Environ Model Assess

DOI 10.1007/s10666-016-9518-2

Colloid-Associated Groundwater Contaminant Transport

in Homogeneous Saturated Porous Media: Mathematical
and Numerical Modeling
Mohsen Kheirabadi 1 & Mohammad Hossein Niksokhan 1 & Babak Omidvar 1

Received: 30 June 2015 / Accepted: 26 May 2016

# Springer International Publishing Switzerland 2016

Abstract During the past two decades, significant efforts 1 Introduction

have been made to study contaminant transport in the pres-
ence of colloids. Several researchers reported that colloidal There is no doubt that rapidly growing concern about environ-
particles could enhance the migration of contaminants in mental issues in recent decades has played a significant role in
groundwater by reducing retardation factor. When the colloi- colloid-associated groundwater-contaminant transport stud-
dal particles are present in the aquifer, the subsurface system ies. The importance of understanding colloid-facilitated/re-
can be considered as a three-phase system with two solid tarded transport phenomenon in water management and reme-
phases and an aqueous phase. The interaction between con- diation has been identified. Colloidal particles, ranging from
taminants, colloids, and solid matrix should be considered in 1.0 nm to 10.0 μm, have a small size and large surface area per
assessing the fate and transport of the contaminant in the unit volume of a colloid. Thus, they have a large number of
groundwater flow system. In this study, a one-dimensional reactive surfaces, and they could easily absorb contaminants
numerical model is developed by employing a fully implicit and enhance their mobility [1].
finite difference method. This model is based on mass balance McCarthy and Zachara [2] reported that because of the
equations and mass partition mechanisms between the carriers extraction of colloidal particles from the solid matrix, their
and solid matrix, as well as between the carriers and contam- composition is similar to immobile soil particles and can ab-
inants in a saturated homogeneous porous medium. This phe- sorb contaminants in the same way. They noted [2] that failure
nomenon is presented by two approaches: equilibrium ap- to account for the role of colloids in facilitating contaminant
proach and fully kinetic first-order approach. The formulation transport can lead to serious underestimation of the distances
of the model can be simplified by employing equilibrium that contaminants can travel.
partitioning of particles. However, contaminant transport can With regard to major pathways and mechanisms for trans-
be predicted more accurately in realistic situations by kinetic port of contaminants in the subsurface environment, two dis-
modeling. To test the sensitivity of the model, the effect of the tinct zones can be visible: the vadose zone and the saturated
various chemical and physical coefficients on the migration of zone [3]. Comparison between these two zones characteristics
contaminant was investigated. The results of numerical indicates that vadose zone has a gas–water interface, higher
modeling matched favorably with experimental data reported oxygen content, relatively higher concentration of organic
in the literature. matter, and subsequently higher microbial activity than satu-
rated zone. Contaminant transport in the vadose zone and
Keywords Colloids . Contaminant transport . Equilibrium saturated zone may be mainly vertical and horizontal, respec-
approach . Kinetic approach . Numerical modeling tively [3].
Various forces involved in the release process of fine par-
ticles adhering to the pore surfaces of the porous media. Static
* Mohsen Kheirabadi forces like London-van der Waals attraction and electrical
mohsen.kheirabadi@gmail.com
double layer repulsion and hydrodynamic forces of the
flowing fluid contribute to the total energy of interactions,
1
Graduate Faculty of Environment, University of Tehran, Tehran, Iran and if the net is repulsive, the fine particle may be released [4].
 M. Kheirabadi et al.

The main source of colloidal particles in the soil is in situ demonstrated that pore exclusion occurred when T/C ratio
particle release. Changes in solution chemistry have a pivotal was less than 1.5. The transport rate for spherical particle
role in particle release in soils and subsurface areas. Many depends on the ratio of particle size to the capillary radius.
research studies indicate that high pH, high Na + saturation, This phenomenon, which is known as the hydrodynamic chro-
and low ionic strength are favorable for dispersion and release matography (HDC), was described for the first time by Small
of fine particles in soils. Saiers and Hornberger [5] experimen- [12]. Although other researchers developed this theory, it can-
tally modeled the transport of 137Cs in the presence of kaolin- not explain the role of colloidal particles in the transport of
ite colloids and reported that the kaolinite colloids could en- groundwater contaminants [8]. In addition to the experimental
hance 137Cs transport through a column packed with quartz models, for a closer look and predicting the possible conse-
sand under conditions of low pore-water ionic strength. They quences of this phenomenon, we need mathematical models.
have also noted that the mobility and affinity of kaolinite for The first attempt to create and solve the colloid-facilitated
sorbing 137Cs decrease with increasing ionic strength and the groundwater-contaminant transport model was made by
mobility of kaolinite colloids increase with increasing pH and Corapcioglu and Jiang [8]. They solved their model numeri-
increasing flow rate. Roy and Dzombak [6] conducted exper- cally under homogenous condition. Corapcioglu and Choi
iments and reported a decrease in ionic strength could cause [13] reported a mathematical model based on mass balance
colloid mobilization. equations while colloids are present in the vadose zone, and
One of the most important sources of colloidal particles in the system can be treated as consisting of four phases. They
groundwater is human activity. Hydrogeochemical perturba- developed a one-dimensional homogeneous numerical model
tions can be induced by some activities such as waste disposal, for colloid transport in unsaturated porous media. Then they
groundwater pumping, and well-drilling. These operations verified their developed model against experimental data.
can provoke colloidal particles formation through precipita- They concluded the presence of air–water interface could cap-
tion or extraction from the solid matrix. ture colloids and retard their transport. Ibaraki and Sudicky
Microbial contaminants (protozoa, bacteria, and viruses) [14, 15] developed a two-dimensional numerical model of
also fall into the colloidal size range and called biocolloids colloid-facilitated solute transport in discretely fractured po-
[3]. They could pose a high risk to water resources. Several rous media. In their modeling, both the equilibrium and kinet-
studies have shown that contrary to popular belief, groundwa- ic approaches for solute sorption reactions were used simulta-
ter supplies are not free of pathogenic microorganism, and a neously. They showed that the slow kinetic desorption of con-
significant fraction of groundwater supplies are a source of taminants from colloids and high value of solid matrix poros-
waterborne diseases [7]. Therefore, for safe disposal of waste- ity could enhance the colloid-associated contaminant
water and adopt the appropriate bioremediation strategies, transport.
knowledge about biocolloids transport in porous media is Dual-porosity approach for groundwater-contaminant
necessary. transport in the presence of colloids has been presented by
Groundwater contaminants can be adhered to colloidal par- Corapcioglu and Wang [16]. They used fully implicit and
ticles by sorption, ion exchange, or other means and migrate to Crank-Nicholson schemes to solve the one-dimensional col-
far distances that are expected by using the conventional loid-facilitated contaminant transport. Sun et al. [17] devel-
advection/diffusion equation with normal retardation values oped a two-dimensional model for colloid transport in hetero-
[8]. Kan and Tomson [9] conducted experimental investiga- geneous porous media. They used the finite element method
tions to understand the role of dissolved organic matter for numerical solution of coupled fluid-flow equation and col-
(DOM) on the transport of hydrophobic organic compounds. loid transport equation.
They reported that the migration of DDT could increase in the Where the colloids have a high affinity for solid matrix
present of DOM by a factor of a thousand or more. Magee surfaces, these fine particles can retard contaminants’ migra-
et al. [10] reported that the retardation factor of phenanthrene tion. This opposite hypothesis has not been thoroughly inves-
in a dark quarry sand column reduced by an average factor of tigated during the past two decades. A few researchers such as
1.8 in the present of silty clay loam (soil DOM). Their exper- Knabner et al. [18] and Totsche et al. [19] considered this
iments show that colloids concentration and distribution coef- aspect of colloid-associated transport in their research papers.
ficients between contaminants, colloids, and solid matrix play They demonstrated that contaminant transport in the presence
a significant role in the migration of contaminants. of colloids could be enhanced or retarded, and this issue de-
Other parameters such as colloid size, soil structure, and pends on the sorption reaction rate of both colloids and
aqueous phase velocity are important factors affecting the fine contaminants.
particles migration in the porous media. Sirivithayapakorn and Bekhit and Hassan [1] developed a two-dimensional nu-
Keller [11] experimentally studied the phenomenon of size merical model of contaminant transport in the presence of
exclusion, the effect of pore throat to colloid diameter ratio colloids. They adopted finite differences with the total
(T/C ratio), and flow velocity on colloid transport. They variance-diminishing (TVD) scheme for solving the
Colloid-Associated Groundwater Contaminant Transport

