See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245407141

Upscaling transportof adsorbing solutes in porous media

Article  in  Journal of Porous Media · January 2010

DOI: 10.1615/JPorMedia.v13.i5.10

CITATIONS READS
21 168

2 authors:

Amir Raoof S. M. Hassanizadeh

Utrecht University Utrecht University
68 PUBLICATIONS   563 CITATIONS    281 PUBLICATIONS   8,613 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Simulation of fluid flow through thin porous layers and their interfaces View project

Sustainability Strategic Theme View project

All content following this page was uploaded by Amir Raoof on 07 January 2014.

The user has requested enhancement of the downloaded file.

 Journal of Porous Media 13(5), 395–408 (2010)

UPSCALING TRANSPORT OF ADSORBING SOLUTES

IN POROUS MEDIA

Amir Raoof ∗ & S. Majid Hassanizadeh

Universiteit Utrecht, Budapestlaan 4, 3584 CD, Utrecht, Netherlands

∗
Address all corespondence to A. Raoof E-mail: raoof@geo.uu.nl

Original Manuscript Submitted: 12/11/2008; Final Draft Received: 4/13/2009

Adsorption of solutes in porous media is commonly modeled as an equilibrium process. Indeed, one may safely assume
that within the pore space, the concentration of adsorbed solute at a point on the grain surface is algebraically related
to the concentration in the fluid next to the grain. The same, however, cannot be said about average concentrations. In
fact, during solute transport, concentration gradients develop within the pore space, and these could potentially give
rise to a scale-dependent adsorption process. The main objective of this research is to develop relationship between pore-
scale adsorption coefficient and corresponding upscaled adsorption parameters. Two approaches are used: Theoretical
averaging and numerical upscaling. In the averaging approach, equilibrium adsorption is assumed at the pore-scale
and solute transport equations are averaged over REV. This leads to explicit expressions for macro-scale adsorption rate
constants as a function of micro-scale parameters. In the numerical approach, first we simulate solute transport within
a single tube undergoing equilibrium adsorption at the pore wall, and then flux averaged concentration breakthrough
curves are obtained. These are used to determine the upscaled adsorption rate constants as functions of pore-scale
hydraulic and adsorption parameters. Results of the two approaches agree very well.

KEY WORDS: pore scale, kinetic adsorption, equilibrium adsorption, upscaling, solute transport

1. INTRODUCTION 1988; Pagitsas et al., 1986; Ryan et al., 1980; Sahimi,

The first continuum scale for the description of flow and
transport processes is the pore scale. Commonly, physic-
ochemical processes are reasonably well understood and
described by simple relationships at the pore scale. For
example, mixing due to diffusion is described by Fick’s
law, equilibrium dissolution is explained by Raoult’s law,
and adsorption at the grain surface is assumed to fol-
low an equilibrium relationship. In practice, however,
we make measurements, and model transport phenom-
ena, on scales much larger than the pore scale. Much
effort is being made to understand and quantify links be-
tween these scales; in particular, the dependence of mass
transfer on hydrodynamic properties of the system (e.g.,
Quintard and Whitaker, 1995; Wood et al., 2000, 2004,
2007; Edwards and Davis, 1995; Edwards et al., 1993;
Kechagia et al., 2002; Mauri, 1991; Mojaradi and Sahimi,
1988; Shapiro and Brenner, 1986, 1987, 1988). Often,
discrepancies between observation and theory arise be-
cause the large-scale description of transport does not ac-
count for some important aspects of small-scale behavior
(Bryant and Thompson, 2001; Raje and Kapoor, 2000;
Gramling et al., 2002). Commonly, small-scale effects are
lumped into empirical terms or coefficients that depend on
porous media properties and structure (Guo and Thomp-
son, 2001). Hence, inclusion of subscale mass transfer
effects in the description of porous systems is essential
to develop theoretically sound equations for describing
modeling mass transfer at the larger scales.
Because of limitations in measurements at small scale
and in observation frequency, experiments alone cannot
provide sufficient qualitative and quantitative understand-
ing of transport; theoretical work is also needed. In
porous media, upscaling may start either at the molecular

