

ductivity. When a rock-salt crystal was put under a series of

pressures rising by steps from 20 to 700 kg./cm.2, each
successive step caused a momentary increase in the conduc-
tivity, while releasing the pressure resulted in no new effect.
When the pressure was again applied there was no further
conductivity jump until the previous maximum pressure was
exceeded, when a momentary increase in conductivity ap-
peared once more. The jump in the conductivity was shown
by Stepajiow(39) to be an increase in the true conductivity
rather than a decrease in the counter-electromotive force due
to polarisation. I t was later found that a crystal, if put under
pressure and then annealed, would give a conductivity jump
when it was subjected to a second compression, even when
this compression did not exceed the initial load. The suggestion
by JoffS (40) that these phenomena result not from an increase
in the number of centres of disorder, but from a displacement
of the charge in the crystal was contradicted by Gyulai(4i).
The experiments reviewed show the multiplicity of effects
which can influence the non-reversible disorder of crystals and
so the conductivity or diffusion. The conductivity depends
upon the few mobile ions in the crystalline mass which are
perhaps 10~4 or less of the total number of ions (42). The energy
for loosening these ions is considered by Smekal to be only 0-4
of the energy for loosening a lattice ion. This estimate may be
compared with the values given in Chap. VIII, p. 363, for the
ratio of the activation energy for volume and surface diffusion,
which may vary from 0-2 to 0-5. The latter is the ratio for
the thorium-tungsten system, the former for caesium on
tungsten.
 327

STRUCTURE-SENSITIVE DIFFUSION PROCESSES

Structure-sensitive diffusion in metallic systems is of fairly

common occurrence. The self-diffusion of bismuth, while
strongly anisotropic, is to a certain extent dependent upon the
bismuth crystal(43) employed:
Z>(t| to 111 plane) = (1-33-16-3) x lQ*e-mM°/RTom* sec."1,
D(± to 111 plane) = (2-22- 6-5) x io~*e-S0>0Q°!RT cm.2 sec.-?.
The variation in the above equations for D is in the temperature
independent factor, Do, but only a small range of values of Do
is found. Bugahow and Rybalko(44) made a study of the
diffusion of zinc and copper in brass in which they found that
the diffusion constants increased when one passed from single
crystals to polycrystalline masses because the diffusion con-
stant depended on grain size. On passing from a single crystal
to a polycrystalline mass, both E and Z>0 (in the equation
D = DQe~EIRT) changed; but in all polycrystalline samples the
E values were the same and only the values of DQ altered.
It has proved possible to measure changes in the state of
tungsten surfaces very readily by following the thermionic
emission, which is extremely sensitive to adsorbed films.
Molybdenum or tungsten filaments are used containing a
certain amount of thorium (in the intergranular boundaries).*
By flashing the filaments at very high temperatures the
surface may be momentarily freed of thorium; and if the
filaments are then kept at some lower temperature (2050° K.
is usual for tungsten), the thorium diffuses slowly from inner to
outer surfaces and the process may be followed by measuring
the thermionic emissioii. It was found that the rate at
which this diffusion outwards occurred depended on the
size of the crystallites which comprised the filaments (45,46)
(Fig. 120), the extreme variation in D being in the instance
cited 300:1. It is to be noted, however, that the slopes of the
three lines of Fig. 120 are all the same—the activation energy
* The solubility of thorium in a tungsten lattice is negligible.
328 STRUCTURE-SENSITIVE DIFFUSION

does not depend on grain size. By mechanical treatment of a

tungsten single crystal (46) it was possible to increase the
diffusion rate without appreciably changing other properties.
This phenomenon can be contrasted with the behaviour of
the malleable metal lead in which mechanical working caused
no change in the self-diffusion co-
efficient (47).
Gehrts (48) showed that the therm-
ionic activation of tungsten and
molybdenum filaments, by diffu- J
sion of thorium from inside to the
surface, obeyed the law
Q = l _ Ce-W******* = 1 - Ce~kt,
;
where ax = 2*406 (the first root of
the Bessel function
of zero order),
& = the fraction of the
Fig. 120. Diffusion of thorium
surface covered b y i n t u n g 8 t e n crystallites of
jbhorium after flash- various sizes(46).
A. Particle diameter 5-3/*;
ing, B. Particle diameter 7*3/4;
D = the diffusion constant, C. Particle diameter 3000/*.

r = the radius of the crystallites composing the

filament,
C = a constant.
From the data of Fonda, Young and Walker (46) he calculated
the intergranular diffusion constant at 2050° K. to be
D = 0-5 x lO-^cm^sec.-1.
Langmuir's(49) data gave
D = 1-1 x 10-10 cm.2 sec.-1 at 2055° K.
Mehl(50) gave for the volume diffusion of thorium in tungsten
the equation
D = l-oOe-120000^2' cm.2 sec.-1,
and for grain-boundary diffusion
D = o-74e-94»oo°//2T cm.2 sec.-1.
 DIFFUSION PROCESSES 329
Some numerical values of DQ and E, showing the influence of
grain size, are given below:

TABLE 80. Constants in the equation D =Doe~EIRT

for a grain-boundary diffusion
Particle E
System radius in p cal./atom cm.2 sec."1
ThinW 3000 94,400 3-0 x 10- 13
7-3 95,600 4-8 x 10- 1
5-3 93,600 7-9 x 10"
94,600 8-4 x 10- 1

The diffusion of a number of elements through tungsten has

been followed by the thermionic emission method (51). None of
the films formed at the tungsten surface is as stable as a
thorium film, but the results are analogous. They demonstrate
that a large energy of activation is necessary for diffusion, and
that the velocity of diffusion depends upon the grain size of
the tungsten. The diffusion data for these metals are collected
in Table 81. Since the values for D and Do are dependent upon
TABLE 81. The constants D, Do and E in the equation
D = Doe~EIRT for diffusion in particular samples of tungsten
D x 1011
Diffusing cm.2 sec.- 1 E Atomic
metal at 2000° K. cm.2 sec.- 1 cal./atom weight

U 1-3 1-0 100,000 238-5

Th 5-9 0-75 94,000 232
Ce 95 10 83,000 140-3
Zr 324 10 78,000 91
Yt 1820 0-46 62,000 89
C in W2C(56) -108,000 12
C in single W crystal 5x10* 72,000 12
at 2460° K. (55,56)

the grain size, they are not to be taken as more than a

measure of these constants for particular specimens of tung-
sten. The diffusion of carbon in tungsten (52) was followed
by measuring the conductivity of the wire whose surface was
maintained saturated with carbon. The conductivity fell
linearly with its carbon content until at the composition W2C it
330 STRUCTURE-SENSITIVE DIFFUSION

was only 7 % of that of pure tungsten. Further diffusion in the

carbide W2C resulted in the formation of WC. The process was
reversible, when the surface carbon was removed by evapora-
tion, or with oxygen as carbon monoxide. The diffusion
constants in tungsten were those in Table 82, when a constant
concentration at the surface of 0-002 % of carbon was
TABLE 82. Diffusion of carbon in tungsten
(a) for 7 mil. pure W

D x 10 cm. sec."1
7 2
5 10
T°C. 2185 2355

(6) for 4 mil. W, 0-5% ThO2

D x 10 cm.2 sec.-1
7
1-6 4-8 7-8 18
T°G. 2070 2188 2300 2400

assumed at all temperatures. Zwikker's(53) data on the same

system emphasise the influence of grain boundaries, since he
found values of the diffusion constant varying in the ratio
30:1 at 1970° K. for different tungsten specimens.
Van Liempt (54) made a study of the diffusion of molybdenum
in tungsten single crystals* and poly crystals, and found once
again a dependence upon the size of the individual crystallites.
His data for the two cases may be expressed by
(a) "Single crystal": D = l-6x
(6) PolycrystaUine mass: D = 2 x lO^
The energy of activation is the same in the "single crystal"
and the polycrystalline mass, but the temperature indepen-
dent factor is different.
Preferential penetration down grain boundaries may some-
times be shown by taking microphotographs of the crystal in
which diffusion has occurred (50). Fig. 121 gives a cross-section
of a bi-crystal of brass, stained so that the preferential loss of
zinc from the grain boundary is very clearly indicated.
• Since E was the same for the supposed single crystal and for the poly-
crystal, and as diffusion occurred down grain boundaries for the latter, it may
be inferred that grain boundary diffusion occurred also in the former, which did
not therefore remain a single crystal.
 DIFFUSION PROCESSES 331

I t might be thought that all metals showing well-defined

grain boundaries would show preferential penetration down
those boundaries, but the photographic evidence is often very
decisively against such an hypothesis. The photographs show
that carburising and nitriding of iron or the penetration of
zinc into a copper bi-crystal do not occur preferentially
down the grain boundaries. I t is therefore rather remarkable
that evaporation of zinc from brass can occur preferentially

Fig. 121. Bi-crystal of brass held for 1 hour at 790° C. in vacuum.

The loss of Zn has occurred around the grain boundary.

down a grain boundary (Fig. 121), and that Bugahow and

Rybalko(44) found that both zinc and copper diffuse in brass
more rapidly when grain boundaries are present (p. 327). I t
is also noteworthy that structure-sensitive diffusion processes
in metals occur most often when the metals are hard, and
have a high melting-point. The diffusion of metals in lead
for example (melting-point 327° C.) is always a true lattice
diffusion, while in tungsten one finds predominantly a grain-
boundary diffusion. The malleability of lead makes it capable
of being deformed without actually breaking the crystals into
332 STRUCTURE-SENSITIVE DIFFUSION

small crystallites, while the effect of mechanical working upon

any hard single crystal is to cause it to change into a poly-
crystalline mass.
It has been observed in a number of studies of the oxidation
rate of metals (55) that, after sintering the oxide film, further
oxidation of the metal obeys a law

where x denotes the thickness of the oxide film, and k and G

are constants. The form of the above equation suggests that

Fig. 122. The velocity constant for oxidation of some metals as a

function of temperature (Dunn)(55).

oxygen attacks the underlying metal by diffusion of oxygen

or of metal ions and electrons through the intervening oxide
layer. The constant k obeys the usual exponential formula
j . = k^-E/RT (pig 122), and on the hypothesis of diffusion as
a rate-controlling factor the slopes of these logk-l/T curves
give the activation energies for diffusion. I t is noted that
a break occurs in the oxidation velocity of copper at about
660° C , although the corresponding curves for the samples
of brass are linear down to 580° C. Wilkins and Rideal(56)
suggested that the break in the curve for copper was due to
 TABLE 83. Systems which may slvow structure-sensitive diffusion

Process Examples Remarks Author

RT 1 w
Gas in solid 0 a -Cu,0 D=D0'e-W + D0<r * ">iRT Dunn (55), Wilkins(56)
Ha, He, Ne, Na, Ar, 02-Si0t1 Barrer(57)
He-pyrex j Urry(59)
Air-porcelain Chap. I l l Roeser(63>
Ha-Fe Poulter and Uffelmanuu)
Ion in ionic Inorganic salts (e.g. NaNO8, Electrical conductivity increased by cold Seith(35), Hevesy(45),
lattice KC1, NaCl) work, and by pressure. Pressed powders, Gyulai(38), Lehfeldt(S4)
or rapidly cooled melts, have conduc- and many others (39,42,29,30)
tivities much greater than single crystals
Metal in ThinW See Fig. 120 Gehrts (48), Langmuir(49),
metallic CinW Ten to thirty times as fast in imperfect Fonda, Young and Walker (46)
lattice crystal as in vapour-grown single crystal and others(45), Andrews (52),
Ce, Th, U, Fe, Yt in W Diffusion rate increases as grain size Dushman, Dennison and
decreases Reynolds(5i), Giess and
van Liempt(64)
MoinW Ten times as fast at 1600° K. in polycrystal Van Liempt(54)
as in single crystal
Self-diffusion in Bi Increased by cold work; decreased by Seith(43) and others (4&)
Zn in a-brass Diffusion 40-fold faster in polycrystal than Bugahow and Rybalko(44)
in single crystal
Cu in y-Fe Preferential penetration along grain Cf. Mehl(50), Sakharovai65)
Cr from Cr-Ni to pure Cr boundaries
Cu from duralumin to pure Al
Cu in zinc Diffusion six times as rapid in polycry- Cf. Mehl(50), Sakharova(65)
stalline mass as in single crystal
334 STRUCTURE-SENSITIVE DIFFUSION

a low-temperature grain-boundary diffusion through cuprous

oxide merging into a high-temperature lattice diffusion.
When gases (He, H 2 , Ar, N 2 , 0 2 ) diffuse through silica glass,
one sometimes gets a continual change in slope of log (per-
meation rate)-l/T curves at low temperatures ((57,58,59); see
also Table 23 and Chap. I l l ) , thought to be due to grain-
boundary diffusion.
Diffusion through metals may pass from being predominantly
lattice diffusion to grain-boundary diffusion by the action of
the diffusing gas upon the metal. Steel becomes brittle when
exposed to the continued action of hydrogen. Similar obser-
vations upon brittleness created in metals by diffusion have
been collected by McBain(60). Copper became brittle and
fissured after diffusion experiments (61), and the diffusion rate
increased rapidly even at constant temperatures and pressure,
suggesting that grain boundaries have developed, and even
become macroscopic channels. Similarly palladium in hydro-
gen becomes disintegrated to a considerable extent, developing -
a thready structure, with longitudinal fissures (62).
One may conclude this chapter by giving in Table 83 a list
of those systems in which grain-boundary diffusion can play
an important part. There are undoubtedly many others which
have not been studied, or have been inadequately studied, and
a number of properties of structure-sensitive diffusion not yet
revealed.

REFERENCES
(1) Chapter VI, pp. 247 et seq., also 292 et seq.
Wagner, C. and Schottky, W. Z. phys. Che?n. 11B, 163 (1930).
Frenkel, J. Z. Phys. 35, 652 (1926).
Jost, W. J. chem. Phys. 1, 466 (1933); Z.<phys.Chem. 169A, 129
(1934).
Wagner, C. Z. phys. Chem. 22B, 181 (1933).
(2) Baumbach, H. and Wagner, C. Z. phys. Chem. 22B, 199 (1933).
Dunwald, H. and Wagner, C. Z. phys. Cliem. 22B, 212 (1933).
(3) Orowan, E. Int. Conf. Phys. 2, 81 (1934).
(4) McBain, J. W. Sorption of Gases by Solids, p. 279. Routledge
(1932).
(5) Smekal, A. Int. Conf. Phys. 2, 93 (1934).
 REFERENCES 335
(6) Lennard-Jones, J. E. and Dent, B. Proc. Boy. Soc. 121 A, 247
(1928).
(7) Lovell, A. Proc. Roy. Soc. 166A, 270 (1938).
Appleyard, E. Proc. Phys. Soc. 49, 118 (1937) (extra part).
(8) Poulter, T. and Wilson, R. Phys. Rev. 40, 877 (1932).
(9)' Bridgman, P. Rec. Trav. chim. Pays-Bos, 42, 568 (1923); Proc.
Amer. Acad. Arts Sci. 59, 173 (1924).
(10) Poulter, T. and Uffelman, L. Physics, 3, 147 (1932).
(11) Joffe, A. Int. Conf. Phys. 2, 77 (1934).
(12) Andrade, E. Int. Conf. Phys. 2, 112 (1934).
Andrade, E. and Martindale, J. Philos. Trans. 235, 69 (1935).
(13) Hilsch, R. and Pohl, R. Trans. Faraday Soc. 34, 883 (1938),
where numerous other references may be found. Also Pohl, R.
Proc. Phys. Soc. 49, 1 (1937) (extra part).
(14) Smekal A. Handbuch d. Phys. 24/2, 835. Berlin: Julius
Springer (1933).
(15) Renninger, M. Z. Kristallogr. 89, 344 (1934).
(16) Darwin, C. G. Phil. Mag. -27, 315, 675 (1914).
(17) Prins, I. Z. Phys. 63, 477 (1930).
(18) Zwicky, F. Rev. Mod. Phys. 6, 193 (1934).
(19) Goetz, A. Int. Conf. Phys. 2, 62 (1934); Z. KristaUogr. Sonder-
heft, 1934.
(20) Orowan, E. Z. Phys. 79, 573 (1932); 89, 774 (1934).
(21) Smith, D. and Derge, G. Trans. Amer. Electrochem. Soc. 669 253
(1934); J. Amer. chem. Soc. 56, 2513 (1934).
(22) Barrer, R. M. To be published.
(23) Buerger, M. J. Z. Kristallogr. 89, 195 (1934).
(24) de Boer, J. H. Rec. Trav. chim. Pays-Bos, 56, 301 (1937).
(25) Mott, N. F. Trans. Faraday Soc. 34, 822 (1938).
(26) Smekal, A. Handbuch d. Phys. 24/2, 883 (1933).
(27) v. Hevesy, G. Z. phys. Chem. 101, 337 (1922).
(28) Tammann, G. and Veszi, G. Zeit. anorg. Chem. 150, 355 (1926).
v. Seelen, D. Z. Phys. 29, 125 (1924).
(29) Goethals, C. Rec. Trav. chim. Pays-Bas, 49, 357 (1930).
(30) Smekal, A. (with Quittner, F.). Z. Phys. 55, 298 (1929).
(31) Fritsch, C. Ann. Phys., Lpz., 60, 300 (1897).
(32) Ketzer, R. Z. Elektrochem. 26, 77 (1920).
Le Blanc, M. Z. Elektrochem. 18, 549 (1912).*
(33) Gyulai, Z. Z. Phys. 67, 812 (1931).
(34) Lehfeldt, W. Z. Phys. 85, 717 (1933).
(35) Seith,W. Z. Phys. 56, 802 (1929).
(36) Tubandt, C. and Reinhold, H. Z. ElektrocHem. 29, 313 (1923).
(37) Beran, O. and Quittner, F. Z. Phys. 64, 760 (1930).,
Wenderowitsch, A. and Drisina, R. Z. Phys. 98, 108 (1936).
(38) Gyulai, Z. and Hartley, D. Z. Phys. 51, 378 (1928).
(39) Stepanow, A. Z. Phys. 81, 560 (1933).
(40) Joffe, A. Z. Phys. 62, 73b (1930).
336 STBUCTURE-SENSITIVE DIFFUSION
(41) Gyulai, Z. Z. phys. 78, 630 (1932).
(42) Smekal, A. Z. Techn. Phys. 8, 561 (1927).
(43) Seith, W. Z. Elektrochem. 39, 538 (1933).
(44) Bugahow, W. and Rybalko, F. Tech. Phys. U.S.S.R. 2, 617
(1935).
(45) v. Hevesy, G. Z. Elektrochem. 39, 490 (1933).
(46) Fonda, G., Young, A. and Walker, A. Physics, 4, 1 (1933)..
(47) v. Hevesy, G., Seith, W. and Keil, A. Z. Phys. 79, 197 (1932).
Seith, W. and Keil, A. Z. Metallic. 25, 104 (1933).
(48) Gehrts, A. Z. Techn. Phys. 15, 456 (1934).
(49) Langmuir, I. Phys. Rev. 22, 357 (1923).
(50) Mehl, R. Trans. Amer. Inst. min. (metall.) Engrs, 122, 11 (1936);
J. Appl. Phys. 8, 174 (1937).
(51) Dushman, S., Dennison, D. and Reynolds, N. Phys. Rev. 29, 903
(1927).
(52) Andrews, M. J. phys. Cliem. 27, 270 (1923).
Andrews, M. and Dushman, S. J. phys. Chem. 29, 462 (1925).
(53) Zwikker, C. Physica, 7, 189 (1927).
(54) v. Liempt, J. Rec. Trav. chim. Pays-Bas, 51, 117 (1932).
(55) Feitknecht, W. Z. Elekrochem. 35, 142 (1929).
Pilling, N. and Bedworth, R. J. Inst. Met. 29, 529 (1923).
Dunn, J. Proc. Roy. Soc. I l l A, 203, 210 (1926).
(56) Wilkins, F. and Rideal, E. K. Proc. Roy. Soc. 128 A, 394 (1930).
(57) Barrer, R. M. J. chem. Soc. p. 378 (1934).
(58) Burton, F., Braaten, E. and Wilhelm, J. Canad. J. Res. 8, 463
(1933).
(59) Urry, W. J. Amer. chem. Soc. 54, 3887 (1932).
(60) McBain, J. W. Sorption of Gases by Solids, p. 264 (1932).
(61) Deming, H. and Hendricks, B. J . Amer. chem. Soc. 45, 2857
(1923).
Hendricks, B. and Ralston, R. J. Amer. chem. Soc. 51, 3278
(1929).
(62) Graham, T. J. chem. Soc. Series n, 7, 419 (1869); Proc. Roy. Soc.
17, 212, 500 (1869); Chemical and Physical Researches, p. 269
(1876).
(63) Roeser, W. Bur. Stand. J. Res., Wash., 7, 485 (1931).
(64) Giess, W. and v. Liempt, J. Z. anorg. Chem. 168, 107 (1927).
(65) Sakharova, M. Tzveinuie Metallui (Non-Ferrous Metals), No. 4
(1932).
 CHAPTER VIII

MIGRATION IN THE SURFACE LAYER

OF SOLIDS

INTRODUCTION
Since molecules, ions and atoms can move in solid lattices (as
when ammonia is sorbed by natroiite, or two metals or salts
interdiffuse), it is not difficult to visualise a similar migration
of particles along external surfaces. Grain-boundary diffusion
occurs more readily than lattice diffusion (for example, the
activation energies are 90 and 120k.cal. respectively for
thorium diffusing in tungsten (i)), so that surface migration
might be expected to occur more readily still. The main lines
of experiment which have led to the present knowledge of
surface migration are:
(1) The study of the growth and dissolution of single
crystals.
(2) Phenomena of condensation and aggregation of con-
tinuous films on solid surfaces.
(3) Examination of stable monolayer or multi-layer systems
by photoelectric and thermionic methods.
Much of the early evidence of the reality of surface migration
came from the first source, and this more or less classical
evidence may now be reviewed.

EVIDENCE OF MOBILITY FROM GROWTH AND

DISSOLUTION OF CRYSTALS
Volmer and Estermann(2), when studying the rate of growth
of mercury crystals at — 63° C. from mercury vapour at
—10° C, noted a remarkably rapid rate of growth of thin
hexagonal crystals in the directions of the plane of the hexagon.
The linear growth of a hexagon was 3 x 10~2cm./min., which
was 1000-fold greater than could be explained by the kinetic
338 MIGRATION IN SURFACE LAYER OF SOLIDS
theory. This growth must occur either by surface migration,
or because impacting molecules entered the lattice and ex-
panded it laterally. Volmer and Adhikari(3) then pointed out
that surface mobility was implied in certain phenomena of
crystal growth from melts. Crystal needles often project above
the surface of the melt, and the needle can only form in this
way if lateral diffusion of ions on its surface takes place.
Even more definite evidence was forthcoming in a number
of studies with benzophenone (3,4) and with phthalic anhydride,
coumarin, salol and diphenylamine(5). In the earliest experi-
ments (3) with benzophenone a succession of mercury drops
was allowed to brush a long crystal of the organic solid which
was slowly worn away. Not only was benzophenone removed
at the point of contact of the mercury, but also for some
distance away. In a later series of experiments (4) a stream of
mercury brushed the edge of a glass plate on which was
benzophenone 0*1-1 mm. away from this edge. The benzo-
phenone was removed by the mercury although it was un-
touched by it. Moll's (5) experiments were made by depositing
a very thin film on glass, the edge of which was rinsed by
dropping mercury. The films were shown to diminish in
thickness by noting changes in the interference colours
observed with transmitted light. Moll could find no evidence
of mobility with paraffin and cetyl alcohol.
Richter and Volmer (6) attempted to measure the diffusion
rates of benzophenone over a mica surface as a function of
temperature. The quantity of benzophenone in units of
10~8g. moving per hour over 1 cm.2 of mica increased as the
temperature rose, although the sorption decreased. Thus the
diffusion has a large temperature coefficient which suggests
that it is an activated process.
An interesting new method for studying the growth of
crystals from solution, which suggests that lateral migration
occurs, has been developed by Berg (6a). Two plane glass plates
form a wedge, so that transmitted light will give interference
colours. The crystallising solution is placed in the wedge, and
lateral growth of a sodium chlorate crystal takes place.
 MOBILITY FROM GROWTH AND DISSOLUTION 339
Concentration gradients caused distortion of the interference
bands, which Berg succeeded in interpreting in terms of the
concentration gradients established. The concentration dis-
tribution was not uniform, being greatest at the edges of a
square plate of sodium chlorate. Thus the highest rate of flow
occurs at the middle of each edge, and to explain the continued
production of plane faces it was argued that a mobile surface
film must have formed.

EVIDENCE OF MOBILITY FROM THE CONDENSATION AND

AGGREGATION OF METAL FILMS

The structure of condensed films

There can be two types of film on solids, those which are stable
in monolayers and those which tend to aggregate into three-
dimensional structures. The conditions for stability of mono-
layers or aggregates are similar to those governing the
stability of films at liquid surface, which either give stable
monolayers (e.g. fatty acids with long chains on water) or
gather into lenses (paraffin, or ethylene dibromide on water).
The monolayer is stable if the spreading occasions a nett
decrease in free energy, when the various interfacial free
energies are considered. Another necessary condition for the
reorganisation of a film is surface or bulk mobility.
There are a great number of metal films which are thermo-
dynamically unstable in this sense. Such films can only be
maintained as glassy deposits if the temperature is so low
that no migration can occur. The predilection which various
sputtered or evaporated metallic deposits have for aggregation
into micro- or macro-crystals is illustrated by the collection
of observations in Table 84. The evidence from X-ray, electron
diffraction, and optical experiments reveals that the three-
dimensional crystalline state is very readily formed. One notes
also that a rise in temperature can cause an orientation of
crystallites to agree with that of the underlying solid lattice.
Sometimes alloy systems may occur, evidencing mobility and
interdiffusion of atoms, or certain crystal parameters may be
340 MIGBATION IN SURFACE LAYER OF SOLIDS

derived from those of the underlying solid, i.e. the micro-

crystal may continue the pattern of its substrate. In a great
many of the systems studied, of which those in the table are
only a few(i7), the crystal grains grow very rapidly as the
temperature is raised, due to more rapid migrations of atoms.
Such a crystallisation of silver, for example, has been observed
from 250° C. (18) to -173° C. (19).
TABLE 84. Some observations (obtained by X-ray and electron-
diffraction methods) on crystalline structure developed in
thin metal films
Nature of systems Observed properties Authors
Mirrors of Ni, Fe, Crystallites. Parameters same Gen, Zelmanov and
Cd, Hg deposited as for metal in bulk. Heating Schalnikow(T)
from the vapour causes growth of crystallites
on cold surfaces
Various thin eva- Crystalline. Regular orientation Kirschner(8)
porated films of crystal grains to match
orientation of substrate
Sputtered Pt Crystalline. Sometimes with Thomson, Stuart
lattice planes parallel to surface, and Murison(9)
sometimes irregularly deposited
crystallites
Ag on Au Crystalline. Same orientation as Farnsworth(io)
underlying solid. Parameters
as for bulk silver
Ag evaporated on Crystals on Au, "amorphous" Deubner(ii)
to cold Cu and Au on Cu
Sputtered films on Crystals, with orientation paral- Swamy <i2)
quartz lel to quartz base; heat treat-
ment increases grain size
Evaporated Bi films Crystals. Orientation of 111 plane Lane (i3)
parallel to base
Au, Ni, Co, Cu, Cr, Above characteristic tempera- Bruck(M)
Pd and Ag films ture, mosaic of small crystallites,
on rock-salt similarly oriented on crystalline
base
Al on Pt Crystalline, thick and thin films. Finch and
Some thin films show para- Quarrellos)
meters corresponding to the Pt
base
Very thin films Pt, Films formed of crystalline alloy Nattaus)
Pd, Ag on Cu and
Ag

While the true criterion of film stability is that the free

energy of film formation should be greater than the free energy
of crystallite formation, another approximate criterion is also
useful. It is usually noted that if the heat of condensation of
 CONDENSATION AND AGGREGATION 341
a metal as a monolayer upon a substrate (AHX) is much less
than its heat of condensation on itself (AH2), the metal fails
to form a stable monolayer on the substrate, while if AHX is
greater than AH2, the monolayer is stable. This relationship
is illustrated for some unstable films in Table 85. On the other
TABLE 85. A comparison of heats of condensation in uniform
layers on a foreign substrate (AHX), and heats of condensa-
tion of the metal on itself (AH2)
AHt
Author System k.cal./atom k.cal./atom
Estermann(2<» Cd-glass 3-5 28
Cd-Cu 30 28
Cd-Ag 5-0 28
Hg-Ag 2-5 18-5
Cockcroft(ai) Cd-Cu) 5-7 28
Cd-Agf

hand, the following metal-substrate systems give large heats

of sorption (AHX), and form relatively stable monolayer
systems:
Cs-W(22), Na-OW(26),
Cs-glass(23), K-OW(27),
Th-W(24), CS-WO3(28),
Tl,In,Ga-W(25), Cs-Cs2O(29).
It is systems belonging to the second group which have
provided a great deal of quantitative information on surface
diffusion. The group will be considered later, and one may now
discuss the behaviour of unstable films.

Some properties of unstable films

The essential instability of some metal film-solid systems may
be shared by films of organic and inorganic solids, which
crystallise readily. The observations on metal films are,
however, more numerous, and their behaviour has been
studied on surfaces of metals, mica, quartz, and diamond.
When silver was evaporated on to polished quartz (30) in
amounts corresponding to less than a monolayer, optical
342 MIGRATION IN SURFACE LAYER OF SOLIDS
examination showed that the film was not homogeneous but
consisted of small crystalline islands. Aggregation into these
islands could only occur by migration over the quartz surface.
Similarly, when a cadmium atom beam was directed on to
a cadmium-sensitised copper surface (21), and a wire was
placed in the path of the beam, the shadow thrown by the
wire on the surface had a diffuse edge. This phenomenon was
attributed to the creeping of cadmium atoms along the under-
lying surface. The same observation (31) was made when the
cadmium beam was replaced by a mercury atom beam.
Surface mobility can explain the experiments of Ditch-
burn (32) on the deposition by sputtering of cadmium films on
glass and metal surfaces. He found that when the cadmium
particles were directed on to the surfaces through fine slits,
no deposit could be observed if the width of the slit was below
5 x 10~2mm., a phenomenon which he attributed to a loss by
surface diffusion of particles which could not be sufficiently
rapidly replaced by condensation.
An investigation of processes of aggregation (18) of thin films
of silver and gold led to definite conclusions concerning the
mobility of atoms on surfaces. The films were initially about
fifty atomic layers thick, and were heated to various tem-
peratures before optical examination. It was found that after
heating at temperatures from 250 to 280° C. the thin silver
films had gathered into small spherulites (Fig. 123), at 300° C.
more spherulites formed, while at 345° C. the size of the
particle increased, and a crystalline outline began to show
(Pig. 124). At 500° C. a crop of small crystallites appeared in
hitherto optically empty areas. It was concluded that the
most freely mobile part of the film was the surface layer, which
for a silver layer on a silver film (of 50 layers) was mobile at
temperatures 700° C. below the melting-point of the metal.
This observation suggests an activation energy for migration
of the surface layer of atoms of a clean metal very much smaller
than the latent heat of evaporation, or of the activation energy
for the self-diffusion constant of silver. Gold films showed their
first gathering into spherulites at 400° C. The nature of the
 CONDENSATION AND AGGREGATION 343
spherulites is not quite clear, but they were supposed to
consist of an aggregate of uniaxial crystalline fibres radiating

Fig. 123. Ag-particles, about 1/x across, which developed on heating

an Ag-film 50 atoms thick to 280° C. (Magnification 1000.)