governing equations numerically. They tested their model model should be modified as a three-phase model [3]. The
against different situations to evaluate the impact of kinetic interactions between colloidal particles and contaminants in
reactions on facilitating or retarding contaminant plumes in saturated homogeneous porous media can be treated as a
two-dimensional domains. Bekhit and Hassan [20] considered three-phase, six unknown constituents system. Figure 1 dem-
different sorption isotherms and investigated the effect of non- onstrates a diagram for this system.
linear and nonequilibrium interactions. They showed that The colloidal particles are considered to be suspended in
employing linear isotherms in any situation for colloid- the aqueous phase or attached to the solid matrix. Corapcioglu
associated contaminant transport may lead to inaccurate and Jiang [8] assumed that the amount of contaminant mass
results. adsorbed on particles is negligible in comparison to the mass
Recently, contaminant transport in the presence of of the colloidal particles. From this perspective, the mass bal-
biocolloids has been significantly studied. Bekhit et al. [21] ance equation for colloidal particles suspended in the aqueous
developed a conceptual and mathematical model and consid- phase is given by:
ered contaminant transport in groundwater in the presence of
∂½ðn−σc ÞC c 
colloids and bacteria. Sen [22] has reviewed a large volume of ¼ −∇: J cf −Qcm ð1Þ
work that has already been performed towards understanding ∂t s

the processes in pathogenic biocolloidal contaminants trans- The mass balance equation for colloidal particles captured
port in saturated and unsaturated porous media. Syngouna and on the solid matrix is expressed by:
Chrysikopoulos [23] conducted experiments to investigate the
effects of cotransport of colloidal clay and viruses in saturated ∂ð ρ c σ c Þ
¼ Qcm ð2Þ
porous media. These experiments have been done in ∂t s

laboratory-packed columns. They examined the effects of

Six unknown constituents exist in the system. The first two
presence/absence of colloids and pore water velocity in the
constituents include the mass concentration of suspended col-
colloid and virus cotransport. Katzourakis and Chrysiko
loids in the aqueous phase, Cc (M/L3) and volume of deposit-
poulos [24] developed a conceptual and mathematical model
ed colloids per unit volume of the porous medium, σc. In
for colloid and virus cotransport in porous media. They solved
Eqs. (1) and (2), n is the porosity, elapsed time is given by
the governing equations numerically using finite difference
t(T), ρc is the density of colloidal particles (M/L 3), Qcm
s is the
methods and tested their results with experimental data col- 3
net rate of colloid capture onto the solid matrix (M/L T), and
lected by Syngouna and Chrysikopoulos [23]. The results
Jfc is the specific mass discharge of suspended particles (L/T).
showed that interactions between colloids and viruses can
The specific mass discharge of suspended colloidal particles is
dramatically affect the fate and transport of viruses in porous
formulated as:
media.
The important aspects of this paper are as follows: (I) It J cf ¼ J B þ J MD þ qW C c ð3Þ
introduces governing equations for colloid-associated con-
taminant transport based on a three-phase medium in saturated J cf ¼ −DB ∇½ðn−σc ÞC c −DMD ∇½ðn−σc ÞC c  þ qW C c ð4Þ
homogeneous porous media. (II) The three-phase cotransport
Where JB is the discharge due to Brownian diffusion (L/T),
model has been presented by two approaches: equilibrium
JMD is the discharge due to mechanical dispersion (L/T), DMD
approach and fully kinetic first-order approach. (III) The mod-
is the mechanical dispersion coefficient (L2/T), DB is the
el equations for each approach have been solved numerically
by a fully implicit finite difference method. (IV) Numerical
results of the models have been verified by experimental data.
(V) Finally, sensitivity analyzes for equilibrium model and
first-order kinetic model have been employed.

