Inuence of polyaluminum chloride self-aggregation on occulation

performance
a a,∗ a
Karen A Swetland , Monroe L Weber-Shirk , Leonard W Lion

a School of Civil and Environmental Engineering, Cornell University, Ithaca, NY 14853, USA

Abstract

Polyaluminum chloride (PACl) is a commonly used coagulant for water treatment. One mode of action of
7+
PACl ((AlO4 Al12 (OH)24 (H2 O)12 ) ) is reported to be through formation of aggregates that bridge between
colloids. Although many studies have considered the eectiveness of PACl under a spectrum of conditions
for inuent suspensions, little is known about the rate of formation of PACl aggregates upon mixing with
inuent water and the eect of aggregate size on the subsequent formation of ocs that can be readily
removed by sedimentation. PACl aggregates larger than 0.18 µm in diameter were formed under controlled
conditions. A kinetic aggregation model was developed and validated to give the aggregate size before mixing
with a colloidal suspension and entering a occulator. After occulation, the ensuing oc sedimentation
velocity and residual turbidity were non-destructively observed. Experimental and modeling results combined
with geometric analysis show that, under the experimental conditions tested in this research, PACl self-
aggregation consistently lowers attachment eciency of the colloidal suspension, reduces the eectiveness of
the occulator, and reduces turbidity removal. Minimization of PACl aggregate size is best accomplished by
immediate, rapid, and ecient mixing of PACl with the inuent water. The relationship between residual
turbidity and PACl dose is consistent with expectations based on a geometric adhesive model of coagulation.

Keywords: Rapid mix; PACl; coagulation; occulation; aggregation; surface coverage;

1. Introduction

In water treatment, coagulation and occulation are used to form particle aggregates or ocs that can
be subsequently removed by sedimentation or ltration. Coagulation increases attachment eciency, i.e.,
the fraction of particle collisions that result in aggregation.. Flocculation is the transport phase in which
the particles collide and aggregate [2]. While coagulation in drinking water treatment always involves the
addition of a coagulant, the particular chemical, dose, and other relevant conditions determine the attachment
eciency and the ensuing success of coagulation. .

1.1. PACl Structure and Precipitation

Polyaluminum chloride (PACl) is an inorganic polymer coagulant that has gained wide acceptance for use
in water treatment as a result of its ecacy over a broader pH range and at lower temperatures than the com-
monly used alternative, aluminum sulfate (alum) [3]. PACl is commonly prepared by the controlled neutral-
ization of aluminum chloride. The dominant stable species in dissolved PACl, (AlO4 Al12 (OH)24 (H2 O)12 )7+ ,
commonly referred to as Al13 , has a Keggin-13 structure and a hydrated radius of 1.2 nm [28]. When the
dose of coagulant exceeds the solubility limit, precipitation occurs. PACl precipitates, or aggregates, are
composed of Al13 subunits. Benschoten and Edzwald [3] proved through timed spectrophotometry that the
precipitation products of alum and PACl were characteristically dierent and that PACl precipitates retained
their polymeric structure.
Many mechanisms have been proposed for coagulation of suspensions by PACl including: (1) Destabi-
lization of colloidal particles by the nucleation of positively charged Al13 precipitates on negatively charged
colloidal surfaces, or precipitation charge neutralization (PCN) [? ]. PCN is based on the observation that,

∗ Email: [email protected] Phone: 1 607 255 8445 Fax: 1 607 255 9004

Preprint submitted to Elsevier August 22, 2012

under the circum-neutral pH of most natural waters, the precipitates of aluminum salts and polymers are
positively charged. These species are thought to destabilize negatively charged colloids by attaching and
neutralizing their surface charge, thereby decreasing electrostatic repulsion between colloids and inducing
aggregation [28]. (2) As a consequence of the relatively large size and stability of PACl aggregates and their
positive charge, electrostatic patch coagulation (EPC) has also been proposed as a potential coagulation
mechanism [28]. EPC is characterized as a localized precipitation charge neutralization mechanism; patches
of positive charge are created by aggregates of coagulant adsorbing to a small fraction of the surface of
the colloid. Attractive electrostatic forces then attach the positively charged PACl patches to the naked
surface of other colloids upon collision. It is hypothesized that large ocs can be formed in this manner
[10, 28, 11]. (3) At high coagulant doses, colloidal particles can be removed when they become enmeshed
in the voluminous self-aggregated coagulant precipitate, a process called sweep occulation [2]. The term
sweep occulation is best thought of as a description of the colloid and coagulant suspension and does not
provide a mechanistic understanding of why precipitated coagulant should interact with itself or colloids in
the observed manner. Sweep occulation is observed at coagulant doses that are higher than are commonly
used in water treatment. Consequently, EPC and PCN are explored in more detail below.

1.2. DLVO Theory and Model

The interaction between charged particles in a suspension is cited as the driving mechanism in PCN
and EPC and has been modeled by the Derjaguin-Landau-Verwey-Overbeek (DLVO) theory. In this model,
Van der Waals interactions are responsible for the attractive force between particles, VA . VA is inversely
proportional to separation distance between particles and is therefore eective at small separation distances
(Equation 1) [21].

−AH a
VA = (1)
12h
where AH is the Hamaker constant, a is the radius of the particle, and h is the distance between the particles.
Repulsive forces in the DLVO model, VR , originate when the similarly charged electrical double layers of
particles overlap and are proportional to the surface charge squared (Equation 2).

VR = 2πaζ 2 exp (−κh) (2)

where  is the dielectric constant of water at 298K, ζ is the zeta potential of the particle (see Section 1.3 for
more detail), and κ is the inverse of the Debye-Huckel length and is given by Equation 3.