Fig. 124. Ag-particles showing their additional growth when the

Ag-film was heated to 345° C. for 2 hr. (Magnification 1000.)

from a centre. Evidence of lateral diffusion is supplied by

measurements of the electrical conductivity of thin films (23;.
When an alkali metal is deposited upon thoroughly outgassed
344 MIGRATION IN SURFACE LAYER OF SOLIDS
pyrex at low temperatures (90° K.), a measurable conductivity
is observed when the deposit is only 10 % of a monolayer. The
conductivity rises as the deposit is increased, but when the
deposition is stopped, the conductivity diminishes again.
However, at a certain critical thickness the conductivity rises
rapidly from 10~3 to 10~7 of that of the metal in bulk to a figure
of the same order (e.g. Hg on pyrex, Fig. 125). In all cases, large
irreversible changes in the conductivity may be effected by
warming. Zahn and Kramer (33) attempted to explain the high
resistivity of very thin films by assuming that a glass-like
non-conducting deposit of metal is formed, but Tammann(34)
has adversely criticised this suggestion. An amorphous state
seems improbable in view of the numerous studies (Table 84)
which indicate a crystalline form developed by aggregation
even in films of high resistivity, and with films of very small
'' nominal thickness ". *
The probable explanation of the resistivity changes of thin-
films can be given in terms of surface tension forces, and of
lateral mobility. That a film of nominal thickness 10% of a
monolayer can conduct, suggests aggregation into rays of
atoms, or islands of atoms, which touch other aggregates, and
so provide a few bridges for conduction. The increase in
resistivity of films on ageing, at 64-90° K., has been ascribed
to a cracking of the film under surface tension forces, to give
a discrete film. The disrupted film aggregates into spheru-
lites or crystallites, especially if warmed, by processes of
surface migration. As the deposition continues, there comes
a time when crystallites, developing under surface tension
forces, lateral migration, and bombardment of the crystallite
surfaces by the impinging atomic stream, begin to touch one
another more and more frequently, so that the conductivity,
at a certain critical "nominal thickness", rises rapidly to a
value very nearly that for the metal in bulk. Fig. 125 shows
the family of curves of nominal thickness against resistivity
* By "nominal thickness" is meant the thickness calculated from the time of
deposition of a calibrated atom stream, neglecting all subsequent aggregation,
and supposing the deposit to be uniform.
 Deposition temperature

20

o
X

Bulk »u»ta!64°K-
500
Thickness {A.)
Fig. 125. The change in resistivity at a critical nominal film
thickness for mercury on pyrex.

600r

s
I
I
• « i
20 40 60 80 100.

Deposition temperature (°K.)

Fig. 126. The critical nominal-thickness as a function of

temperature for mercury films on pyrex.
346 MIGRATION IN SURFACE LAYER OF SOLIDS
for mercury deposited on pyrex, while Fig. 126 shows the
critical thickness at which the large decrease in resistivity
indicated in Fig. 125 occurs, as a function of the temperature
of deposition. This critical thickness depends on temperature,
and increases in the same direction with temperature as
crystallite size.
Unfortunately, most of the evidence of surface mobility
inferred from these experiments is qualitative only, and while
with development of the theory of the processes of de-
formation and surface migration in the films one may hope
that figures such as those above may give quantitative in-
formation concerning surface diffusion, one must at present
obtain this information from the study of types of film which
give stable monolayers. It is, however, possible to make an
estimate of the activation energies for diffusion in a number of
these systems, if the heat of sorption is known. In cases where
the energy of activation for migration has been measured (35,36)
or calculated (37,38) it has been found to be one-third to one-
sixth of the heat of vaporisation. Accordingly, for the systems
given in Table 85 the activation energies for migration of
cadmium on glass, copper, or silver, or of mercury on silver,
are from 500 to 1500cal./atom, and mobility must persist to
very low temperatures. Applying the same rule to the alkali
metals, the activation energy for diffusion in the surface layer
should be about 5000cal./atom. Andrade's(i8) data (p. 342)
on multi-atomic silver and gold films show that the surface
layer migrates at temperatures about 700° C. below the
melting-point, so that this surface layer is again very mobile.
In a multi-atomic film it would be natural to assume a
different mobility in the layer next the solid substrate* a layer
in the middle of the film, and a layer on the surface of the film.
The evidence of the previous paragraph suggests that this is
true for the surface and interior. The interpretation placed by
Dixit on his results (39) on the orientation of crystalline aggre-
gates, was that the layer next the substrate may also be very
mobile. By assuming that this layer on heating to moderate
temperatures could assume the properties of a two-dimensional
 CONDENSATION AND AGGREGATION 347
gas, Dixit was able to show how the orientation of crystallites
occurred to conform to the pattern of the substrate. It is
interesting that when zinc is deposited on molybdenum the
film aggregates into crystallites already orientated at 10° C.
If there are Ns adatoms/cm.2 of surface, and each adatom
requires an activation energy E to become mobile, the number
NA which is mobile is given by

Ncr-Nt" fa '
o A. JS
where fA and/g are the partition functions for the mobile and
immobile atoms. The fraction NA/(NS — NA) approaches unity
at high temperatures, and the system behaves as a two-
dimensional gas. At low temperatures NA/(NS-NA) tends to
zero, and one has an immobile film. Since mercury still
aggregates on pyrex at 20° K., according to Appleyard and
LovelFs(23) electrical conductivity measurements (p. 344),
mercury-on-pyrex must have a very small value of E. It is
interesting that mercury shows a high mobility on a crystal
of mercury at — 63° C. (p. 337) and a very small sorption heat
on silver (Table 85). The activation energy for diffusion of
mercury on the surface of tin, as an amalgam (p. 369), was only
1920cal./atom, and occurred rapidly at room temperatures.
MEASUREMENTS OF SURFACE MIGRATION IN
SOME STABLE FILMS

Properties of stable films

When atoms of barium, caesium, potassium, thorium or
similar metals are deposited on a surface of metallic tungsten,
stable films may be built up varying in nominal thickness
frpm a fraction of a monolayer to many monolayers. The new
composite surface has contact potentials and thermionic or
photoelectric work functions different from those of the clean
metal. The movements of atoms in these films can therefore
be followed by the variation in the thermionic or photoelectric
currents, i, which alter as the fraction d of the surface covered
alters. It is important to find how the current i depends on 6.
348 MIGRATION IN SURFACE LAYER OF SOLIDS
The thermionic emission from the clean surface is giveu by
the Richardson equation
i0
in which b x R is the energy needed (in cal./g.ion of electrons)
to remove an electron from the metal into free space. Since
this work function enters as an exponent, small changes in
b x R cause large changes in i. At 1500° K., the value of i for
a tungsten surface covered by a monolayer of thorium is 105
times the value for clean tungsten (40). To determine the
connection between 6 and i, Langmuir(io) originally supposed
that the change in contact potential, V, was proportional to
the electric moment Nfi per unit area, where N is the number
of sorbed atoms and fi is the electric moment per atom. He
assumed that /i did not depend on N, and therefore that
V = kd,
and that the work function for a clean surface becomes for the
composite surface ,
r
b
where kx may be either positive or negative. If i0 denotes the
current from a clean surface, and ie the current from the
surface when a fraction 6 is covered, one has

(bxR + W)

Similarly, if im is the maximum electron emission, supposed

to occur at or near 6 = 1, a similar pair of equations may be
written, and by combining the four, one obtains

In this expression & would be the fraction of the surface

covered if the following assumptions held:
(1) that /i does not depend on N,
(2) that maximum emission occurs at d = 1.
 MEASUREMENTS OF SURFACE MIGRATION 349
Langmuir showed that for eaesium-on-tungsten im occurs
when 6 = 0-67, so that the second assumption is not always
true, while Becker (41) found for barium-on-tungsten that
Langmuir's relation was better replaced by

-[M(l-er^)] (O<0<O-85).
gl0\lmllo)
This type of deviation may be attributed to the variations
in /i with 6 or N, so that the first assumption need be true
only for dilute films.
Becker and Brattain(24) condensed thorium at a constant
rate upon a tungsten strip, and measured the growth of log i
as a function of time t. If the time needed to build up a film
showing maximum electron emission is tm, the following
relationship holds: ^
/

6m — 1 if maximum emission occurs when the surface is just

covered with a monolayer. When log ijim was plotted against/,
the curve of Fig. 127 was found, which up to dm may be
expressed by Iog10i/i0 = a(l—e-^), for l > / > 0 , and where
a = 6-54 and c = 2-38. This curve may be regarded as typical
of the variation of log i/i0 with 0 or / for most stable alkali,
or alkali earth, metal films. The thermionic emission rises to
a maximum and then falls asymptotically to its value for the
metal in bulk. For thorium-on-tungsten the relation between
Langmuir's 6' and Becker and Brattain's / is
61 = 1-135(1-e*88')-
One thus sees that to measure 6 from logi one must first
measure log i/i0 as a function of / = tjtm = 0/6m9 and then
evaluate 6m by allowing atoms to fall at a known rate on to a
known area of tungsten surface, or alternatively by measuring
the integrated positive ion current when the total deposit is
evaporated in ionic form (42). The specific surface of tungsten
may with some measure of certainty be taken as 1-4 x the
geometrical area. The surface, which is made up of crystal
350 MIGRATION IN SURFACE LAYER OF SOLIDS

faces somewhat tilted from the horizontal, thus consists of an

intersecting pattern of the surfaces of tungsten crystallites.
Unit area of a crystal contains 1-425 x 1014 tungsten atoms and
it is thought that adatoms occupy the surface so that the ratio
Number of surface tungsten atoms
Number of surface adatoms

•X

1
1i
NE TEST (DEPOSITION VERY U NIFORM)
° ?AKEN
p IN TUBE NO.I
IFFERENT TESTS ALL TAKEN

/
• OX A f
SI TUBE NO.2
O V'ALUE FOR CLEAN TUNGSTE M RIBBON
MINIMUM VALUE LOGiO 1 /.
•J BTAINEO / l M

MPIRICAL EQUATION
F»ROBABLC CURVE FOR f >0- SO

f
Fig. 127. The thermionic activity of a tungsten ribbon versus the
time of deposition from thorium wire.

is an integer. For example, the integer for thorium on tungsten

is two(22), and for caesium on tungsten is four (42).
To bring the filament into a condition suitable for ther-
mionic and photoelectric studies of films, it must be flashed
at 2800° K, in order to remove a tenaciously held monolayer
of oxygen (A 44). Tungstic oxide itself distils off the surface at
 MEASUREMENTS OF SURFACE MIGRATION 351
a much lower temperature, and 2200° K. is sufficient for its
removal (26). Thus one may prepare clean tungsten surfaces,
and surfaces of W-O, both of which lend themselves admirably
to migration experiments.

Methods used in measurements of surface diffusion

A group of methods depending on thermionic or photoelectric
properties has been worked out for the measurement of
surface migration rates. These methods and the way the data
are analysed will now be discussed, taking in order thermionic
and photoelectric studies.
A. Surface diffusion by measurement of thermionic emission.
(i) The moyements of caesium, barium, and thorium in
films on tungsten (24,45) have been followed by evaporating the
metal so as to give a uniform deposit on a tungsten strip,
suitably pre-treated (p. 350). From calibration curves giving
logi/im as a function of 0, or / = 6jdm (Fig. 127), the move-
ments of a deposited film down concentration gradients may
be followed as a function of time. If all the film is originally
on one side only of the strip, the growth of log i/im on the bare
side and the decay of log i/im on the side where the film was
deposited give a mean value of / or d as a function of time
on each side. The strip, after the deposition of a film where
2 > / > 0, is raised to a temperature where migration but not
evaporation proceeds. To interpret the data it is necessary to
idealise the process of diffusion by assuming that the diffusion
constant D is not dependent upon the concentration. One
may then write ^

Take the i/-axis to be parallel to the tungsten ribbon and in

the middle of its front face, so that df/dt = 0 = D(d2f/dy2) in
the ^-direction, since the strip is very long. Let x be any
distance normal to the t/-axis in the surface of the ribbon,
352 MIGRATION IN SURFACE LAYER,OF SOLIDS

whose width is w, so that the mid-point of the back face is

then at x = to. The boundary conditions are:
(1) aU = 0 : / = / 0 fora; = %wto -\wj= Ofor# = \wtow,
and — \w to — w;
(2) for all values of t, jj- = 0 at x = 0 and x = w.

The solution of the Pick law is then

/ = i/o +—"if ( - e-»"W cos — sin \mn\,

and since the series converges rapidly one may compute/for

any value of x and t. Brattain and Becker, for thorium on
tungsten, computed / as a function of x9 and of t (for x = 0
and x = £w>), when the time t was expressed in units ofw2/Dn2.
Then from the experimental logi versus / curve, and the
computed/versus x curve, they found logi as a function of #,
and finally by graphical integration logi was found as a
function of t for the whole front surface of the ribbon. Next
values of w21Dn2 were chosen until the calculated log i versus t
curve fitted the experimental one.
'" (ii) Langmuir and Taylor (46) followed the movement of
caesium along a tungsten wire by depositing a uniform film
along the wire, and removing caesium from the central portion.
They then measured the amount which flowed in from the
ends by evaporating it, once again only from the central part
of the filament, as a measured positive ion current. To carry
out these measurements three metal cylinders were arranged
along the length of the wire in series, the wire passing axially
through them, parallel to their length. The caesium was
deposited with the cylinders at + 22 V. The caesium from the
wire in the central cylinder was withdrawn as positive ions
by altering the potential of the cylinder to — 44 V., after which
the potential was again returned to + 22 V., and the wire held
at a temperature suitably high for migration without evapora-
tion. Finally the total caesium that had moved into the centre
was estimated, as positive ion current, by again altering the
 MEASUREMENTS OF SURFACE MIGRATION 353
potential of the central cylinder and flashing. Prom this
quantity of caesium the authors evaluated the diffusion
constant.
(iii) Another method due to Langmuir and Taylor (46) is
based on the discovery that caesium films on tungsten may
exist as two phases, a and ft, in equilibrium. From the con-
densed a-phase caesium escapes as atoms, and from the
/?-phase as ions, both rates being equal in the equilibrium
condition to the rates of arrival from the gas. At a given
temperature there is only one critical pressure, p0, of caesium
vapour at which the two phases can coexist. When p0 is
altered to pl9 one phase must disappear, and the phase
boundary moves along the wire with a velocity v. The resulting
surface migration is balanced by differences in the rates of
evaporation {va and vp for atoms and ions) and of condensation,
/iv va and vp are functions of T and 6, and /ix of 6 and the
vapour pressure of caesium. If v = {va + vv), as ordinate, is
plotted against N, the number of atoms/cm.2, as abscissa, the
curve rises from iVr = 0toiV = 7x 1012, falls from N = 7x 1012
to N = 43 x 1012 and rises when N> 43 x 1(X12, inversely as
the p-v curve of an imperfect gas. The curve (v—/i) versus
N behaves similarly and intersects the JV-axis at three
points, thus enclosing two areas Ax and A2. The condition for
the stationary phase boundary (46,47) is Ax = A2, while if [i
becomes /il9 the velocity v of the movement of the phase
boundary along the wire is

when D is regarded as independent of N and very small. The

experiments did indeed suggest that at equilibrium Ax = A2
and that v was proportional to {fix -~/i), thus giving a value of
D = 6 x 10~4 cm.2 sec.- 1 at 967° K., where N = 7-3 x 1013.

B. By the measurement of photoelectric emission.

That the photoelectric effect could be used to follow surface
migration was first suggested by work of Ives(48), but the
subsequent developments of the method and technique were
BD 23
354 MIGRATION IN SURFACE LAYER OF SOLIDS
carried out by various workers (26,27,28). The method has
provided some detailed information concerning the forces
governing the diffusions. It has one great advantage over the
previous thermionic method, in which one measures only the
integrated current over the whole filament. The photoelectric
method, on the other hand, permits one to measure the actual
concentration gradients along the surface, and their change
with time, since a small well-defined spot of light may be made
to traverse the surface on which the film is deposited.

(i) Bosworth deposited a small patch of sodium (26) or potas-

sium(27) on the centre of a tungsten strip, and by traversing
the patch with a spot of light measured the concentrations and
concentration gradients as a function of time. Three methods
were used to interpret the data, based on solutions of the
diffusion equation and on a calibration curve giving log i as
a function of N. The strip of tungsten, which had been
out-gassed at a temperature sufficient to leave the oxygen
monolayer but to remove oxides of tungsten (26), absorbed the
sodium or potassium, a portion of which reappeared on the
surface when the strip was heated to a suitably high tem-
perature. The alkali metal thus migrated over the surface of
crystallites and then between grain boundaries down into the
tungsten. The capacity of tungsten to absorb the alkali metal
was limited, and when the limit was reached the deposit of
alkali metal simply spread over the surface to give a uniform
layer and Fick's law could be applied only to the centre of
the deposit. This spreading took 1 or 2 hr. at 300° K., or
5-10 sec. at 800° K. The families of curves of Fig. 128
show the processes of absorption of sodium in sodium-free
tungsten.
The family of concentration-tiine-distance curves in Fig.
128 corresponds very nearly to the'solution of the diffusion
equation (26) when Co is the amount of substance per unit
area at time t = 0, concentrated at the zero plane, and
when diffusion occurs in the direction +x, inwards down
inter-crystalline surfaces (Chap. I, p. 44). The diffusion
 MEASUREMENTS OF SURFACE MIGRATION 355
constant for this process is denoted by A, and the solution
at x = 0 is

1-6 20 2-4
Distance along* strip, cm
Fig. 128. The absorption of sodium by a sodium-free tungsten strip.
A. 5 mins.
F. 1 min. \
B. 10 „
C. 20 „ fat 295° K. G. 3 mins. I
at 415° K.
D. 40 H. 5 „
E. 60 I. 11 ,, i

Accordingly 1/C2 should be a linear function of t, as indeed

was found to be the case (Fig. 129). The curves do not always
pass through the origin because there is an uncertainty in
fixing the time of the beginning of the experiment.
(ii) The analysis of the data in which the diffusion laws
could* be applied only to the central part of the curves (27)
23-2
356 MIGBATION IN SURFACE LAYER OF SOLIDS

(p. 367) was carried out by employing the operational notation

of Heaviside, with p =^djdt\ V2 = d2jdx29 in the case of one
dimensional diffusion.

(a)

20 40 60
Time in minutes
Fig. 129. The concentration of sodium at the peak of the
curves of Fig. 128, as a function of time.
Curve (a) 415° K. Slope = 0022.
Curve (b) 293° K. Slope = 0-00067.

Then, since Co = f(x) only,

or

and thus

or C

The total amount of sodium which has diffused from within

 MEASUREMENTS OF SURFACE MIGRATION 357
two ordinates a and 6 of concentration-distance curves such
as shown in Fig. 128 is then

If the derivatives of higher order than d?jdx* are neglected

and the ordinates a and 6 are chosen so that dzCojdx* is zero,
the expression reduces to

where the bar implies an average value of dGjdx over the

time interval t. This method suffers from one weakness—
the assumption that the derivatives of Co form a rapidly
converging series. It gave values of D = Doe~E/RT and of E
which were, however, in fair agreement with those obtained
by the previous method.
(iii) A third method of analysis of data obtained by using
the photoelectric exploration of the surface with a spot of
light depends on the setting up of an equilibrium state of film
concentrations along a wire with a temperature gradient. In
this gradient the cooler parts of the wire are more densely
covered than the hotter parts. At equilibrium,

dt dx\ dx
dCdD ndC .
%
dx dx dx
Also, since D = Doe~b/T,
d D d l
b D (

giving by substitution in the first expression

E dldx[loge(dCldx)]
b
~R~ dldx(iJT)
358 MIGRATION IN SURFACE LAYER OF SOLIDS
The values of (dC/dx) were measured photoelectrically, and of
T as a function of x with an optical pyrometer. The results
obtained by this method were comparable with those obtained
by the other two methods.
(iv) A further method of obtaining the activation energy E,
which was independent of the amounts of metal deposited, or
of the thickness of the substrate into which diffusion occurred,
was employed by Frank (28). He used the simple Fick law
1 fi
T)~?t = "a""*' a n c * s o n e gl e c t e cl the variations of D with C. The
method consisted in measuring the relative beam intensities
nx and n2 for which deposition-time curves taken at tem-
peratures Tx and T2 coincide. Under these conditions

so that one may readily compute E. The Fick law is to be

solved for the conditions:
(1) At t = 0, inside the substrate C = 0 for all x.
(2) At the surface x = 0, the concentration depends on
time t as determined by the deposition curve: Cx==0 =f(t).
Also, the total deposit is given by
d

J,
o
C(x, t) dx = nt,

where d is the thickness of the substrate into which the film

diffuses. At the temperatures Tx and T2 one may then write
(1) (2)

A 8* ~ te2 ' D2dt ~ dx2 *

Cyar,O) = O, C,(a:,0) = 0,
rd rd
I C^ix^dx = nxt, C2(x,t)dx = n2t.
Jo Jo
Each solution of the equations (1) can be transformed into
 MEASUREMENTS OF SURFACE MIGRATION 359
a solution of the equations (2) by simply reducing the time
scale. If OCT = t, by substitution in (2) one obtains
1 dC2_d202
OLD2 dr " 3a;2 '

>C2(x, ocT)dx = a?i 2 r,

Jo
and the two sets of equations are equivalent if
a = D1/D2 = nx\n2.
The methods based upon the thermionicand photoelectric
properties of surfaces form an interesting study in themselves.
The sequel will show that the surface diffusion constant D is
not independent of the surface concentration, and this fact
constitutes the major objection to these analyses. Only by
using the Fick law in the form

dt ~ dx\ dxj
can one allow for variations in D with C. This law cannot be
satisfactorily applied by using thermionic methods, which
give only integrated-effects over the filament. The photo-
electric method, however, is potentially capable of application
to measure D as a function of C. If, by using the method A (ii)
of Taylor and Langmuir (p. 352), the central part of a tungsten
filament were cleared of caesium, for example, and the caesium
from the sides then diffused inwards to the centre, the actual
concentration gradients could be measured by a photoelectric
exploration of the wire with a spot of light. The concentration-
distance curves can then be submitted to Matano's analysis
(Chap. I, p. 47) of the equation

dt ~ dx\ dxj
to give D as a function of 0. No such experiment has yet been
made, and until this has been done the surface diffusion data
must suffer in accuracy, although the main properties of
surface flow are reasonably well established.
360 MIGRATION IN SURFACE LAYER OF SOLIDS

Migration in films of caesium on tungsten

The composite surface Cs-W has been the subject of numerous
studies (35,45,46). Tungsten after ageing at 2800° K. provides
a surface, homogeneous save for about 0-5 %, on which a
caesium monolayer is completed when N, the number of
atoms per cm.2, is 3*56 x 1014, an atom density giving a ratio
caesium: tungsten = 1:4 on the surface. The sorption heat of
caesium on the inhomogeneous 0-5 % of the surface is 80 k.caL,
compared with AH = 63-5 k.caL for dilute films on the rest of
the surface. All the properties of the Cs-W surface—rates of
evaporation of atoms, ions, and electrons, heats of sorption,
velocities and activation energies of migration—depend very
strongly upon the fraction of the surface covered. The electron
emission reaches its maximum at 6 = 0*67 of a monolayer.
The migration of caesium on the surface of tungsten was
first observed by Becker (45). It is interesting to compare values
of the surface diffusion constant D obtained by different
methods. By the method which depends on freeing the centre of
the filament from caesium (p. 352), Langmuir and Taylor (35,46)
found values of D in cm.2 sec."1 at N = 2-73 x 1013 atoms/cm.2
and at temperatures of 654, 702, 746 and 812° K. which
conformed to the equation
logl02> = -0-70-3060/7 7 . (1)
The method depending upon the movement of the phase
boundary between the dense and dilute surface phases (which
can coexist at appropriate temperatures and pressures of
caesium (p. 353)) gave a value of D of 6 x 10~4 cm.2 sec.-1 when
N = 7-3 x 1013 at T = 967° K., whilst the extrapolation of
equation (1) gives D = M x 10-4cm.2sec.-x at 967° K. and
with N = 2-73 x 1013. The discrepancy may be due to the
variation in D with surface concentration (p. 372) for when
N = 2-73 x 1013 and 1-74 x 1013 respectively the first method
gives D8I2OK. = 3*4 x 10~5 and 1-4 x lO^cm^sec."1.
As the surface concentration of caesium increases the heat
of sorption diminishes, being about 41,000 cal./atom when a
 MEASUREMENTS OF SURFACE MIGRATION 361

monolayer is nearly completed. The formation of multilayers

results in a further decrease in the heat of sorption, until the
film assumes the properties of caesium in bulk, for which the
heat of condensation is 18,240 cal./atom. The activation energy
for migration in the first layer, when Nm= 2-73 x 1013, is about
20 % of the heat of sorption, so that if the same ratio exists
in a multi-atomic film, or for the surface of the metal in bulk,
the energy of activation would be 4600 cal./atom (35). Using
this assumption, Taylor and Langmuir calculated the values
of Dx and J92, *^e diffusion constants in a monolayer, and in
the surface layer of a thick deposit, respectively (Table 86).
The diffusion constant D2 is greater than the average diffusion
constant in liquids at room temperature, and the assumed
activation energy (4'6k.cal.) compares with the energy of
activation of 5-3k.cal. when D2O diffuses into H2O(50). One
TABLE 86. Surface diffusion constants fat* caesium in a mono-
layer, and in the surface layer of a thick deposit, of caesium
on tungsten
Temp. ° K. Dx (cm.2 sec."1) D 2 (cm.* sec."1)
300 1-2 x lO"11 0-00034
400 4-3 x 10-*7 000134
500 1-5 x 10- 0$K)22
600 1-6 x lO"6 00027
700 8 xl0-« 00032

may thus regard the surface layer of a thick caesium deposit

as being in a condition not very different from that of a liquid.
On the other hand, the mobility of the caesium in the first
layer is 106-fold less at room temperature than the mobility
in a liquid, and does not approach the mobility of a liquid
until a temperature of 700-800° K. is reached.

Migration of thorium on tungsten

Thorium-coated tungsten filaments have a high thermionic
emissivity and, like caesium-coated tungsten filaments, have
been widely studied(45,24,1,22). It is possible to prepare fila-
ments of tungsten with thoria incorporated, and it is with
362 MIGRATION IN SURFACE LAYER OF SOLIDS

thoriated filaments of this type that much of the available

information has been obtained. These studies have con-
siderably extended knowledge of the grain-boundary diffusion
processes already discussed (Chap. VII, p. 327), and will now
be considered in relation to surface mobility.
Quantitative measurements on the surface diffusion constant
are somewhat scanty, the most complete analysis being that
of Brattain and Becker (24), who followed the diffusion of
thorium evaporated on to tungsten from the covered to the
uncovered side of the tungsten strip, by measuring the changes
in thermionic emission with time. The theory of their method
has been described earlier (p. 351).

^ < ?K -1
, — • * "

• *—-
• •
• EXPfIRIMENTAL POINTS

TIME OF FLASHING IN HOURS

Fig. 130. Comparison of experimental and calculated migration

curves for front side of ribbon.

At the start of the migration experiment/= 6j6m (p. 349)

was 1-77. The strip was then flashed at 1535 or at 1655° K.,
and its thermionic emission measured periodically at a lower
temperature of 1261° K. Pig. 130 shows the observed values
of log i/im plotted against time after flashing at 1535 and at
1655° K. In the same figure the full curve has been calculated
according to the equation (p. 352):
/ o , 2 /o Wn=oo
^°° / n
l
mWDtlw* m7TX
• m7T
\
7T
w 2
Very clearly the experimental and theoretical curves are a
poor fit, and lack of agreement must be ascribed to the
 MEASUREMENTS OF SURFACE MIGRATION 363
variation in D with/(p. 371). Since the initial value of / was
1-77 on the front and zero on the back of the tungsten strip, the
final value on back and front w a s / = 0-885, as was verified
by the thermionic emission. Brattain and Becker made a
rough calculation of the energy of activation, E, for thorium
over tungsten, arriving at E = 110 k.cal./atom, but this
value cannot be accurate since / was not the same at both
the temperatures used (1535 and 1655° K.). The values of D
(in cm.2 sec"1) were
at 1535° K. D = 1-84 x 10"9, .
at 1655° K. D = 2-44 x 10"8,
but perhaps half this variation had its origin in the different
/ values at which the observations were made. Using the
Dushman-Langmuir equation D = (E/N0h)d2e~EIRT as an
empirical guide (Chap. VI, p. 298), one finds agreement
between the observed and calculated values of D at 1655° K.,
when E = 66*4 k.cal./atom. This is a much more likely value
than 110 k.cal./atom, since it gives a satisfactory sequence
with the expressions for lattice and slip-plane diffusion already
given (Chap. VII, p. 328), as the following set of equations
demonstrates:
Volume diffusion: log10D = 0-0- 26,200/T.
Grain-boundary diffusion: log10D = - 0 - 1 3 - 19,700/T.
Surface diffusion: log1QD = - 0-33 - 14,500/7.
The extent to which D varies as the surface concentration
of the thorium is increased will be discussed later when the
spreading pressure in films is considered (p. 372). At the
moment it will be sufficient to comment that when
0 = 0, D/Do = 1,
0 = 0-49, D/Do= 11-3,
0 = 0-98, D/Do = 99-5, if Do = D at 0 = 0,
according to the calculations of Langmuir(22).
364 MIGRATION IN SURFACE LAYER OF SOLIDS

The mobility of sodium on tungsten-oxygen surfaces

Bosworth(26) using the three photoelectric methods described
on pp. 353-9 measured the migration of sodium over and into
tungsten which had been out-gassed at 2200° K. and therefore
retained a monolayer of oxygen. These methods were:
(i) The rate of diminution of sodium concentration at the
centre of an island of sodium deposited on sodium free
tungsten, by one dimensional migration into the tungsten.
(ii) The rate of diminution of sodium concentration in an
island of sodium deposited on sodium-saturated tungsten,
using Heaviside's solution of the diffusion equation, in terms
of dCjdx and d2C/dx*, etc.
(iii) The equilibrium distribution of sodium over the surface
of a strip, with a temperature gradient along it.
It is noteworthy that the values of D = Doe~EfRT or of E
obtained by the three methods, treated as one-dimensional
diffusions, were in moderate agreement, although the diffusion
processes must have varied among themselves. For instance,
in (i) the diffusion is of sodium into the body of the tungsten,
and in (ii) and (iii) is a spreading over the surface. The inference
would be that in the strip of tungsten used the impedance to
migration down grain boundaries is comparable to the im-
pedance to migration over the surface. This is not true of the
analogous thorium-tungsten systems where Langmuir's inter-
Dretation of available data gave
^grain boundary = 90'Ok.caL/atom,
and ^surface = 66-4 k.cal./atom.
Typical values of E computed by the method (i) for a
number of experiments between 290 and 455° K. are:
6950; 6260; 5800; 5330; 6730; and 5800cal./atom,
the mean of these data being 6260 cal./atom. The values
obtained for E vary considerably with temperature, being as
low as 3200 cal. between 76 and 200° K., and as high as
 MEASUREMENTS OF SURFACE MIGRATION 365
8600 cal./atom between 400 to 550° K. The variation in E with
temperature could be regarded as a specific heat effect—the
specific heat of the sodium in the mobile state being less than
that in the immobile state. Finally, it should be observed that
the data take no account of the variation of D with 6.
As soon as the body of the tungsten was fully charged with
sodium, the spreading of sodium over the surface could be
observed at room temperatures. Using the method (ii), and
choosing two ordinates, a and b, such that dPCJdz* is zero
where they cut the Co, one has

Then one may read off values of dC/dx from the graph, and
rb
find (Co— C)dx graphically, and so compute D. The results
Ja
are indicated by the following set of data:
$=32min. $=27 min. $ = 60 n
(*£) 79 39 60

(dC\I -62 -31 -55

\dx) b
/•215
(C0-C)dx 2-22 0•68 2-90
Jl-85
Dooon 0-8 x 10- 5 0-6 x 10" 5 0-7 x lO" 5
For this set of data the surface concentration was 2-1 x 1015
atoms/cm.2
In the manner outlined, and for initial concentrations
ranging from 2-8 x 1015 to 6-6 x 1015 atoms/cm.2 of apparent
surface a mean value of AWOK. = 0 # 8 x lO^cm^sec." 1 was
computed. It was, however, noted that the larger the value
of the surface concentration, the greater was D. The values
of D at a number of temperatures are given in Table 87.
From the data of this table one may plot the curve log/)/?7*
against \jT and from the slope of this curve compute the
activation energy for diffusion as 5*5 k.cal./atom.
366 MIGRATION IN SURFACE LAYER OF SOLIDS
Method (iii), depending on the equilibrium distribution of
sodium along the strip when a temperature gradient existed
along it, gave values of E varying from 7-4 to 4-4k.cal./atom,
and which were therefore similar to the values of the activa-
tion energy obtained by the other methods.