2 Fine Particles Transport in Porous Media

2.1 Mathematical Model and Governing Equations

Models for contaminant transport in porous media have been

traditionally based on a two-phase approach: the mobile aque-
ous phase and the immobile solid phase. However, due to the
presence of colloidal particles in subsurface environments and Fig. 1 Conceptual schematic illustration of the six concentration
their role in contaminant transport, the traditional two-phase components in the three-phase system
 M. Kheirabadi et al.

Brownian diffusion coefficient (L2/T), and qw is the specific (1/T). Kinetic expressions for the simultaneous contaminant
discharge of water (L/T) [8]. capture/release mechanism are expressed by
The remaining four constituents include mass fraction of
contaminant adhered to mobile colloid surfaces per unit mass Qacm ¼ K m
a ðn−σc ÞC d −K sm ðn−σc ÞC c σcm
of mobile colloids, σcm, mass fraction of contaminant adhered þ K 2 ρc σc σcc −K 1 ðn−σc ÞC c σcm ð11Þ
to immobile colloid surfaces per unit mass of immobile col-
loids, σcc, mass concentration of contaminant dissolved in the Qacc ¼ K im
a ðn−σc ÞC d −K sc ρc σc σcc −K 2 ρc σc σcc
aqueous phase Cd (M/L3) and finally mass fraction of contam-
þ K 1 ðn−σc ÞC c σcm ð12Þ
inant adsorbed to the stationary solid matrix, σd. These re-
maining four constituents related to the forms that contami- Qad ¼ K f r ðn−σc ÞC d −K br ρb σd ð13Þ
nant can take in the groundwater in the presence of colloids.
The mass balance equation for contaminant adhered to mo- Where Km im
a and Ka are adsorption rate coefficients of con-

bile colloid surfaces can be written by: taminant on mobile and immobile colloids, respectively. Ksm
is the contaminant desorption rate coefficient of mobile col-
∂½ðn−σc Þσcm C c 
 loids (1/T) and Ksc is the contaminant desorption rate coeffi-
¼ −∇: σcm J cf þ Qacm ð5Þ
∂t cient of immobile colloids (1/T). Kfr and Kbr are adsorption
and desorption rate coefficient of contaminant onto and from
The mass balance equation for contaminant adhered to col-
the solid matrix, respectively (1/T).
loids captured on solid matrix is expressed by:
When equilibrium conditions exist, the net rate of colloids
∂ðσcc ρc σc Þ captured on the solid matrix is equal to zero, Qcms = 0 [25]. It
¼ Qacc ð6Þ occurs when the sorption of the colloids to the solid matrix is
∂t
reversible and instantaneous; in this situation, Eq. (10) can be
The mass balance equation for contaminant dissolved in simplified to:
the aqueous phase is given by:  
K 1 ðn−σc Þ
∂½ðn−σc ÞC d  σc ¼ Cc ð14Þ
¼ −∇: J df −Qacm −Qacc −Qad ð7Þ K 2 ρc
∂t
Equilibrium distribution coefficient of colloids with the
And finally the mass balance equation for contaminant ad- stationary solid matrix can be defined as:
hered to the solid matrix can be written by:
 
K 1 ðn−σc Þ
∂ð ρ b σ d Þ K6 ¼ ð15Þ
¼ Qad ð8Þ K 2 ρc
∂t
Moreover, the equilibrium sorption of colloids is expressed
In Eqs. (5) to (8), Qacm is the net rate of contaminant sorp-
by:
tion to mobile colloid surfaces (M/L3T), Qacc is the net rate of
contaminant sorption to immobile colloid surfaces (M/L3T), σc ¼ K 6 C c ð16Þ
Qad is the net rate of contaminant capture onto the solid matrix
(M/L3T), and Jfd is the specific mass discharge of dissolved When the sorption of contaminant to mobile and immobile
contaminant (L/T). colloids is in equilibrium, the mass transfer rates of the con-
taminant adsorbing and desorbing from the mobile and immo-
J df ¼ −D∇½ðn−σc ÞC d  þ qW C d ð9Þ bile colloids are equal. When the sorption of contaminant to
immobile colloids is in equilibrium,
D is the hydrodynamic dispersion coefficient of contami-
nant (L2/T), where D* = DB + DMD is the hydrodynamic dis- a ðn−σc ÞC d ¼ K sc ρc σc σcc
K im ð17Þ
persion coefficient of colloids (L2/T), and qW = (n − σc)V. V is  im 
K a ðn−σc Þ
the aqueous phase velocity (L/T). Eqs. (1), (2) and (5) to (8) σcc ¼ Cd ð18Þ
are governed on this phenomenon, and for a description of K sc ρc σc
a a a
each approach, we should define Qcm s , Qcc, Qcm and Qd spe- The equilibrium sorption of dissolved contaminants to im-
cifically. Kinetic expression for the simultaneous colloid mobile colloids is expressed by:
capture/release mechanism is given by:
σcc ¼ K 4 C d ð19Þ
s ¼ K 1 ðn−σc ÞC c −K 2 ρc σc
Qcm ð10Þ

Where K1 and K2 are the first-order adsorption and release Where K4 is the equilibrium distribution coefficient of dis-
rate coefficients onto and from the solid matrix, respectively solved contaminant with immobile colloids (L3/M). Likewise,
Colloid-Associated Groundwater Contaminant Transport

when the sorption of contaminant to mobile colloids is in

equilibrium,

Kma ðn−σc ÞC d ¼ K sm ðn−σc Þσcm C c ð20Þ

 m 
Ka
σcm ¼ Cd ð21Þ
K sm C c

The equilibrium sorption of dissolved contaminant to mo-

bile colloids is given by:

σcm ¼ K 5 C d ð22Þ

Where K5 is the equilibrium distribution coefficient of dis-

solved contaminant with mobile colloids (L3/M). When equi-
librium conditions exist, the net rate of contaminant capture on
the solid matrix, Qad, is equal to zero [25]. It occurs when the Fig. 2 Schematic diagram of the equilibrium distribution processes
sorption of the contaminant to the solid matrix is reversible between contaminant, colloids, and solid matrix
and instantaneous; in this situation, Eq. (13) can be simplified
to:
adsorption isotherm, substitution Qacm, Qacc, and Qad from
Qad ¼ K f r ðn−σc ÞC d −K br ρb σd ¼ 0 ð23Þ Eqs. (5), (6), and (8) into (7) yields an equation that can be
  simplified by substituting (9), (16), (19), (22) and (25). Where
K f r ðn−σc Þ
σd ¼ Cd ð24Þ D* = DB + DMD, the use of constant aqueous phase velocity
K br ρb and rearranging yields:
σd ¼ K 3 C d ð25Þ  
∂C d ρ ðK 4 −K 5 ÞK 6 ∂C c
R*c þ c Cd ¼
Where K3 is the equilibrium distribution coefficient of dis- ∂t θð1 þ K 5 C c Þ ∂t
solved contaminant with the stationary solid matrix (L3/M). 1  2 
D∇ C d −ð1 þ K 5 C c ÞV :∇C d þ D* K 5 ∇C c :∇C d
1 þ K 5Cc
ð29Þ
2.2 Equilibrium Approach
The retardation coefficient is expressed by:
A common approach to modeling the transport of contaminant  0
ρb K 3 þ K 6 K 4 C c
in the presence of colloidal particles is equilibrium approach. Rc ¼ 1 þ
*
ð30Þ
θ 1 þ K 5 Cc
In this approach, the mass partitioning mechanisms between
the colloids, solid matrix, and contaminant are assumed to be Where K′6 is the equilibrium distribution coefficient of col-
instantaneous [8]. Magee et al. [10] reported that if the colloids loids with the stationary solid matrix defined in terms of vol-
are poor sorbates and the retardation of colloidal particles by umetric fraction of deposition, equals to K6ρc/ρb. Under ex-
the solid matrix is dramatically lower than the retardation of perimental conditions, for overcoming to the low hydropho-
the contaminant by colloids, we could assume bicity of the test contaminants such as phenanthrene, it is
that(n − σc) = θ = const. The sum of (1) and (2) would yield: reasonable to have a constant and high level of colloids [10].
∂ðθC c Þ ∂ðρc σc Þ In this case (e.g., constant carrier concentrationCc), Eq. (29)
þ ¼ −∇: J cf ð26Þ reduces to:
∂t ∂t
The substitution of (16) into (26) and rearranging yields: ∂C d D
R*c ¼ ∇2 C d −V :∇C d ð31Þ
∂t 1 þ K 5Cc
∂C c
θRc ¼ −∇:J cf ð27Þ
∂t
Magee et al. [10] noted that since Cdt = Cd + σcmCc = (1 +
Where Rc is the colloids retardation coefficient K5Cc)Cd and the term (1 + K5Cc) is present in Eq. (30), this
ρc K 6 term represents the effect of mobile colloids on contaminant
Rc ¼ 1 þ ð28Þ transport. In addition to the equilibrium approach, a more
θ
realistic approach to this phenomenon was given by
Figure 2 shows a diagram for this approach. With the as- Corapcioglu and Jiang [8]. In this paper, the kinetic approach
sumption of (n − σ c ) = θ = const and regarding linear will be studied based on first-order kinetic expressions for the
 M. Kheirabadi et al.

mass partitioning between colloids, solid matrix, and Likewise, the substitution of Eqs. (4), (11), and (12) into (5)
contaminant. and (6) yields:

    
∂ C *c σcm ∂C * ∂σcm ∂2 C *c ∂C *c
¼ D* c −V C *c þ D* −V −ðK sm þ K 1 ÞC *c σcm
∂t ∂x ∂x ∂x 2 ∂x
2.3 Kinetic Approach þK m
a C d þ K 2 σc σcc
* *

ð34Þ
A set of model equations for colloid-associated contaminant

transport in saturated homogeneous porous media is presented ∂ σ*c σcc
¼ K im
a C d −ðK sc þ K 2 Þσc σcc þ K 1 C c σcm
* * *
ð35Þ
by assuming linear kinetic isotherm for mass partitioning be- ∂t
tween contaminant, colloids, and solid matrix. This assump-
Furthermore, upon substitution of Eqs. (9), (11), (12), and
tion implicitly expresses that the sorption sites for solutes are
(13) into Eqs. (7) and (8), the mass balance equation for dis-
abundant in the system. Failure to consider proper kinetic
solved contaminant and contaminant sorbed on solid matrix
isotherm assumption may lead to dramatic error in the model
can be expressed by Eqs. (36) and (37), respectively
prediction. Figure 3 shows a schematic diagram for the kinetic
approach. This figure is a conceptual schematic diagram and ∂C *d ∂2 C * ∂C *d
illustrates the reaction processes between contaminant, col- ¼ D 2d −V −K m
a C d −K a C d þ K sc σc σcc
* im * *
∂t ∂x ∂x
loids, and solid matrix.
þ K sm C *c σcm −K f r C *d þ K br σ*d ð36Þ
For expression of this model in a one-dimensional form, we
assume that (n − σc) = θ = const and aqueous phase velocity is ∂σ*d
constant. Where C*c = θCc, σ*c = ρcσc, C*d = θCd, and σ*d = ρbσd, ¼ K f r C *d −K br σ*d ð37Þ
∂t
the substitution of Eqs. (4) and (10) into Eqs. (1) and (2)
yields: Equations (32) to (37) form a set of equations which is almost
impossible to be solved analytically. For solving this problem, we
∂C *c ∂2 C *c ∂C *c should use numerical methods. The next section describes a
¼ D* −V −K 1 C *c þ K 2 σ*c ð32Þ
∂t ∂x 2 ∂x method which has been used to solve the set of equations.
∂σ*c
¼ K 1 C *c −K 2 σ*c ð33Þ
∂t
3 Numerical Modeling