1/2
2e2 NA I

κ= (3)
kB T
where e is the elementary charge of an electron, NA is Avogadro's number, I is the ionic strength of the
solution, kB is Boltzmann's constant, and T is the temperature.
Attractive and repulsive forces between particles can be summed to nd the total force, VT (Equation
4). When VT is positive (VR dominates), the area under the curve of VT as a function of separation distance
represents the activation energy, the energy that must be provided to allow two particles to get close enough
so that the attractive Van der Waals force dominates.

VT = VR + VA (4)

Application of the DLVO theory to a colloidal suspension suggests that the electrostatic destabilization
of colloidal suspensions occurs when the activation energy is reduced so colloids can approach one another.
According to PCN and EPC, the goal of coagulation is the electrostatic destabilization of the colloidal
suspension. Electrostatic destabilization can be achieved by changing the surface charge of the colloids,
which is incorporated into the DLVO theory by zeta potential, ζ, or through reducing the distance over
which repulsive forces act by increasing the ionic strength, I, of the solution. The addition of positively
charged coagulant aggregates that adsorb/attach to the surface of the colloid would reduce the surface
charge and the resulting energy barrier. Electrostatic attraction would hold the PACl aggregates to the
colloidal surface, and the Van der Waals forces accounted for in DLVO theory would be responsible for the
inter-particle bond that holds the charge-neutralized colloidal particles together [13].

2
 Figure 1 depicts the interaction energy of a suspension of kaolin clay at the solution conditions used for
experiments in this research. DLVO theory suggests that kaolin clay particles in Cornell tap water would
be able to aggregate with a spacing between particles of 10 nm because the van der Waals attractive force
exceeds the electrostatic repulsive force at that separation distance. However, a kaolin suspension in Cornell
tap water does not create ocs with sedimentation velocities that enable removal in a sedimentation tank
without the addition of a coagulant (Figure 5 at 0 mg
L coagulant dose). The inability of the DLVO theory
to predict the failure of kaolin clay to aggregate suggests that an additional mechanism is important in
determining the aggregation behavior of colloids undergoing occulation.

Figure 1: Interaction energy for 30 mg

L kaolin clay in Cornell tap water, κ = 0.237 nm−1 , a = 1 µm, ζ =
−25 mV , 0 µM Al.

1.3. Measuring charge

Precipitation-charge neutralization (PCN) is hypothesized to facilitate aggregation when the colloid's
negative surface charge is near neutral. Surface charge, the electrical charge present at the surface, should
be measured as a surface potential, the dierence between the electrical potential of the surface and the bulk
uid (outside the electrical double layer). However, counter-ions and water molecules that are strongly bound
to the surface of the charged particle interfere with a direct measurement. Zeta potential is the dierence
in electrical potential between the shear plane and the bulk uid and is used as a surrogate parameter for
surface potential [21]. The shear plane is dened as the position where bulk uid can slip by a moving
particle; uid within the shear plane moves with the particle due to intermolecular forces. It has long been
assumed that the shear plane lies near the Stern layer, the monolayer of positive ions from the solution
that are attracted to the colloid surface, approximately 0.5 nm from the surface of the colloid. Li et al. [8]
determined the shear plane's distance from the surface to be approximately the the same as the Guoy plane.
The Guoy plane is dened by electrochemical composition and is located one Debye length from the colloid
surface. The volume created one Debye length (20 − 30 nm in a 10−4 M 1:1 electrolyte solution, and 4.2 nm
in the Cornell tap water used for this study) from the surface of the colloid is the volume containing sucient
excess ions of opposite charge to counter act the charge of the particle. It should be noted that while EPC
and PCN are based on electrostatic charge neutralization, zeta potential is not an appropriate parameter for
determining the coagulant dose at which EPC will be eective because zeta potential is averaged over the
whole surface of the colloid.

1.4. Diusion Limited & Reaction Limited Colloid Aggregation (DLCA & RLCA)
Aggregation often begins with a suspension of mono-disperse particles that collide and form clusters due
to Brownian motion [1, 13]. These clusters continue colliding due to both Brownian motion and velocity
gradients resulting in a suspension of polydisperse clusters. Through static and dynamic light scattering
experiments, researchers have found that the aggregation kinetics of many dierent colloids follow the same

3
pattern [13]. As a result, colloidal aggregation kinetics have been conceptually divided into two distinct stages
- diusion limited (DLCA) and reaction limited colloid aggregation (RLCA). In DLCA, the aggregation rate
is limited only by the rate of collisions between the particles; this assumes an attachment eciency near unity.
To predict an attachment eciency near unity for DLCA, the repulsive barrier between two approaching
particles must be reduced to much less than kB T , an approximation of the Brownian energy. If the energy
barrier is reduced to near zero, the full extent of DLCA can be achieved. In RLCA, collisions are driven
by velocity gradients and multiple collisions are required before two particles can stick together (attachment
eciency less than one) resulting in a slower aggregation rate. In all stages of colloid aggregation, the
clusters of original colloids (primary particles) take on a fractal structure, with a fractal dimension df ≈ 1.8
for DLCA and df ≈ 2.1 for RLCA [13, 1? ].

1.5. PACl Self-Aggregation Model (DLCA & RLCA)

To interpret the impact of PACl self aggregation on occulation, a physically-based model was generated
to capture the kinetics of Al13 self-aggregation in a circum-neutral pH suspension in the absence of other
7+
colloids. The primary PACl molecule ((AlO4 Al12 (OH)24 (H2 O)12 ) ) is estimated to be 1.2 nm in diameter
[17]. However, because of its propensity to self-aggregate, PACl is rarely observed as individual molecules.
PACl obtained from suppliers has already undergone some self aggregation. Therefore, the initial diameter,
dinitial , used in the model was not the diameter of a single PACl molecule, but rather the observed average
particle size, 180 nm (determined using a Malvern Zetasizer Nano-ZS, for the PACl used in this research
obtained from Zhengzhou City Jintai Water Treatment Raw Material Co., Ltd. ). The model provides an
estimate of the nal size of a PACl aggregate after a specied period of mixing due to collisions resulting
from both diusion and shear (Equation 20).
To account for the fractal nature of aggregates [11, 1, 16, 23], a fractal growth equation was used to
determine size of the aggregate particle after each consecutive collision (Equation 5). Inherent in this
equation is the assumption that all collisions occur between identical particles; such that collisions double
the number of the primary particles within the resulting aggregate.

n
dn = dinitial 2 DF ractal (5)

where n is the number of sequential collisions, DF ractal is the fractal dimension, dn is the aggregate diameter
after n collisions, and dinitial is initial aggregate diameter. The initial oc volume fraction, φinitial is
calculated by (Equation 6).