1091-028X/10/$35.00 Copyright °
c 2010 Begell House, Inc. 395
396
level (e.g., Murdoch and Hassanizadeh, 2002) or at the
pore scale (e.g., Whitaker, 1969, 1986; Hassanizadeh and
Gray, 1979). The most common approach is upscaling
from the pore scale, where the principal idea is to aver-
age the pore-scale transport and reaction processes over a
representative support volume to produce a macroscopic
model of the reactive transport.
Reaction processes such as adsorption, mineral disso-
lution, or homogeneous reactions could greatly influence
transport of dissolved matter in the soil and groundwa-
ter (Serrano, 2003; Bolt, 1979; Weber et al., 1991; van
der Zee, 1990; Acharya et al., 2005). These processes
typically occur at solid-fluid boundaries or fluid-fluid in-
terfaces. Commonly, adsorption is modeled as a (linear)
equilibrium process (Weber et al., 1991). The equilib-
rium assumption means that the chemical potential of the
solute in the fluid next to the solid grain is equal to the
chemical potential of the solute attached to the grain. This
assumption is probably acceptable in most interactions at
the pore scale, that is, at the pore boundaries. The same,
however, cannot be said at the larger scales, where we
work with average solute concentrations. Owing to vari-
ation of concentration within the pore and mass transfer
at the grain surface, the solute concentration close to the
wall will be different from the average concentration in
the bulk fluid (Binning and Celia, 2008; Meile and Tun-
cay, 2006; Li et al., 2007; Kechagia et al., 2002). Consid-
ering this issue, the adsorbed solute concentration, which
could be in equilibrium with the solution concentration
close to the adsorbed site, may not be in equilibrium with
the average concentration of solution. Instead, one often
has to use a so-called kinetic relationship between average
concentration of dissolved and adsorbed solutes. This is
particularly the case if the porous medium has a micro-
scopically heterogeneous structure such as in aggregated
porous media or fractured media.
Studies of reactive solute transport commonly involve
a series of batch experiments, through which reaction pa-
rameters are obtained. Given the fact that in batch experi-
ments, there are no spatial variations in the concentration
field (as it is a well-mixed system), the reaction models
and reaction parameters actually pertain to the pore scale.
In an open system, where diffusion and/or hydrodynamic
dispersion occur, gradients in concentration exist, as ob-
served both in experiments (e.g., Rashidi et al., 1996;
Kapoor et al., 1998; Taylor 1953; Aris, 2005) and through
numerical simulations (e.g., Li et al., 2006; Cao and Ki-
tanidis, 1998; Shapiro and Adler, 1997). As such, the
reaction rates are an outcome of coupling between the re-
action and hydrodynamic processes (Li et al., 2007); this
Raoof & Hassanizadeh
coupling can be the reason for much faster laboratory-
measured reaction rates of many minerals than those ob-
served in the field (White and Brantley, 2003; Maher et
al., 2004). Then the question arises as to what extent pore-
scale reaction models and parameters are applicable at the
macro scale.
Recently, Meile and Tuncay (2006) addressed this
question for the case of mineral dissolution and homoge-
nous reaction with the aid of a pore-scale numerical
model. They found that macro-scale descriptions of these
processes are different from pore-scale descriptions be-
cause of the effect of small-scale gradients in concentra-
tion fields. To investigate these effects, they numerically
generated virtual porous media using random placement
of identical spherical particles and solved diffusion and
reaction in the resulting pore spaces. They showed that
upscaled values of reaction and dissolution rates depend
on type of reaction, pore geometry, and macroscopic con-
centration gradient. They found that differences between
two scales become more significant for surface reactions
as compared to homogeneous reactions. A limitation in
the work of Meile and Tuncay (2006) is that they con-
sidered only diffusion transport and neglected advection.
Other modeling studies have shown that the role of advec-
tion on distribution of chemicals at the pore level is very
important (e.g., Bryant and Thompson, 2001; Knutson et
al., 2001; Robinson and Viswanathan, 2003; Szecsody et
al., 1998).
Li et al. (2007) studied the effect of pore-scale concen-
tration gradients on a mineral dissolution rate influenced
by advection. They introduced two kinds of models for
minerals that could dissolve at different rates. First, they
developed a Poiseuille flow model that coupled the re-
action rate to both advection and diffusion within a pore
space. Next they developed a well-mixed reactor model
that assumed complete mixing within the pore. They have
shown that concentration gradients could cause scale de-
pendence of reaction rates. Significant concentration gra-
dients would develop when diffusion is slower than the
advection process, provided that rates of advection and re-
action are comparable. This shows the effect of pore-scale
gradients and residence times on transport of reactive so-
lutes. The effect of residence times on reactive transport
was also addressed by Robinson and Viswanathan (2003),
who showed the importance of pore-scale gradients, espe-
cially for nonlinear reactions; solute pulses of short dura-