In this section, one-dimensional modified advection-diffusion

equations are solved by the backward-time center space
(BTCS) scheme. This method, which is also called as the fully
implicit method, is employed for solving Eq. (31) and the set
of Eqs. (32) to (37). In this method, the unknown variables are
evaluated at the time step (n + 1), where (n) is the previous
time step, and all the variables are known at this time step.
Solving partial differential equations are subjected to the
boundary and initial conditions. The following initial and
boundary conditions are employed for the contaminant trans-
port problem in equilibrium approach

C d ðx; t Þ ¼ C d0 at x ¼ 0 ð38Þ

∂C d ðx; t Þ
¼ 0 at x ¼ L ð39Þ
∂x
C d ðx; t Þ ¼ 0 at t ¼ 0 ð40Þ

In equilibrium approach, numerical solution of (31) is

expressed by:

 ðnÞ ðnþ1Þ
 ðnþ1Þ

−R*c C d i ¼ ðS þ PÞC d i−1 − R*c þ 2S C d i
Fig. 3 Schematic diagram of the adsorption and release processes ðnþ1Þ
between contaminant, colloids, and solid matrix in kinetic approach þ ðS−PÞC d iþ1 ð41Þ
Colloid-Associated Groundwater Contaminant Transport

Where σ*(n
c
+ 1)
d , σd , σcm and σcc , respectively [8]. The finite
, C(n) (n) (n) (n)

 difference equations for Eqs. (34) to (37) are expressed by

D Δt
S¼  Eqs. (55) to (58), respectively.
1 þ K 5 C c Δx2


ðnþ1Þ  * ðnþ1Þ ðnþ1Þ
ð−S 2 −P2 ÞC *d i−1 þ 1 þ 2S 2 þ K m
a þ K a Δt þ K f r Δt C d i
im
−ðS 2 −P2 ÞC *d iþ1 ¼
and ðnÞ ðnÞ ð nÞ ðnÞ
C *d i þ K br σ*d i Δt þ K sc σ*c ðnþ1Þ i σcc i Δt þ K sm σcm i C *c i
ðnþ1Þ
Δt

V Δt ð55Þ
P¼ :
2Δx ðnþ1Þ 1 ðnÞ K f r Δt ðnÞ
σ*d i ¼ σ*d i þ C* ð56Þ
1 þ K br Δt 1 þ K br Δt d i
In kinetic approach, Eqs. (32) and (33) can be solved inde-
ðnþ1Þ 
pendently of the rest of the equations. The following initial ðnþ1Þ ðnþ1Þ ðnþ1Þ
S 3 σcm i−1 þ C *c i −P3 σcm i −S 3 σcm iþ1
and boundary conditions are employed for the colloids trans-
port problem: ðnþ1Þ ðnÞ ðnÞ ðnþ1Þ ðnÞ
¼ C *c i σcm i þ K m a C d i Δt þ K 2 σc i
* *
σcc i Δt ð57Þ
 
C *c ðx; t Þ ¼ C *c0 at x ¼ 0 ð42Þ ðnþ1Þ 
σcc i ¼ 1
ðσ*c ði nþ1Þ þðK sc þK 2 Þσ*c ði nþ1Þ ΔtÞ 
∂C *c ðx; t Þ
ðnþ1Þ  ð58Þ
¼0 at x ¼ L ð43Þ σ*c i
ðnÞ
σcc i þ K im C * ðnÞ
Δt þ K C * ðnþ1Þ
σ
ðnÞ
Δt
∂x a di 1 ci cm i

C *c ðx; t Þ ¼ 0 at t ¼ 0 ð44Þ
Where
σ*c ðx; t Þ ¼ 0 at t ¼ 0 ð45Þ
DΔt V Δt ST :Δt
S2 ¼ ; P2 ¼ ; S3 ¼ ; P3 ¼ SQ:Δt; ST
L is the length of the column, and is the initial mass Cc0* Δx 2 2Δx 2Δx
concentration of colloids times θ at the column boundary. " ! #
* ðnþ1Þ * ðnþ1Þ
* C c iþ1 −C c i−1 * ðnþ1Þ
Numerical solution of Eqs. (32) and (33) are expressed by: ¼ D −V C c i
2Δx
!
ðnþ1Þ K 1 K 2 ðΔt Þ2 ðnþ1Þ ðnþ1Þ
ð−S 1 −P1 ÞC *c i−1 þ 1 þ 2S 1 þ K 1 Δt− C *c i −ðS 1 −P1 ÞC *c iþ1 ¼
1 þ K 2 Δt and
ðnÞ K 2 Δt ðnÞ
C *c i þ σ* " ðnþ1Þ ðnþ1Þ ðnþ1Þ ! ðnþ1Þ ðnþ1Þ ! #
1 þ K 2 Δt c i C *c iþ1 −2C *c i þ C *c i−1 C *c iþ1 −C *c i−1 ðnþ1Þ
SQ ¼ D* −V −ðK sm þ K 1 ÞC *c i :
ð46Þ 2Δx2 2Δx