CP ACl
φinitial = (6)
ρinitial
where CP ACl is the concentration of PACl in the PACl aggregation tube used in this research, and ρinitial
gm
is the observed density of PACl, 1.138 mL (see Section 1.6). The oc volume fraction for a given collision,
φn , is then calculated as (Equation 7):

 3−DF ractal
dn
φn = φinitial (7)
dinitial
The eective particle density number, Nn , is determined by dividing the oc volume fraction by the
volume of a single aggregate (Equation 8).

φn
Nn = π 3 (8)
6 dn
Given Nn , the model of Meibodi et al. [15] that assumes Brownian motion and uses the Smoluchowski
approach for the collision of particles in a dilute suspension is used to calculate the average time for a given
collision, n, (tn,dif f usive ) (Equation 9).

3ν
tn,dif f usive = (9)
8kB T Nn
where ν is the dynamic viscosity of water at 298 K . Because of the initial 180 nm aggregate size and con-
centration used in experiments, shear induced collisions also occur in the reactor and contribute signicantly

4
to the nal aggregate size. Shear induced collisions are considered RLCA in nature and have an attachment
eciency less than unity. The time for a shear induced collision can be modeled by Equation 10, as derived
by Weber-Shirk and Lion [23].

  13   1
1 6 ν 2 1
tn,shear = 2 (10)
6 π ε φn3
In the transitional range between DLCA and RLCA, diusion and shear transport processes act in concert
to cause collisions. Equation 11 calculates the collision time when both transport mechanisms are operative.

1
tn = 1 1 (11)
tn,dif f usive + tn,shear

Each collision time is subtracted in the model from the total experimental reaction time available. Total
reaction time was controlled in experiments through laminar ow in a microbore tube of dened length and
diameter. When the available reaction time is depleted, the number of collisions is used in the fractal growth
equation (Equation 5) to give the nal aggregate size. The number of sequential collisions was not limited
to integer values to obtain the best estimate of the average aggregate size.
The model for diusion limited PACl aggregate growth was validated experimentally using a Malvern
Zetasizer Nano-ZS. A 5 mM Al suspension of PACl aggregates was ltered through a 0.2 µm syringe lter to
isolate a narrow size distribution so as to reduce variability in the observed sizes during aggregation. After
ltration, initial aggregates were 55 nm in diameter. As expected, this is less than the average aggregate
size in the unltered suspension, 180 nm. The suspension pH was adjusted to 7.5 by addition of dilute
N a2 CO3 and a series of size measurements was taken (nal PACl concentration 2.5 mM Al). The sample
cell was not mixed during the measurement phase and thus the only transport mechanism for collisions was
Brownian motion. Observations of aggregate size immediately after pH neutralization were not obtained
because initial collisions occurred at a time scale faster than the start-up detection time required for the
Zetasizer (10 s). However, even with the diculties inherent in observing particle collisions over very short
time and length scales, the results (Figure 2) suggest that the model based on DLCA is consistent with the
observed self-aggregation for PACl.

Figure 2: Diusion limited PACl self-aggregation as a function of time at a neutral pH. PACl concentration
was 2.5 mM Al. Measurements exhibit variability because of the polydisperse character of the suspension and
because averaging of size measurements was eliminated to decrease the time interval between measurements.
The model assumes a fractal dimension of 1.85 for DLCA in the quiescent sample and an initial particle size
of 55 nm. R2 = 0.69, n = 20

1.6. Coagulation Geometries

A careful geometric analysis for the attachment of coagulant particles to colloid surfaces can enlighten
our understanding of the role of the coagulant on subsequent colloid aggregation. The eectiveness of PACl

5
as a coagulant for negatively charged kaolin clay particles is generally attributed to charge neutralization
based on its high positive charge density [26, 10, 3, 28, 25, 12]. However, if the diameter of PACl coagulant
aggregates exceeds the Debye length, there is no reason to expect charge neutralization to be a prerequisite
of aggregation. This is consistent with observations by Chu et al. [4], Wu et al. [25] of occulation with
negative zeta potential (see Section 1.7).
As an initial approximation, kaolin clay platelets were assumed to have the volume of a sphere with a
diameter of 2 µm [28, 10]. The platelets were assumed to be cylinders with a 10:1 diameter to height ratio,
resulting in a diameter of 3.8 µm, height of 0.38 µm, and an initial surface area of 27 µm2 . The geometries
used in Figure 3a correspond to a turbidity of 15 NTU (30 mg/L clay, relationship determined through
laboratory observations) and a PACl concentration of 14.4 µM Al, at a pH of 7.5. The number of clay
gm
particles present assumes the density of kaolin clay, ρClay , is 2.65 mL . If initial PACl aggregate diameters
are close to 180 nm with a fractal dimension of 2.9 due to aggregation and dehydration that occurred during
the manufacturing process (as was the case for the PACl used in this research) the precipitated PACl would
cover 14% of the clay surface on average (Figure 3a) [14, 3]. This calculation assumes that all PACl molecules
above the solubility limit precipitate and attach to a clay platelet. At pH 7.5 the solubility limit of PACl is
82 nM Al [3]. If the coagulant is allowed to self-aggregate at circum-neutral pH for 6.2 s under the conditions
in this study, 1 µm aggregates are formed (see aggregation model below), covering only 4% of the colloid
surface (Figure 3b). For comparison, the Debye length corresponding to the ionic strength (I = 5.2 mol
m3 ) of
Cornell University tap water is 4.2 nm. This coupled geometric and DLVO analysis suggests that charge
neutralization may not be relevant for PACl dosages typically used in drinking water treatment.

a) b)

Figure 3: Visual geometries, approximately to scale for the experimental conditions in this study, 15 NTU
mg
(30
L ) and 14.4 µM Al. a) Small aggregates (180 nm) cover 14% of the clay surface with 113 PACl aggregates
per clay particle representing the control experiment described in section 2, b) Large PACl aggregates (1076
nm) cover 2.4% of the clay surface with 0.65 PACl aggregates per clay particle representing the worst case
observed in this study.