TABLE 87. Diffusion constants D for the migration of sodium

into and over tungsten with a monolayer of oxygen
10*xZ)in
Temp. ° K. cm.8 sec."1
293 0-8
350 3-2
375 60
410 13
420 20
430 30
450 34
500 50
520 77
555 128
620 200
690 270
740 310
800 330

The mobility of potassium on tungsten-oxygen surfaces

The same photoelectric methods were applied by Bosworth (27)
to this system as he had earlier used in the analogous sodium-
oxygen-tungsten surface films (i.c). The behaviour of the two
systems was closely analogous. As for sodium, the first effect
observed was a uniform fading out of the photo-emission,
corresponding to an absorption of the potassium film by the
potassium-free tungsten. Then when the tungsten was filled
with potassium, the potassium deposit spread over the surface,
by a process analogous to two-dimensional evaporation rather
than as a true diffusion, save in the central part of the original
island of potassium.
The application of the method (i) of the previous section
led to a value of the activation energy of 6960 cal./atom—a
similar value to that observed for sodium, but as with the
 MEASUREMENTS OF SURFACE MIGRATION 367

sodium-tungsten system this figure depended upon the surface

concentration. Method (ii) of the previous section employs the

where the ordinates a and b are so chosen that <PC0/dz? is

zero in a concentration-distance curve (Fig. 128). The values
of D obtained were, for very dilute films:
Temp.° K. 480 510 590 710 780
D cm2. sec."1 0-57 x 10~5 1-4 x 10"* 10 x 10~5 140 x 10~6 280 x 10~5
When the curve logD versus \jT was plotted, the activation
energy was found to be 15,300 cal./atom. Table 88 shows that
the slope of the logD versus l/T curves changes as the surface
concentration grows, and the numerical values of E are given
for various values of surface concentration in this table.
TABLE 88. The variation in activation energy E with JV,
the number of atomsjcm.2
N x 10- 14 E
atoms/cm. 2 cal./atom
[0] [16,700]
0-06 16,000
012 15,500
0-24 14,600
0-48 13,700
0-60 13,200
1-2 12,100
1-5 10,900
2-4 8,100
30 7,700
4-8 6,750

Mobility of caesium on tungstic oxide

Frank (28) measured the changes in photoelectric emissivity
of tungstic oxide (p. 358), as caesium in measured quantity
was deposited, and then allowed to diffuse away. He found
that at high temperatures the deposit decayed more slowly,
but this result was due in some way to the caesium which
fyad collected below the surface by migration, during the
deposition, or in earlier experiments.
368 M I G R A T I O N I N S U R F A C E L A Y E R OF S O L I D S

THE MIGRATION OF OTHER FILM-FORMING SUBSTANCES

The method of Brattain and Becker (p. 351) has been used
to show that migration of barium and caesium can occur.
When barium (or caesium) was deposited on one side of the
tungsten strip, which was then raised to 1000° K., the therm-
ionic emissions from the front and back slowly became equal.
The value o f / = 6jdm on the front was initially 0-80, while
after the flashing at 1000°K., the final value of/ on both
back and front was 0*4.
Becker (51) remarked that while most of the results described
relate to electropositive films, they should also apply to
electronegative ones, such as oxygen, save that here the
electron emission is decreased as the quantity of oxygen sorbed
is increased. An experiment suggested that oxygen migrated
very rapidly at 1400° K. These statements would merit further
study, since the great readiness with which tungsten chemi-
sorbs oxygen from the surroundings may vitiate many results.
In a study of the properties of indium, thallium and gallium
films on tungsten oxide, Powell and Mercer (25) observed that,
at temperatures 200° C. below the temperature of evaporation
of ions, there was a gradual decay with time in the positive ion
current (tested by momentarily raising the temperature of
the system). These effects may be understood by assuming a
migration into the oxide or over its surface. Similar obser-
vations (29) were made on a Cs-Fe2O3 system, in which it was
shown that the photoelectric current decayed with time, and
that an inward or lateral spreading of caesium in the oxide
was a possible explanation.*
K611er(29) deposited caesium on a silver surface, to give a
multimolecular layer, and then exposed the composite surface
to the action of oxygen. Simultaneous observations of the
photoelectric properties of the Cs-Cs2O-Ag surface showed that
as fast as caesium oxide was formed it was covered by a
polyatomic layer by processes of readjustment by diffusion in
* In this case the contamination of the metal by gas, or its re-evaporation, were
not positively excluded, so that the evidence is not conclusive.
 MIGRATION OF OTHEJt SUBSTANCES 369
the film, so that the photoelectric properties remained almost
unaltered until nearly all the caesium was used up.
Two other methods of obtaining information concerning
surface migration are worthy of mention. The first is the
method of the radioactive indicator. If polonium is deposited
on a silver foil, at one end only, and the temperature is raised
to 300° C, a creeping of the polonium along the silver could
be noted, the velocity of which increased as the temperature
was raised (52). No volume diffusion of polonium through the
foil took place up to 500° C, an interesting commentary upon
the relative ease with which surface and volume diffusion
occur.
The second method is one which may have some general
applicability to amalgams. It consists in measuring the rate at
which mercury will spread over metal surfaces. Spiers (53)
found that a drop of mercury spreads over tin foil in circular
or elliptical areas in which, when diffusion has ceased, there
is a uniform mercury content (11*8 %Hg). There is thus a
concentration discontinuity from 11*8% to 0% mercury at
the edge of the area, and the edge may be easily observed.
Alty and Clark (54) made quantitative measurements on rates
of spreading, which they found to be sensitive to the pre-
treatment of the surface and the nature of the medium (water,
oil, or air) in contact with it. The surface diffusion was much
more rapid than the volume diffusion, for after a surface
diffusion of several centimetres the mercury had penetrated
into a tin block by a fraction of a millimetre only.
The spreading of mercury up the surface of tin rods dipping
into the mercury was a one-dimensional diffusion which was
assumed to obey the following conditions:

() D

where n is the number of mercury atoms/cm.2 at time t and

height x above the surface of the liquid mercury.

(ii) n = 0 at t = 0 and x > 0, n = n0 at x = 0 for all t.

BD 24
370 MIGRATION IN SURFACE LAYER OF SOLIDS
This gives as the solution of (i)

and since at the upper edge according to Spiers n = nx for

all t where nx corresponds to 11'8 % of mercury, the progress
of this boundary is given by

Thus — = erf I , / r ,- I = constant,

and so , = constant, C.
I *sJ\JJt)
Accordingly D = #2/402tf and #2 is a linear function of t. From
the slopes of #2 — t curves at various temperatures,an activation
energy for surface diffusion of 1920 caL/atom was calculated.

A COMPARISON OF THE DATA

The numerical values of the diffusion constants are compared

at a few selected temperatures for some of the systems showing
surface and intergranular diffusion, in the data following:
Temp. ° K.
500 550 600 650 700 800
Na-OW 2>cm*./.ec.
=59xlO-5 102xl0-5 177xlO- 5 232 x 10~5 278 x 10"5 330 x 10"5
All for (9^1
K-0WDcm.«/.ec.
= 1 1 x 1 0 - 5 4-6x10-5 19x10-5 66xlO- 5 126 x 10~B 340 x 10~5
All for 0 c - 0
Cfl-W Dcm.Vwc.
= 0 0 1 5 x 10-5 0055 x 10-5 0 . 1 6 x 10-s o-4 x 10~5 0-85 x 10"5 3 0 x 10"5

Th- /.
= 1-84 x lO"9 at 1535° K. for 0 ^ 1

The data show considerable differences, those in the Na-OW

system comparing with data calculated by Langmuir(i,22) for
a surface of caesium metal:
Temp. ° K . = 500 600 700
Cs-Csmet»i Z)cm.^ec. = 220x10-5 270x10-5 320 x 10"6
 COMPARISON OF THE DATA 371
Again the potassium in the K-OW system at low temperatures
is much less mobile than in the Na-0W system, as corresponds
to the difference in 6 (~0, and ~ 1 respectively); but the
mobility of the potassium passes that of the sodium at 800° K.
TABLE 89. The variation in heat of sorption with surface,
concentration and valence
N (atoms/cm.2) AH
System f=d/6mt ord cal./atom Reference

Cs-W N^O ' 65,100 Taylor and

N=2-17S x 10" (/ = 1, 0=0-67) 44,500 Langmuir (35)
i\T=3-56xlOM(0 = l) 40,800
Na-OW A r c*0,/^0 32,000 Bosworth(26>
/=0-2 28,500
/=0-4 27,000
/=0-6 23,000
/=io 17,000
Th-W A'=O-87xlO 14 178,000 Langmuir (i.22)
iY = 2-3xlO 14 174,000
A' = 5-0xlO M 172,000

So great is the effect of surface concentration upon mobility,

however, that errors in estimating the concentration could
cause the observed trends. Monovalent metals are more mobile
than bivalent metals and bivalent metals than tetravalent
metals. Thus mobility in alkali metal monolayers on tungsten
can be observed at 300° K.; barium migrates measurably only
at 1000° K.; and thorium at 1500° K. The trend shown m E,
both in respect to surface concentration and valency, is
reflected in corresponding trends in the heats of sorption AH
(Table 89).

THE VARIATION IN THE DIFFUSION CONSTANTS WITH

SURFACE CONCENTRATION

The cause of the increase in D or decrease in E, as the surface

concentration increases, is considered to be a powerful lateral
interaction of the dipoles of moment //, formed by each adatom
and its electrical image. The lateral repulsion between two
such dipole systems whose centres are a distance r apart is
given by Force = (3/2) i^
24-2
372 MIGRATION IN SURFACE LAYER OF SOLIDS

Langmuir(42) considered a metallic surface covered with

adatoms at a surface concentration N, which interacted with
the force given above. He was then able to show that such an
array of dipoles, if the short range forces of repulsion were also
considered (so that no two adatoms can simultaneously occupy
the same site), would obey the equation
F = NkT/(l - 0 ) + 3-34JVV+1-53 x 10-* N*T*p* I.
Here / is an integral whose numerical value can be obtained
from values of fi, N, and 6, and is never greater than 0-89,
and F is the spreading force in dynes/cm.
Where these repulsive forces operate in a system in which
there is a concentration gradient, there is a nett force operating
in the direction of the concentration gradient, a force which
rises rapidly with surface concentration. Thus Langmuir was
able to show that the force which operates on a given adatom,
due to the concentration gradient, was

This means that the height of the energy barriers for an atom
moving in the direction of increasing N is greater than that
for an atom moving in the opposite direction, and leads to the
expression r- 2p n

where Do is the diffusion constant when N~Q. This analysis,

originally made for caesium films (42), was applied also to
thorium (22) films, both on tungsten. Values of F and D/Do are
given below for a number of values of 0, for Th-W systems:

TABLE 90. Effect of the spreading force F upon

diffusion constants
6 000 005 01 0-2 0-3 0-4
F (dynes cm." 1 ) 0-0 7-7 16-8 390 65-6 96-1
i)/D 0 (Th-W) 1-0 1-25 1-46 1-85 2-20 2-25
0 0-5 0-7 0-9 10 1-2 1-4
F (dynes cm." 1 ) 132 226 377 502 1069 9630
D/Do (Th-W) 2-86 3-71 513 6-35 120 99-5
 DIFFUSION AND SURFACE CONCENTRATION 373
An analysis of the spreading forces for sodium on oxygen-
tungsten was made by Bosworth^55). There is a relation be-
tween the spreading force F and the vapour pressure, p,
obtained from Gibb's equation:

F=

Bosworth then found the experimental connection between

i i * A^I T «/ 7000-3200/
log 2> = log/+ 0-71 - 1 - 6 / y S

so that by inserting this value of log p in the first equation, and

using numerical values, one finds
F = 0037T./+ 140/2(l -0-0004T) dynes/cm.,
since F = 0 when/ = 0.
Topping (56) deduced the relationship F" = 4-51/*2JV* for the
electrical force between an array of dipoles. His equation was
equated to the second term in the right-hand side of the above
equation for F in terms of/. The first term is considered as
due to Gibbsian thermal pressure, F'. The observed values of
the spreading forces due to dipole interaction and those
calculated from the Topping equation are given in Table 91,
in which a quite reasonable agreement is found.

TABLE 91. Values of spreading force due to dipole repulsion

from experimental data and from Topping's equation, for
the system Na-OW

(1-0-0004T)
f (dynes cm."1), (degrees cm.-1)
when T is small
002 0-06 015
01 1-4 3-6
0-2 5-6 130
0-4 220 420
0-8 900 1100
10 1400 1500
374 MIGRATION IN SURFACE LAYER OF SOLIDS

A somewhat different method of evaluating the spreading

force is based upon the variation in the energy of activation,
E, with N. If the value of E at N — 0 is Eo, one may write
F = (E0-E)N. 1-58 x 10-12dynes cm.-1.
Bosworth's (27) data on K-0 W lead to the following values of F :
NxlO1* 0 0 6 0 1 2 0-24 0-48 0-60 1-2 1-5 2-4 3 0 4-8
atoms/cm.2
^(dynes/cm.) 0-23 0-96 3-5 10-0 14-3 38 60 142 187 322
The recorded values of F are for a range of surface concen-
trations comparable with those in Table 91 for Na-OW.

PHASE CHANGES IN STABLE MONOLAYERS

The equation of state of a gas shows that under suitable

conditions gaseous and condensed states may coexist. When
hydrogen gas dissolves in palladium, dilute and condensed
phases may exist in equilibrium (57), or when a film of myristic,
palmitic, or similar fatty acids is spread upon water, com-
pressed and expanded states can occur together at suitable
temperature or pH of the underlying liquid (58). It is therefore
interesting to inquire whether two phases can occur on stable
monolayers on tungsten.
Using the idea of thermal and electrostatic spreading
pressures developed in the previous section, one may write:
F = thermal spreading force -f electrostatic spreading force

A-Ao
where A and Ao denote the areas occupied per adatom in a
film of surface concentration N(N = 1/-4), and at saturation
(N = No = l/A0) respectively. When Ao = 11*7 A.2 one may
employ the values of the electrostatic spreading pressure in
Table 91, and so plot F-A curves for the Na-OW system.
Below 700° K. the curves obtained resemble those for a con-
densible gas, so that the spreading forces calculated from
experiment should lead to the formation of two phases.
Probably the two phases of caesium on tungsten detected
 PHASE CHANGES IN STABLE MONOLAYERS 375
by Langmuir and Taylor (46,35,47) were of this kind (p. 353).
Langmuir(47) attempted a kinetic analysis of conditions
favouring equilibrium between two such phases.
The phases discussed here are coexistent in a monolayer.
Their constitution therefore differs from that of films of
cadmium or mercury on glass (pp. 341 et seq.), where mono-
layer systems rearrange themselves into three-dimensional
aggregates. An exact analysis, employing statistical mechanics,
of conditions yielding a two-dimensional condensed phase
or a three-dimensional aggregate would be of importance.
Preliminary studies of the behaviour of double layer films
have been made by Cernuschi(59) and by Dube(60).

CALCULATION OF THE SURFACE DIFFUSION CONSTANT

The surface diffusion constant has been calculated by
several authors (37,61,35,62). Lennard-Jones's theory (37) was
used by Ward (63) to interpret the slow sorption of hydrogen
by copper.
The most recent expression (6i) takes the form

D
T + T*'
In this expression a denotes a constant (| or J), v denotes the
average velocity of an adsorbed atom for the period T during
which it is activated and T* is the time between successive
activations. The attempt to calculate T and T* has been made
for two cases:
(1) When the activation energy is received by atomic
vibrations from the underlying solid (64). It was concluded
that, when RT ^E,T was of the order 10"1* sec, and nearly
independent of temperature.
(2) When the activation energy is received by collisions
between metallic electrons from the adsorbent, and the
adatom(65). It was again found that T was approximately
10~12 sec, and nearly independent of the temperature. Since
metallic electrons may have as much as 50k.cal. of energy,
376 MIGRATION IN SURFACE LAYER OF SOLIDS
and there may be 1015 collisions/second with the adatom, it is
evident that an ample reservoir of energy is available, and
that strongly bound adatoms may be activated.
The corresponding values of r* are of course strongly de-
pendent on temperature, since they contain the Boltzmann
factor e~EIRT. It can also be shown (61) that

where F* and F are the partition functions of migrating and

vibrating states respectively. Thus

These equations may as an example be applied to a simple

type of potential energy field. The field is supposed to consist
of cylindrical potential energy holes separated from each other
by walls of height Eo occupying a fraction <j> of the whole
surface, whose area is A. Then
/2nmkT\
hr-)e ">

and D =

This when E0^>RT reduces to

and if En<gRT to D =
<f>(E0/RTy
Finally, when Eo tends to zero, D = avH, where r is the
interval between successive collisions in a two-dimensional
gas. In this case simple theories give to a the value J. Applica-
tions of the theory to the diffusion of sodium on oxygen-
tungsten surfaces (26) led to a mean free path in the activated
state of - 1 0 - 7 cm.
 377

APPLICATIONS OF SURFACE MOBILITY IN

PHYSICO-CHEMICAL THEORY
Assuming that certain adsorbed films are mobile, gas laws
such as

may be proposed, by analogy with three-dimensional equations

of state. Here F denotes the surface pressure, A the area per
molecule, and Ao the area per molecule at saturation. If F is
proportional to p, the gas pressure, and A to l/x, when x
denotes the amount adsorbed, an adsorption isotherm may be
derived rather like Langmuir's isotherm. Ao, like the "co-
volume" in van der Waals's equation, may depend on tem-
perature, and so explain the experimentally observed variation
in the saturation value of the sorption with temperature.
Langmuir's original isotherm could not do this. Similarly, the
swelling of charcoals when they sorb vapours (66) may be
explained as a penetration due to two-dimensional pressure
which thrusts apart the interpenetrating graphitic flakes.
Maxted's(67) studies of the catalytic homogeneity of certain
surfaces could be reconciled with Taylor's (68) theory of active
points if it is assumed that a migration of the reacting atoms,
or atoms of the catalyst poison, could take place over the
surface.
Various workers (69,70,71) have reported that when atom
streams of cadmium, mercury, or similar metals are directed
on to cooled surfaces, there is a critical stream density for
each temperature below which aggregation into three-dimen-
sional micro-cryBtals cannot occur. The application of an
equation

to the sorbed atoms, which gives condensed phases, analogously

to van der Waals's equation of state, has been used to explain
the critical stream density and temperature. On this view the
three-dimensional aggregate must build on the top of a
378 MIGRATION IN SURFACE LAYER OF SOLIDS

relatively immobile condensed phase in two dimensions, but

not on the mobile gaseous phase. One should note, however,
that the critical conditions are not sharply defined (72) so that
aggregates may form over a range of stream densities of the
impinging beam of atoms. This is illustrated in Fig 131.
60

50

40

I 30
20

10

—4
-60 -80 -100 -120 -140 -160
Temperature (° C.)
Fig. 131. Critical condensation phenomena for cadmium.
• , on glass; 0 , on sulphur; O, on naphthalene.

An interesting application of the theory was made by

Devonshire (73) in an attempt to explain the anomalous
diffraction of helium beams at crystal surfaces observed by
Frisch and Stern (74). It was found that at suitable angles of
incidence impinging helium atoms need not be reflected from
the surface, but could move along it in a mobile state for some
distance before being emitted. It could also be shown that
 APPLICATIONS OF SURFACE MOBILITY 379
when helium is sorbed on crystal surfaces (e.g. LiF), the zero
point energy of the helium is so large that it can always pass
over the energy barriers produced by the periodicity of
the crystal surface, and so would be in the state of a two-
dimensional gas even at 0° K.
Calculations of the energy periodicity of the crystal surface
for argon on KC1 were made by Lennard-Jones and Dent (37).
Similar calculations were made by Barrer(38) for sorption on
the basal planes of graphite. These calculations showed that
gases would be mobile at quite low temperatures, in the case
of hydrogen even at liquid air temperatures. More detailed
calculations of the same kind were made by OrrW, on argon-
KC1, and argon-CsI systems, allowing for van der Waals's,
repulsive, and electrostatic energies. In the case of KC1, the
energy periodicity gave the following data for the fractional
number of atoms rendered mobile:
Temp. ° K. 10 20 40 60 80
0000012 00017 00306 00933 01667

On Csl the energy periodicity of the surface was far more

marked, and there was a strong preferential adsorption above
the centres of lattice cells of layers of caesium or iodine ions.

REFERENCES
(1) E.g. Langmuir, I. J. Franklin Inst. 217, 543
(2) Volmer, M. and Estermann, J. Z. Phys. 7, 13 (1921).
(3) Volmer, M. and Adhikari, G. Z. Phys. 35, 170 (1925).
(4) Z. phys, Chem. 119, 46 (1926).
(5) Moll, F. Z. phys. Chem. 136, 183 (1928).
(6) Richter,M. Diss.T.H., Berlin, 1931.
Volmer, M. Trans. Faraday Soc. 28, 359 (1932).
(6a) Berg, W. Proc. Boy. Soc. 164A, 79 (1938).
(7) Gen, M., Zelmanov, I. and Schalnikow, A. Phys. Z. Sowjet. 4, 825
(1933).
(8) Kirschner, F. Z. Phys. 76, 576 (1932).
(9) Thomson, G. P., Stuart, N. and Murison, C. A. Proc. phys. Soc.
45, 381 (1933).
(10) Farnsworth, H. E. Phys. Rev. 42, 588 (1932); 43, 900 (1933);
47, 331 (1935).
(11) Deubner, A. Naturwissenschaften, 23, 557 (1935).
380 MIGRATION IN SURFACE LAYER OF SOLIDS
(12) Swamy, R. S. Proc. phys. Soc. 46, 739 (1934).
(13) Lane, C. T. Nature, Lond., 130, 999 (1932).
(14) Briick, L. Ann. Phys., Lpz., 26, 233 (1936).
(15) Finch, G. and Quarrell, A» G. Nature, Lond., 131, 877 (1933);
Proc. Roy. Soc. 141 A, 398 (1933).
(16) Natta, G. Naturwissenschaften, 23, 527 (1935).
(17) See Lange, H., Kolloidzschr. 78, 109, 231 (1937) for a review.
(18) E.g. Andrade, E. Trans. Faraday Soc. 31, 1157 (1935).
Liepus, T. Glastechn. Ber. 13, 270 (1935).
(19) Hass, G. Naturwissenschaften, 25, 232 (1937).
(20) Estermann, J. Z. phys. Chem. 106, 403 (1923).
(21) Cockcroft, J. Proc. Roy. Soc. 119A, 293 (1928).
(22) E.g. Langmuir, I. Ada Phys.-chim. 1, 371 (1934).
(23) Appleyard, E. Proc. phys. Soc. 49, 118 (1937); Discussion on
Conductivity Electricity in Solids.
Appleyard, E. and Lovell, A. Proc. Roy. Soc. 158 A, 718 (1937).
Loyell, A. Proc. Roy. Soc. 157A, 311 (1936); 166A, 270 (1938).
(24) Brattain, W. and Becker, J. A. Phys. Rev. 43, 428 (1933).
(25) Powell, C. F. and Mercer, R. L. Philos. Trans. 235 A, 101
(1935-36).
(26) Bosworth, R. C. Proc. Roy. Soc. 150 A, 58 (1935).
(27) Proc. Roy. Soc. 154A, 112 (1936).
(28) Frank, L. Trans. Faraday Soc. 32, 1402 (1936).
(29) Roller, L. Phys. Rev. 36, 1643 (1930).
(30) Estermann, J. Z. Phys. 33, 320 (1925).
(31) Knauer, F. and Stern, O. Z. Phys. 39, 774 (1926).
(32) Ditchbum, R. Proc. Camb. phil. Soc. 29, 131 (1933).
(33) Zahn, H. and Kramer, J. Z. Phys. 86, 413 (1933).
Kramer, J. Ann. Phys., Lpz., 19, 37 (1934).
(34) Tammann, G. Ann. Phys., Lpz., 22, 73 (1935).
(35) Taylor, J. B. and Langmuir, I. Phys. Rev. 44, 423 (1933).
(36) Orr, W. J. C. Trans. Faraday Soc. 35, 1247 (1939).
(37) E.g. Lennard-Jones, J. E. Trains. Faraday Soc. 28, 333 (1932).
(38) Barrer, R. M. Proc. Roy. Soc. 161 A, 476 (1937).
(39) Dixit, K. R. Phil. Mag. 16, 1049 (1933).
(40) Langmuir, I. Phys. Rev. 22, 357 (1923).
(41) Becker, J. A. Phys. Rev. 33, 1082 (Abstract) (1929).
(42) Langmuir, I. J. Amer. chem. Soc. 54, 1252 (1932).
See also Langmuir, I. and Kingdon, K. H. Proc. Roy. Soc. 107 A,
61 (1925).
(43) Langmuir, I. J. Amer. chem. Soc. 35, 105 (1913).
(44) Roberts, J. K. Proc. Roy. Soc. 152 A, 445 (1935).
Langmuir, I. and Villars, D. J. Amer. chem. Soc. 53, 495 (1931).
(45) Becker, J. A. Trans. Amer. electrochem. Soc. 55, 153 (1929); Phys.
Rev. 28, 341 (1926); Phys. Rev. 34, 1323 (1929).
(46) Langmuir, I. and Taylor, J. B. Phys. Rev. 40, 463 (1932).
(47) Langmuir, I. J. chem. Phys. 1, 3 (1933).
 REFERENCES 381
(48) Ives, H. Astrophys. J. 60, 4 (1924).
Ives, H. and Olpin, A. Phys. Rev. 34, 117 (1929).
(49) Jeffreys, H. Camb. Math. Tracts, No. 23, 10 (1927).
(50) Orr, W. J. C. and Butler, J. J. chem. Soc. p. 1273 (1935).
(61) Becker, J. A. Trans. Faraday Soc. 28, 148 (1932).
(52) Schwartz, K. Z. phys. Chem. 168 A, 241 (1934).
(53) Spiers, F. Phil. Mag. 15, 1048 (1933).
(54) Alty, T. and Clark, A. Trans. Faraday Soc. 31, 648 (1935).
(55) Bosworth, R. C. Proc. Boy. Soc. 162A, 32 (1937).
(56) Topping, J. Proc. Roy. Soc. 114A, 67 (1927).
(57) Lacher, J. Proc. Roy. Soc. 161 A, 525 (1937).
(58) Rideal, E. K. Surface Chemistry, chap, in (1930).
(59) Cernuschi, F. Proc. Camb. phil. Soc. 34/392 (1938).
(60) Dube, G. Proc. Camb. phil. Soc. 34, 587 (1938).
(61) Lennard-Jones, J. E. Proc. phys. Soc. 49, 140 (1937); Discussion
on Conduction of Electricity in Solids.
(62) Alty, T. Phil. Mag. 15, 1035 (1933).
(63) Ward, A. F. Proc. Roy. Soc. 133A, 506 (1931).
(64) Lennard-Jones, J. E. and Strachan, C. Proc. Roy. Soc. 150A, 442
(1935).
(65) Lennard-Jones, J. E. and Goodwin, E. Proc. Roy. Soc. 163 A,
101 (1937).
(66) E.g. Bangham, D. and Fakhoury, N. J. chem. Soc. p. 1324 (1931).
(67) Maxted, E. B. and co-workers. J. chem. Soc. p. 502 (1933);
pp. 26, 672 (1934); pp. 393, 1190 (1935).
(68) E.g. Taylor, H. S. Trans. Faraday Soc. 28, 247 (1932).
(69) Knudsen, M. Ann. Phys., Lpz., 50, 472 (1916).
(70) Wood, R. W. Phil. Mag. 30, 300 (1915); 32, 364 (1916).
(71) Semenoff, N. Z. phys. Chem. 7B, 471 (1930).
(72) Chariton, J., Semenoff, N. and Schalnikow, A. Trans. Faraday
Soc. 28, 169 (1932).
(73) Devonshire, A. Proc. Roy. Soc. 156A, 37 (1937).
(74) Frisch, R. and Stern, O. Z. Phys. 84, 430 (1933).
 C H A P T E R IX

P E R M E A T I O N , SOLUTION AND D I F F U S I O N
OF G A S E S I N O R G A N I C S O L I D S

PERMEABILITY SPECTRUM
Membrane-forming organic solids include waxes, fats, rubbers,
proteins and protein derivatives, cellulose and cellulose
derivatives, resins and alkyl sulphide polymers. Many syn-
thetic polymers have valuable properties of plasticity, rigidity
or elasticity. One finds every type of gas flow through them,
and a diversity of permeabilities which receives a number of
practical applications. In technological journals there are
numerous studies of gas flow through rubbers (especially
helium, air and hydrogen); of the flow of water., air and carbon
dioxide through fruit and food wrappings and cartons; and of
air and water through gutta percha and paragutta insulators,
leathers, and paint and varnish films. On the theoretical side
these polymers provide material for the study of diffusion
kinetics, and of various types of gas flow in solids, such as
activated diffusion, molecular flow, streamline flow, or orifice
flow. Of special interest are transition regions between one
type of flow and another, as yet little studied.
The process of diffusion in polymers may conveniently be
discussed in two parts: first, the flow of gases and not easily
condensible vapours (CO2, SO2, NH 3 ) in organic solids; and
second, the flow of water and organic liquids and vapours
through the membranes. This is because, as the sequel will
show, gas diffusion obeys the simple law
3C_ 82C
8* &r2
fairly rigorously, while vapour diffusion does not usually do so.
It is possible to arrange the permeabilities of organic mem-
branes to air, for example, in a permeability spectrum (i) as
indicated in Fig. 132.
 PERMEABILITY SPECTRUM 383
The permeabilities in Fig. 132 are expressed in c.c./sec./
cm.2/cm. of Hg pressure and are therefore not absolute, since
thickness (which may vary from several millimetres to a
fraction of a millimetre) has not been included. Generally a
given type of membrane will give somewhat different per-
meabilities for different specimens. Thus we have border-line
cases such as vegetable parchment listed in sections B and C;
or cellulose compounds most of which could be grouped in
section A as well as section B, according to the variable degrees
of permeability which may be encountered.
A B C D
Includes Includes papers, fibre- Textile
semi-papers boards and leather fabrics
Balloon Cellulose Filter paper Tagboard
fabrics nitrate Blotting Bond paper
"Cellophane" Cellulose paper Railroad
Regenerated acetate Insulating, board
cellulose Cellulose board Solid binders
Rubbers esters Newsprint board
Neoprene Glassine paper Pressboard
Polysulphides Vegetable Antique Vegetable
Resins parchment book paper parchment
Lapquers Asphalt Lined straw Leathers
Paints saturated board
paper Super-
calendered
book paper
-8 -7 -6 5 -4 -3 -2 -1 0
log (permeability)
Fig. 132. Showing the logarithm of the permeability towards
air of groups of organic membranes.