ðnþ1Þ 1 ðnÞ K 1 Δt ðnþ1Þ

σ*c i ¼ σ* þ C* ð47Þ
1 þ K 2 Δt c i 1 þ K 2 Δt c i

Where S1 = D*Δt/Δx2 and P1 = VΔt/2Δx. As mentioned 4 Models Verification and Sensitivity Analysis
above, two last equations (e.g., (46) and (47)) can be solved
independently. A similar fully finite difference scheme with The accuracy of the proposed numerical models should be
the following initial and boundary conditions is employed to evaluated. The results of numerical modeling for the equilib-
solve the contaminant transport problem: rium and kinetic approaches have been examined with two
different types of laboratory experiments. Parameters of these
C *d ðx; t Þ ¼ C *d0 at x ¼ 0; t ¼ pulse duration ð48Þ two models are listed in Tables 1 and 2. We used the proposed
∂C *d ðx; t Þ equilibrium model to simulate the transport of phenanthrene in
¼ 0 at x ¼ L ð49Þ porous media in the presence of dissolved organic matter
∂x
(DOM) as reported by Magee et al. [10]. Moreover, we
C *d ðx; t Þ ¼ 0 at t ¼ 0 ð50Þ
employed the proposed kinetic model to simulate the transport
σcm ðx; t Þ ¼ 0 at x ¼ 0 ð51Þ of cesium in porous media in the presence of amorphous silica
σcm ðx; t Þ ¼ 0 at t ¼ 0 ð52Þ colloids as reported by Noell et al. [25]. In the equilibrium
model experiment, the high and constant level of colloids
σcc ðx; t Þ ¼ 0 at t ¼ 0 ð53Þ concentration was employed into the finite column, and con-
σ*d ðx; t Þ ¼ 0 at t ¼ 0 ð54Þ stant concentration of contaminant continuously was injected
with constant velocity.
C*d0 is the initial mass concentration of dissolved contami- Magee et al. [10] conducted experiments to predict the
nant times θ at the column boundary. Eqs. (34) to (37) are phenanthrene transport in the presence of DOM. They
solved as a set of equations simultaneously. Solutions for employed a 5.5-cm-long glass column with 25-mm diameter.
C(n
d
+ 1)
, σ(n
d
+ 1)
, σ(n
cm , σcc
+ 1) (n + 1)
are obtained by using C*(n
c
+ 1)
, They used the dark quarry sand as porous media and silty clay
 M. Kheirabadi et al.

Table 1 Equilibrium model parameters sorption of DOM by the dark sand can justify the constant
Description Parameter Value colloids concentration assumption. Figure 4 shows the per-
centage of effluent phenanthrene with and without the pres-
Column length L 5.5 cm ence of DOM. Duplicate experimental data reported by
Column diameter Di 2.5 cm Magee et al. [10] shows phenanthrene eluted faster in the
Dry bulk density of the solid matrix ρb 1.61 × 106 mg/
l presence of colloids. It can be seen the breakthrough curves
Density of colloidal particles ρc 1.94 × 106 mg/ obtained from proposed model, A and B, match with the ex-
l perimental data favorably. The results were reported for total
Equilibrium distribution coefficient of dissolved K4 0.00438 l/mg
contaminant with immobile colloids mobile contaminant, Cdt i.e.
Equilibrium distribution coefficient of dissolved K5 0.00438 l/mg
contaminant with mobile colloids C d t ¼ C d þ σcm C c ð59Þ
Equilibrium distribution coefficient of colloids K′6 0.601 × 10−6
with the stationary solid matrix defined in l/mg For sensitivity analyzes, we employed previous models
terms of volumetric fraction of deposition, parameters [10, 25]. Sensitivity analyzes to various rate coef-
equal to K6ρc/ρb ficients have been presented for equilibrium approach in
Mass concentration of suspended colloids in the Cc 99 mg/l
aqueous phase Figs. 5, 6, and 7. The range of rate constants depends on the
Initial mass concentration of contaminant Cd0 0.0392 mg/l chemical characteristics of colloids and soil grains surface [8].
dissolved in the aqueous phase at x = 0 In Fig. 5, we used R*c as a combination of equilibrium model
Porosity (without – with DOM presence)
parameters, i.e., K3, K4, K5, K′6, and Cc to investigate the
n 0.299–0.397
Equilibrium distribution coefficient of dissolved K3 6.29 × 10−5
contaminant with the stationary solid matrix l/mg overall sensitivity of the model.
Hydrodynamic dispersion coefficient of D 0.0287 cm2/s As can be seen in Fig. 5 when R*c increases, the total mobile
contaminant
contaminant decrease significantly. In Fig. 6, we fixed all the
aqueous phase velocity V 0.000278 cm/s
reaction rates, and with changing the mass concentration of
suspended colloids, we have examined the effect of changes in
loam (Rhinebeck) as DOM or colloidal particles. The solution the mass concentration of the total mobile contaminant.
had a pH in the range of 6.4–7.5 and 99 mg/L colloids con- In addition to chemical parameters, sensitivity analysis has
centration. High level of colloids concentration and low been fulfilled for investigating the effects of changes in hydro-
dynamic dispersion coefficient of contaminant, D. Figure 7
Table 2 Kinetic model parameters illustrates that changes in hydrodynamic dispersion coefficient
Description Parameter Value of contaminant make significant changes in total mobile con-
(column 1) taminant transport.
To test the accuracy of the proposed numerical kinetic
Column length L 49.5 cm model, we have used the experiment results of Noell et al.
Column diameter Di 0.635 cm
Dry bulk density of the solid matrix ρb 1.51 g/ml [25]. They conducted experiments to predict the cesium
Density of colloidal particles ρc 2.2 g/ml
Initial mass concentration of suspended colloids Cc0 200 mg/l
in the aqueous phase at x = 0
Initial mass concentration of contaminant Cd0 0.398 μm/l
dissolved in the aqueous phase at x = 0
Porosity n 0.4
Longitudinal dispersivity αL 0.1 cm
Aqueous phase velocity V 0.0044 cm/
s
Adsorption rate coefficient of colloids K1 4 × 10−4 1/s
onto the solid matrix
Release rate coefficient of colloids K2 1.6 × 10−3
from the solid matrix 1/s
Adsorption rate coefficient of contaminant Km
a 1.0 × 10−3
on mobile colloids 1/s
Adsorption rate coefficient of contaminant Kim
a 1.0 × 10−3
on immobile colloids 1/s
Contaminant desorption rate coefficient Ksm 3.33 × 10−3
of mobile colloids 1/s
Contaminant desorption rate coefficient Ksc 3.33 × 10−3
of immobile colloids 1/s
Adsorption rate coefficient of contaminant onto the Kfr 2.70 × 10−3
solid matrix 1/s Fig. 4 Comparison between the simulated breakthrough curves with
Contaminant desorption rate coefficient Kbr 3.98 × 10−4 experimental data of Magee et al. [10]: (A) predicted curve with the
of the solid matrix 1/s equilibrium model without colloids and (B) predicted curve with the
equilibrium model with colloids
Colloid-Associated Groundwater Contaminant Transport