The geometric analysis accounts for the observation that industrial grade PACl is not a monodisperse
suspension of primary particles, but rather a suspension of preformed PACl aggregates (Malvern Zetasizer
Nano-ZS). The size of these aggregates can vary greatly based on the chemical preparation techniques at the
factory including aging times and handling [3, 27, 7]. The industrial grade powdered PACl used in this study
(source: Zhengzhou City Jintai Water Treatment Raw Material Co., Ltd) was found to produce a suspension
with an initial mean particle size of 180 nm when mixed with distilled water at a concentration of 5 mM Al
(Malvern Zetasizer Nano-ZS). The fractal dimension for the preformed aggregates was estimated from the
bulk density of the PACl granules as described below.
Assuming a porosity, ε, of 0.40 for random packing of spherical particles, Equation 12 was used to nd
the density of the preformed aggregates, ρinitial [5].

M obs
V obs
ρinitial = (12)
1−ε
where Mobs is the observed mass of a dry sample of the preformed aggregates, and Vobs is the observed
gm
volume of the sample of preformed aggregates. The observed density, ρinitial , was 1.138 mL . Next, Equation

6
13 was used to calculate the number of primary particles in a single preformed aggregate, Ninitial .
ρinitial ∗ Vinitial
Ninitial = (13)
ρ0 ∗ V 0
where Vinitial is the volume of a spherical preformed aggregate with a 180 nm diameter, ρ0 is the density of
gm
PACl primary particles, 1.907 mL , and V0 is the volume of a single primary particle with a 1.2 nm diameter.
The number of binary collisions that must have occurred to create the preformed aggregate was found by
Equation 14 to be 20.9.

Collisions = log2 (Ninitial ) (14)

Finally, the fractal dimension for the factory based aggregation is determined by Equation 15.

Collisions
DF ractal =   (15)
log2 dinitial
d0

The resulting fractal dimension determined for the conditions observed in this study is 2.9, signifying
that the PACl density was nearly independent of aggregate size in the preliminary aggregation stage during
production of the PACl granules at the manufacturer. With this information, it is possible to model the
fractional coverage or depth of coagulant on the surface of each clay platelet for a given set of PACl concen-
trations and geometric considerations. Over the range of doses in this study, increased clay surface coverage
causes an increase in attachment eciency during occulation, which could easily explain the improvement
in turbidity removal (Figure 7b). Equations 16-19 outline the calculations in the geometric model for PACl
aggregate coverage of clay.

mg 1
NClay = Inf luent T urbidity ∗ 2 ∗ (16)
L ∗ N T U MClay
where NClay is the number of clay platelets per unit volume, Inf luent T urbidity is the inuent turbidity in
mg
NTU, 2 L∗N TU is a conversion based on laboratory observations, MClay is the mass of each clay platelet,
gm
calculated from the volume given above and the density of kaolin clay, ρClay , 2.65
mL . The number of
binary collisions that must have occurred to create an aggregate of the size dictated by the model, df inal , is
calculated in Equation 17.

   
dinitial df inal
Collisions = DF,initial log2 + DF,f inal log2 (17)
d0 dinitial
where DF,initial is the fractal dimension of the preformed aggregate, 2.9, d0 is the diameter of the primary
PACl molecule, 1.2 nm, dinitial is the diameter of the preformed aggregate, 180 nm, DF,f inal is the fractal
dimension of the aggregate, and df inal is the nal diameter of the aggregate determined using Equation 5.
DF,f inal is assumed to be equal to DF,initial , 2.9, throughout the mixing chamber due to the high velocity
gradients. The number of PACl aggregates per unit volume in the occulator, NP ACl , is calculated by
Equation 18.

CP ACl,SRW ∗ N A
NP ACl = (18)
M WP ACl ∗ 2Collisions
where CP ACl,SRW is the concentration of PACl in the occulator, NA is Avogadro's number, and M WP ACl
is the molecular weight of PACl. The total volume of PACl aggregates on the surface of a single clay,
VP AClperClay is found by (Equation 19).

VP ACl,f inal ∗ NP ACl

VP AClperClay = (19)
NClay ∗ (1 − ε)
where VP ACl,f inal is the volume of a spherical PACl aggregate of diameter df inal . The total volume of
the PACl aggregates is then used to determine the fractional coverage of the clay platelets. The model
output is shown below (see section 3). Because PACl aggregates are assumed to be spherical, as opposed
to chain-like in structure, larger aggregates have a lower surface area to volume ratio, and will require a

7
larger mass of coagulant to provide sucient areal coverage of the clay platelets to obtain a high attachment
eciency. Therefore, as PACl aggregates increase in size they will be spaced further apart on the clay
platelets, the attachment eciency will decrease, and the overall particle removal eciency in occulation
and sedimentation will decrease.