The experimental problem involved differs according to the

section being studied. For section A, for example, where very
small permeabilities are being encountered, various types of
diffusion apparatus have been described, all of which have as
their object the making and maintaining of gastight junctions
between the membrane and the gas chambers in contact with
its ingoing and outgoing surfaces. Dewar(2) in one form of cell
fastened his rubber membranes by tying and then waxing.
Daynes(3) found that, with glycerine lubricant between the
flanges of the two chambers and the surfaces of the rubber
membranes, the joints were satisfactorily airtight. Schumacher
384 PERMEATION, SOLUTION, DIFFUSION OF GASES

and Ferguson (4) described a mercury sealed cell (Fig. 133)

which they considered suitable for any membrane. Rayleigh (5)
used membranes supported between funnels clamped together
and waxed along the flanges to prevent lateral diffusion.
It is usual in these systems, in
which diffusion occurs from the high-
pressure side into a vacuum, to support
the membrane with a metal gauze, or
another rigid porous support, in order
to prevent distortion of the mem-
brane. Other types of cell have been
employed (Edwards and Pickering (6)),
in which the total pressure on either
side was kept the same, or nearly the
same, but on one side was hydrogen,
whose permeability was to be studied,
and on the other was air. The hydro-
gen diffusing into the air stream was
estimated by means of a Rayleigh
interferometer. These types of appa-
ratus exemplify the two methods of
measuring permeability, i.e. diffusion
into vacuum, with manometricestima- Fig. 133. Schumacher and
tion Of the diffusing gas, and diffusion Ferguson's diffusion cell.
into an air stream, with interferometric gas analysis.
The diffusion cells used in studying section A can readily be
used to study section B of the permeability ctart (Fig. 132).
Similarly the apparatus for section C may be extended in its
application to section B.
Apparatus has been developed in the Bureau of Standards
(F. Carson (i); also (7)) for measuring the permeability of papers,
fibreboards, and leathers (section C of Fig. 132). Instruments
have also been described (8) for measuring permeabilities of
fabrics and textiles, of even smaller impedance to air flow.
Fig. 134 illustrates the permeameter of Schiefer and Best (9).
The fabric pressure gauge is inclined at a slope of 1 in 10, which
allows the difference in pressure between the chamber A and
 PERMEABILITY SPECTRUM 385
the atmosphere to be measured to 2^0 inch. The difference in
pressure between the chambers A and B is registered by the
air orifice gauge, and gives at once the rate of flow of gas
through the calibrated air orifice, and hence through the fabric.
Air is drawn from the atmosphere through the fabric and air
orifice by a suction fan.
The instruments mentioned cover the groups A-D of the
permeability spectrum (Fig. 132); and so cover a 1012-fold
variation in air permeability shown by organic membranes.

Fig. 134. Permeameter for textiles.

STRUCTURES OF MEMBRANE-FORMING SUBSTANCES

Considerable progress has been made towards understanding
the chemical and physical structures of many of the polymers
or condensation polymers whose permeability will be dis-
cussed. One method of attack, which has been purely chemical,
has been to investigate the stages in which the polymer can
be synthesised or broken down. It is found that one may have
polymerisation or condensation polymerisation into chains,
plates, or three-dimensional networks. The family of linear or
chain polymers is a large one including
Natural and synthetic rubbers,
Certain proteins, e.g. silk, wool,
Cellulose and cellulose esters,
Fusible and soluble resins, e.g. novolaks,
Polyesters, -amides, and -anhydrides,
Polysulphides.
BD 25
 386 PERMEATION, SOLUTION, DIFFUSION OF GASES

Among platy substances one may include such naturally

occurring inorganic condensation polymers as
Mica, Stilbite and heulandite,
whose structures have been considered elsewhere (Chap. I l l ,
pp. 93 and 95), and which cannot yet be synthesised. Platy
polymers among organic highly polymerised substances seem
to be rare, since graphite can hardly be regarded as a polymer.
The three-dimensional networks include polymers and con-
densation polymers such as
Poly-p-divinyl benzene,
Styrene-p-divinyl benzene interpolymer,
Bakelites and similar infusible insoluble resins,
Certain urea formaldehyde resins.
Polyamide
Fusible Polysulphide Polyester chain (sUk)
Polystyrene resin chain chain (from chain (from 'Jellulose chain - (R=alkyi
chain <novolaks) dihalides) hydroxy-acid) skeleton group) '

1 NR
k
(CH t ) B
CO
CO CHtOH

A i
(CH S ) B
c'o
CO CH2OH

A i
(CH 2 ) n c'o
CO
CH2OH
o

CH2OH

Fig. 135. Some linear polymers and condensation polymers.

 MEMBRANE-FORMING SUBSTANCES 387
In Figs. 135 and 136 are illustrated diagrammatically the
manner in which certain linear and three-dimensional polymers
are built up. These diagrams have of course nothing to say
concerning the spatial relationships of the various chains with
•—CH.CHj— —CH—CHa

!H—CH8—CH—CH8—CH—CH2CH. CH8

CH—CH8—CH8—CH8— —CH—CH8

Poly-divinyl benzene

OH
CH.CH8-Y'\-CHJ

CH8

dk I I d

Possible structure for bakelite (infusible resin)

Fig. 136. Some three-dimensional networks.
25-2
388 PERMEATION, SOLUTION, DIFFUSION OF GASES

one another, for the elucidation of which one must employ

the X-ray method.
From the X-ray diffraction patterns of the polymers one
can very often decide the spatial arrangement of the whole
macro-molecule. The unit cell of cellulose is indicated in
Fig. 137. When cellulose is nitrated (io) it
is found that the glucose rings of Fig. 137
still lie in the same parallel planes and
that the dimensions along the chain are
nearly unaltered. Normal to the length of
the chain, however, there is a big increase
in the distance between the chains (in the
ratio 1*7 to 1)*accompanying the nitration.
Similarly, when cellulose is esterified the
distance between the chains grows as the
aliphatic side chains become longer(ii). Fig. 137. The unit cell
r
_. . , . of cellulose^).
This is reflected in a progressive lowering
of the melting-point, as shown in Table 92, the interaction
between chains becoming progressively less.
TABLE 92. The physical properties of some cellulose esters

Tensile
Ester Specific Melting-point strength (14)
weight (°C.)(13) (kg./mm.s)
Acetate 1-377 245 9-12
Propionate 1-268 239 6-7
Butyrate 1178 183 5-6
Valerate 1178 160 4-5
Capronate 1-110 87 2-5

The structure of certain fibrous proteins (wool and silk) has

also been studied by the X-ray method, Fig. 138 giving
Astbury's (15) model for part of the unit cell in wool. Usually
the proteins become denatured when one attempts to de-
hydrate them, a fact which has prevented successful elucida-
tion of protein structure in many instances, although in a
recent study it was found that there is a great deal of
crystalline order in certain highly hydrated proteins (16).
The structures of cellulose', hair, wool, and silk are plainly
 MEMBRANE-FORMING SUBSTANCES 389

fibrous ones, and the tensile strengths might be expected to

be those of chemical bonds, if the chains were continuous
throughout the length of the fibre. These chain strengths may
be calculated (it), and it appears that the tensile strengths are
not of the anticipated magnitude This, with other evidence,
leads one to the micellar view of the structure in such fibres
According to this theory, bundles of molecular threads are
supposed to compose a block, and the aggregate of a number

Fig. 138. Portion of space model of the structure of wool or human hairua)

of such blocks or micelles then builds up the visible thread.

The micellar theory of solids has been extended to include
rubbers, most organic highly polymerised substances, and
inorganic crystals Thus one might a priori expect inter- and
intracellular diffusion processes in organic solids analogous
to grain-boundary and intracrystalline diffusion in silica glass
(Chap. I l l , p 125) or copper oxide (Chap. VII, p. 332). There
jare two types of micro-crystalline structure possible in organic
polymers (18,19,20,21,22,23) illustrated by Figs 139 and 140.
390 PERMEATION, SOLUTION, DIFFUSION OF GASES

Fig. 139 shows the discontinuous or block structure for rigid

inembranes such as cellulose or inorganic solids; while Fig. 140
gives a model for elastic long-chain polymers such as rubbers.
The arrangement of chains is lattice-like, but the chains may
be too far from or too near to neighbouring chains, and so the
structure tends to be that of disorder. But in Fig. 140 ordered
regions may be distinguished (marked by thicker lines),
although they are not the self-contained units of Fig. 139.
Stretching rubber by putting the carbon chains under tension
tends to increase the degree of order by drawing them into

Fig. 139. Discontinuous micellar structure postulated for cellulose (23). a, Haupt-
valenzketten; b, intramicellar regions; c, intermicellar holes; d, intermicellar
long spaces.

Fig. 140. Continuous micellar structure postulated for rubber (23).

alignment. The stretched rubber is then crystalline, with

eight parallel isoprene residues per unit cell, although there
is still some doubt as to their exact relative positions (18,24).
Staudinger(25,26,27) characterised the chain length of linear
polymers by the effect upon them of suitable solvents. The
characteristic behaviour was
(1) Chain-length 50-250 A. Mol.wt. > 10,000. Such colloids,
which dispersed easily to give true solutions, were called
"hemicolloids".
 MEMBRANE-FORMING SUBSTANCES 391
(2) Chain-length 250-2500 A. These colloids dispersed after
swelling to give highly viscous solutions. They were described
as "mesocolloids".
(3) Chain-length > 2500 A. Polymers of this chain length
dispersed only after very intense swelling to give solutions of
anomalous viscosity even at great dilutions. Staudinger called
these polymers "eucolloids".
The solubility may also be used to determine whether cross-
linking of polymer chains to a three-dimensional network has
occurred. A network cannot actually disperse, although it may
undergo intense swelling. Polystyrene is a linear polymer, and
therefore soluble; but when a small amount of ^-divinyl
benzene is added, the interpolymer, due to cross-linking of
the polystyrene chains by p-divinyl benzene bridges, becomes
insoluble (27).

PERMEABILITY CONSTANTS OF GROUPS A AND JB OF FIG. 132

In 1866 Graham (28) studied the diffusion of gases through
rubber. He regarded the permeation process as solution,
diflPusion and re-evaporation of the diflFusing gas, a viewpoint
which in essential details is held to-day. Wroblewski (29) in 1879
considered that Fick's laws of diffusion applied. This was
later verified by experiment in many instances. Wroblewski
also made some of the earliest measurements of the solubility
of gases in rubber, and his work in this field was followed by
that of Hufner(30) and Reychler (31). Kanata(32) extended the
study of membrane permeability from rubbers to celluloid and
gelatin. As the importance of the permeability of leathers,
balloon fabrics, packaging materials and textiles was realised,
more and more studies of membrane permeability were made.
The literature is so scattered in technological journals that no
previous attempts have been made to give a comprehensive
survey of the available data from a theoretical standpoint.
Attention has been directed in Chap. II to the possibility
of defining various permeability constants. The permeability
constant used in the present chapter has the dimensions
392 PERMEATION, SOLUTION, DIFFUSION O^ GASES

Pxtxrnr1, and denotes c.c. of gas at 1 atm. pressure and a

standard temperature (293 or 273° K.) passing per second
through a membrane 1 cm.2 in area, 1 mm. or 1 cm. thick, when
the pressure difference is 1 cm. of mercury or 1 atm. When
only comparative data are being considered, the original units
may have been retained. The data presented in Tables 94-99
represent some of the more trustworthy of the published data,
and are suitable for reference material.

TABLE 93. The chemical nature of rubber-like

membranes used in Table 94
Main Formulae of simple or
Name constituent polymerised molecules
Rubber (vulcanised) Polyisoprene, CH 2 =C—CH=CH 2
1 ,rt, ley? 111 • * "L"1 i
by sulphur
Rubber (unvulcanised) Polyisoprene „
"Neoprene" (vulcanised Polychloroprene, 0112=0—CH=K3H2
commercial) cross-linked
by sulphur
"Neoprene" (unvulcanised Polychloroprene „
commercial)
Polychloroprene (pure) —
"Vulcaplas" Polysulphide R—S—S—R—S
_S_R_S—S—R
Butadiene-methyl- — CH^C—CH=CH t ,
methacrylate inter-
polymer CH8
0Ht=O~C00CH3
CH8
Butadiene-acrylo- _ CH2==C—CH==CH2,
nitrile interpolymer 1
CH3
CHj^CH—CN
Butadiene-styrene — CH2=C—CH=*CH2,
interpolymer
L
v>»xx8
CgH.5—CH=CH2
Ethylene polymer Ethylene CH2—CHa—CH2
—CH2—CHt

In Table 94 are given permeability constants for the rubber-

like polymers of Table 93. The data summarised may be
taken as typical of rubber-like membranes. There are minor
variations only in permeability towards a given gas over the
whole grouj), with the exception of the polysulphide rubbers,
 PERMEABILITY CONSTANTS 393

which are much less permeable (~ 100-fold) than the hydro-

carbon rubbers. Small amounts of cross-Unking of the poly-
mer chains by sulphur cause no appreciable changes in the
permeability of neoprene or natural rubber.
The question of the effect on the permeability of cross-
linking of polymer chains by sulphur or oxygen still remains
uncertain. Edwards and Pickering (6) studied the influence of
ageing and of vulcanisation of rubber upon its permeability,
using as membranes rubber-coated balloon fabrics. It was
observed:
(1) That ageing of the rubber was accompanied by a
characteristic decrease in permeability, and usually by a
decrease in the free sulphur.
(2) In a series of experiments where the percentage of
combined sulphur varied from 0-3 to 2-5, no change in per-
meability was found; but in a second series where the combined
sulphur varied from 1-5 to 10 % a decrease in permeability
occurred. In each case the acetone extract—which gives a
measure of resinification and oxidation—was about the
same.
As the amount of combined sulphur is further increased, the
rubber becomes dark and hard and finally forms a compound
with 32 % of combined sulphur. This polymer of high sulphur
content is called ebonite, and it will be seen on referring to
Table 96 that the permeability of ebonite to helium is about
^ of the permeability of rubbers; to hydrogen it is about j$;
and to nitrogen < 2^0 • Thus the increased rigidity has caused
the permeability to become less, and much more selective.
The impermeability (33,34) of polysulphide rubbers may be
of technical importance. The permeability of polysulphide
rubbers of the two following types (35) has been the subject of
investigation by Sager(34):
—R

11
394 PERMEATION, SOLUTION, DIFFUSION OF GASES

TABLE 94. Permeability constants, P, for rubber-like polymers(%S)

(c.c. at 293° K./sec./cm.a/mm. thick/cm. Hg pressure)

System Temp. ° C. PxlO6

He-"neoprene" 0 00022
(vulcanised and with 30-4 0-0078
fillers) 41-5 0-0158
570 0035
730 0-048
101-3 0094
He-"neoprene" (raw, I 21-6 00039
unvulcanised) 37-6 00116
(material softening)
II 18-8 00023
340 0-0055
(material softening)
III 0 0-0006
18-8 0-0025
24-6 0-0036
He-rubber (2% S, 5 min. 19-2 0-0051
vulcanised) 30-8 0-0078
42-8 00116
570 0-0179
62-5 00195
79-2 0-0300
He-rubber (2% S, 45 min. 19-5 0-0086
vulcanised)
He-"vulcaplas" 500 0000174
^59-0 0-000342
68-5 000045
He-polyethylene rubber 170 00067
22-6 0-0082
30-8 0-0115
38-5 00184
H2-neoprene (vulcanised 17-5 00085
commercial) 18-2 00090
26-9 00128
34-6 00201
44-4 00302
520 0-0370
63-7 0-0534
Hj-butadiene-acrylonitrile 0 00032
interpolymer 200 0-0085
290 00128
41-5 0-0200
50-2 00315
65-3 0-055
78-1 0075
H2-butadiene-methyl-metha- 20 0023
crylate interpolymer
H2-polyethylene rubber 50 0053
37-2 0021
34-8 0019
22-6 0011
 PERMEABILITY CONSTANTS 395

TABLE 94 (continued)
System Temp. °C. PxlO*
H2-polystyrene-butadiene I 19-9 0-0084
polymer 11 210 00112
Hj-chloroprene polymer 31-8 00049
(pure) 39-7 00086
41-3 0-0088
49-9 00116
58-7 00179
69-5 00214
73-5 00411
N2-"neoprene" 271 000137
35-4 00023
441 0-0032
541 00058
65-4 00106
84-7 00222
Ng-butadiene-acrylonitrile 200 0-00061
interpolymer 381 00019
48-5 00029
59-5 0-0048
70-5 00070
78-6 00178
Nt-butadiene-methyl- 21-2 0-0028
methacrylate interpolymer 44-6 00057
540 0-0087
61-9 00132
770 0023
Ng-polystyrene-butadiene 200 00029
interpolymer 35-5 00057
500 00102
64-2 00161
A-"neoprene" 361 0-0068
52-2 00144
61-8 00224
73-7 00311
86-2 00655
A-butadiene-methyl- 200 00059
methacrylate interpolymer 30-8 00111
39-2 00162
51-8 0027
62-3 00395
A-polystyrene-butadiene I 19-5 00109
polymer 30-3 00195
40-7 0-0287
51-2 0041
64-6 0074
II 640 0036
396 PERMEATION, SOLUTION, DIFFUSION OF 1&ASES
Sager found that the rubber with four sulphur atoms per
primary molecule was the less permeable of the two, but in
agreement with Table 94 both were much less permeable than
polyisoprene rubber (Table 95). Barrer(35a) extended his work
on rubber-like polymers to rigid or inelastic membranes of

TABLE 95. Hydrogen permeability of polysulphide* and

polyisoprene rubbers at room temperature
(P in c.c. at N.T.p./cm.2/mm. thick/cm. Hg pressure)

P (polyisoprene P (for polydisulphide P (for polytetrasulphide

rubber) x 107 rubber) x 107 rubber) x 107
Sample 1 0*46 Sample 1 0*028 Sample 1 0*010
Sample 2 0*53 Sample 2 0031 Sample 2 0*015
Sample 3 0*54 Sample 3 0041 Sample 3 0*018
Sample 4 0*034 Sample 4 0*020
Sample 5 0*033 Sample 5 0*017

bakelite, ebonite and " cellophane " (Table 96). For these mem-
branes the permeability is considerably smaller than that of
elastic membranes, with the exception of vulcaplas or poly-
ethylene sulphide rubbers. This result is true for rigid mem-
branes of inorganic (SiO2, B2O3, glass) as well as of organic
substances (Table 97). Not only are rigid membranes usually
less permeable, but they are also more selective in preventing
the diffusion of inert gases of high molecular weight (N 2 ,0 2 , A),
while allowing light gases (He, H2, Ne) to diffuse (Chap. III).
Sager (36) measured the hydrogen permeability of a large
number of film-forming substances supported on a closely
woven cotton fabric. The membranes included
Inelastic:
CH 8 0H

Regenerated cellulose-

CH 8 0H
Poly vinyl alcohol CH—CH2—CH—CHt—CH
OH OH OH

Prepared from 2. 2'-dichlorethyl sulphide and sodium polysulphide.

 PERMEABILITY CONSTANTS 397
CH2OR

Cellulose esters- —o—

Polyvinyl alcohol acetate CH—CH2—CH—CH8—CH

O.CO.CH, O.CO.CH3 O.CO.CH3
Poly vinyl acetal CH—CH2—CH—CH,—CH—CHa—CH
O—C2H4—O O—CtH4— O
Polyhydric alcohol-polybasic acid resin
HO. OC—R—CO—OCH2—CH2—O—CO—R—COOCH8-CH2
Elastic:
Polychloroprene rubber (Table 93)
Polyethylene sulphide rubber (CH2)SX(CH2)2SX(CH2)2SX
Polyisoprene rubber (Table 93)

TABLE 96. The permeability, P, of organic membranes

of various kinds to gases
PxlO6
System Temp. ° C. c.c. at 293° K./sec./cm.2/
mm. thick/cm. Hg
H2-bakelite 75-5 000052
57-0 000035
450 000026
34-2 0000164
200 0000095
N2-bakelite 750 000027
56-6 0000115
47-2 0000083
37-5 0-000048
361 0-000047
200 00000095
180 0-0000085
He-ebonite 82-5 00072
670 0-0047
54-5 000305
43-7 0-00255
170 000105
H2-ebonite 830 000192
670 000141
490 000091
340 0-00045
N2-ebonite 67-2 0000025
He-cellophane 850 000050
740 000035
52-5 000012
 Table 97. Hydrogen permeabilities of film-forming materials (after Sager(SQ))

10 9 x per- 107 x permeability

Material Solvent meation rate Weight const*, jr x aensrcy u
(c.c./cm.2/sec./ (g./cm.a) (P in c.c. at N.T.P./
atm. at 25° C.) cm.2/sec./cm. thick/
atm. at 25° C.)
Regenerated cellulose sheet (cellophane) 012 00051 0-0059
Polyvinyl alcohol Water 012 00068 0-0079
Poly vinyl alcohol Water 0-23 0-0064 0015
Polyvinyl acetal Ethanol 0-81 00098 0030
Polyvinyl acetal Ethanol 0-46 0-0088 0041
Polyhydric alcohol-polybasic acid resins:
Ethylene glycol citrate Ethanol 012 00109 00126
Diethylene glycol citrate Ethanol 012 00119 0-0138
Glycerol phthalate 0-23 00122 0-028
Ethylene glycol phthalate 0-35 00125 0044
Diethylene glycol phthalate Ethyl acetate 0-58 0-0105 0061
Glycerol succinate (10%) and 116 0-0081 0094
Diethylene glycol succinate acetone (90%) 104 0-0109 0114
Glycerol sebacate 1-23 00112 0143
Diethylene glycol sebacate 1-86 00102 0190
Polyvinyl cellulose esters:
Polyvinyl acetate (high viscosity) Ethanol 5-5 00092 0-51
Polyvinyl acetate (high viscosity) Ethanol 3-6 00098 0-35
Polyvinyl acetate (low viscosity) Ethanol 11-6 0-0095 1-10
Polyvinyl acetate (low viscosity) Ethanol 151 0-0098 1-48
Cellulose nitrate Mixture 2 8-6 00122 105
Cellulose nitrate Mixture 2 110 00112 1-22
Cellulose acetate Mixture 1 17-4 0-0078 1-36
Cellulose acetate Mixture 1 24-4 00061 1-49
Cellulose acetostearate Benzol 19-7 00092 1-81
Rubber substitutes:
Polyethylene sulphide (Thiokol A) — 0-23 00322 0-074
Polyethylene sulphide (Thiokol A) — 0-58 00176 0-101
Polychloroprene (Duprene) 6-85 00122 0-84
Ether-soluble rubber Toluol 10-8 00095 103
Smoked sheet rubber 22 0-0067 1-47
Smoked sheet rubber — 25-6 00057 1-46
Solvent mixture 1: ethylene dichloride 60%, ethyl alcohol 10%, methyl cellosolve 10%, cellosolve acetate 5%, ethyl acetate 15%.
Solvent mixture 2: toluol 20%, ethyl alcohol 60%, ethyl acetate 10%, cellosolve 10%.
 PERMEABILITY CONSTANTS 399

Sager's data do not give a strict comparison or measure of

the permeabilities, since it was not possible to obtain films of
identical thickness, nor were the actual thicknesses stated.
The permeation rate defined as c.c./cm.2/sec./atm. at 25° C. is
not an absolute measure of permeability. If, however, one
uses the quantity
Permeation rate x weight of film per unit area,
one has a good approximation to comparable and absolute
data, since Variations in density are less important. The true
permeability constant , P (in c.c. at N.T.p./cm.2/sec./cm,
thick/atm. at 25° C.) is given by the relation
Permeation rate x weight of film in g. per sq. cm.
~~ Density of film (g. per unit volume)
It will then be remembered that the units of Tables 94 and 97
are different (cm. thickness for mm. thickness, and atm. for
cm. pressure), and that the constant P of Table 94 must be
multiplied by 7-6 to compare with the constant P of Table 97.
The second column of the table gives the solvent used in
forming the film on the cotton weave support.
Sager (36) considered that the data showed that permeability
was governed by the chemical nature of the films. The films
included crystalline, fibrous and amorphous substances, and
the degree of polymerisation varied over a wide range, but
no simple connection could be observed between permeability
and crystalline structure or degree of polymerisation. This
aspect of the permeability requires examination of a limited
number of chemically different film types and a controlled
degree of polymerisation. There is, however, a general corre-
spondence between the hydrogen permeability in a given
chemical type, and the hydrogen solubility in analogous un-
polymerised molecules. The permeability of materials rich in
hydroxyl groups ("cellophane", cellulose, polyvinyl alcohol,
and hygroscopic resins) is very low, corresponding to the small
solubility of hydrogen in glycerol or water. As soon as ester
groups are introduced (polyvinyl and cellulose esters), the
400 PERMEATION, SOLUTION, DIFFUSION OF GASES

permeability towards hydrogen increases, to parallel the

higher solubility of hydrogen in esters. Similarly, the hydrogen
solubility in liquid hydrocarbons is considerable and so one
finds a large permeability of rubber to hydrogen. Hydrogen
is very sparingly soluble in carbon disulphide; therefore it
should diffuse with difficulty through polysulphide rubbers.
This analogy should not be carried too far, since the per-
meability is conditioned both by the solubility of the gas
and by its diffusion constant within the polymer.
De Boer and Fast (37) studied the hydrogen permeability of a
number of derivatives of cellulose. The permeability constants
at 0°C. (c.c. at N.T.p./sec./cm.2/mm. thick/cm. Hg) were:
Regenerated cellulose P x 107 = 0-000047
Triacetyl cellulose Px 107 = 00140
Nitro-cellulose P x 107 = 0-0092
Celluloid P x 107 = 00164
These permeability constants are consistent with those in
Table 97, and emphasise again the small permeability of
cellulose and its derivatives. From the data of Table 98 one
may construct a table showing the range of permeabilities to
hydrogen of different chemical types *of polymer.

TABLE 98. Range in relative permeability in

various polymers
Relative permeability
Type of substance towards hydrogen
at 20° C.
Rubbers (natural and synthetic) l-(M)-6
Polyvinyl and cellulose esters 1-8-0-2
Polyhydric alcohol-polybasic acid resins (1-9-1-3) x 10- 11
Polysulphide resins (Thiokol A) (1-0-0-7) x 10"2
Polysulphide rubbers (6-3-31) x 10-
"Cellophane", polyvinyl alcohol and (8-0-0-05) x 10-*
acetal, and regenerated cellulose

There is a 103-fold variation among the permeability con-

stants, the rubbers being among the most permeable of the
membranes.
 TABLE 99. Relative permeabilities of membranes^) (hydrogen permeability as standard)
Gas
Membrane
NH8 W
H, He AT N8 o, CO coa
Rubber (vulcanised), 25° C. 1-00 0-62 016 0-44 2-88 8-00
Rubber (unvulcanised), 20° C. 1-00 0-30 019 011 0-35 016 2-50
"Neoprene" (vulcanised), 20° C. 1-00 0-61 0-29 010
Chloroprene polymer, 20° C. 100 0-22
Butadiene-acrylonitrile interpolymer, 1-00 008
9rt° C
Butadiene-methyl-methacrylate inter- 1-00 0-26 011 — — — —
polymer, 20° C. Q
O
Butadiene-polystyrene interpolymer, 1-00 — 0-95 0-27 — — — — izj
ono /i
Ethylene polymer, 20° C. 1-00 0-75 -— -— — —
— H
Polysulphide rubbers (at room tern-
"Vulcaplas" 0-0026 (P for H2-neoprene as standard)
Sager's disulphide polymer 0066 (Sager's value of PHa-rubber as standard)
Sager's tetrasulphide polymer 0033
402 PERMEATION, SOLUTION, DIFFUSION OF GASES

Another interesting comparison can be made by considering

the relative permeabilities in a given membrane of a series of
gases. The relative permeation velocities bear no simple
relation to the molecular weights of the diffusing gases. The
ratio of the permeation velocities (6,28,38,2) for CO2 and H 2
prove to be almost independent of the sample of polyisoprene
rubber employed, since for nine samples the average ratio
°f ^coi/^Hi w a s 2-76, and the extreme variations were from
3«03 to 2*48. The permeation rate ratios for each of a series of
gases in rubbers of different chemical types are found to show
minor variations only (Table 99).

TABLE 100. Relative permeation velocities in membranes

(O2 as standard)
Membrane
Gas
Celluloid Rubber Gelatin
2-47 2-86 100
100 100 100
CO* 8-98 906 413
250 31-9
so 2 61-6 360 95-2

The same absence of any connection of relative velocities

with molecular weights (as required in effusion or molecular
streaming) is found when gases diffuse through widely differing
membranes such as celluloid, rubber and gelatin (32) (Table 100).
Once more rather striking regularities in relative velocities are
revealed, although irregularities and specific effects are be-
ginning to creep in.

The variables in permeation through membranes

of groups A and B
The pressure, area of membrane, thickness of membrane, and
the temperature are the possible variables in the permeation
kinetics. If (as is shown later, p. 413) Henry's law, S = kp,
applies to the solubility S of gas in the polymer substance,
dC d2C
and the Fick law — = D ^-^ applies to the diffusion process,
 PERMEABILITY CONSTANTS 403

within the material, in the steady ^tate of flow, one would

find that the permeation rate was proportional to the pressure
difference, the area of the membrane, and inversely to the
thickness of the membrane.