Fig. 7 Sensitivity of equilibrium model to changes in hydrodynamic

Fig. 5 Sensitivity of the equilibrium model to R*c as a combination of
dispersion coefficient, D
model parameters, i.e., K3, K4, K5, K′6, and Cc

transport in the presence of amorphous silica colloids. They colloidal particles flowing through these columns were con-
employed a 45.9-cm-long acrylic glass column with 6.35-mm tinuous with constant velocity.
diameter. They employed replicate columns with two different Sensitivity analyzes to various rate coefficients have
sized glass beads. To simulate our proposed kinetic model, we been presented for the kinetic approach in Figs. 9, 10,
used physical parameters determined for column 1. They used 11, and 12. We considered constant feed of colloids and
synthetic water, which had a pH of 8.9 and an alkalinity of contaminant solutions in sensitivity analyzes forK1, Ka,
11.44 meq/l. They named this water synthetic P-1 water. At andD. Figure 9 illustrates the sensitivity of total mobile
high pH, amorphous silica colloids could remain stable in contaminant concentration to the adsorption rate coeffi-
solution. Figure 6 shows the percentage of effluent cesium cient of colloids onto the solid matrix, K1. It shows that
with the presence of colloids. Experimental data reported by a decrease in deposition rate coefficient can make a sig-
Noell et al. [25] show that cesium eluted faster in the presence nificant effect on total mobile contaminant breakthrough
of colloids. As seen in Fig. 8, the breakthrough curve matches curves and an increase in total mobile contaminant trans-
favorably with the experimental data. In kinetic model exper- port. Obviously, it can be recognized that K2 has a similar
iments, the influence of colloids on contaminant migration effect, but reversely. It indicates that contaminants with
was tested by injecting 0.09 pore volume slugs of an equili- high adsorption rate coefficient will be more mobile in
brated suspension of cesium and colloids into the columns. the presence of mobile colloids.
This pulse has been employed for each experiment, and the

Fig. 6 Sensitivity of equilibrium model to changes in mass concentration Fig. 8 Comparison between the simulated breakthrough curve with
of suspended colloids in the aqueous phase experimental data of Noell et al. [25]
 M. Kheirabadi et al.

Fig. 9 Sensitivity of the kinetic model to changes in adsorption rate Fig. 11 Sensitivity of the kinetic model to changes in hydrodynamic
coefficient of colloids onto the solid matrix, K1 dispersion coefficient, D

We assume that the adsorption rate coefficient of contam- physical parameters of the model, i.e., D and Cc. In numerical
inant on mobile colloids is equal to the adsorption rate coeffi- modeling, we assumed hydrodynamic dispersion coefficient
cient of contaminant on immobile colloids, i.e., Km im
a = Ka = Ka. of colloids, D*is equal to hydrodynamic dispersion coefficient
Figure 10 shows the effects of changes in Ka on the total of contaminant, D. Figure 11 illustrates the sensitivity of the
mobile contaminant concentration. As seen in Fig. 10, at small kinetic model to hydrodynamic dispersion coefficient, D.
pore volumes, increasing Ka has an adverse effect on the total Changes in hydrodynamic dispersion coefficient of contami-
mobile contaminant transport. It can be attributed to increas- nant have a significant effect on total mobile contaminant
ing in Kim
a . However, finally, because of larger K2 than K1, as transport. This parameter is highly influential, and error in
expected, by increasing Ka the total mobile contaminant con- its calculation can make a tremendous error in the model
centration increases. prediction.
Changes in contaminant desorption rate coefficient for mo- Lastly, we investigated the sensitivity of the kinetic model
bile and immobile colloids, Ksm and Ksc, could create opposite to changes in Cc. Figure 12 shows the sensitivity of the kinetic
effects to Ka. Changes in the two other remaining factors, i.e., model to changes in initial colloid concentration, Cc0. For a
Kfr and Kbr, have significant effects on the total contaminant better illustration of sensitivity analysis, we used experimental
concentration, but there is no interaction between these rate model parameter exactly. Unexpectedly, changes in Cc do not
constants and absence or presence of colloids. In addition to have any impact on the behavior of the model. It is very
chemical parameters of the model, we investigated changes in

Fig. 10 Sensitivity of the kinetic model to changes in adsorption rate Fig. 12 Sensitivity of the kinetic model to changes in initial colloid
coefficient of contaminant on colloids, Ka concentration, Cc
Colloid-Associated Groundwater Contaminant Transport