1.7. Plausible coagulation mechanisms for aggregated coagulant at low doses

The inuence of coagulant dose on zeta potential is well documented in studies of the hypothesized PCN
mechanism [26, 2, 9, 3]. The published data, shows that the zeta potential is not zero at the lowest residual
turbidity. Results by Chu et al. [4], Wu et al. [25] show that residual turbidities drop precipitously at very low
doses while the zeta potential remains negative. Similarly, ecient occulation continues at highly positive
zeta potential and the complete lack of correlation between zeta potential and residual turbidity are well
documented [4]. Thus, charge neutralization does not explain the observed results and there must be another
mechanism leading to enhanced attachment eciency.
PACl self-aggregation and the absence of restabilization at positive zeta potentials demonstrates that
electrostatic neutrality is not required for aggregation and conrms that the PACl-PACl attachment mech-
anism is stronger than the electrostatic repulsion. Therefore, although the zeta potential of the colloids
becomes more positive as PACl is added to a suspension, there is no evidence that the change in zeta poten-
tial or electrostatic charge is what causes the aggregation to be successful. Therefore a dierent attractive
force must be operative to cause aggregation of either negatively or positively charged colloids. The precise
nature of the attractive force between the coagulant precipitate and other colloids is not the focus of this
paper. However, it is hypothesized that trivalent cations bonded to hydroxides within the Al13 molecule have
a strong polarity due to the large dierences in electronegativity between the trivalent cation and oxygen.
The atomic dierences in electronegativity produce molecular level charge distributions that would provide
strong intermolecular bonding. The attractive force would enable PACl to bond with itself and with clay
even when there is electrostatic repulsion. If valid, this hypothesis suggests that the goal of coagulation may
be to apply sucient PACl adhesive to the colloidal particles so that a high fraction of the collision impact
sites are coated with coagulant. The observed change in residual turbidity and oc size as the coagulant dose
is increased may therefore be the result of the changing fractional coverage of the colloidal surfaces with the
PACl adhesive. The geometric analysis given above shows that the PACl aggregates are approximately 40
(180 nm/4.2 nm) times larger than the Debye layer thickness and hence PACl aggregates extend beyond the
electric double layer and electrostatic repulsion may not signicantly inuence the interaction between clay
platelets with attached PACl aggregates.
As the geometric analysis in Figure 3 illustrates, self aggregation of PACl has a signicant inuence on the
resulting geometry of the colloidal surfaces. If the PACl aggregate size can be kept small, then the coagulant
can cover more of the colloid surface and potentially improve the attachment eciency and turbidity removal.
If the PACl is an adhesive, then the geometry of the adhesive on the colloidal surfaces is expected to inuence
attachment eciency and there should be no improvement in attachment after the colloids are completely
coated. The primary gain in attachment eciency is expected to occur at low fractional coverage of the
colloids.

2. Materials and methods

Experiments were conducted using an apparatus comprised of synthetic raw water and coagulant metering
systems, a coiled tube hydraulic occulator, and a occulation residual turbidity analyzer (FReTA) (see
Figure 4). Tse et al. [20] provide a complete description of the experimental apparatus and methods; only
the method for coagulant addition was changed for the experimental data presented here.
Briey, the synthetic raw water (SRW) metering system consisted of a concentrated stock suspension
of kaolinite clay (R.T. Vanderbilt Co., Inc., Norwalk, CT) mixed with tap water to produce a feedback-
regulated constant turbidity raw water source [22]. Reported Cornell University tap water characteristics
are: total hardness ≈ 150 mg
L as CaCO3 , total alkalinity ≈ 136
mg
L as CaCO3 , pH ≈ 8.2, and dissolved
mg
organic carbon ≈ 1.8
L [19]. The concentrated clay stock and the SRW feedstock were each stirred to ensure
homogeneous suspensions. For all of the experiments performed in this study, the SRW was maintained at a
constant turbidity of 15 ±
1 NTU, which corresponded to a clay concentration of approximately 30
mg
L , and
◦
a constant temperature, 25 C . All inuent chemicals were metered with computer controlled Cole Parmer

8
 Figure 4: Schematic of the experimental assembly.

MasterFlex L/S digital peristaltic pumps. Industry grade (31% as Al2 O3 ) polyaluminum chloride (PACl),
(AlO4 Al12 (OH)24 (H2 O)12 )7+ , was used as the coagulant for all experiments (Zhengzhou City Jintai Water
Treatment Raw Material Co., Ltd). The PACl was diluted with distilled water to give a stock concentration
of 5 mM Al. A N a2 CO3 stock was used to neutralize the acidic PACl stock and initiate precipitation. The
N a2 CO3 stock was prepared with distilled water at a concentration of 3.43 mM ; this concentration was
chosen because with 1 to 1 mixing it adjusted the PACl stock to pH 7.5. Dilution water was prepared with
distilled water and was adjusted to pH 7.5 with 75 µM N a2 CO3 . The dilution water and N a2 CO3 stock
were combined rst and then combined with the PACl stock in 0.8128 mm inner diameter micro-bore tubing
(Cole Parmer). The ow rate of the dilution stock was varied between experiments to alter the residence
time and PACl concentration in the PACl aggregation tube (center of Figure 4).
The mixing time provided for PACl self-aggregation was determined by Equation 20.

AT ube ∗ LT ube
tM ixing = (20)
QP ACl + QN a2 CO3 + QDilution
where AT ube 0.519 mm2 , LT ube is the length of the micro-
is the cross-sectional area of the micro-bore tubing,
µL
bore tubing, 5 cm or 60 cm, QP ACl is the ow rate of coagulant stock, 14.4
s , found by using Equation 21,
µL
QN a2 CO3 is the ow rate of base stock, 14.4 s , and QDilution is the ow rate of pH adjusted distilled water
stock, which was varied to produce a range of mixing times and concentrations. The micro-bore tubing
length was either 5 cm or 60 cm to achieve the full range of mixing times without creating excessive head
loss through the tube.