The pressure difference

In the steady state the velocity of diffusion for a given con-
stituent of a gas mixture is proportional to the difference in
pressure of that constituent between the ingoing and outgoing

Q. 200

o
a ISO
/

A T 0°C

PRESSURES IN ATMOSPHERES

Fig. 141. The effect of high pressure upon permeation rates.

surfaces of the membrane (33,37,6,39,40). This law holds for

pressures which are not too great (2), and is a characteristic of
non-specific types of activated diffusion (see Chap. III). The
velocity of diffusion is the same when hydrogen diffuses into
a vacuum or into an atmosphere of air (41). Similarly, when a
mixture of oxygen and nitrogen diffuses into hydrogen, the
rate of permeation of each gas is practically unaltered by the
presence of the other gases (^i). At high partial pressure
differences (2) (up to many atmospheres) the rates of diffusion
through pure rubber both for hydrogen and carbon dioxide
increased with pressure, the increases being shown in Fig. 141.
The departure from linearity does not appear, in the case of
hydrogen at any rate, to be due to breakdown of the perfect
26-2
404 PERMEATION, SOLUTION, DIFFUSION OF GASES

gas laws at these pressures, since Wustner (see Chap. Ill) found
linear permeation rate-pressure curves for hydrogen and silica
up to 800 atm. Rather the effect might be ascribed to an
influence of pressure upon the membrane itself (compression
or distension) and suggests that deformation can modify
permeabilities.

O-002J OOO3O 0-0033 0-0036.

{Temperature in
Fig. 142. Influence of temperature upon the permeability constants of
elastic polymer-gas systems.
Curve 1. He-vulcanised rubber
Curve 2. H2-vulcanised neoprene.
Curve 3. He-vulcanised neoprene.
Curve 4. Argon-vulcanised neoprene.
Curve 5. N2-vulcanised neoprene.
Curve 6. He-vulcaplas (polysulphide rubber).

The thickness of the membrane

In the steady state of diffusion the velocity of diffusion is
inversely proportional to the thickness of the membrane.
This is shown by the results of Edwards and Pickering (6) who
allowed hydrogen to diffuse through rubber membranes of
varying thickness, and plotted the reciprocal of the per-
 PERMEABILITY CONSTANTS 405
meability (called the impedance) against the thickness. Their
graph was linear.
The influence of temperature
The most interesting relationships were revealed by the study
of the influence of temperature on the permeation rate. I t was
found by Shakespear(40) that the temperature coefficient of
the permeation rate was the same for all differences in

3-3

Fig. 143. Influence of temperature upon the permeability constants of

inelastic polymer-gas systems.

partial pressure, and all thicknesses of membrane. Graham (28)

long ago showed that the temperature coefficient was ab-
normally large. This is also clear from the data of Edwards
and Pickering (6), Taylor, Herrmann and Kemp (42), and
Dewar(2).
It may be shown that the permeabilities conform to the
equation p = p

a relationship first pointed out by Barrer(43). The exponential

temperature coefficient is shown in Figs. 142 and 143(33,43),
406 PERMEATION, SOLUTION, DIFFUSION OF GASES

for elastic and for inelastic membranes. The temperature

coefficients of the permeability constant in cal./mol. of gas
are collected in Table 101. I t is interesting to find that
these coefficients are not very different for the impermeable
inelastic membranes and the permeable elastic membranes.
The main differences in the permeability must therefore
come from the temperature independent factor, Po, in the
equation
P = Poe-WRT.
As Table 101 shows, Po is very much smaller for cellulose and
its derivatives than for elastic polymers. The difference in JF^
for rubbers and for cellulose must be due partly to the
larger solubility of hydrogen in rubber. It is then possible
that in cellulose, because of the low solubility, diffusion
is predominantly of grain-boundary type, while in rubber
it can occur readily inside the micelles composing the
colloid.

THE AIR PERMEABILITY OF GROUP C OF F I G . 132

Previously we have considered coherent membranes without

pin-holes or capillary channels. Most papers, fibreboards, and
leathers have capillaries down which streamline (44) or
Knudsen(45,46) flow may occur, or pin-holes which may act as
orifices for orifice flow(47) or effusion(48) (Chap. II). Stream-
line flow conforms to the equation

where vx = the volume flowing per unit time into the tube at
pressure pl9
v2 = the volume flowing per unit time into the tube at
pressure p2,
Ap=p1-p2>
ri = the viscosity of the gas,
I = the length of the pore, of radius r.
 TABLE 101. Temperature coefficients, E9 of permeability constants (33,35a) in cal./moL

E (caL/mol.)
He H, N, Ar CO co2 H2O
Elastic membrane:
Rubber (vulcanised) 6,300, 6000, — — — — 7600 2800
A ACifi
D,*UU DOUU
Rubber (unvulcanised) 9500
___ 8200 9500 9600
7,800
*' Neoprene'' (vulcanised) 8,800 8300 10,500 10,700 — — — —
Butadiene-acrylonitrile inter- — 8200 9,800 — — — — —
poiymer —
Butadiene-methyl-methacrylate 9,500 8,850
interpolymer
Butadiene-polystyrene inter- — — 7,900 7,900 — — — —
polymer
Chloroprene polymer — 8300 — — — — — —
Ethylene polymer 7,700 8050 — •
— — — — —
Inelastic membrane:
Bakelite 5100 10,500
Ebonite 5,900 6500 — —
"Cellophane" 10,100 — — — — — —
Celluloid 5600
Nitrocellulose 5700
Triacetyl cellulose 7500 —
Regenerated cellulose — 7600 — — — — — —

Hj,-celluloid: P 0 =4-4 x 10~5. Ha-regenerated cellulose: P o =0-59 x 10~5.

Hj-nitrocellulose: P 0 =3-2 x 10~6. H,-neoprene: P 0 =2240 x 10~5.
H,-triacetyl cellulose: P o = 142 x 10"8. H,-butadiene acrylonitrile interpolymer: P 0 = 1020 x 10~5.
408 PERMEATION, SOLUTION, DIFFUSION OF GASES

For orifice flow to occur the capillary must be short enough

to act as a jet or nozzle. Under these conditions the equation
of flow becomes

Pi
2±- = c o n s t a n t ^
Ap
The constant depends upon the area of the nozzle, and not
its length.
The properties of streamline and orifice flow are summarised
in Table 102(49), and can easily be verified by reference to the
equations of flow. The very complete series of experiments by

TABLE 102. Characteristics of high-pressure gas flows

Poiseuille capillary flow Orifice flow
v/Ap is constant for low values v/Ap decreases as Ap increases.
of Ap
v/Ap is proportional to the reciprocal v/Ap is independent of the length
of length (or membrane thickness) (or membrane thickness) !
The rate of flow decreases slightly The rate of flow increases slowly
as the temperature increases, be- with temperature
cause the viscosity of the gas
increases as T increases
v, for small constant values of Ap, v, for constant values of Ap, in-
is independent of px creases slowly as px decreases

Carson (49) has tested these features for numerous papers. For
the thinnest tissues the behaviour with respect to Ap sug-
gested that flow might be intermediate between streamline
and orifice flow. Within the uncertainties of using different
samples an inverse proportionality existed between the per-
meation velocity and the thickness, while a small change in
temperature had a negligible influence upon the rate of flow.
The results are on the whole more in conformity with Poiseuille
flow than with orifice flow. The effect of altering the absolute
pressure pl9 for a constant Ap, however, was not in accord with
either theory, and was attributed by the author to elastic
deformations of the membrane under pressure. In addition,
an approximate proportionality between area and diffusion
 AIR PERMEABILITY 409

rate was observed, and it was shown that the relative humidity
of the air bore no simple relationship to the permeation
velocity. In order for Poiseuille flow to occur in these mem-
branes, the length of the capillaries must be much greater than
their diameter. I t is therefore probable that the actual
thickness of the membranes is less than the length of the
capillaries.
In Table 103 Carson's (49) data have been converted to
absolute permeabilities, and show the range of permeability
constants covered by the paper section of group C of Fig. 132.

TABLE 103. The air 'permeabilities of some membranes

Permeability
Kind of material c.c./sec./cm.2/cm.
Hg/mm. thickness
Insulating board 36-5
Filter paper 3-51
Blotting paper 1-51
Newsprint paper 0047
Antique book board 00298
Lined strawboard 00177
Supercalendered book paper 00059
Tagboard 00102
Bond paper 000141
Railroad board 000116
Solid binders board 0000465
Press board 0-9000714
Vegetable parchment 00000018

It is necessary to add that other types of flow than that of

Poiseuille (e.g. molecular streaming, or even activated
diffusion) may be occurring in the case of the least permeable
membranes cited. The sorting out of flow processes is not yet
at all complete.
Leathers
The air permeability of leather shows the same wide variations
that were observed for papers and fibreboards, and some of
the associated phenomena have been ascribed to Poiseuille
flow (50,51). The leather may be regarded as a network of
capillaries joining larger cavities within the material. These
cavities and capillaries are not constant in size. Thus while in
410 PERMEATION, SOLUTION, DIFFUSION OF GASES

TABLE 104
A. Sole leathers
Permeability x 102
Thickness c.c./sec./cm.2/mjn. thick/cm. Hg
Type of leather cm.
G*-*F F->G
English bend 0-608 910 7-30
0-438 5-76 5-40
0-490 7-40 7-80
0-420 4-65 4-74
0-390 406 3-29
Waterproofed Eng- 0-536 4-76 5-34
lish bend 0-415 1-63 1-50
French bend 0-350 4-42 414
0-370 1:42 1-32
0-420 1-97 2-21
0-382 1-85 1-85
0-624 3-94 4-08

B. Upper leathers
Box calf 0140 15-7 111
0130 14-2 7-55
0120 20-7 17-65
Gorse calf 0195 130 604
0160 110 6-00
0-210 28-2 24-60
Willow calf 0095 70 2-95
0130 28-3 22-6
0130 15-9 8-52
Glace kid 0060 0;734 0-360
0070 1-78 1-24
0050 1-22 0-60
Heavy chrome 0-312 405 2-47
tanned upper 0-310 2-78 0-485
0-280 1-21 0-422
0-320 7-42 4-36
Heavy vegetable 0-240 00128 000413
tanned upper 0-252 00176 00176
(stuffed) 0-240 000428 000418
0-230 000845 00128
Russia calf (vegetable 0-220 59-5 62-5
tanned) 0160 37-9 38-5
0180 62-8 651
Patent leather 0125 0 0
0090 0 0
0120 0 0

* G denotes "grain side"; F denotes "flesh side."

 AIR PERMEABILITY 411

certain cases the rate of flow through the leather was pro-
portional to the pressure difference (51), in others the constant
of proportionality changed with pressure, or with the treat-
ment given the membrane (50). Sometimes the rate of flow
from the flesh to the grain side is different from that in the
converse direction (52) (see Table 104).
In Table 104 are given some absolute permeabilities
for a number of leathers, calculated from data given by
Edwards (50).
Bergmann and Ludewig's(5i) figures for sole leather agree,
when converted to similar units, with those found by Edwards
in range and order of magnitude. Wilson and Lines (53)
made a study of the .air and water permeability of leather in
which they found a parallelism between these permeabilities
when the leather contained neatsfoot oil, or was finished
with collodion or casein.

Membranes in series
When the leather membranes are placed in contact (51), and
the permeabilities of the separate membranes are P1 and P2>
it was found that the resultant permeability, P, was within
6 % given by 1 x x

The reciprocal of the permeability is called the impedance, so

that the above expression means that the impedance is
additive. The same property has been observed for gas-rubber
diffusion systems (3).

THE SOLUTION AND DIFFUSION OF GASES IN

ELASTIC POLYMERS

The passage of a gas through a membrane is governed by the

equation P = —D(dC/dx), where D is the diffusion constant,
and dC/dx the concentration gradient. When the permeability
P is known, and the solubility, the diffusion constant may be
412 PERMEATION, SOLUTION, DIFFUSION OF GASES

calculated. Two methods have been successfully employed to

measure D:
(1) The solubility has been measured directly, and then the
permeability constant. Use of the Fick law in the form

(I denotes the thickness, and Cx and C2 are the concentrations

at the ingoing and the outgoing faces respectively) then allows
the evaluation of D. For measuring the solubility the same
methods may be employed as in studies of sorption equi-
libria (54), and any sensitive apparatus described for that
purpose could be employed.
(2) Another method may be used which allows one to
measure P. D, and the solubility, in one experiment. This
method was introduced by Daynes(3) and developed by
Barrer(33,55,56). It will be seen, by referring to Chap. I, p. 18,
that if one has a membrane with dissolved gas initially at a
concentration Co throughout, and with constant concentra-
tions Cx and C2 at the faces x = 0 and x = I respectively, the
rate of flow of gas through the membrane takes some time to
settle down to a steady state. If one plots the rate of increase
of concentration on the low pressure side against time, the
C-t curve approaches asymptotically a line which intersects
the £-axis at the point
g, o0-|
D 6 + 3 2_T

provided the increase in concentration is small. When Co is

zero, the intercept is

L =—

and when both Co and C2 are negligible compared with Cv the

intercept is L = Z2/6J9. Thus by a single experiment which
follows both non-stationary and stationary states offlowone
 SOLUTION AND DIFFUSION OF GASES 413

measures P and D, and then calculates the solubility, Q. The

equations n

P = D ^ ( F i c k ' s law),

Cx = kp (Henry's law),

/-a} M I T

30'S'C. Jzo'C

100 200
Time (mins.)

Fig. 144. The time lag in setting up a steady state of flow. ©,argonin butadiene-
methyl- methacrylate interpolymer; x , nitrogen in butadiene-met hyl-
methacrylate interpolymer.

give also the relation klp

The predictions of all these expressions were verified by

Daynes (3). The intercepts L were, for membranes about 1 mm.
thick, suitably large for evaluating D, as Fig. 144(33) indicates.
Most early measurements of solubility employed the static
method (i), Wroblewski(29) found that Henry's law was
obeyed: Cx = kp. The temperature dependence of the solu-
414 PERMEATION, SOLUTION, DIFFUSION OF GASES

bility does not agree with some more recent data (p. 418) for
hydrogen and air, although the actual solubilities are similar.
Htifner (30) found that rubber will sorb its own volume of carbon
dioxide at room temperature, but could not measure the
absorption of hydrogen, oxygen, or nitrogen. Reychler(3i)
found that rubber sorbed 1-06 vol. of CO2 at 18° C, and
26 vol. of SO2. The most complete and satisfactory measure-
ments of solubility by the static method were made by Venable
and Puwa(57). They showed that Henry's law was obeyed for

120
110
100 ^
90
\

50
\
40
\
30
20
10

20 49 60 eo loo no HO
c.c.gas at N.P.T. perWOc.c.rubber
Fig. 146. The influence of temperature upon the solubility of CO2 in rubber.

CO2-rubber systems, and also that the solubility of ethylene

and carbon dioxide decreased with temperature (Fig. 145).
The data plotted as log (solubility) against 1/T give exothermal
heats of solution of 3300 for CO2 and 2700 for ethylene.
Venable and Fuwa also showed that adsorption was not
important in determining the amount of gas taken up, since
the observed uptake was not altered by increasing the total
rubber surface.
Tammann and Bochow(58) measured the solubility of
hydrogen in a number of rubbers and at pressures as high
as 1150 kgm./cm.2 Even at these pressures Henry's law
was approximately valid, and the solubilities extrapolated to
 SOLUTION AND DIFFUSION OF GASES 415
1 atm. pressure were of the same order as the solubilities given
in Tables 106 and 107. When the pressure on such a hydrogen-
filled rubber was released, the rubber swelled to several times
its original volume.
Oxygen may dissolve in rubber in two ways. Physical
solution can be followed by chemical reaction. Chemical
reaction occurs autocatalytically(59), and as resinification
proceeds the rubber becomes first tacky, and ultimately hard
and brittle. When the surface area is small the rate of
reaction with oxygen may depend upon the extent of the
surface, but as the surface is increased diffusion becomes no
longer a rate controlling step, and the reaction velocity is
independent of the surface area.
The most complete and satisfactory data on solubility and
diffusion in rubbers have been obtained by the second method
of p. 412, depending on the time lag in setting up the stationary

TABLE 105. A comparison of absorption coefficients

measured by the methods (1) and (2) of p. 412
&2i°c. (c.c. at fci7° o. (c.c. at
N.T.P./C.C. rubber) N.T.P./C.C. rubber)
Gas by sorption by non-stationary
method (VenabFe flow method
and Fuwa) (Daynes)
<0-01 0040
o2 3 0073 0091
NH 9-30 410
-CO, 0-99 0-86
Air 0045 0043

state. Where comparison is possible, the solubility constants

k (from C = kp) agree well when obtained by the two methods
(Table 105). Agreement is not good for ammonia which,
however, approaches its equilibrium sorption value very
slowly The agreement for the other gases provides support for
the assumption made in the time lag method that the con-
centrations just inside the ingoing and outgoing surfaces of
rubber are governed by Henry's law, C = kp.
Barrer(33) used the flow method (2) to measure P , D, and k.
416 PERMEATION, SOLUTION, DIFFUSION OF GASES

It was found that the intercepts L and the permeabilities

obeyed the relationships
L = LoeE'RT (Fig. 146),
E RT
P = Poe- if (Figs. 142, 143).
Thus since
one finds D = Doe~EiRT9

2-0

0-002/5 O'OOSO O-00325 O-OO3Z

(Temperature in °K).
Fig. 146. Curves showing that log (L) is a linear function of l/27(33>.
Curve 1. A-neoprene.
Curve 2. Na-neoprene.
Curve 3. N2-styrene-butadiene interpolymer.
Curve 4. A-styrene-butadiene interpolymer.
Curve 5. Ha-neoprene.
Curve 6. H2-butadiene-acrylonitrile interpolymer.

so that diffusion in polymers is activated. From the Fick law

one obtains = koeAHiRT,

where AH denotes the heat of solution. Although activated
diffusion occurs in the polymers, it is non-specific, since any
gas of suitably small molecular dimensions will diffuse.* It
• And also large molecules which distend the rubber as they difiPuse.
 SOLUTION AND DIFFUSION OF GASES 417

differs from the diffusion of gases in metals, which is specific

to certain gases and certain metals with which the gas can
react chemically (N2-Fe), or form an alloy (H2-Pd).
In Tables 106-108 are given data on the solubility and the
diffusion constants of gases in polymers. The heats of solution

TABLE 106. Solubility, permeability and diffusion constants in

vulcanised rubber at 25° C. (calculated from data of
Daynes®) and Edwards and Pickering (6),)
Solubility
c.c. at N.T.P./C.C. Diffusion constant
Permeability constant
Gas rubber/atm. (25° C.) c.c. at N.T.P./(25° C.)
2
pressure sec/cm. mm./cm. Hg cm.2 sec."1

H2 0-040 0045 x 10-« 0-85 x 10- 5

O2 0070 0-020 x 10-« e 0-21 x 10-«
N, 0035 00071 x 10~ 0 1 5 x 10~56
CO2 0-90 0-132 x 10- 8 011 x 10"

in Table 107 are not as accurate as the actual solubilities,

since the latter decrease only very slowly as the temperature
rises. It is also to be noted that the solubility of argon is
greater in all cases than that of nitrogen, which accounts in
part for the higher permeability of rubbers to argon (Tables 107
and 94).
The solution of gases in rubbers occurs exothermally even
in constant volume systems (33,60), with the possible exceptions
of helium and hydrogen. This must be contrasted with the
endothermic heats of solution of many gases in organic
liquids (61). The solubility constant k for a perfect gas is
related to the standard free energy of solution, AO0:

But since = AH-TA8,

one may calculate both the free energy and entropy of the
solution process (33). The solubility constant of gases in organic
liquids is practically the same as the solubility constant in
organic polymers, and therefore the difference in the heats of
solution is due to the entropy term. When comparable entropy
BD 27
418 PERMEATION, SOLUTION, DIFFUSION OF GASES

TABLE 107. S olubil ities of gases in organic polymers

Solu-
bility k
Polymer Gas Temp. (c.c./c.c. lo Is lo k AH
°C. rubber/ ogio og So 2 - 3 0 # T
atm.)
"Neoprene" H2 0-0 0065 9 7
1 07 °
(vulcanised) 17-0 0051
270 0053
361 0051
46-5 0050
"Neoprene" A 361 0155 f , ,OA 1630
lost K — 1*86 l
(vulcanised) 52-2 0-141
61-8 0117
73-7 0106
86-2 0102
"Neoprene" N2 271 0054
loff k 2*28 4-
(vulcanised) 36-4 0050
541 0044
65-4 0-040
84-7 0038
Chloroprene Ha 31-8 0115 1600
iU
polymer 41-3 0-103 ol0"' *• "** ' A.ftfkifi
49-9 0097
60-8 0090
69-5 0083
73-5 0-082
Butadiene-acrylo- N, 170 0063
log 1 " 1 1 I1 7 0 °
nitrile inter- 38-1 0-050
polymer 48-5 0-048
59-5 0040
70-5 0038
Butadiene-acrylo- Ht 0 0040 loff Jc— l*ftft-4-
nitrile inter- 20 0*037
polymer 29 0036
41-5 0033
50-2 0036
Butadiene-methyl- N, 39-5 0-084 200
methacrylate 550 0-075 lQS»fc=--2-15-1 °
interpolymer 660 0071
78-0 0-060
Butadiene-methyl- A 200 0134 1 4 5 0
• V I A, io k- 1*95
methacrylate 30-8 0-125
interpolymer 39-2 0112
51-8 o-yo
52-3 0099
Butadiene-styrene 200 0094 ! 7 1 Q^ 1000
interpolymer 35*5 0-086 log10A- 1 8 / |4 6 o y
(sample I) 500 0082
64-2 0080
(A 640 0-168)
Butadiene-styrene A 19-5 0-218
log * - 1*63 | 1 1 0 °
interpolymer 30-3 0-210
(sample II) 40-7 0196
51r2 0186
64-6 0170
(N, 64-6 0-101)
 SOLUTION AND DIFFUSION OF GASES 419

data are obtained for liquids (62) and polymers (33), it appears
that there is a larger entropy decrease by 4 or 5 units for the
^ Gas dissolving in polymer
than for the process
Gas dissolving in liquid-
According to the theory of polymer-monomer solutions an
additional entropy is to be expected, there being R entropy
units difference in the two processes(62a). This entropy
difference is configurational; in other words, there are more
ways of mixing flexible long chain molecules with monomer
molecules than there are of mixing two simple molecular species.
The activation energy for diffusions in rubbers is as large
for helium and for hydrogen as the corresponding energies in
rigid inorganic membranes of silica glass (Chap. III). The
energy of activation rises as the molecular weight of the
diffusing molecule increases, although the increases in the
energy with molecular weight are very much smaller than
those found in rigid membranes of silica glass. The influence
of molecular weight upon the activation energy can be
illustrated by reference to published data upon gas-silica (626),
gas-heulandite(63), gas-cellulose (35a, 37), and gas-neoprene(33)
diffusion systems (Table 109).
One feature which has to be explained is the magnitude of
the activation energy in rubbers, since the internal elasticity
on a molecular scale might be expected to reduce the energy
barrier involved in migration. When one calculates the energy
needed to cause a gas atom to pass through an elastic two-
dimensional crystal, it can be shown (33) that the potential
energy barrier becomes very small indeed with only minor
elastic displacements of the components of the two-dimensional
lattice. Barrer advanced the theory that the major part of
the energy of activation was the energy needed to create
'' holes'' in the three-dimensional rubber network. * The energy
• The theory of holes in the rubber substance agrees with the model of a
rubber polymer given in Fig. 140.
27-2
420 PERMEATION, SOLUTION, DIFFUSION OF GASES

TABLE 108. Diffusion constants in polymers

(D = Doe-ElRT cm.2 sec.-1)
D
Svstein Temp. cm,2 sec.~~* A> E
°C. xlO5 cm.2 sec.-1 cal./mol.

H2-"neoprene" 0 0037 90 9,250

" (vulcanised) 17 0103
27 0-180
36-1 0-297
46-5 0-481
A-"neoprene" 361 0-033 54-6 11,700
(vulcanised) 52-2 0-078
61-8 0145
73-7 0-253
86-2 0-484
N2-"neoprene" 271 0019 79 11,900
(vulcanised) 35-4 0034
441 0055
541 0096
65-4 0180
84-7 0-450
H2-chloroprene 31-8 0-33 39-4 9,900
polymer 41-3 0-56
49-9 0-91
60-8 1-44
69-5 21
73-5 2-4
Hj-butadiene-acrylo- 0 0061 5-44 8,700
nitrile interpolymer 200 0177
29-0 0-27
41-5 0-46
50-2 0-66
N2-butadiene-acrylo- 17 00066 28-1 11,500
nitrile interpolymer 381 0029
48-5 0044
59-5 0-088
70-5 014
N2-butadiene-methyl- 39-5 0041 38 11,500
methacrylate 550 0092
interpolymer 660 016
(sample I) 78-0 0-29
A-butadiene-meth\ 1- 20 0034 151 10,300
methacrylate 30-8 0062
interpolymer 39-2 0112
(sample II) 51-8 0-186
62-3 0-309
A-butadiene-styrene 19-5 0038 1-84 9,000
interpolymer 30-3 0-068
(sample I) 40-7 0111
51-2 0175
64-6 0-304
N2-butadiene-styrene 30-Q 0-0237 0-93 8,900
interpolymer 35-5 0-0506
(sample II) 50-0 0-095
64-2 0153
 SOLUTION AND DIFFUSION OF GASES 421
of hole formation would then consist of the van der Waals's
energy of cohesion absorbed when a number of cohering > CH2
groups in adjacent hydrocarbon chains were separated. The
fluctuations of thermal energy in the rubber micelle would
always be sufficient to maintain a certain number of these holes
in the rubber substance, because of the existence of which the
rubber can dissolve gases. There must then be a drift of dis-
solved gas in the direction of decreasing concentration gradient,
by way of these holes. The energy absorbed in producing a hole
large enough to accommodate a helium atom is a little less than
the energy needed to produce a hole large enough to accommo-
date the bigger argon atom, and so E
&rgon

TABLE 109. The influence of molecular weight upon the

activation energy for diffusion in membranes
v, x• * * (Si<>2 Iv /i i ^ J E
* (cellulose I
E cal./mol.
(neoprene) , ^ | E (heulandite)
cal./mol. conJpoundfl) |
caL/mol.
8,000 5,600
9,250 10,000
11,900 26,000
11,700 32,000
6,900 1 5400
(in rubber) (_L 301 face)

* These values of E are approximate since they include the small temperature
coefficient of solubility.

It was also suggested that plastic deformation of a rubber

under stress was due to the same causes. Sections of hydro-
carbon chain become separated from other chains by thermal
agitation, and under the shearing force become somewhat
displaced before coming together again. The temperature
coefficient for the viscosity of rubber has been given as
10,000 cal.(64), in agreement with the values 8700-11,900
cal./mol. observed (Table 108) when gases diffuse through
rubbers. Permanent displacements of the hydrocarbon chains
relatively to one another are only possible, however, if the
chains are not strongly cross-linked by sulphur or oxygen
bonds.
4 2 2 P E R M E A T I O N , SOLUTION, D I F F U S I O N OF GASES

MODELS FOR DIFFUSION IN RUBBER

A number of formulae have been derived for D, the diffusion

constant, based upon various models of the diffusion process.
The formulae all depend upon the premise of a medium in
which the diffusing particle is vibrating, and moving to
successive positions of equilibrium when a sufficient activation
energy has been acquired by the system particle-medium.
The most general expression based upon kinetic theory (Chap.
VI) is

Here v = the vibration frequency of the diffusing particle in

the medium,
n = the number of degrees of freedom in which the
activation energy E is stored,
d = the mean free path of the diffusing particle in the
activated state.
When n = 2, E
(Bradley) (66),

and if n = 1, D
The transition state theory of reaction velocities may be
applied to diffusion systems, the expression for D being (67):

where Fn, F* denote the partition functions of the system

particle-medium in the normal and activated states re-
spectively, excluding from the latter the partition function for
the co-ordinate in which diffusion occurs.
In Table 108 the diffusion constants have been expressed
by the formula D = D0e~EIR2\ and the values of Do may be
used to calculate values of the mean free path, d (Table 110).
 MODELS FOR DIFFUSION IN RUBBER 423
For all formulae involving a few degrees of freedom only
Table 110 shows that the mean free paths are very much larger
than one would expect. In Table 110 are given for comparison
the mean free paths in some liquid-liquid diffusion systems
in which it can be seen that the mean free paths are of
normal molecular magnitudes (excepting water which usually
shows anomalous properties).
An explanation of this behaviour of gas-polymer diffusion
systems, compatible with the diffusion process as pictured
previously (p. 421) and involving only a small mean free path
of the diffusing particle, is that the activation energy is dis-
tributed through many degrees of freedom in the particle-
medium system. Such a distribution of energy is needed for
hole formation in the polymer. Thus one has in these systems
an analogy with "fast" chemical reactions (for example, the
unimolecular decomposition of large organic molecules). The
numbers of degrees of freedom needed to give values of d of
molecular magnitude are given in Table 111.
Eyring's formula (67) relating the diffusion constant D and
the velocity constant for passing over an energy barrier, k, is
D = id 2 ,
where d as before denotes the mean free path in the activated
state. If the formula of Wynne-Jones and Eyring(68) for a
reaction velocity constant is used, one has

k __ e-AH±/RTeAS±/R_

h
or
h J Fn h
so that

where E = E0-BT and Eo and AH± are identified. When a

value of d of 5 x 10~8 cm. is assumed, the entropies of activation
are those of Table 111. A value of d = 10 x 10~8cm. would
reduce all these entropies by 2*76 units. The large increase in
 SO

TABLE 110. Mean free paths calculated using various formulae for D and assuming v = 2-5 x 1012 seer1
hT F*
D J
System Do (cm.2 sec."1) E (cal./mol.)
> lM£r)dt 2)0=2.72^-9- d*
H2-butadiene-acrylonitrile 56 8,700 d = 1160A. (2=480 A.
polymer d / ^ = 182A.
n
N2-butadiene-acrylonitrile 28 11,500 820 316 130 H
M
H2-neoprene 9-4 9,250 476 204 74 O
A-neoprene 55 11,700 1150 410 185
N2-neoprene 78 11,900 1370 490 215
N2- butadiene- methyl- metha- 37 11,500 946 350 150
crylate interpolymer
H2O in D2O 0197 5,300 69 36 11
C6H6OH in CH8OH 3-4 xlO" 3 3,150 91 6-9 1-4 CO
M
C6H5OH in C«H6 316 x 10- 8 3,080 8-7 6-9 14 o
S-C2H2Br4 in S-C2H2C14 1-68 xlO~ 8 3,365 6-4 50 1-0
Br2 in CS2 0-43 x 10- 8 1,540 3-2 3-4 0-4

CO
 MODELS FOR DIFFUSION IN RUBBER 425
±
A8 ifit were all due to the diffusing molecule would corre-
spond to more than its entropy of solution(33). Thus the
medium itself must share in the entropy change, and so both
kinetic theory and transition state theory lead to the same
viewpoint: that the activation energy is shared in part or

TABLE 111. Degrees offreedom in acquiring energy of activation

for diffusion from Wheeler's equation, and the entropy of
activation from Wynne-Jones and Eyring's equation

Diffusion system E d* x 108 n (entropy

cm.8 sec.- 1 cal./mol. cm. units)
H2-butadiene-acrylo- 56 8,700 24 14 17-8
nitrile interpolymer
N^-butadiene-acrylo- 28 11,500 91 12 16-4
nitrile interpolymer
Ht-neoprene 9-4 9,250 6*9 14 14-2
A-neoprene 55 11,700 3-2 13 17-7
N2-neoprene 78 11,900 3-7 13 18*4
N,-butadiene-methyl- 37 11,500 3-3 12 16-9
methacrylate inter-
polymer

* d is calculated from the equation

RT
) (n—1).