controversial, and it seems opposite to our hypothesis which References

says colloids have significant effects on contaminant trans-
port. Bekhit and Hassan [1] studied this issue very carefully. 1. Bekhit, H. M., & Hassan, A. E. (2005). Two-dimensional modeling
They expressed that in the equilibrium approach model, the of contaminant transport in porous media in the presence of col-
loids. Advances in Water Resources, 28(12), 1320–1335.
effect of colloids was incorporated, using a modified effective
2. McCarthy, J. F., & Zachara, J. M. (1989). Subsurface transport of
retardation factor in the conventional advection–dispersion contaminants. Environmental Science & Technology, 23(5), 496–
transport equation, but in the kinetic model, we assumed that 502.
the rate coefficients to be constant independent of colloid con- 3. Sen, T. K., & Khilar, K. C. (2006). Review on subsurface colloids
centration. They explained how the distribution coefficient is and colloid-associated contaminant transport in saturated porous
media. Advances in Colloid and Interface Science, 119(2), 71–96.
normally obtained and concluded changing colloids concen- 4. Khilar, K.C. and Fogler, H.S. (1998). Migrations of fines in porous
tration does not directly reflect on the rate constants and the media. Vol. 12. Netherlands: Springer. doi:10.1007/978-94-015-
amount of facilitation. 9074-7.
5. Saiers, J. E., & Hornberger, G. M. (1996). The role of colloidal
kaolinite in the transport of cesium through laboratory sand col-
umns. Water Resources Research, 32(1), 33–41.
6. Roy, S. B., & Dzombak, D. A. (1998). Sorption nonequilibrium
5 Summary and Conclusions effects on colloid-enhanced transport of hydrophobic organic com-
pounds in porous media. Journal of Contaminant Hydrology, 30(1),
In recent decades, considerable efforts have been made to 179–200.
7. Bhattacharjee, S., Ryan, J. N., & Elimelech, M. (2002). Virus trans-
quantify and understand the phenomenon of contaminant port in physically and geochemically heterogeneous subsurface po-
transport in the presence of colloids. Numerical and experi- rous media. Journal of Contaminant Hydrology, 57(3), 161–187.
mental modeling results demonstrate that conventional 8. Corapcioglu, M. Y., & Jiang, S. (1993). Colloid‐facilitated ground-
advection-diffusion equations could not express this phenom- water contaminant transport. Water Resources Research, 29(7),
enon appropriately. The researchers believe that interaction 2215–2226.
9. Kan, A. T., & Tomson, M. B. (1990). Ground water transport of
between the colloidal particles, contaminant, and solid matrix hydrophobic organic compounds in the presence of dissolved or-
should be considered in colloid-associated contaminant trans- ganic matter. Environmental Toxicology and Chemistry, 9(3), 253–
port equations simultaneously, and advection-diffusion equa- 263.
tions should be corrected accordingly. Governing equations of 10. Magee, B. R., Lion, L. W., & Lemley, A. T. (1991). Transport of
dissolved organic macromolecules and their effect on the transport
this phenomenon were obtained by considering the system as
of phenanthrene in porous media. Environmental Science &
a three-phase medium with two solid phases, i.e., colloidal Technology, 25(2), 323–331.
particles and solid matrix. Equations (1), (2) and (5) to (8) 11. Sirivithayapakorn, S. and Keller, A. (2003). Transport of colloids in
represent a system of equations with 6 unknowns constitutes. saturated porous media: A pore‐scale observation of the size exclu-
Eqs. (1) and (2) could be solved together and independently sion effect and colloid acceleration. Water Resources Research. doi:
10.1029/2002WR001583.
from the rest of the equations to get the concentration of mo- 12. Small, H. (1974). Hydrodynamic chromatography a technique for
bile and immobile colloids. Solutions to Eqs. (5) to (8) are size analysis of colloidal particles. Journal of Colloid and Interface
convoluted, and we have to employ an iterative solution algo- Science, 48(1), 147–161.
rithm. In solving these equations, we have used the fully im- 13. Corapcioglu, M. Y., & Choi, H. (1996). Modeling colloid transport
in unsaturated porous media and validation with laboratory column
plicit scheme. In this field, various detailed examinations have data. Water Resources Research, 32(12), 3437–3449.
been conducted by several researchers. In this study, we have 14. Ibaraki, M., & Sudicky, E. (1995). Colloid‐facilitated contaminant
employed the results of these experiments and we have veri- transport in discretely fractured porous media: 1. Numerical formu-
fied numerical modeling results with the results of these ex- lation and sensitivity analysis. Water Resources Research, 31(12),
periments. The results of the comparison are favorable. The 2945–2960.
15. Ibaraki, M., & Sudicky, E. (1995). Colloid‐facilitated contaminant
results clearly show that contaminants could be transported to transport in discretely fractured porous media: 2. Fracture network
greater distances than conventional advection-diffusion examples. Water Resources Research, 31(12), 2961–2969.
models can predict. Model sensitivity to various parameters 16. Corapcioglu, M. Y., & Wang, S. (1999). Dual‐porosity groundwater
has been presented for the equilibrium and kinetic approaches. contaminant transport in the presence of colloids. Water Resources
Research, 35(11), 3261–3273.
The models are sensitive to changes of linear isotherm
17. Sun, N., et al. (2001). A novel two-dimensional model for colloid
coefficients and equilibrium distribution coefficients. We transport in physically and geochemically heterogeneous porous
know that the usage of equilibrium approach will be media. Journal of Contaminant Hydrology, 49(3), 173–199.
true when the concentration of colloidal particles is sig- 18. Knabner, P., Totsche, K., & Kögel‐Knabner, I. (1996). The model-
nificantly higher than the concentration of contaminant, ing of reactive solute transport with sorption to mobile and immo-
bile sorbents: 1. Experimental evidence and model development.
and it happens so rarely in nature. Therefore, the kinetic Water Resources Research, 32(6), 1611–1622.
approach can be more a realistic approach to this 19. Totsche, K., Knabner, P., & Kögel‐Knabner, L. (1996). The model-
phenomenon. ing of reactive solute transport with sorption to mobile and
 M. Kheirabadi et al.

immobile sorbents: 2. Model discussion and numerical simulation. 23. Syngouna, V. I., & Chrysikopoulos, C. V. (2013). Cotransport of
Water Resources Research, 32(6), 1623–1634. clay colloids and viruses in water saturated porous media. Colloids
20. Bekhit, H.M. and Hassan, A.E. (2007). Subsurface contaminant and Surfaces A: Physicochemical and Engineering Aspects, 416,
transport in the presence of colloids: effect of nonlinear and non- 56–65.
equilibrium interactions. Water Resources Research. doi:10.1029/ 24. Katzourakis, V. E., & Chrysikopoulos, C. V. (2014). Mathematical
2006WR005418. modeling of colloid and virus cotransport in porous media: appli-
21. Bekhit, H. M., El-Kordy, M. A., & Hassan, A. E. (2009). cation to experimental data. Advances in Water Resources, 68, 62–
Contaminant transport in groundwater in the presence of colloids 73.
and bacteria: model development and verification. Journal of 25. Noell, A. L., et al. (1998). The role of silica colloids on facilitated
Contaminant Hydrology, 108(3), 152–167. cesium transport through glass bead columns and modeling.
22. Sen, T. K. (2011). Processes in pathogenic biocolloidal contami- Journal of Contaminant Hydrology, 31(1), 23–56.
nants transport in saturated and unsaturated porous media: a review.
Water, Air, & Soil Pollution, 216(1-4), 239–256.