P AClDose ∗ QP lant
QP ACl = (21)
P AClStock
where QP lant is the total ow rate through the occulator, 5 mL
s , P AClDose is the concentration of PACl in
the occulator, 14.4 µM Al, P AClStock is the concentration of PACl entering the microbore tubing, 5 mM .
The SRW and coagulant was passed through a rapid mix unit comprised of a 120 cm segment of 4.3 mm
(0.17 ) ID tubing coiled around a cylinder with an outer diameter of 5 cm to ensure thorough mixing of the
SRW and the PACl. Reynolds number in the rapid mix tube was approximately 1450. Results from a dye
study showed that adequate mixing was achieved at this ow rate due to the secondary currents induced
by the coiling. The coagulated SRW entered an 84 m coiled tube occulator. The average velocity gradient

9
in the occulator, G, was maintained at 50 s−1 , calculated by Equations 2-9 of Tse et al. [20], and the
overall plant ow rate was maintained at 5 mL
s resulting in a hydraulic residence time, θ, of 1200 s in the
occulator. As Owen et al. [18] note, occulation is frequently studied in batch reactors with oine size
measurements for aggregation processes, resulting in poor control over the energy dissipation rate, reaction
time, and questionable size measurements. A tube occulator was used because it can be idealized as a high
Peclet number reactor much like a baed hydraulic occulator and also because the average velocity gradient
in laminar tube ow is well dened [23]. After two hydraulic residence times in the occulator, 2400 s, the
peristaltic pumps were ramped to a stop. The FReTA actuated ball valve closed, and the turbidity in the
quiescent settling column was measured over a period of 30 minutes. The occulation residual turbidity
analyzer (FReTA) was used to non-destructively measure both the sedimentation velocity and the residual
turbidity of the euent from the occulator. The settling velocity of the particles was calculated by dividing
the 13.64 cm distance between the bottom of the ball valve and the center of the zone illuminated by the
turbidimeter infrared LED by the time elapsed during settling. The residual turbidity is dened as the
average euent turbidity in the fty second interval around the capture velocity of 0.12 mm
s which is a
conservatively designed lamellar settler capture velocity [24].
Based on initial control experiments performed with the tube occulator at QP lant = 5 mL
and Gθ =
s
60, 000, a coagulant dose of 14.4 µM Al was determined to provide approximately 50% turbidity removal with
an initial turbidity of 15 NTU (Figure 5). Residual turbidity is comprised of ocs that settle more slowly
than the selected capture velocity, 0.12 mm
s . The 14.4 µM Al dose was chosen for subsequent experiments.
The sensitivity of residual turbidity to changes in coagulant addition at this dose enhanced the ability to
observe changes in the eectiveness of PACl at dierent PACl aggregate sizes (Figure 5).

Figure 5: Control experiment used to choose the coagulant dose. Error bars represent a 95% condence
interval, n=4.

The fractional coverage of the clay by the PACl was calculated using the geometric model and the
measured size of the PACl aggregates. The majority of the reduction in residual turbidity was obtained by
the rst 10% of clay surface coverage. There was negligible improvement in performance above a fractional
coverage of approximately 40%. These results are consistent with the expectation that the eectiveness of an
adhesive is related to the fractional coverage and that the signicant improvement in attachment eciency,
as indirectly measured by residual turbidity, occurs when the colloids are not completely covered.

10
 Figure 6: Residual turbidity as a function of fractional clay surface coverage by the PACl.

3. Results and discussion

The PACl aggregates (180 nm in diameter) are many times larger than the Debye length (calculated to
be 4.2 nm for the experimental conditions) and thus charge neutralization was not a signicant factor in oc-
culation with powdered PACl. The geometric model and aggregation models were combined to demonstrate
the potential role of clay platelet coverage on occulation performance.
Figure 7a depicts the observed relationship between collision potential and turbidity for the tested condi-
tions. In Figure 7d the y-axis is pC∗, dened by Equation 22. pC∗ is the negative logarithm of the fraction
of turbidity that remains.

 
∗ Residual T urbidity
pC = − log (22)
Inf luent T urbidity
All calculations of size and fractional coverage incorporate a degree of uncertainty because the attachment
eciency, α, is unknown during self aggregation in the PACl mixing tube. The attachment eciency term
is expected to be low to account for curvilinear particle trajectories [6] which were not considered in the

11
collision model. The PACl aggregates have a high fractal dimension and a low porosity and thus ow of
water through the aggregate is small and trajectory deection as particles approach could be signicant.
Attachment eciency values ranging from 0.05 to 0.2 were used to calculate the results in Figures 7b, 7c,
and 7d. Assumption of higher values of α resulted in the calculation of unreasonably sized PACl aggregates
(of the same order of the clay particle size). Attachment eciencies greater than 0.1 resulted in fewer than 1
PACl aggregate per clay particle and is thus inconsistent with the observed occulation and sedimentation.
Because mixing time was varied by the addition of dilution water, concentration of PACl in the PACl
aggregation tube varied in each test. The parameter Gθφ2/3 , where G is the velocity gradient, θ is the mixing
time or residence time in the PACl aggregation tube, and φ is the initial oc volume fraction, was chosen to
represent collision potential in the mixing tube in accordance with Weber-Shirk and Lion [23]. Since longer
mixing times and higher PACl concentrations allow more PACl self-aggregation, the experimental results
can also be expressed as a function of model-calculated PACl aggregate size as shown in Figure 7b. It is
clear that the performance of PACl as a coagulant is impaired by the formation of large PACl aggregates.
Given the diameter of model-calculated PACl aggregates, the surface coverage of kaolin clay colloids can
be calculated (Equation 23) and is shown in Figure 7c. For a constant PACl dose, increased aggregate size
results in fewer PACl aggregates per clay, which in turn correlates with the decrease in fractional removal
shown in Figure 7d.