9
where v =2-5 x 1018.

wholly in various degrees of freedom of the polymer or

polymer-solute system, a viewpoint already implied in the
"hole" theory of solution and diffusion in elastic polymers
(p. 421).
Since DQ =

it is seen that logD 0 is proportional to AS±. Barrer(33)

collected the available diffusion data for a variety of systems,
and plotted the number of diffusion systems having Do within
certain limits against logD0. The frequency curve of Fig. 147
was obtained, in which the full curve represents the periodicity
curve for all diffusion systems, and the dotted curves for
special types of system. It can be seen that the entropy
426 PERMEATION, SOLUTION, DIFFUSION OF GASES

of activation for gas-elastic polymer diffusion systems is

greater than of liquid-liquid or gas- and solid-solid diffusion
systems. The peak of the periodicity curve for liquid-liquid
systems lies between the peaks for the curves for rubber and
solid diffusion systems. The solid diffusion systems cover the

liquids. !cn
r \mtals.
/ ^ glasses

-J /*f Oo.
. 147. Periodicity curve for Do in D = D^RT for activated diffusions.
O Rubbers; x crystals and metals; + glasses; 0 liquids; Q surface
diffusions.

greatest range in values of Do (107-fold). This may be because

the greatest variety of structures occurs in crystals, metals,
and glasses, involving special mechanisms of diffusion.
Rubbers, and liquids, approximate to two single types, in
which structural singularities are missing, so that narrow
bands in the periodicity curve cover the observed range of
values of D«.
 MODELS FOB DIFFUSION IN RTTBBBB 427
The values of DQ and k0 in the expressions
D = Doe-E'BT,
k = koe-Bl«T
have been related by the relation (p. 423)
Do = kod*.
This equation permits a comparison of unimolecular reactions
and activated diffusions. The distribution curve corresponding
to Pig. 147 for unimolecular reactions (69) has its peak at
k0 = 1013"~14, and a spread of 1013-fold. The corresponding
curve for bimolecular reactions (70) has a peak at k0 = 10 u
(litres x g. mol."1 x sec."1) and a spread of 10*6-fold. If a value
of d = 2*1 x 10~8 cm. is assumed, the maxima in Fig. 147
occur at
kQ = 3 x 1016sec.-x for rubbers,
k0 = 3 x lO^sec.-1 for liquids,
i o = l x 1012sec.~"1 for crystals and glasses.
One thus notes the similarity in activated diffusions and in
unimolecular reactions.

REFERENCES
(1) Carson, F. Bur. Stand. J. Res.,'Wash.,.12, 567 (1934).
(2) Dewar, J. Proc. Roy. Instn, 21, 813 (1914-16).
(3) Daynes, H. Proc. Roy. Soc. 97 A, 286 (1920).
(4) Schumacher, E. and Ferguson, L. J. Amer. chem. Soc. 49, 427
(1927).
(5) Rayleigh, Lord. Proc. Roy. Soc. 156 A, 350 (1936).
(6) Edwards, J. and Pickering, S. Sci. Pap. U.S. Bur. Stand. 16,
327 (1920).
(7) E.g. Benton, A. F. Industr. Engng Chem. 11, 623 (1919).
Buckingham, E. Tech. Pap. Bur. Stand. 14 (T 183) (1920).
Doughty, R., Seborg, C. and Baird, P. Tech. Asa. Papers, 15, 287
(1932).
Emanueli, L. Paper Tr. J. 85 (TS 98) (1927).
Gallagher, F. Paper, 33, 5 (1924).
(8) E.g. Barr, G. J. Text. Inst. 23, 206 (1932).
Marsh, M. J. Text. Inst. 22 (T 56) (1931).
(9) Schiefer, H. and Best, A. Bur. Stand. J. Res., Wash., 6, 51 (1931).
428 PERMEATION, SOLUTION, DIFFUSION OF GASES
(10) Mathieu, M. La Nitration de la Cellulose, Actualites Scientifiques,
No. 316. Paris, 1936.
(11) Trillat, J. J. C.R. Acad. Sci., Paris, 197, 1616 (1933).
(12) McBain, J. W. Sorption of Gases and Vapours by Solids, Figs.
110 and 111. Routledge, 1932.
(13) Sheppard, S. E. and Newsome, P. T. J. phys. Chem. 39, 143
(1935).
(14) Hagedorn, M. and Moeller, P. Veroff. ZentLab. Anilin. photogr.
AbU 1, 144 (1930).
(15) Astbury, W. and Woods, H. Nature, Lond., IIS, 913, Fig. 1
(1930).
(16) Bernal, J., Fankuchen, I. and Perutz, M. Nature, Lond., 141,
523 (1938).
(17) E.g. de Boer, J. H. Trans, Faraday Soc. 32, 10 (1936).
(18) Mark, H. and Meyer, K. Der Aufbau der Hochpolymeren
Organischen Naturstojfe. Leipzig, 1930.
(19) Meyer, K. Kolloidzschr. 53, 8 (1930).
(20) Siefriz, W. Protoplasma, 21, 129 (1934).
(21) Frey-Wyssling, A. Protoplasma, 25, 262 (1936).
(22) Guth, E. and Rogowin, S. S.B. Akad. Wiss. Wien, na, 145, 531
(1936).
(23) Clews, C. B. andSchosberger, F. Proc. Roy. Soc. 164A, 491 (1938).
(24) Sauter, E. Z. phys. Chem. 36B, 405, 427 (1937).
(25) Staudinger, H. Ber. dtsch. Chem. Ges. 62B, 2893 (1929).
(26) Staudinger, H. and Husemann, E. Ber. dtsch. Chem. Ges. 68B,
1691 (1935).
(27) Staudinger, H., Heuer, W. and Husemann, E. Trans. Faraday
Soc. 32, 323 (1936).
(28) Graham, T. Phil. Mag. 32, 401 (1866).
(29) Wroblewski, S. Ann. Phys., Lpz., 8, 29 (1879).
(30) Hufner, G. Ann. Phys., Lpz., 34, 1 (1888).
(31) Reychler, A. J. Chim. phys. 8, 617 (1910).
(32) Kanata, K. Bull. Chem. Soc. Japan, 3, 183 (1928).
(33) Barrer, R. Trans. Faraday Soc. 35, 628, 644 (1939).
(34) Sager, T. P. Bur. Stand. Res., Wash., 19, 181 (1937).
(35) Martin, S. and Patrick, J. Industr. Engng Chem. 28, 1144
(1936).
(35a) Barrer, R. Trans. Faraday Soc. 36, 644 (1940).
(36) Sager, T. P. Bur. Stand. J. Res., Wash., 13, 879 (1934).
(37) de Boer, J. H. and Fast, J. Rec. Trav. chim. Pays-Bos, 57, 317
(1938).
(38) Kayser, H. Ann. Phys., Lpz., 43, 544 (1891).
(39) Barr, G. Tech. Rep. Adv. Comm. Aero., Lond. (ref. by fa.
Daynes (3)).
(40) Shakespear, G. Unpublished (ref. by H. Daynes (3)).
(41) Shakespear, G., Daynes, H. and Lambourn. A brief account of
some Experiments on the Permeability of Balloon Fabrics to Air.
Adv. Comm. for Aeronautics (ref. by H. Daynes (3)).
 REFERENCES 429
(42) Taylor, R., Herrmann, D. and Kemp, A. Industr. Engng Chem.
28, 1255 (1936).
(43) Barrer, R. M. Nature, Lond., 140, 107 (1937); Trans. Faraday
Soc. 34, 849 (1938).
(44) Meyer, O. Ann. Phys., Lpz., 127, 253 (1866).
(45) E.g. Knudsen, M. Ann. Phys., Lpz., 41, 289 (1913).
(46) Clausing, P. Ann. Phys., Lpz., 7, 489, 569 (1930).
(47) E.g. Buckingham, E. Tech. Pap. Bur. Stand. T 183 (1920-21).
(48) E.g. Thomson, W. Sci. Papers, 2, 681, 711. Cambridge, 1890.
(49) Carson, F. Bur. Stand. J. Res., Wash., 12, 587 (1934).
(50) Edwards, R. J. Soc. Leath. Tr. Chem. 14, 392 (1930).
(51) Bergmann, M. and Ludewig, S. J. Soc. Leath. Tr. Chsm. 13, 279
(1929).
(52) Bergmann, k . J. Soc. Leath. Tr. Chem. 12, 170 (1928).
(53) Wilson, J. and Lines, Q. Industr. Engng Chem. 17, 570 (1925).
(54) McBain, J. W. The Sorption of Oases and Vapours by Solids.
Routledge, 1932.
(55) Barrer, R. M. Phil. Mag. 28, 148 (1939).
(56) Trans. Faraday Soc. 36, 1235 (1940).
(57) Venable, C. and Fuwa, T. Industr. Engng Chem. 14, 139 (1922).
(58) Tammann, G. and Bochow, K. Z. anorg. Chem. 168, 263 (1928).
(59) Kohman, G. J. phys. Chem. 33, 226 (1929).*
(60) Bekkedahl, N. Bur. Stand. J. Res., Wash., 13, 411 (1934).
(61) Horiuti,J. Sci. Pap. Inst. phys. chem. Res., Tokyo, 17,125 (1931).
Lannung, "A. J. Amer. chem. Soc. 52, 68 (1930).
(62) Bell, R. P. Trans. Faraday Soc. 33, 496 (1937).
(62a) Barrer, R. and Skirrow, G. J. Polymer Sci. 3, 564 (1948).
(626) Barrer, R. M. J. chem. Soc. p. 378 (1934).
(63) Tiselius, A. Z. phys. Chem. 169 A, 425 (1934).
(64) Ewell, R. H. J. appl. Phys. 9, 252 (1938).
(65) Wheeler, T. S. Trans. Nat. Inst. Sci. India, 1, 333 (1938).
(66) Bradley, R. S. Trans. Faraday Soc. 33, 1185 (1937).
(67) Eyring, H. J. chem. Phys. 4, 283 (1936).
(68) Wynne-Jones, W. F. and Eyring, H. J. chem. Phys. 3, 492
(1935).
(69) Polanyi, M. and Wigner, E. Z. phys. Chem. 139A, 439 (1928).
(70) Moelwyn-Hughes, E. A. Kinetics of Reactions in Solutions,
chap, iv, p. 439 (1933).
 CHAPTER X

PERMEATION OF VAPOURS THROUGH, AND

D I F F U S I O N IN, ORGANIC SOLIDS

Diffusion of simple gases in many organic solids obeys Fick's

laws and may therefore be regarded as showing the ideal
behaviour. The diffusion of vapours, however, only approxi-
mately follows these laws, which then become limiting con-
ditions only. It is these non-ideal systems which now remain
to be discussed.

WATER-ORGANIC MEMBRANE DIFFUSION SYSTEMS

The most numerous researches have been carried out upon
water-membrane systems, for which a summarising paper by
Carson (i) has outlined the main methods of measurement. To
prevent lateral diffusion of water through the edges of the
specimen, or between it and its supports, various devices
have been adopted. The edges have been moulded in wax, or
compressed with a mixture of beeswax and resin. Shellac,
petrolatum, and rubber have been used for the same purpose;
or the cell may be mercury sealed. Films of lacquers and paints
may be made by painting the lacquer on suitable surfaces, such
as amalgamated tin-plate (2), and then peeling off the dried
membrane.
A great number of the measurements on permeability have
been made upon the complex membranes of Table 112, in which
the chemical nature of the membranes is briefly indicated.
Most of the membranes sorb water freely. Natural rubber, one
of the least hydrophilic of these substances, may if exposed
to water vary from a non-swelling non-sorbing medium to
complete dispersion in the water, according to the carbo-
hydrate and protein content (3). Nearly all cellulose or protein-
containing substances swell and sorb water strongly (4), and
the accompanying sorption is then influenced by tempera-
 WATER-ORGANIC MEMBRANE SYSTEMS 431
ture(5), vapour pressure, and markedly by saline (6,7) sub-
stances. This has led to the suggestion that sorption and
swelling represent a tendency of water to equalise the salt
concentrations inside or outside the membrane (8). In so far as
the permeability must depend on the concentration gradients
of water within the material, the water permeability will
depend upon the various factors which influence sorption
processes.
TABLE 112

Membrane. Chemical constituents

Technical rubber and its Rubber hydrocarbon (polyisopcene)
derivatives (gutta-percha, Sulphur
paragutta, insulating com- Resins )
pounds) Fillers
Carbohydrates V present as impurity
Protein 1
Salts J
Paints, varnishes, lacquers Synthetic resins (phenol formalde-
hydes, glycerol phthalates, glycol
sebacates, etc.)
Nitrocellulose and cellulose esters
Drying oils
Paraffin, aromatic and terpene hydro-
carbons
Alcohols and esters
PbCO8, Al, or C black
Leathers Natural protein membranes
Tannins
Oils
Salts
Solids such as Cr2O8, A12O8, Fe2O,
Papers Carbohydrates (cellulose, lignin)
Sizing agents (alum)
Protein
Textiles Carbohydrates (cellulose, lignin)
Proteins
Dyes
Mordants

On the basis of chemical composition one may construct a

permeability spectrum for water (Pig. 148). The permeabilities
cover a 106-fold range and may arbitrarily be regarded as
comprising four main groups. In constructing the permeability
chart the permeability constant was derived for a thickness
of 1 mm., wherever possible. Unfortunately, in many references
the thickness was not given. The range of permeabilities
432 PERMEATION AND DIFFUSION OF VAPOURS

(10 -fold) is 106-fold less than the range of air permeabilities

6

previously given (p. 383). A number of permeability data will

be given later (p. 440) and the comparison with air or hydrogen
permeabilities made in more detail (p. 443).

Wax Rubber Moisture-proofed Paints, varnishes, Cellulose

Rubber Asphalt or waxed fibre- lacquers Textiles
board, cellulose, Cellulose acetate Metal
papers, waxed Cellulose ethers gauzes
glassine Cellulose nitrates Leathers
Paints, varnishes, Parchments, Open
lacquers glassine cells
Cellulose acetate Leathers
Cellulose
-7 -5 -4 -3 -2
log permeability
Fig. 148. A water-permeability spectrum.
(Permeability in c.c./sec./cm.2/cm. Hg.)

Dependence of ^permeability upon thickness and

vapour-pressure difference
Membranes which sorb little water (e.g. purified polystyrene (9))
behave at low relative vapour pressure as indicated by Jack's
law, and the permeation velocity is proportional to I/I
(I denotes the membrane thickness). However, Fig. 149 in-
dicates that at high humidities the permeation velocity
constant rises with increasing thickness. Certain paint and
varnish films (io,ii) also obeyed the law p oc Ijl; but in other
cases marked divergences arise (12,13).
Similarly a number of researches (14,15,16,17,18,19,20) at low
humidity showed that the permeation velocity was pro-
portional to the vapour-pressure difference. The more hygro-
scopic the substance, however, the lower is the humidity above
which this relationship breaks down. Pure rubbers and waxes
sorb little water, and the permeability "constants " calculated
from Fick's law are little afiFected by variations of humidity at
the surfaces of the membrane. Many rubbers, however, contain
hydrophilic impurity and sorption occurs freely. Figs. 150(6)
and 151(9) show that the permeation rate and the sorption
isotherms follow similar courses.
 WATER-ORGANIC MEMBRANE SYSTEMS 433
Cellulose and cellulose compounds are more hydrophilic
than rubbers, and departures from the relation Permeation
velocity oc Vapour pressure difference quickly appear. In

-UK
p,= 23-6 MM.Hg
P2»OOMM.H 9

p, = 7-66 MM.Hg
P2=0O MM.Hg

•08 12 ••« '20

THICKNESS IN CENTIMETERS

Fig. 149. Permeation rate constants at high and low

humidities as a function of thicknessO).

V M»Ofi PRESSU R E - WATER ^B50RPTI0N

1
TEMI »ERAr ;PE- - —i •5-J
VtyOR > R t S SURE EATER AT 2 •c.
24
»
- •• — •ii

-«-
22

20
r Jt
? « f
a
* 18

| 14 / / JI
| 12 i\l
$ 10 It I I——
6
Ih
// —
SVTTA P

RCHJ
'
BLE COfi WNC

|
—FILL! 0 CC UNO ds)UO

y\ 0-4 0-8 1-2 1-8 2*0

PER CENT WATER
2-4 2-8

Fig. 150. Absorption of water vapour by gutta percha at 25° C. (6).

membranes made of resins, paints, and varnishes, the permea-

tion velocity again follows roughly the sorption isothermal (16)
(Fig. 152). Leathers show a very great range in their water
BD 28
434 PERMEATION AND DIFFUSION OF VAPOURS

and air permeability, but again on the whole conform to the

general principle of increased permeability "constants'* at
high humidity.

1
t«25O*C.
'CO MN

1 1 I
8 12 16 20 23 6
VAPOR PRESSURE p, IN MM.Hg
O-2O 0-40 O-60 0-90 100
RELATIVE VAPOR PRE5SURC

Fig. 151. Permeation rate constant as a function

of vapour pressure for rubber^).

35 2

1-30 /
/ •oio
<A/
MO st jre abso rption.-v / 005
^« *••
1*20 /

S-is ir
-M tur e f en »tr atiOft
|
T

10 20 30 40 50 00 70 60 90 100

RELATIVE HUMIDITY AT eo*e -PERCENT

Fig. 152. The influence of varying the water-vapour pressure upon the quantity
sorbed and upon the permeation velocity through aluminium paints (io>.

The concentration gradients established in rubber

during permeation
In the steady state, if the Fick law were valid, linear gradients
would be established. By using thin laminae (9) of rubber
joined in series, one can determine the actual concentration
gradients existing in the membranes. That there is not a linear
gradient at high relative humidities is shown by Fig. 153, in
 WATER-ORGANIC 435
MEMBRANE SYSTEMS
which one sees an actual concentration gradient {or soft
vulcanised rubber under diffusion equilibrium. The dotted
curve gives the concentration gradient calculated from the
c6ncentration-pressure isothermals of Lowry and Kohman(6).
The two curves lie close together and indicate that while
Henry's law holds at low pressure, an abnormal sorption occurs
at higher humidities.

1
U— THESE DOTTEO LINES REPRESENT LAYERS
1 | IN THE EXPERIMENTAL SAMPLE

1 I 1 1 l
1 1 1 1 I i
\ 1 1 i
1 - ACTUAL DISTRIBUTION Of WATER
(EXPERIMENTAL VALUES)

1 T 1 1 1 1
1 1 1 1
1
V ""^
1 1
-CALCULATED FROM PRESSURE VS.
CONCENTRATION DATA OF

X LOWRY AND KOHMAN

j
JS 1
1

1
1
1
1
1

1
1
i

1
I 1
«O «0 00 100 120
THICKNESS IN MILS

2M ?2 W 14 12 10 tt • 2 0
•
WET SIOE
VAPOR PRESSURE IN MM.H9 t
ORY StOE
or SAMPLE OF SAMPLE

Fig. 153.

The influence of temperature upon the

permeability constants
The data of previous sections indicates that at low humidities
one may use the linear Fick law* as an approximation which
breaks down as the humidity increases. The influence of
temperature upon permeability constants is two-fold. There
is first the effect upon the diffusion constant and secondly the
effect upon the absorption coefficient, k. Sometimes in colloidal
systems the coefficient k increases with temperature even when
* P^=D(C1- Cz)jl = Dk {px -p2)/l, where k denotes the absorption coefficient,
and the other symbols have their usual significance.
28-2
436 PERMEATION AND DIFFUSION OF VAPOURS
sorption is exothermal, due to irreversible changes in the
colloid. The data on the influence of temperature upon
permeability constants may be summarised as follows:

Impermeable membranes Permeable membranes

Rubbers, silk, and rubber-like Cellulose and cellulose compounds
polymers 0) gave permeability gave permeability constants which
constants which increased with did not depend on temperature in
temperature, often exponentially many cases a 4.15,16, J 8,19.22)
Gelatin latex gas-cell fabrics (2) The permeability constants for
gave permeabiEty constants in- porous leathers did not depend on
creasing rapidly with temperature temperature < 10)
Synthetic resins gave permeability Paint films (16) containing alu-
constants increasing exponenti- minium gave permeability con-
ally with temperature stants not appreciably depending
on temperature

A rule which was found valid in rubber-gas diffusion systems

again emerges, that the permeability constants of the least
permeable membranes are highly sensitive to temperature
changes, but that in porous membranes the constants are
independent, or slightly dependent upon, temperature.
Some data giving permeability constants and permeation
rates at different temperatures are presented in Tables 113
and 114, for systems with high temperature coefficients, and
ilegligible coefficients respectively. For a number of resin, and
rubber, membranes the curves of log P (P =• permeability
constant) against l/T are linear, and from the slopes one may
evaluate E in the equation P = Poe~EIRT. Some of these values
of E are collected in Table 115; their values are on an average
somewhat smaller than E values for the permeability of gas-
rubber systems. Diffusion of water in rubber is activated
(p. 445); and it seems that diffusion is also activated in the
resin membranes of Table 115.

Miscellaneous effects in permeation rate studies

Soluble salts, either in the membrane or in the solution in
contact with the membrane, alter the amount of water
absorbed, and therefore the concentration gradients and
velocities of permeation. The addition of salts to water in
 WATER-ORGANIC MEMBRANE SYSTEMS 437
TABLE 113. The influence of temperature upon permeability
constants far water diffusion {9)
Permeability
Temp. constant x 108
Substance °C. g./cm.2/cm./
hr./mm. Hg
Polystyrene 211 40
350 4-5
Varnished silk 300 3-4
350 4-7
Vulcanised rubber 00 50
210 6-9
250 7-3
350 8-5

TABLE 114. The influence of temperature upon permeation

rate through leather (20)
Permeation rates
mg. H2O/24 hr./l-267 cm.2
Temp. Ratio
°C. Through Through of rates
leather air space
5 71 98 0-72
20 192 286 0-67
25 236 360 0-66
30 328 505 0-65
35 430 660 0-65
40 560 863 0-65
45 765 1214 0-63

TABLE 115. Values of E in P = PoeEtRT (E in cal/mol.)

Composition of membrane
E
Non-volatile Volatile (solvents)
Bakelite and glyceryl phtha- Butyl alcohol and hydro- 8200
late(2) carbons
Cellulose nitrate, glycol seba- Esters, alcohols and aromatic 4700
cate, glyceryl phthalate(2) hydrocarbons
Bakelite, China-wood oil, Butyl alcohol, turpentine, di- 7500
castor oil and linseed oil (2) pentine, mineral spirits
»> »» 5800
99 6200
Glyceryl phthalate(2) Aliphatic and terpene hydro- 6400
carbons
9* 99 4300
Rubber O) —- 2800
438 PERMEATION AND DIFFUSION OF VAPOURS

which leather was immersed decreased the amounts of water

sorbed(7) in the order:
SO4 > citrate > acetate > Cl' > NO3 > CNS\
and Na+>K+>Li+.
It was similarly established (21) that salt solutions decreased
the permeation rates. The leather was almost impermeable
at sodium chloride concentrations of 2*5 normal.
Coatings and fillers also influence permeability. Waxing
diminished the permeability of celluloses, cellulose compounds,
and synthetic resins (2). Collodion and casein coatings decreased
the air and water permeabilities of leather (20). Sager (23) found
that carbon black incorporated in polysulphide rubbers
increased their permeability to hydrogen. On the other
hand (ii), the water permeability of paint films was diminished
by incorporating aluminium.
Sometimes irreversible sorption effects arise. For example,
the permeability of a certain rubber sample was twice as great
in contact with liquid water as when saturated vapour was in
contact with the membrane (24). Occasionally, diffusion does
not occur so readily in one direction as in the converse
direction (20).

THE PERMEABILITY CONSTANTS TO WATER OF

VARIOUS MEMBRANES

Based on the tables of Carson (i), Table 116 provides numerical

values .of the permeability constants, in c.c./sec./cm.2/cm. Hg,
and per mm. thick, when the thickness has been given. The
final column of Table 116 states whether or not the thickness
was known. The values of the permeability constants are not
strictly comparable unless this is so, especially since the thick-
ness of paper and of cellulosic membranes, paint films, or
varnish films will usually be less than 1 mm. There are, how-
ever, sufficient references for which thicknesses were given to
permit comparisons to be made. Taking Levey's (28) value for
paper (29 x 10~6 c.c./sec./cm.2/mm. thick/cm, of Hg) as an index
figure, one finds values of the permeability of the same order
 PERMEABILITY CONSTANTS 439

or less for cellulose ethers and esters, or regenerated cellulose.

It is noteworthy that certain textiles offer a resistance to the
flow of water vapour comparable with that for cellulosic com-
pounds. Moisture-proofed and waxed cellulosic substances are
about one hundred times as impermeable (when comparable in
thickness) as paper; and the paint, varnish and lacquer films
vary from one five-hundredth to one-quarter of Levey's value
for paper. Wax is more than one thousand times less permeable,
as are certain members of the rubber group. The permeability
is greater the more hydrophilic the membrane substance. One
notes the extremely low permeability of the rubber-like
polymer polyethylene tetrasulphide, only four times the
permeability of wax, and four hundred times less permeable
than paper. The synthetic hydrocarbon polymers, such as
polystyrene, show permeabilities between those of soft
vulcanised and of hard rubber, the permeability of the latter
being one-quarter that of the former.
It has already been observed (p. 432) that the range of
values of 106-fold for water permeability is much less than the
corresponding range of values for air of 1012-fold. At one end
of the permeability spectrum of Fig. 148, that of the least
permeable substances in Table 116, the water permeability
is 10- to 1000-fold greater than the air permeability.- Similarly,
for some members of the cellulosic group (e.g. ordinary re-
generated cellulose) the water permeability is 103-fold greater
than the air permeability. The hydrogen permeability, like the
air permeability, is smaller than that for water, there being
a 30-100-fold difference for rubber, chloroprene, and poly-
ethylene sulphide; but a 3000-1500-fold difference for the
cellulosic substances cellophane and cellulose acetate. Once
more it is the rule that like is more permeable to like. Hydrogen
by analogy with its solubility in hydrocarbons should dissolve
more freely in rubber, and water less freely. Typical data
are collected in Table 117. On the other hand, at the high
permeability end of the spectrum (the leathers and textiles)
the air permeability may be 10-1000 times greater for
leathers, and even 105-fold greater for textiles, than the
water permeability.
 TABLE 116. Water vapour permeabilities in cc/sec/cmf/cm. Hgjmm. thick;
and in cx.Jsec.jcm.2jcm. Hg when the thickness is not stated
Experimental details
Type of membrane 106 x per- Reference Pressure Thickness
meability Moisture Temp. difference given
gradient °C. mm. Hg or not
Wax 0021 Taylor, Herrmann and KempO) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Rubber group:
Polyethylene sulphide 0076 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Polystyrene 1-26-1-50 Taylor, Herrmann and KempO) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Balata 0-58-0-63 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Hard rubber 0-52 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Gutta-percha 0-50-0-52 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Paragutta 0-60-0-67 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Vulcanised chloroprene 0-91 Taylor, Herrmann and Kemp O) W-A-S-A-Oh 21 to 25 6 to 24 Yes
polymer
Soft rubber 2-3-2-7 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Various rubbers 0-09 Edwards and Pickering (27) W-A-S-A-PaO5 25 23-8 Yes
0-014-1-4 Schumacher and Ferguson(io> W-V-S-V-P2O5 25 23-8 Yes
0-29 Levey (28) W-A-S-50h 21 9-3 Yes
3-45 Abrams and Chilsona-*) W-A-S-50h 21 9-3 Yes
Moisture-proofed group:
Asphalt 0-40 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
Regenerated moisture- 0-864-5-75 Charch and Seroggie(io) W-A-S-lh 40 53 No
TvpnofiRH ofil 1 n 1 n s A
Regenerated moisture- 2-9 Birdseye(29) W-A-S-A-CaCl2 38 50-6 No
proofed cellulose (in- 2-9 Hydenoo) W-A-S-A-H2SO4 38 49-2 No
cluding cellophane) 2-9 Fabel(3i) W-A-S-A-CaCl2 30 32 No
2-9-8-6 Tressler and Evers(32) W-A-S-50h - 1 4 , 21 9-3 No
10 Abrams and Brabenden33> W-A-S-50h 21 9-3 No
14-4 Abrams and Chilson(i4) W-A-S-50h 21 9-3 No
23 Staedel(is) W-A-S-65h 20 6-1 No
Moisture-proofed 2-9 Abrams and Brabender(33) W-A-S-50h 21 9-3 No
fibreboard 2-9 Tressler and Evers(32) W-A-S-50h - 1 4 , 21 9-3 No
2-9 Harvey (34) W-A-S-A-CaCl2 25 23-8 No
See foot of Tabled p. 442, for key to column four.
 TABLE 116 (continued)

Experimental details
Type of membrane 106 x per- Reference Moisture Temp. Pressure Thickness
meabihty gradient °C. difference given
mm.Hg or not