VP AClperClay
F ractional Coverage = (23)
SAClay ∗ df inal
where SAClay is the surface area of a single clay platelet and VP AClperClay is the total volume of PACl
aggregates per clay platelet given by Equation 19.
The data collected from FReTA in each of these experiments can be used to calculate the distribution
of sedimentation velocities for the ocs formed under the given conditions. Figure 8 shows turbidity as a
function of sedimentation velocity for a sample of the aggregate sizes in the study, assuming α = 0.1 to
estimate the PACl aggregate sizes given in the legend. The sedimentation velocity is strongly correlated to
the nal oc size and determines the probability that a oc is removed. The slope of the curve represents
the fraction of the ocs with the corresponding sedimentation velocity. Residual turbidity is comprised of
ocs that settle more slowly than the capture velocity of a conservatively designed lamellar settler in a
sedimentation tank, 0.12 mm
s in these experiments. Figure 8 shows that the fraction of particles with settling
mm
velocities lower than 0.12
s increased at larger PACl aggregate sizes.
Starting with 180 nm initial aggregates, formation of a 1076 nm PACl aggregate requires approximately
6.2 seconds in the mixing chamber assuming a self-aggregation attachment eciency of 0.1. As shown
in Figure 7a, providing PACl time for self-aggregation will decrease the performance of the subsequent
occulation process by decreasing the attachment eciency in the occualtion phase andincreasing residual
turbidity. Therefore, rapid mixing of coagulant with colloid suspensions after coagulant addition is needed
to ensure PACl does not self-aggregate before contacting colloids in the water to be treated. PACl will
begin to self-aggregate as soon as the PACl stock blends with sucient raw water to neutralize the pH of
the mixture. The dilution required to begin self-aggregation is a function of the acid neutralizing capacity
(ANC) of the PACl stock and the ANC and pH of the raw water. Self-aggregation will occur most readily for
high pH and high ANC waters such as raw waters in equilibrium with calcium carbonate. Self-aggregation
is also favored by high PACl dose and low raw water colloid concentration because the distance between
PACl aggregates would be small and the probability of rst colliding with a PACl over a clay platelet would
be enhanced. If mixing does not occur at the point of coagulant application, the pH of the PACl stock will
be neutralized upon entering the water and the local concentration of coagulant will remain high leading to
self-aggregation. Self-aggregation after complete mixing with the raw water is expected to be insignicant
because PACl aggregate interactions with clay platelets become highly favored over self-aggregation.
While rapid mix of coagulant with raw water has been a prerequisite to occulation for decades, the
results of this study suggest that coagulation with PACl requires immediate, rapid mixing to decrease the
local concentration of coagulant at the point of addition. Although beyond the scope of the present study,
mixing also aids in the uniform application of the PACl aggregates on the raw water colloid surfaces.

12
4. Conclusions

1. Electrostatic charge neutralization is not required for PACl coagulation because the PACl aggregates
are much larger than the Debye length scale.
2. Geometric considerations show the reduction in residual turbidity as a function of PACl dose occurs
over a dose range that is consistent with partial clay platelet coverage with adhesive PACl aggregates.
3. PACl will readily precipitate and aggregate at circum-neutral pH in short time scales (~5 s) to sizes
that are signicant to the performance of the whole water treatment plant.
4. At a constant dose, increased size of PACl aggregates decreases clay surface coverage and negatively
aects the performance of subsequent occulation and sedimentation processes.
5. PACl self-aggregation is not expected to be a factor in municipal water treatment plants once the
coagulant is uniformly mixed with the raw water.

The quantitative turbidity removal - PACl aggregation relationships observed in this research are specic
to the experimental test conditions and are expected to vary in magnitude with inuent water quality and
coagulant dose. However, the qualitative eect of coagulant self-aggregation can be generalized and it is
clear that PACl aggregate size is a characteristic that should be considered when evaluating the results
of occulation performance experiments. From the perspective of the practical application of PACl to
water treatment, this research shows that turbidity removal by occulation and sedimentation is improved
if vigorous mixing occurs at the point of PACl addition. Immediate mixing with a high energy dissipation
rate, ε, will increase the separation distance between PACl aggregates and increase the probability that the
closest particle which the PACl will attach to is a clay particle, as opposed to another PACl aggregate. Some
methods of coagulant addition such as a drip feed into an entrance tank of standing water are expected
to not be eective means for rapid dispersal of coagulant. These dosing methods are anticipated to allow
self-aggregation of coagulant and necessitate a higher coagulant dose than would be needed if ecient mixing
were achieved.

5. References

[1] D. Asnaghi, M. Carpineti, M. Giglio, and M. Sozzi. Coagulation kinetics and aggregate morphology in
the intermediate regimes between diusion limited and reaction limited cluster aggregation. Physical
Review A (Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics), 45(2):101823,
January 1992.

[2] David Bache and Ross Gregory. Flocs in Water Treatment. IWA Publishing, London, UK, 2007.

[3] J. E. Van Benschoten and J. K. Edzwald. Chemical aspects of coagulation using aluminum salts - i.
hydrolytic reactions of alum and polyaluminum chloride. Water research, 24(12):15191526, 1990.
[4] YongBao Chu, BaoYu Gao, QinYan Yue, and Yan Wang. Investigation of dynamic processing on
aluminum oc aggregation: Cyclic shearing recovery and eect of sulfate ion. Science in China Series
B: Chemistry, 51(4):386392, 2008.
[5] R. M. German. Particle packing characteristics. Technical report, Metal Powder Industries Federation,
1989.

[6] D. W. Hendricks. Fundamentals of Water Treatment Unit Processes: Physical,

Chemical, and Biological. Taylor and Francis, 2009. ISBN 9781420061918. URL
http://books.google.com/books?id=9QNtn0P6fkwC. 2010036943.

[7] Chengzhi Hu, Huijuan Liu, and Jiuhui Qu. Preparation and characterization of polyaluminum chloride
containing high content of al13 and active chlorine. Colloids and Surfaces A- Physicochemical and
Engineering Aspects, 260(1-3):109117, 6 2005.
[8] Hang Li, Shiqiang Wei, Changle Qing, and Jinshan Yang. Discussion on the position of the shear plane.
Journal of colloid and interface science, 258(1):4044, February 2003.

13
 [9] Tao Li, Zhe Zhu, Dongsheng Wang, Chonghua Yao, and Hongxiao Tang. The strength and fractal
dimension characteristics of alum - kaolin ocs. International Journal of Mineral Processing, 82(1):
2329, 2 2007.

[10] Jr-Lin Lin, Ching-Ju M. Chin, Chihpin Huang, Jill R. Pan, and Dongsheng Wang. Coagulation behavior
of al13 aggregates. Water research, 42(16):42814290, 2008.
[11] Jr-Lin Lin, Chihpin Huang, Ching-Ju M. Chin, and Jill R. Pan. Coagulation dynamics of fractal ocs
induced by enmeshment and electrostatic patch mechanisms. Water research, 42(17):44574466, 11
2008.

[12] Jr-Lin Lin, Chihpin Huang, Ching-Ju M. Chin, and Jill R. Pan. The origin of al(oh)3-rich and al13-
aggregate ocs composition in pacl coagulation. Water research, 43(17):42854295, 9 2009.
[13] M. Y. Lin, H. M. Lindsay, D. A. Weitz, R. C. Ball, R. Klein, and P. Meakin. Universal reaction-limited
colloid aggregation. Physical Review A (Statistical Physics, Plasmas, Fluids, and Related Interdisci-
plinary Topics), 41(4):200520, 02/15 1990.
[14] Po-Hsun Lin, Monroe L. Weber-Shirk, and Leonard W. Lion. Comparison of the ability of three
coagulants to enhance lter performance. ASCE Journal of Environmental Engineering, (accepted
for publication), 2010.

[15] Majid Emami Meibodi, Mohsen Vafaie-Sefti, Ali Morad Rashidi, Azadeh Amrollahi, Mohsen Tabasi,
and Hossein Sid Kalal. Simple model for thermal conductivity of nanouids using resistance model
approach. International Communications in Heat and Mass Transfer, 37(5):555559, 5 2010.
[16] Jun Nan, Weipeng He, Juanjuan Song, and Xinxin Song. Fractal growth characteristics of ocs in
occulation process in water treatment. In 2009 International Conference on Energy and Environment
Technology (ICEET 2009), volume 2, pages 5828, Piscataway, NJ, USA, October 2009. State Key Lab.
of Urban Water Resource Environ., Harbin Inst. of Technol., Harbin, China, IEEE.

[17] C. R. O'Melia, K. A. Gray, and C. Yao. Polymeric inorganic coagulants. Technical report, American
Water Works Association, December 1989 1989.

[18] A. T. Owen, P. D. Fawell, J. D. Swift, D. M. Labbett, F. A. Benn, and J. B. Farrow. Using turbulent
pipe ow to study the factors aecting polymer-bridging occulation of mineral systems. International
Journal of Mineral Processing, 87(3-4):9099, July 2008.
[19] Bolton Point Water System. Drinking water quality report 2011. Technical report, Southern Cayuga
Lake Intermunicipal Water Commission, 2011.

[20] Ian C. Tse, Karen Swetland, Monroe L. Weber-Shirk, and Leonard W. Lion. Method for quantitative
analysis of occulation performance. Water research, 45(10):30753084, 5 2011.
[21] C. J. van Oss, R. F. Giese, and P. M. Costanzo. Dlvo and non-dlvo interactions in hectorite. Clays and
Clay Minerals, 38(2):151159, 1990.
[22] M. L. Weber-Shirk. An automated method for testing process parameters, 2008.

[23] Monroe L. Weber-Shirk and Leonard W. Lion. Flocculation model and collision potential for reactors
with ows characterized by high peclet numbers. Water research, 44(18):51805187, 10 2010.
[24] Roderick M. Willis. Tubular settlers a technical review. Journal AWWA, June:331335, 1978.
[25] Xiaohong Wu, Xiaopeng Ge, Dongsheng Wang, and Hongxiao Tang. Distinct coagulation mechanism
and model between alum and high al13-pacl. Colloids and Surfaces A: Physicochemical and Engineering
Aspects, 305(1-3):8996, September 2007.
[26] Xiaohong Wu, Xiaopeng Ge, Dongsheng Wang, and Hongxiao Tang. Distinct mechanisms of particle
aggregation induced by alum and pacl: Floc structure and dlvo evaluation. Colloids and Surfaces A:
Physicochemical and Engineering Aspects, 347(1-3):5663, 2009.

14
[27] Mingquan Yan, Dongsheng Wang, Jinren Ni, Jiuhui Qu, Christopher W. K. Chow, and Hailong Liu.
Mechanism of natural organic matter removal by polyaluminum chloride: Eect of coagulant particle
size and hydrolysis kinetics. Water research, 42(13):33613370, 2008.
[28] Changqing Ye, Dongsheng Wang, Baoyou Shi, Jianfeng Yu, Jiuhui Qu, Marc Edwards, and Hongxiao
Tang. Alkalinity eect of coagulation with polyaluminum chlorides: Role of electrostatic patch. Colloids
and Surfaces A: Physicochemical and Engineering Aspects, 294(1-3):163173, 2007.

15
(a) Residual turbidity as a function of 2/3
collision poten- (b) Residual Turbidity as a function of PACl
tial in the PACl aggregation tube, Gθφ . aggregate size for 4 assumed attachment e-
ciencies.

(c) Surface coverage as a function of PACl aggregate size. (d) Observed pC* as a function of the number of PACl aggre-
gates per clay.
Figure 7: Inuent turbidity was 15 NTU, PACl dose was 14.4 µM Al for all points. Initial PACl aggregate
size of 180 nm estimated for the model.

16
Figure 8: Sedimentation velocity from PACl aggregates of various sizes. The vertical line represents the
capture velocity, the sedimentation velocity at which the residual turbidity measurement is taken. Data
was collected at 1 second intervals for 30 minutes of settling. Each plot consists of 50 data points, each an
average of 36 seconds of data.

17