Moisture-proofed group:
Waxed paper 0-58-1730 Thomas and Reboulet(35) 100h-S-A-CaCl2 30 31-9 No
3-45-62*3 Birdseye(29> W-A-S-A-CaCla 38 50-6 No
3-60-202 Hyden<30) W-A-S-A-H2SO4 38 49-2 No
14-4 Tressler and Evers<32) W-A-S-50h - 1 4 , 21 9-3 No
21-6-260 Abrams and Chilson U4) W-A-S-50h 21 9-3 No
23-244 Abrams and Brabender(33> W-A-S-50h 21 9-3 No
Waxed glassine 4-9-105 Birdseye (29) W-A-S-A-CaCl2 38 50-6 No
Miscellaneous water- 115-115 Tressler and Evers(32) W-A-S-50h - 1 4 , 21 9-3 No
proofed papers 576 Staedel(i8) W-A-S-65h 20 61 No
Synthetic resins group:
Aluminium paints 0-068-2-12 Wray and van Vorstd2) 95h-S-A-Al2O8 27 25-4 Yes
0-076-2-9 Edwards and WrayOG) 95h-S-A-Al2O8 27 25 Yes
101-374 Hunt and Lansing (17) 95h-S-A-Al2O3 27 25 No
Varnishes, lacquers, 0-26-9-4 Kline (2) W-A-S-65h 21 to 38 6 to 14 Yes
aircraft finishes
Cellulose and cellulose
corn/pound group: 0-41 21 to 25 6 to 24 Yes
Plasticised cellulose Taylor, Herrmann and Kemp <<»> W-A-S-A-Oh
acetate
Cellulose acetate 0-58-28-8 Tressler and Everso?) W-A-S-50h - 1 4 , 21 9-3 No
54 Taylor, Herrmann and Kemp <9> W-A-S-A-Oh 21 to 25 6 to 24 Yes
317 Abrams and Brabender(33) W-A-S-50h 21 9-3 No
634 Abrams and Chilson U4> W-A-S-50h 21 9-3 No
24-5 Levey (28) W-A-S-50h 21 9-3 Yes
8-1-4-5 Wosnessenski and Dubnikowdi) High h-S-Low h 20 14-9 Yes
Cellulose nitrate 20-86-3 Wing oi) W-S-A-CaCl2 40 55-4 No
5-4 Levey (28) W-A-S-50h 21 9-3 Yes
4-5 Wosnessenski and Dubnikowd3) High h-S-Low h 20 17-6 Yes
Cellulose ethers 3-6 Taylor, Herrmann and Kemp (9) W-A-S-A-Oh 21 to 25 6 to 24 Yes
7-2-151 Levey (28) W-A-S-50h 21 9-3 Yes*
See foot of Table, p. 442, for key to column four.
 TABLE 116 (continued)
Experimental details
^ype of membrane 10* x per- Reference Pressure Thickness
meability Moisture Temp. difference given
gradient °C. mm. Hg or not
Cellulose and cellulose
compound group:
Regenerated cellulose 546 Birdseye(29) W-A-S-A-CaCl2 38 50-6 No
560 Charch and Scroggie <i5> W-A-S-lh 40 5-3 No
965 Abrams and Chilson(H) W-A-S-50h 21 9-3 No
33 Levey (28) W-A-S-50h 21 9-3 Yes
1655 Gregory (36) W-A-S-63h 38,21 37 No
3240 StaedeldS) W-A-S-65h 20 61 No
5190 Trillat arid Matricon(37) lOOh-S-V-Oh 20 17-5 No
Paper 1050 Abrams and Chilsonu*) W-A-S-50h 21 9-3 No
1070 Abrams and Brabender(33) W-A-S-50h 21 9-3 No
29 Levey (28) W-A-S-50h, 21 9-3 Yes
1800-3600 StaedeldS) W-A-S-65h 20 61 No
Vegetable parchment 25 l^evey(28) W-A-S-50h 21 9-3 Yes
360 Birdseye(29) W-A-S-A-CaCl2 38 50-6 No
Glassine 23-4 Levey (28) W-A-S-50h 21 9-3 Yes
504 Fabel(si) W-A-S-A-CaCl2 30 32 No
680 Abrams and ChilsonvH) W-A-S-50h 21 9-3 No
806 Abrams and Brabender(33 W-A-S-50h 21 9-3 No
Various membranes:
Leather 14-4 Edwards (38) High h-S-Low h 15 to 21 6 to 17 No
72-1440 Wilson and Lines (20) W-A-S-A-H2SO4 5 to 45 0to72 No
Textile fabrics 45-3 Levey (28) W-A-S-50h 21 9-3 Yes
1080-1510 Sale and Hedrick(39) W-A-S-65h 38,21 37 No
1440 Barr (40) W-A-S-Warm air 35 37 No
1650-1870 Gregory (36) W-A-S-63h 38,21 37 No
Copper gauze 1730 Gregory (36) W-A-S-63h 38,21 37 No
Open cell 3140-19,700 Abrams and Chilson(i4) — — —
3620 Levey (28) — — —
4840 Gregory (36) — — — —

The key to column four is: W = water surface. A = air space. V ~ evacuated space.
S = specimen. h=
 PERMEABILITY CONSTANTS 443

TABLE 117. A comparison of some hydrogen and water

permeabilities of membranes

106 x approximate* 106 x water

hydrogen permeability permeability
Membrane at 25° C. (c.c./cm.2/ (c.c./cm.2/sec./
sec./mm./cm. of Hg) mm./cm. of Hg)
" Cellophane "f 0-00008 012-0-24 at 38° C.
Cellulose acetate 0018 540 at 25° C.
Polyethylene sulphide 0001 0076 at 21-1° C.
Polychloroprene 0011 0-91 at 21-1° C.
Rubbers 0019 0-52 at 25° C.
(smoked sheet rubber) (hard rubber)
0019 2-3-2-7 at 25° C.
(ether soluble rubber) (soft vulcanised rubber)

* In computing the hydrogen permeability a density of unity had to be

assumed, since Sager's(4i) original paper did not give his film derisities, needed
to compute the actual thickness of his membranes.
t ''Cellophane" according to Hydenoo) is manufactured in thicknesses of
0-025-0-05 mm. His permeabilities, for which the membrane thickness was not
stated, have been computed on this basis.

SORPTION KINETICS IN ORGANIC SOLIDS

Fick's linear law

dx
has been shown to be a limiting law only, for vapour-membrane
diffusion systems. This is also true of the law -^- = D-j-%
when applied to systems which swell during sorption of
vapours, and give sigmoid isotherms. Such isotherms are
given by cotton (43), wool (44), leather (41), or wood (45). More
generalised laws such as

ot ox
(see also p. 47) should be used in such systems. However, it
has been customary to employ" the simpler law, and to use
slabs of leather (42), rubber (2), bakelite(48) or cellulose deri-
vatives (46) in computing the diffusion constants D. A suitable
solution is then
444 PERMEATION AND DIFFUSION OF VAPOURS

where 6 = nWt/l2, I = the thickness of the slab, and Q, #«,

denote the amounts sorbed at time t and at equilibrium
respectively.
When water is sorbed by rigid non-swelling membranes of
bakelite, the solution given above holds with some accuracy (48).

1
iNCtICAS4 Mr
1-00 TMI CKNE ES IS
—— — —o
f\.Of\

A MOtSTURC *NO CELLULOS! : ACETATE

0*5 0-60 f EFFECT OF THICKNESS

ON REDUCCe ABSORPTION RATE .

0-50

V1
0*30

0-20

0-10
1
0
2Q 40 60 80 100 120 140 160 180 200 220 240 .260 280 300 320 340 360 380
T I M E (M!N.)
Fig. 154. Curves of relative sorption against time for water-acetyl cellulose
systems (46>. The form of the curve yf- against t is dependent on the mem-
Hfoo
brane thickness.
On the other hand, in non-rigid swelling membranes of
rubber (47) or cellulose acetate (Fig. 154) the departures were
considerable.
The sorption velocity of water in rubber samples increases
as the temperature rises. The amount sorbed after a period of
20 hr. (#20) I s given below for a typical case:
Temp. ° C. Amount sorbed in 20 hr. {Q20)
24 00110
60 00265
70 00390
100 0-079

* When ^ - > 0 - 4 , all terms in 0 save the first can be neglected, so giving
H»0O

a linear plot of In ( l - -~-\ against t. Also, for ^—<0-6, the full equation can be

Q
well approximated by ^— =
 SORPTION KINETICS IN ORGANIC SOLIDS 445

Assuming that the parabolic diffusion law {Q/Qoo)2 =

applies, the slope of the curve log Q(20) versus ijT (T denotes
°K.) gives an activation energy of 6900 cal./mol., a figure
which may be compared with 6300 cal./atom for He-rubber
to 11,900 cal./mol. for N2-neoprene. Smoked rubber, para-
rubber, and pale crepe rubber similarly gave values of the
activation energy of 7000, 9800 and 9600 cal./mol. of water
respectively.
When rubber swelled in benzene (49), or when paraffin (50
was absorbed by cured rubber and by rubber gum, the positive
temperature coefficients of the sorption velocity suggested
that in these and other organic solids vapours are sorbed by
processes of activated diffusion. The occurrence of activated
diffusion has already been established for gas-rubber (52),
cellulose, -bakelite and similar systems (Chap. IX). In liquids
also, diffusion constants (53) may be expressed by the equation
D = Doe~E/RT, although the values of E are usually smaller.

TABLE 118. Diffusion constants in miscellaneous systems

System D (cm.2 sec."1) Temp. ° C.
8
H20-cellulose acetate OD 017-1-7 x lO" 7 25
H2O-leathers(4i) 0-41-19*5 x lO" 25
DaO-H2O(46) 2-5 x 10-* 25
Ha-rubber 0-85 x lO"55 25
N2-rubber 011 x 10-6 25
CH8OH-CaH6OH 2-7 x 10- 19
C^H^Cl^CjgHaB^ 0-56 x 10-* 19

The data in Table 118 permit a qualitative comparison of

the diffusion constants of water, gases, and liquids in organic
polymers and liquids.

A MODIFIED DIFFUSION LAW FOR SORPTION OF

WATER BY RUBBERS

The presence of salts in rubbers and leathers alters the sorption

of water by these colloids (p. 436). For this reason Daynes(8)
advanced a theory of diffusion as an osmotic phenomenon, the
water being sorbed to equalise salt concentration differences
446 PERMEATION AND DIFFUSION OF VAPOURS

inside and outside the membrane. The law was then for-
mulated as dC d%

where h denotes the humidity of .the atmosphere which would

be in equilibrium with the solution at x. The relation between
C and h is given by the sorption isothermal. The diffusion
equation may now be written
dh I dh

So long as dC/dh is a constant (i.e. at low humidities), a simple

Dxi~\ may be used. The tempo of
4Z2
diffusion is then governed by a single parameter 6 = —— ,

n
TIME

Fig. 155. Absorption and desorption rate curves in rubber(57).

but in the more complex case by the parameter 6' = —=- -^r.
nD^ on,
Because dC/dh, and therefore d', increase rapidly at high
humidity, sorption must proceed more and more slowly as
humidity increases, or saturation is approached. In desorption,
as humidity decreases, the converse is true, i.e. the desorption
velocity steadily increases. The first prediction was fulfilled
when water was sorbed by vulcanised and unvulcanised
rubber(54,8), gutta-percha, and paragutta(54); and the second
 VAPOURS OTHER THAN WATER 447

was indicated by the data of Fry (55) and of Cooper and

Scott (56) (Fig. 155(57)). The simple law-^- = D - ^ would on the
other hand require the ordinates of the asymptotes to the
curves A and B to stand in the ratio 2 : 1 , while the curves
A and C should be symmetrical.

THE PASSAGE OF VAPOURS OTHER THAN WATER

THROUGH MEMBRANES

Although few studies have been made of the permeability of

membranes to vapours other than water, the data are of
considerable interest theoretically and practically. Typical
data are those of Dewar(58) and Edwards and Pickering (27)
(Table 119).

TABLE 119. Permeability of rubber to vapours at

room temperature
10* x permeability Ratio of permeability
Vapour constant (c.c./sec./cm.2/nam. constants for vapour
thick/cin. Hg pressure) and for hydrogen (23)
H20 2-41 (23) 55
1-28 (56) 291
CaH5OH 1-14 (56) 25-9
CH8C1 0*81 (23) 18-5
C2H6C1 8-75 (23) 198

Kahlenberg (60) showed that if a cell H2O/rubber/ethyl

alcohol was set up the alcohol diffuses into the water more
rapidly than water into alcohol. At 15° C. the vapour pressures
are 12 mm. for water and 32 mm. for alcohol, so that, with the
permeability constants of Dewar in Table 119, one could
predict the direction of diffusion actually observed by Kahlen-
berg. As usual the permeation velocities bear no simple
relationship to molecular size or mass, the large molecule,
©thyl chloride, being transmitted ten times as fast as the
smaller molecule, methyl chloride. The solubility of the
diffusing ^substance in the membrane is one important factor
in diffusion. Another must be the extent to which the mem-
448 PERMEATION AND DIFFUSION OF VAPOURS
brane swells. The swelling, which increases as the amount
sorbed increases, causes the polyisoprene chains to separate
more and more. One might then imagine that the looser the
structure becomes the less its resistance to diffusion would be.
Payne and Gardner (59) also showed that the permeability
of a membrane to a vapour depended upon the solvent
capacity of the membrane for the vapour. They used smooth
cellulose paper of thickness 0*1 mm. as support for the films,
which in nearly every instance, in the form of a suitable
varnish, penetrated the paper completely. The paper itself had
only a slight impedance to diffusion of the vapour. Fig. 156
gives a summary of their relative permeability data for a large
number of films, and diffusing substances.
The data of Fig. 156 give the following series for the relative
permeabilities of a number of vapours:
Glue [water-soluble, but with macropores]:
H2O > C6H6 > C2H5OH > CH3COCH3 > C2H2OeOCH3.
Rubber [soluble in C6H6, and partially in CH3COCH3]:
C6H6 > CH3COCH3 > C2H5OH > H2O.
Paraffin max [soluble in C6H6]:
C6H6 > CH3COCH3 > C2H5OCOCH3 > C a H 5 0H > H2O.
Cellulose nitrate [soluble in CH3COCH3, C2H5OCOCH3]:
CH3COCH3 > C2H5OH > C2H5OCOCH3 > C6H6 > H2O.
Saw linseed oil [soluble in CH3COCH3, C2H6OCOCH3]:
C2H5OH, CH3COCH3 > C2H6OCOCH3 > C6H6 > H2O.

REMARKS ON THE PERMEATION AND DIFFUSION PROCESSES

The nature of the movement of water, organic vapours or

gases depends primarily upon the manner in which the diffusing
substance is held inside the solid, and on the nature of the
channels. The sorption may be:
A. Van der Waals's sorption,
B. Dipole sorption in a monolayer,
 CftMEABIUTY

1. Raw linseed oil 10. Rubber

2. Uodied hoseed oil 11. Paraffin wa*
3. Rosin spar varnish 13. Glue
4. Phenolic varnuh 14. Gelatin
5. Shellac 15. Cellulose nitrate
6 Rosin 16. Plasticised cellulose nitrate
7. Alkyd ream 17. Eater gum lacquer
8 MoS.fied alkyd 18. \luminum lacquer
<* Resin-oil modified alkyd 19. Porous lacquer

Fig. 156. Comparison of permeabilities of films.

 450 PERMEATION AND DIFFUSION OF VAPOURS

C. Multilayer sorption at higher humidities, with possible

orientation,
D. Capillary condensation in pores.
The sorbed substance may be in true solution within the
solid, or it may be non-homogeneously distributed or per-
sorbed. Very often, for example in cellulose (61), van der Waals
sorption, or dipole sorption is superseded at higher pressures
by capillary condensation. Occasionally, as in certain proteins,
chemical effects are suspected (62). The mercerising of natural
cellulose is thought to cause a re-orientation of the pyranose
chains (see Fig. 137, Chap. IX, for a diagram of the unit cell)
in the manner indicated in Fig. 157 (63), the hydration causing

A B
Fig. 157. A. Section of cell of native cellulose.
B. Section of cell of hydrated cellulose.
The fr-axis is perpendicular to the plane of the diagram.

a contraction parallel to the cell direction and an expansion

in directions normal to it. Ordinary sorption does not alter
the cell, but may still affect membrane permeability since
it was observed that while desiccated cellulose membranes
were impermeable to air, the same membranes after sorbing
water became permeable to air(t>4).
The dispersion of rubber in organic liquids is dependent upon
the extent to which the isoprene chains are cross-linked by
sulphur. Excessive vulcanisation causes a complete netting of
rubber, and as would be anticipated, the extent of vulcanisa-
tion conditions the extent of sorption of benzene (65,66) or a
 PERMEATION AND DIFFUSION PROCESSES 451
similar organic liquid (Fig. 158). The limiting case is repre-
sented by a membrane of polymerised p-divinyl benzene
which neither swells in nor sorbs benzene appreciably. These
phenomena must influence profoundly
the permeability of the membrane, for
it is not likely that netted membranes
will be able to transmit large organic
vapours readily, unless the transmission
is intermicellar. Experiments on these
aspects of permeability do not appear 6 /2
Time >n hovrs
to have been made.
The peculiar nature of some of the ^g-158. Dependence ofthe
r
amount and velocity
water-protein systems is best brought of sorption upon the
out by considering X-ray data. While degree of vulcanisa-
certain natural proteins such as wool,
silk, or horn are obtained, when desiccated, in crystalline
form (67), others become denatured completely if they are
dehydrated. They then fail to give X-ray patterns corre-
sponding to a crystalline arrangement; but the X-ray patterns
of the same forms when highly hydrated may indicate a
remarkable degree of crystalline order (68). Thus the water
is not only sorbed but sometimes plays a vital part in the
protein structure. The diffusion of water in such gelatinous
membranes might have interesting features owing to the
unique role of the water in the gel.
It has been noted that the stronger the sorption of a vapour
by an organic colloid the more quickly does the permeation
process deviate from the Fick laws P = D-^- and-x- = D-r-g,
as the humidity increases. The classification of Wilson and
Fuwa(69), which gives a rough measure of the capacity of
various types of organic solid to sorb water, may therefore be
used as a guide to the degree of applicability which the Fick
laws might be expected to have. Some of the substances listed
(finely divided inorganic solids-, carbon black, silica gel, etc.)
are permeable by hydrodynamic flow, but other membranes
which have been considered in this chapter show permeation
29-2
452 PEBMEATION AND DIFFUSION OF VAPOURS

rates which are independent of the hydrostatic head of

pressure, and therefore do not transmit liquids by hydro-
dynamic flow.

REFERENCES
(1) Carson, F. Misc. Publ. U.S. Bur. Stand. M. 127 (1937).
(2) Kline, G. Bur. Stand. J. Res., Wash., 18, 235 (1937).
(3) Boggs, C. and Blake, J. Industr. Engng Chem. 18, 224 (1926).
(4) For a summary of the data see J. W. McBain, The Sorption of
Gases and Vapours by Solids, chap. XH. Routledge, 1932.
(5) McBain, J. W. The Sorption of Gases and Vapours by Solids,
p. 372. Routledge, 1932.
(6) Lowry, H. and Kohman, G. J. phys. Chem. 31, 23 (1927).
(7) Kubelka, V. Kolloidzschr. 51, 331-6 (1930); also M. Bergmann,
J. Soc. Leath. Tr. Chem. 14, 307 (1930).
(8) E.g. Daynes, H. Trans. Faraday Soc. 33, 531 (1937).
(9) Taylor, R., Herrmann, D. and Kemp, A. Industr. Engng Chem.
28, 1255 (1936).
(10) Schumacher* E. and Ferguson, L. Industr. Engng Chem. 21, 159
(1929).
(11) Wing, H. Industr. £ngng Chem. 28, 786 (1936).
(12) Wray, R. and van Vorst, A. Industr. Engng Chem. 25, 842 (1933).
(13) Wosnessenski, S. and Dubnikow, L. M. Kolloidzschr. 74, 183
(1936).
(14) Abrams, A. and Chilson, W. Paper Tr. J. 91, T.S. 193 (1930).
(15) Charch, W. and Scroggie, A. G. Paper Tr. J. 101, T.S. 201
(1935).
(16) Edwards, J. and Wray, R. Industr. Engng Chem. 28, 549 (1936).
(17) Hunt, J. and Lansing, D. Industr. Engng Chem. 27, 26 (1935).
(18) Staedel, W. Papier Fabr. 31, 535 (1933).
(19) Stillwell, S. Tech. Pap. For. Prod. Res., Lond., 1 (publ. 1926-36).
(20) Wilson, J. and Lines, G. Industr. Engng Chem. 17, 570 (1925).
(21) Bergmann, M. J. Soc. Leath. Tr. Chem. 13, 161 (1929).
(22) Martley, J. Tech. Pap. For. Prod. Res., Lond., 2 (publ. 1926-36).
(23) Sager, T. P. Bur. Stand. J. Res., Wash., 19, 181 (1937).
(24) Schroeder, P. N. Z. phys. Chem. 45, 76 (1903).
(25) Daynes, H. Proc. Roy. Soc. 97A, 286 (1920).
(26) Bergmann, M. and Ludewig, S. J. Soc. Leath. Tr. Chem. 13, 279
(1929).
(27) Edwards, J. and Pickering, S. Sci. Pap. Bur. Stand. 16, 327
(1920).
(28) Levey, H. Plastic Prod. 11, 52 (1934).
(29) Birdseye, C. Industr. Engng Chem. 21, 573 (1929).
(30) Hyden, W. Industr. Engng Chem. 21, 405 (1929).
 REFERENCES 453
(31) Fabel, K. Kunstseide, 15, 383 (1933).
(32) Tressler, D. and Evers, C. Paper Tr. J. 101, T.S. 113 (1935).
(33) Abrams,A.andBrabender,G. Paper Tr.J. 102, T.S. 204(1936).
(34) Harvey, A. Paper Tr. J. 78, T.S. 256 (1924).
(35) Thomas, C. and Reboulet, H. Industr. Engng Chem. Anal, ed.,
2,390(1930).
(36) Gregory, J. J. Text. Inst. 21, T. 66 (1930).
(37) Trillat, J. and Matricon, M. J. chim. Phys. 32, 101 (1935).
(38) Edwards, R. J. Soc. Leath. Tr. Chem. 16, 439 (1932).
(39) Sale, P. and Hedrick, A. Bur. Stand. Tech. Paper, 18, 540(1924).
(40) Barr, G. Second Rep. Fabrics Co-ord. Res. Committee, D.S.I.R.
GtBrit.p. 113 (1930).
(41) Sager, T. P. Bur. Stand. J. Res. J., Wash., 13, 879 (1934).
(42) Bradley, H., McKay, A. and Worswick, B. J. Soc. Leath. Tr.
Chem. 13, 10, 87 (1929).
(43) Masson, O. and Richards, E. Proc. Roy. Soc. 78 A, 412 (1907).
(44) Hedges, J. Trans. Faraday Soc. 22, 178 (1926).
(45) Tech. Pap. For. Prod. Res., Lond., 1 and 2.
(46) Sheppard, S. and Newsome, P. J. phys. Chem. 34, i, 1160 (1930).
(47) Andrews, D. and Johnston, J. J. Amer. Chem. Soc. 46, 640
(1924).
(48) Leopold, H. and Johnston, J. J. phys. Chem. 32, 876 (1928).
(49) Kirchof, F. Kolloidchem. Beih. 6, 1 (1914).
(50) Lundal, A. E. Ann. Phys., Lpz., 66, 741 (1898).
(51) Ostwald, W. Orundiss der KoUoidchemie, p. 370.
(52) Barrer, R. M. Trans. Faraday Soc. 35, 628 (1939).
(53) Trans. Faraday Soc. 35, 644 (1939).
(54) Daynes, H. India Rubb. J. 84, 376 (1932).
(55) Fry, J. India Rubb. J. 73, 513 (1927).
(56) Cooper and Scott. Lab. Circ. Res. Ass. Brit. Rubber Manuf.
no. 65 (1930).
(57) Daynes, H. Rubber Tech. Conf., London, May 23-5, 1938.
(58) Dewar, J. Proc. Roy. Instn, 21, 813 (1914-16).
(59) Payne, H. and Gardner, W. Industr. Engng Chem. 299 893 (1937).
(60) Kahlenberg, L. J. phys. Chem. 10, 141 (1906).
(61) Sheppard, S. Trans. Faraday Soc. 29, 77 (1933).
(62) Bancroft, W. and Barnett, C. J. phys. Chem. 34, 449, 753, 1217,
2433 (1930).
(63) Meyer, K. and Mark, H. Ber. dtsch. chem. Ges. 61, 593 (1928).
(64) Alexejev, A. and Matalski, V. J. chim. Phys. 24, 737 (1927).
(65) Kirchof, F. Kolloidzschr. 35, 367 (1924).
(66) Stamberger, P. Kolloidzschr. 45, 239 (1928).
(67) Astbury, W. and Woods, H. J. Nature, Lond., 126, 913 (1930);
127, 663 (1931).
(68) Bernal, J., Fankuchen, I. and Perutz, M. Nature, Lond., 141,
528 (1938).
(69) Wilson, R. and Fuwa, T. Industr. Engng Chem. 14, 915(1922).
 AUTHOR INDEX
Abrams, A. and Chilson, W., 432, Beetz, W., 240
436, 440, 441, 442 Bekkedahl, N., 417
— and Brabender, G., 440, 441, 442 Bell, R. P., 419
Ackerel, T., 57, 58 Bellati and Lussana, 144, 145
Adzumi, H., 60, 65, 66 Benton, A. F., 384
Alexejev, A. and Matalski, V., 450 -— and Elgin, J., 230
Alexejew, D. and Polukarew, O., 145, — and White, T., 230
204 Beran, O. and Quittner, F., 325
Alty, T., 117, 142, 375 Berg, W., 338
— and Clark, A., 369 Bergmann, M., 411, 438
Andrade, E., 316, 340, 342, 346 — and Ludewig, S., 409, 411
— and Martindale, J., 316 Bernal, J., 292
Andrews, D. and Johnston, J., 444 — Fankuchen, I. and Perutz, M.,
Andrews, M., 229, 329, 333 388, 451
— and Dushman, S., 329 Berthelot, M., 117
Appleyard, E., 314, 341, 343 Birdseye, C, 440, 441, 442
— and Lovell, A., 341, 343, 347 Blasins, H., 57, 58
Arkel, A. van, 241 Blythswoocl, Lord, and Allen, H. S., 88
Arnold, J. and M'William, A., 239 Bodenstein, M., 144, 201
Arrhenius, S., 240 — and Kranendieck, F., 117
Astbury, W. and Woods, H., 388, 451 Boer, J. H. de, 320, 389
Aten, A. and Zieren, M., 145, 203 — and Fast, J., 157, 158, 212, 400,
403, 419
Baedeker, K., 263 Boggs, C. and Blake, J., 430
Bancroft, W. and Barnett, C , 450 Boltzmann, L., 48
Bangham, D. and Fakhoury, N., 377 Borelius, G., 291
Bansen, H., 71 — and Lindblom, S., 161, 162, 168,
Barnes, C , 25 171, 176, 178, 201, 202
Barr, G., 384, 403, 442 Bose, E., 240
Barrer, R. M., 41, 87, 93, 105, 106, Bosworth, R. C , 44, 245, 341, 351,
117, 121, 122, 125, 126, 127, 129, 354,355, 364,366, 371, 373, 374, 376
130, 131, 137, 162, 168, 174, 177, Braaten, E. O. and Clark, G., 120,126,
178, 179, 180, 181, 182, 184, 188, 129, 168
194, 196, 199, 201, 203, 217, 218, Bradley, H., McKay, A. and Worswick,
221, 231, 233, 318, 333, 334, 346, B., 443
379, 393, 394, 396, 401, 403, 405, Bradley, R. S., 115, 299, 305, 422
407, 412, 413, 415, 416, 417, 419, Bragg, W., Sykes, C. and Bradley, A.,
425, 445 291
—- and Rideal, E. K., 87 Bramley, A., 164, 208, 224, 225, 241,
Bartell, F., 73 245, 275
— and Carpenter, D. C , 73 — and Allen, K., 226
— and Miller, F., 73 — and Beeby, G., 208, 224, 225, 241,
— and Osterhof, H., 73 245, 275
Barwich, H., 79 — and Heywood, F., 241, 245, 275
Baukloh, W. and Guthman, H., 196,234 — Heywood, F., Cooper, A. and
— and Kayser, H., 192, 193, 196, 198 Watts, J., 208, 224, 225, 241, 245,
Baumbach, H. v. and Wagner, C, 212, 275
249, 267, 312 — and Jinkings, A., 208, 224, 225,
Becker, J. A.,44,349,351,360,361, 368 241, 245, 275
Beebe, R., Low, G., Wildner, E. and — and Lawton, G., 208, 224, 241,
Goldwasser, S., 230 245, 275
 AUTHOR INDEX 455
Bramley, A. and Lord, H., 208, 209, Damkohler, G., 87
224, 241, 245, 275 Darwin, C. G., 318
— and Turner, G., 208, 224, 241, 245, Daynes, H., 217, 383, 411, 412, 413,
275 417, 431, 445, 446, 447
Brattain, W. and Becker, J. A., 341, Deming, H. and Hendricks, B., 168,
349, 351, 361, 362, 363, 368 191, 334
Braunbek, W., 297, 305 Deubner, A., 340
Braune, H., 269, 272, 275, 285, 301 Deville, H. and Troost, L., 144
— and Kahn, 0., 274 Devonshire, A., 378
Bremond, P., 72 Dewar, J., 383, 402, 403, 405, 415, 447
Brick, M. and Philips, A., 241, 275 Diergarten, H., 239
Bridgman, P., 315 Dietl, A., 88
Brillouin, M., 55, 56 Ditchburn, R., 342
Brower, T., Larsen, B. and Schenk, Dixit, K. R., 346
W., 241 Doehlemann, E., 182
Briick, L., 340 Donnan, F., 65
Bruni, G. and Meneghini, D., 241 Dorn, J. and Harder, O., 297
— and Scarpa, G., 266 Doughty, R., Seborg, C. and Baird, P.,
Bryce, G., 230 384
Buckingham, E., 53, 58, 66, 384, 406 Dube, G., 375
Buerger, M. J., 318 Duhm, B., 220
Bugahow, W. and Rybalko, F., 278, Dunn, J., 241, 329, 332, 333
327, 331, 333 Dunwald, H. and Wagner, C, 212,
— and Breschnewa, N., 278 250, 252, 267, 312
Burmeister, W. and Schloetter, M., 227 Durau, F. and Schratz, V., 115
Burton, E., Braaten, E. O. and Dushman, S., Dennison, D. and
Wilhelm, J. O., 121, 129, 334 Reynolds, N., 275, 329, 333

Cailletet, L., 144, 145 Eason, A. B., 57, 58

Capron, P., Delfosse, J., Hernptinne, Edwards, C, 162, 198, 202, 203, 222
M. de and Taylor, H. S., 79, 80 Edwards, J. and Pickering, S., 384,
Carslaw, H., 4, 10,15, 20, 35,36, 38,40 393, 402, 403, 404, 405, 417, 440
Carson, F., 382, 384,408,409, 413,430, — and Wray, R., 432,433,434,436,441
438 Edwards, R., 409, 411, 442
Cernu8chi, F., 375 Eilender, W. and Meyer, O., 225
Chalmers, J., Taliaferro, D. and Emitein, A., 268
Rawlins, E., 74 Elam, C, 276
Chapman, S., 80 Eltzin, I. and Jewlew, A., 228
Charch, W. and Scroggie, A. G., 432, Emanueli, L., 384
436, 440, 442 Emeleus, H. and Anderson, J., 147,
Chariton, J., Semenoff, N. and 148, 149
Schalnikow, A., 378 Engelhardt, G. and Wagner, C, 182
Charpy, G. and Bonnerot, S., 144 Estermann, J., 341
Chilton, J. and Colbourn, A., 74 Euringer, G., 214, 222
Cichocki, J., 245, 297, 305 Ewell, R. H., 421
Clausing, P., 53, 64, 82, 85, 86, 406 Eyring, H., 274, 302, 305, 422, 423
Clews, C. B. and Schosberger, F., 389, — and Sherman, J., 188
390 — and Wynne-Jones, W., 274,303,305
Clews, F. and Green, A., 70, 71
Clusius, K. and Dickel, G., 80, 82 Fabel, K., 440, 442
Cockcroft, J., 341, 342 Fajans, K., 295
Coehn, A. and Jiirgens, H., 212 Fancher, G. and Lewis, J., 73, 74,76, 77
— and Specht, W., 212, 220, 221, 268 Faraday, M., 240
— and Sperling, K., 213, 221, 268 — and Stodart, 239
Cooper and Scott, 447 Farkas, A., 78, 93, 185, 186
Cremer, E. and Polanyi, M., 188 — and Farkas, L., 184
456 AUTHOR INDEX
Farnsworth, H. E., 340 Harvey, A., 440
Feitknecht, W., 329, 332 Hass, G., 340
Finch, G. and Quarrell, A. G., 340 Hedges, J., 443
Fonda, G., Young, A. and Walker, A., Hemptinne, M. de and Capron, P., 79,
275, 327, 328, 333 80
Foussereau, 257 Hendricks, B. and Ralston, R., 334
Frank, L., 341, 354, 358, 367 Herbert, J., 115
Franzini, T., 213 Herbst, H., 88
Frazer, J. and Heard, L., 231 Hertz, G., 78, 79
Frenkel, J., 85, 248, 289, 293, 294, Herzfeld, K. and Smallwood, M., 54
305, 311 Hessenbruch, W., 227
Frey-Wyasling, A., 389 Hevesy, G. v., 241, 248, 269, 272, 274,
Frisch, R. and Stern, O., 378 275, 287,289,292,300, 322, 327, 333
Fritsch, C , 322 — and Obrutcheva, A., 241
Fry, J., 447 — and Paneth, F., 241, 242
— and Seith, W., 244, 258, 269, 274,
Gaede, W., 53, 55, 64 275, 276, 277
Gallagher, F., 384 — Seith, W.and Keil, A., 244, 258,
Gehrts, A., 328, 333 269, 274, 275, 277, 328
Gen, M., Zelmanov, I. and Schalnikow, Hey, M., 87, 103, 106, 108
A., 340 Hicks, L., 246
Giess, W. and Liempt, J. van, 333 Hilsch, R., 108, 112, 250
Giolotti, F. and Tavanti, G., 239 — and Pohl, R., 250, 316, 319, 320,
Goethals, C , 322, 333 321
Goetz, A., 317 Hollings, H., Griffith, R. and Bruce,
Graetz, L., 240 R., 231
Graham, T., 144, 334, 391, 402, 405 Holt, A., 170
Gray, A., Mathews, G. and MacRobert, Horiuti, J., 417
T., 36 Hiifner, G., 391, 414
Green, H. and Ampt, G., 76, 78 Hunt, J. and Lansing, D., 432, 441
Gregory, J., 442 Hyden, W., 440, 441, 443
Griffith, R. and Hill, S., 231
Grimshaw, L., 239 Iijima, S., 230
Grinten, W., 80 Ingersoll, L. and Zobel, O., 4, 38
Groh, J. and Hevesy, G. v., 241 Ivea, H., 353
Groth, W. and Harteck, P., 82 — and Olpin, A., 353
Grube, G. and Jedele, A., 245, 246
Guillet, L. and Roux, A., 227 Jacquerod, A. and Perrot, F., 117
Guth, E. and Rogowin, S., 389 Jeffreys, H., 356
Gyulai, Z., 262, 274, 323, 326, 333 Jette, G. and Foote, F., 250
— and Hartley, D., 325 Joffe, A., 277, 315, 326
Johnson, F. and Larose, P., 164, 168
Haber, F. and Tolloczko, St., 266 Johnson, J. and Burt, R., 126, 137
Hagedorn, M. and Moeller, P., 388 Jost, W., 10, 12, 248, 254, 256, 263,
Hagenbach, E., 56 268, 274, 275, 287, 289, 293, 294,
Hagg, G., 145, 157, 158, 210 295, 305, 311
— and Kindstrom, A. C 250 — and Linke, R., 268
— and Sucksdorff J., 250 — and Nehlep, G., 254, 256, 287
Ham, W., 161, 168, 173, 178 — and Widmann, A., 184, 221, 233
— and Rast, W., 198 Jouan, R., 168, 184
— and Sauter, J. D., 170,193,197, 234
Hammett, P. and Brunauer, S., 230 Kahlenberg, L., 447
Hanawalt, J. D., 210 Kalberer, W. and Mark, H., 85
Harkness, R. and Emmett, P., 230 Kanata, K., 391, 402
Harmsen, H., 78, 79 Kanz, A., 70, 71
— Hertz, G. and Schutze, W., 78, 79 Kawalki, W., 14
 AUTHOR INDEX 457
Kayser, H., 402 Lombard, V., Eichner, C. and Albert,
Ketelaar, J., 249, 263 M., 164, 168, 191, 194, 196
Ketzer, R., 323 Lovell, A., 314, 341, 343
Kirchof, F., 445, 450 Lowry, H. and Kohman, G., 431, 432,
Kirschfeld, L. and Sieverts, A., 159 433, 435
Kirschner, F., 340 Lundal, A. E., 445
Klaiber, F., 263
Kline, G., 430, 436, 437, 438, 441, 443 Magnus, A. and Sartori, G., 230
Klose, W., 53, 55, 64 Manegold, E., 60, 78
Knauer, F. and Stem, O., 342 March, H. and Weaver, W., 22, 24, 39
Knudsen, M., 53, 54, 55, 63, 64, 377, Mark, H. and Meyer, K., 389, 390
406 Marsh, M., 384
Koch, E. and Wagner, C , 256 Martin, S. and Patrick, J., 393
Kohlrausch, W., 240 Martley, J., 436
Kohlschutter, H., 231 Masing, G., 239
Kohman, G., 415 — and Overlach, H., 239
Koller, L., 341, 368 Masson, O. and Richards, E., 443
Konigsberger, J., 258 Matano, C, 47, 49, 275, 279, 280
Korsching, H. and Wirtz, G., 80, 82 Mathewson, C, Spire, E. and Milligan,
Kraft, H., 116 W., 157
Kramer, J., 344 Mathieu, M., 388
Kriiger, F. and Gehm, G., 145,210,241 Maxted, E. B. and co-workers, 377
Kubelka, V., 431 Mayer, E., 126
McBain, J. W., 145,161, 313, 334, 388,
Lacher, J., 150, 152, 154, 189, 374 389, 431
Lane, C. T., 340 McKay, H., 244, 275
Lange, H., 340 Mehl, R., 239, 275, 280, 283, 285, 287,
Langmuir, I., 245, 275, 328, 333, 337, 328, 330, 333
341, 348, 349, 350, 353, 361, 363, Melville, H., 188
370, 371, 372, 375 — and Rideal, E. K., 168, 179, 186,
— and Dushman, S., 298, 305 188
— and Kingdon, H., 349, 350, 372 Merica, P. and Waltenburg, R., 157
— and Taylor, J. B., 44, 352, 353, Meyer, K., 389
360, 375 — and Mark, H., 450
— and Villars, D., 350 Meyer, L., 210
Lannung, A., 417 Meyer, 0., 406
Le Blanc, M., 323 Moelwyn-Hughes, E. A., 427
Lehfeldt, W., 264, 265, 324, 333 Moll, F., 338
Lennard-Jones, J. E., 117, 142, 234, Mollwo, E., 108, 110, 111, 250
346, 375, 379 Morosov, N. M., 230
— and Dent, B., 314 Morris, T., 200
— and Goodwin, E., 375 Mott, N. F., 320
— and Strachan, C , 375 Muskat, M. and Botset, H., 76
Leopold, H. and Johnston, J., 443, 444
Levey, H., 438, 440, 441, 442 Nagelschmidt, G., 93, 98
Lewkonja, G. and Baukloh, W., 196, Narayanamurti, O., 78
234 Natta, G., 340
Leypunsky, O., 230 Nehlep, G., Jost, W. and Linke, R.,
Liempt, J. van, 215, 229, 275, 285, £68
301, 305, 330, 333 Nernst, W., 240, 268
Liepus, T., 340, 342 — and Lessing, A., 144, 145
Linde, J. and Borelius, G., 145, 210, Norton, A. and Marshall, F., 227, 229
241
Lombard, V., 164, 168 Orowan, E., 313, 317
— and Eichner, C , 169, 170, 175, Orr, W. J. C , 346, 379
182. 196 — and Butler, J., 361
458 AUTHOR INDEX
Ostwald, W., 445 Sauter, E., 390
Owen, E. A. and Jones, J., 210 Scherr, R., 81
Schiefer, H. and Best, A., 384
Pace, J. and Taylor, H. S., 231 Schlichter, C, 73
Paneth, F. and Peters, K., 122 Schmidt, G., 170
Paschke, M. and Hauttmann, A., 226, Schottky, W., 248, 249
275 Schroeder, P. N., 438
Payne, H. and Gardner, W., 448 Schumacher, E. and Ferguson, . L.,
Phipps, T., Lansing, W. and Cooke, 384, 432, 436, 440
T., 258 Schwartz, K., 369
Pilling, N. and Bedworth, R., 329, 332 Seelen, D. v., 297, 322, 325
Piutti, A. and Boggiolera, E., 126 Seith, W., 245, 250, 259, 260, 261, 264,
Pohl, R., 316, 319, 320, 321 266, 267, 275, 277, 278, 324, 327, 333
Polanyi, M. and Wigner, E., 298, 427 — and Etzold, H., 268, 275
Post, C. and Ham, W., 161, 168, 169, — Hofer, E. and Etzold, H., 275
173, 191 — and Keil, A., 275, 328
Poulter, T. and Uffelman, L., 199, 315, — and Kubaschewski, O., 268
333 — and Peretti, E., 241, 258, 259, 275,
— and Wilson, R., 314 287
Preston, E., 69, 71 Semenoff, N., 377
Prins, I., 318 Sen, B., 288
Seybold, A. and Mathewson, C, 157
Rahlfs, P., 249 Shakespear, G., 403, 405
Raisch, E., 78 — Daynes, H. and Lambourn, 403
— and Steger, H., 78 Sheppard, S., 450
Ramsay, W., 144 — and Newsome, P. T., 388, 444, 445
— and Collie, N., 65 Shishacow, N. A., 93, 125
Rasch, E. and Hinrichsen, F., 258 Siefriz, W., 389
Rawdon, H., 239 Sieverts, A., 159
Rayleigh, Lord, 116, 119, 132, 384 — and Briining, K., 157, 159, 212
Reis, A., 295 — and Danz, W., 149, 189
Renninger, M., 247, 316, 317, 318 — and Hagen, H., 159, 212
Reychler, A., 391, 414 — and Hagenacker, J., 155
Reynolds, 0., 57, 144, 240 — and Krumbharr, W., 158
Rhines, F. and Mathewson, C , 157 — and Zapf, G., 149, 189, 190
— and Mehl, R., 275, 280, 282, 283 — Zapf, G. and Moritz, H., 158, Ifc9
Richardson, O., 169 Simons, J. H., 155
— Nicol, J. and Parnell, T., 144, 164, Smekal, A., 258, 263, 273, 274, 275,
168 295, 313, 316, 321, 323, 326, 333
— and Richardson, R. C , 117 Smith, C, 276
Richter, M., 338 Smith, D. and Derge, G., 200, 318
Rideal, E. K., 374 Smithells, C, 145, 151, 157, 158, 159,
Roberts, J. K., 350 161, 162, 191, 223
Roberts-Austen, W., 239 — and Fowler, R. H., 151, 152, 154
Roeser, W., 130, 333 — and Ransley, €. E., 122, 164, 168,
Rogener, H., 108, 250 169, 171, 175, 176, 179, 195, 201,
Rontgen, P. and Braun, H., 158 227, 228, 229
— and Moller, H., 227 Smoluchowski, M. v., 53, 55, 56
Runge, B., 209, 224, 226 Spencer, L., 168
Ryder, 168 Spiers, F., 279, 369
Spring, W., 239
Sager, T. P., 393, 396, 398, 399, 438, Staedel, W., 432, 436, 440, 441, 442
443, 445 Stamberger, P., 450
Sakharova, M., 333 Staudinger, H., 390
Sale, P. and Hedrick, A., 442 — Heuer, W. and Husemann, E., 390,
Sameshima, J., 66 391
 AUTHOR INDEX 459
Staudinger, H. and Husemann, E., 390 Venable, C. and Fuwa, T., 414
Steacie, E. and Johnson, F., 155 Villachon, A. and Chaudron, G., 227
— and Stovel, H., 231 Villard, P., 117, 122
Stefan, J., 14 Volmer, M., 338
Steigman, J., Shockley, W. and Nix, — and Adhikari, G., 338
F., 244, 275 — and Esterman, I., 337
Stepanow, A., 326, 333 Voorhis, C. C. v., 126, 129
Stillwell, S., 432, 436
Stodola, A. and Lowenstein, L., 57, 58 Wagner, C , 182, 212, 248, 250, 252,
Strock, L., 249, 263 253, 256, 263, 267, 269, 271, 311,
Swamy, R. S., 340 312
Swan, E. and Urquhart, A., 88 — and Schottky, W., 248, 267, 311
Sykes, C. and Evans, H., 291 Waldmann, L., 80
Wang, J. S., 179, 180
Takei, T. and Murakami, T., 159 Warburg, E., 53, 61
Tammann, G., 344 — and Tegetmeier, F., 277
— and Bochow, K., 414 Ward, A., 117, 230, 233, 375
— and Schneider, J., 220, 221 Watson, G., 36
— and Schonert, K., 224, 226 Watson, W^ 117
— and Veszi, G., 322, 325 Wells, C , 276
Tanaka, S. and Matano, C, 241 — and Mehl, R., 225, 226
Taylor, H. S., 232, 377 Wenderowitsch, A. and Drisina, R.,
— and Diamond, H., 231 325
Taylor, J. B. and Langmuir, I., 346, Wheeler, C , 300, 305
360, 361, 371, 375 Wheeler, T. S., 422
Taylor, N. W. and Rast, W., 130 Wicke, E., 87
Taylor, R., Herrmann, D. and Kemp, Wiedmann, E., 240
A., 405, 432, 433, 434, 436, 437, 440, Wiener, O., 48
441 Wilde, H. and Moore, T., 74
Taylor, W. H., 92, 93 Wilkins, F. and Rideal, E. K., 329,
Thomas, C. and Reboulet, H., 441 332
Thomson, G. P., Stuart, N. and Williams, G. A. and Ferguson, J., 120,
Murison, C. A., 340 121, 126, 139, 140
Thomson, S. P., 240 Williams, J. and Cady, L., 44
Thomson, W., 406 Wilson, J. and Lines, G., 411, 432,
Tiselius, A., 87,96,97,101,106,276,419 437, 438, 442, 445
Tompkins, F. C , 115 Wilson, R. and Fuwa, T., 451
Toole, F. and Johnson, F., 159 Wing, H., 432, 438, 441
Topping, J., 373 Winkelmann, A., 144, 164, 170
Tressler, D. and Evers, C , 440, 441 Wirtz, K., 80
Trillat, J., 388 Wood, R. W., 377
— and Matricon, M., 442 Wooldridge, D. and Smythe, W., 78
Troost, L., 144 Wosnessenski, S. and Dubnikow,
T'sai, L. S. and Hogness, T., 121, 126, L. M., 432, 441
129, 137 Wray, R. and Vorst, A. van, 432,
Tubandt, C, 245, 250, 260, 261, 266 441
— Eggert, S. and Schibbe, G., 245, Wroblewski, S., 391, 413
250, 261, 263, 266 Wiistner, H., 117, 118, 120, 121, 126,
— and Reinhold, H., 263, 324 139, 140, 175
— Reinhold, H. and Jost, W., 269, Wyckoff, R., Botset, H., Muskat, M.
271, 272, 274 and Reed, D., 77
and Liebold, G., 24Q, 250, 261, Wynne-Jones, W. and Eyring, H., 423
266
Zahn, H. and Kramer, J., 344
Urry, W., 116, 118, 120, 121, 126, 129, Zwicky, F., 317
137, 142, 333, 334 Zwikker, C , 330, 333
 SUBJECT INDEX
Absorption, see Sorption, Solution Classification of some diffusion sys-
— of alkali metals by alkali halides, tems, 123
109-10, 111-15, 312, 316 Colour centres in alkali halides, 109-
— of alkali metals in tungsten, 354-5 10, 319-21
— of ammonia, sulphur dioxide, and Condensation and aggregation of atom
hydrochloric acid by alkali halides, beams on surfaces, 339-47
115 Conductivity of salts, 240, 247-50,
— and diffusion of gases in finely 251-3, 256, 257-65, 265-7, 268-72,
divided metals, 230-4 272-3, 274, 276-8, 283, 289, 294-7,
— of halogens by alkali halides, 109- 305. See also Structure-sensitive
10, 111-14, 312 conductivity
— of hydrogen by alkali halides, — of solid solutions of salts, 256,
109-10, 111-14, 312 261-3, 323-5
— of thallium by alkali halides, 113 — and space-charge redistribution,
Activated adsorption, 232-3 325
Activated diffusion, 71, 73,106,121-3, — of thin films, 343-6
132-3, 169, 207-8, 222-3, 382, 403, Criteria of stability in films, 339, 340-1
409, 416, 436, 445 Critical nominal thickness of thin
Activation energy for diffusion, 103, films, 344-6
141,221,222,223,225,274,275,420, Critical streaming density for aggrega-
421, 424, 425 tion of atom beams on solids, 377-8
for structure-sensitive diffusion, Cross-linking and permeability of
327-9 polymers, 393
for surface diffusion, 346,360-1, — and solubility of polymers, 391
363, 364-6, 366-7, 370 — and sorption by polymers, 450-1
Activation energy, radius and polari- Crystal growth, 337-9, 339-41, 341-3,
sability in alkali halides, 264-5 375, 377-8.
Adatoms, 347, 350
Ammoniate formation in natrolite, Degassing of metals, 31, 226-30
105-6 Degrees of freedom in diffusion pro-
Analogy between activated diffusion cesses, 299-300, 422-3, 425
and chemical reaction, 427 Dependance of activation energy for
Anionic conductors, 267 diffusion upon concentration, 282-3,
367, 372, 374
Base exchange, 240 — of tensile strength upon sorption,
Berthollide compounds, 249-50. See etching and fibre diameter, 315-16
also Interstitial compounds Derivation of diffusion constants from
solutions of Fick's laws, 47-9, 50,
Calculation of conductivity, 293-7 97-8, 112-14, 214-19, 351-2, 355.
— of diffusion constant, 293-6, 298- 356-9,412-13. See also Measurement
303. See also Models of conduction of diffusion constants
and diffusion processes, etc. Differential forms of the diffusion
— of energy periodicity of crystal equation, 1-5
surfaces, 379 Diffusion anistropy, 98, 102-3, 193,
— of surface diffusion constant, 375- 276-9, 327
6 Diffusion equation when the diffusion
Capillary condensation, 450 constant depends upon concentra-
Chemical composition of gases de- tion, 47-9, 101, 443, 446
sorbed from metals, 227 — coupled with interface reaction^,
of glasses in relation to per- 37-43, 174-5, 176-8, 179-83, 191
meability, 129-30, 137-9 — in cylinders, 31-7
 SUBJECT INDEX 461
Diffusion between finite layers, 14 Effect of spreading pressure on diffu-
— in finite solids, 13 sion constants, 363, 372-4
— of gases in glasses, 139-41 — of temperature upon conductivity,
— of gases in metals, 207-25 257-65, 321-5
— of gases and alkali metals in — of temperature upon diffusion, 115,
alkali halides, 108-16 133, 141, 207, 221-6, 258, 278, 28B,
— of gases in organic membranes, 328,444-5. See also Activation
411-27 energy for diffusion, Influence of
— and grain size, 328-30 temperature on permeability
— of instantaneous plane source of — of temperature on the solubility of
solute, 44 gases in solids, 111, 140,150-2,155-
— of instantaneous point source of 7, 414, 418
solute, 46 — of temperature on time lag in
— of instantaneous spherical surface establishing steady states of flow,
source of solute, 47 203, 218, 413, 416
— of interstitial ions, 292, 293-6 Effusion, 53-4, 61, 65-6, 78, 406
— of nascent hydrogen through Electrolysis of sodium ions through
metals, 144-5, 200-4 glass, 96, 240
— of non-metals in metals, 224-5 Electronic conductors, 267
— through a permeable membrane Energy of disorder in crystals, 254-7
separating two stirred fluids, 24-8 Entropy of activation for diffusion,
— by place exchange, 29, 96, 98, 274, 276, 287, 423-6
292-3, 300, 303 — of solution of gases in polymers,
— of salts through metal foils, 245 417, 419
— in semi-infinite solids, 11-12 Equilibrium disorder in crystals, 247-
— across sharp boundaries, 8-9 50, 254-7, 293, 311
— in spheres, 28-31 Equilibrium in a monolayer under -a
— by spontaneous gliding, 292-3 temperature gradient, 357-8, 364,
— by spreading, 96. See also Zeolitic 366
diffusion Evidence of surface mobility from
— to and from a stirred fluid in con- properties of unstable films, 339-47
tact with a quiescent medium, 21-4
— with surface concentration a func- Factors governing relative permeabi-
tion of time, 19-20, 35 lity of metals to hydrogen isotopes,
— in two different media, 10-11 187-8
— of vacant lattice sites, 294, 297 — influencing diffusion constants,
— through the wall of a hollow 279-83, 283-91, 304
cylinder, 35-7 Farbzentren, see Colour centres
— in wires when the surface concen- .F-centres, see Colour centres
tration is fixed, 32-5 JT-centres, 320
Dipole adsorption, 449-50 Fick's laws, see Differential forms of
Displacement of oil from oil-bearing the diffusion equation
sand, 73 Flow of fluids through capillary
systems, 53-78, 82-8, 406-11
Effect of diffusion on mechanical — of gases in consolidated and un-
strength, 204, 334 consolidated sands, 73-8
— of finishes and fillers on the through miscellaneous solids,
permeability of organic membranes, 78
411, 438 through porous plates, 65-9
— of gas pressure on the conductivity through refractories, 69-73
of crystals, 251-3, 267 Fractionation of gases by activated
— of mechanical deformation on diffusion, 120-1, 393, 396
conductivity of crystals, 325-6 Free energy of solution of gases in
— of soluble salts on the permeability organic polymers, 417
to water of organic membranes, 436, Frenkel disorder in crystals, 254, 257,
438s 293-5, 300, 303
462 SUBJECT INDEX
Glass electrode, 240 Influence of thickness upon permea-
Grain-boundary diffusion, 127-30, tion rates, 120, 404-5, 408, 433
131, 141, 197-200, 245, 311-12, Interdiffusion of metals, 8, 10, 31,
327-34,337,389. See also Structure- 44, 47, 49, 239, 241, 244-5, 245-7,
sensitive diffusion 272-6, 278-9, 279-91, 292, 298-305,
327-34
Heats of adsorption by non-stationary — of salts, 8, 240, 245, 272, 274. See
streaming, 85-7 also Conductivity of salts
— of condensation of metals on Internal surfaces, 313-19
solids, 341 Interstitial compounds, 92, 96, 156-7,
— of solution of alkali metals in 249-50, 292
alkali halides, 110-11 Interstitial ions, 92, 248-57, 293
of gases in alkali halides, 110- Intra-crystalline diffusion, see Grain
11 boundary diffusion
of gases in metals, 153 Investigations of gas, flow in capillaries,
of gases in polymers, 418 61-5
— of sorption of metal monolayers Irreversible permeation velocities
on tungsten, 360-1, 371 through glasses, 125-31
Hemicolloids, 300 Irreversible sorption by organic col-
Hertz method of fractionating gas loids, 438
mixtures, 78-80
Hysteresis effects in gas-metal sys- Kinetic theory of diffusion in zeolites,
tems, 150, 192, 194-5, 199 106-8
Kinetics of sorption and desorption,
Identity of current carrying ions, 265- 86-8, 228-9, 230-4, 443-5, 445-7
8 Knudsen flow, see Molecular streaming
Impedance, 405, 411
Influence of acid strength upon per- Lag in establishing steady states of
meation rate of nascent hydrogen flow, 18-19, 31, 37, 203, 217-19,
through metals, 203 221, 222-3, 412-13, 415-16
— of concentration upon diffusion Lateral contraction of surfaces, 314
constants, 47, 101-2, 225-6, 279- Lattice diffusion, 127-8, 130, 141,
83, 371-4, 443, 445-7 197-200, 331, 334, 389. See also
— of current density upon permea- Interdiffusion of metals, Inter-
tion rate of nascent hydrogen diffusion of salts, Volume diffusion
through metals, 182-3, 201 Lifetime of adatoms in the activated
— of hydration upon cellulose struc- state, 375
ture, 450 — of atoms adsorbed on glass, 85-6
— of hydroxyl groups upon per- Lineage structures, 318-19
meability of organic membranes, Linear polymers, 385-6
399-400
— of impurity upon diffusion con- Measurement of concentration gra-
stants, 234r-6 dients, 97, 111, 208-10, 241, 244,
— of mechanical working upon diffu- 354-5, 359, 434-5
sion constants, 328, 331-2 — of diffusion constants, 47-9, 50, 97,
— of phase changes upon permeabi- 111-14, 139-40, 208-19, 241-5, 272,
lity, 191-2 351-9, 369-70, 411-13
— of pressure upon permeation — of permeability, 117-20, 161-2,
velocity, 66-9, 74, 76, 120-1, 169- 163, 383-5, 430
83, 403-4, 406, 408-9, 411, 432-4 — of sorption and desorption ve-
— of pre-treatment upon permeabi- locities, see Measurement of diffu-
lity, 192-7 sion constants
— of temperature upon permeability, — of transport numbers in crystals,
71-3, 120-5, 125-31, 126-9, 133-7, 266, 267-8
162-9, 202-3, 394-5, 397, 405-6, Mechanism of activation of adsorbed
408, 435-6, 437 atoms, 375-6
 SUBJECT INDEX 463
Mechanism of flow through metals, Permeability constants, definitions
166-7, 170-1, 178-83 and dimensions, 7, 60-1, 391-2
— of flow through rubbers, 391, 420, Permeability of membranes in series,
421 411
Membrane forming polymers, 382, Permeability spectrum, 383, 432
385-91, 392, £96-7, 431 Permeation rates at high pressures,
Mesocolloids, 391 117, 175-8, 403
Micellar structures of chain polymers, Phase boundary processes, 37-43, 177,
389-90 179-83, 187, 218-19
Micrographic evidence of grain boun- Phase changes in monolayers, 353,
daries, 314, 315, 317,319 374-5
Mixed conductors, 267 Photochemical processes in alkali
Mobility of ions, 255, 268-72, 272 halides, 108-10, 319-21
Models of conduction and diffusion Photoelectric emission, 347, 353-9,
processes in crystals, 291-305 364, 366, 367, 368
— for diffusion in rubber, 422-7 Physical properties of cellulose esters,
Molecular streaming, 53, 54-5, 60, 64, 388
65, 69, 71, 82-8, 130, 131, 133, 142, Platelike polymers, 385-6
402, 406 Poiseuille flow, see Streamline flow
Multilayer adsorption, 450 Primary flaws in crystals, 313-14
Properties of films on tungsten, 347-
Nature of hydrogen-metal systems, 51
149-55
— of metallic crystals, 147-9 Radioactive indicators, 241-5, 265,
— of permeation processes through 271, 272, 304, 369
organic solids, 448, 450-2 Relation between diffusion and con-
Nernst lamp, 240 ductivity constants, 268-72, 294r-5
Non-equilibrium disorder in crystals, — between permeability and chemi-
247-8, 311-12, 313-21 cal nature of organic membranes,
Non-stationary states of flow in 399-400
capillary systems, 82-8 — between types of molecular flow,
Numerical values of conductivity con- 131-3
stants, 264 Relative activation energies for volume,
of diffusion constants, 98, 101, grain-boundary and surface diffu-
103, 104, 108, 113, 116, 141, 217, sion, 326, 337, 346, 363
221, 222, 223, 224, 225, 229, 271, Relative behaviour of hydrogen iso-
274-5, 278, 281, 284, 328, 329, 330, topes in solution and diffusion in
360, 361, 363,365, 367, 370,420,445 metals, 183-91
of permeability constants, 70, Relative permeabilities of membranes,
76-7, 116, 133-7, 138, 168, 394-5, 137, 400, 402, 439, 443, 447-8, 449
396, 397, 398, 400, 409, 410, 417, Removal of benzophenone by surface
440-2, 443, 447 diffusion, 338
of solubility constants, 140, Role of water in proteins, 450, 451
415, 417, 418
Schottky disorder in crystals, 254, 257,
Optical absorption of alkali halides, 294, 300, 303
109-10, 111, 319-21 Selective adsorption on ionic crystals,
Order-disorder transformation in me- 379
tals, 290-1 Self-diffusion, 244, 271, 273, 278, 290,
Orifice flow, 53, 58-60, 382, 406, 408-9 300-1
Osmotic theory of diffusion in organic Sigmoid sorption isotherms, 443
solids, 445-7 Slip-plane diffusion, 200
Oxidation of metals, 332 Solution of gases in alloys, 158-61
in glasses, 139-41
Periodicity curves for activated diffu- in organic membranes, 411-19
sion systems, 426 in metals, 145-61
464 SUBJECT INDEX
Solution of gases and vapours in ionic Summary of reactivity of metals to
crystals, 109, 110-11 gases, 145-7
Solutions for the steady state of flow, Surface diffusion, 3, 44, 245, 337-47,
5-6 351-79
Sorption and desorption from a hollow Surface migration, see Surface diffusion
cylinder, 35-7 Surface mobility, see Surface diffusion
from a membrane, 14-17,443-5 Sweep curves, 316-17, 318
from a solid cylinder, 33-5,
214-15 Thermionic emission, 327,329,347-53,
from a spherical shell, 29 359, 361-3, 368
from a solid sphere, 29 Thermo-diffo8ion method for frac-
— and diffusion in tungsten oxide,367 tionating gas mixtures, 80-2
— in swelling media, 93, 430, 432-4, Three-dimensional polymers, 385-7
435, 443-5, 450-1 Turbulent flow, 53, 57-8, 73, 74, 75
Spreading pressure in films, 363, 372-4 Two-dimensional gas, 346, 377-9
Streamline flow, 53, 55-6, 58, 60, 61, Types of concentration gradient, 210-
63, 64, 65, 69, 71, 73, 74,130-1,133, 11, 245-7, 434-5
382, 406, 407-9
Structure of cellulose, 388 Unsolved diffusion problems, 50-1
— of crystalline rubber, 390
— of proteins, 388-9 van der Waale adsorption, 230, 448,
Structure-sensitive conductivity, 110, 450
311-13, 321-6, 333 Variation of electrical resistance with
Structure-sensitive diffusion, 127-30, amount sorbed, 212, 329-30
245, 311-12, 327-34, 389. See also Volume diffusion, 311, 326, 363, 369.
Grain-boundary diffusion See also Lattice diffusion
Structure-sensitive properties, 313
Structures of some silicates and glasses, Zeolitic diffusion, 96, 223, 292, 300,
91-6 303