tion; and systems with broad residence time distribution
curves. Characteristic timescales of reaction processes
pose constraints for transport models (Mo and Friedly,
2000; Cao and Kitanidis, 1998).
Journal of Porous Media
Upscaling Transport of Adsorbing Solutes
Experimental studies (e.g., Guo and Thompson, 2001)
as well as pore-scale numerical models (Knutson et al.,
2001) have shown the dependence of mass transfer coef-
ficients (e.g., in dissolution process) on the hydrodynam-
ics of porous media. For example, Knutson et al. (2001)
found that the dimensionless mass transfer coefficient of
dissolution increased with Peclet number, Pe.
Another method for the upscale adsorption process in
porous media is the homogenization technique. Auriault
and Lewandowska (1996) have used the homogenization
technique (double asymptotic developments) to derive a
macroscopic equation for the average concentration field
of a solute that, at micro scale, is undergoing adsorp-
tion. They found effective parameters (e.g., dispersion-
adsorption tensor) that characterize the medium at the
macro scale. Through various characteristic dimension-
less parameters, they have shown the practical importance
of processes in different flow/transport regimes. The
macroscopic parameters were found to be dependent only
on the microscopic transport parameters and microscopic
geometrical properties. However, they found conditions
under which the macroscopic
¡ ¢ model does not exist, for ex-
ample, P e ¿ O ε−1 where ε, the homogenization pa-
rameter, is the ratio of the micro-scale heterogeneities size
to the macro-scale domain size. Because under some con-
ditions the problem could not be homogenized, they de-
scribed nonhomogenizable and homogenizable domains
in terms of characteristic dimensionless parameters. The
existence of nonhomogenizable domains demands cau-
tion in applying their method to real situations.
Van Duijn et al. (2008) applied the homogeniza-
tion technique to study adsorptive solute transport in
a capillary slit. They found upscaled equations using
the asymptotic expansion technique in terms of the ra-
tio of characteristic transversal and longitudinal lengths
under dominant Peclet and Domkohler numbers. They
have distinguished different characteristic timescales such
as longitudinal, transversal, desorption, and adsorption
timescales. They have derived effective models for vari-
ous conditions such as linear adsorption-desorption, non-
linear reactions, and equilibrium adsorption. They have
verified their method through comparison of solutions
with numerical solutions of original problems with and
without adsorption. They did simulations with high Peclet
numbers (larger than 104 ). They found excellent agree-
ment between results of their models and original prob-
lems.
This research is aimed at developing upscaled rela-
tionships for adsorption in terms of pore-scale properties
through both theoretical and numerical averaging. We
Volume 13, Number 5, 2010
397
wish to find relations for upscaled transport coefficients
for a solute that, at the pore scale, undergoes diffusion
and advection as well as first-order equilibrium reaction at
the solid surface. First, the problem of solute transport is
formulated at the pore scale, considering equilibrium ad-
sorption. Then the equations are averaged, and upscaled
equations are obtained. As a result, explicit expressions
are derived for the mass exchange between fluid and solid
phases in terms of average concentrations. Next, steady
state flow and transient adsorption in a single tube are
simulated numerically. The resulting breakthrough curves
from this single-tube model are compared to the solution
of the 1-D continuum scale transport equation to estimate
upscaled adsorption parameters.
2. THEORETICAL UPSCALING OF
ADSORPTION IN POROUS MEDIA
2.1 Formulation of the Pore-Scale Transport
Problem
Consider the transport of a solute in the pore space of a
granular soil. Processes affecting transport are considered
to be advection, diffusion, and chemical reaction within
the water phase plus adsorption to the solid phase at the
pore boundaries. The general form of the equation of
mass balance for the solute reads:
∂ci ¡ ¢
+ ∇. ci v + ∇.ji = rb i (1)
∂t
where ci (ML−3 ) is mass concentration of solute i
in a pore; v (LT−1 ) is interstitial water velocity; rb i
(ML−3 T−1 ) is the rate of chemical reactions with other
solutes, and ji (ML−2 T−1 ) is the diffusive flux of the so-
lute. The diffusive flux ji is given by Fick’s first law:
ji = −D0i ∇ci (2)
where D0i (L2 T−1 ) is the molecular diffusion coefficient
of solute i in water.
In principle, one should solve this equation within the
pore space of the soil subject to boundary conditions. At
a point on the pore boundary, adsorption causes a flux of
solute from the fluid to the solid phase; this gives rise to
an increase of the mass density of adsorbed solutes. The
rate of adsorption is equal
¡ to the¢solute mass flux normal
to the pore boundary, ci v + ji · n. So the following
condition at the pore boundary holds:
∂si
= (ci v + ji ) · n|s (3)
∂t
398 Raoof & Hassanizadeh

where si (ML−2 ) is the mass of adsorbed solute per unit Note that this volume-averaged definition of the flow ve-
area of the solid grains, n is a unit vector normal to the locity is admissible only if the fluid mass density vari-
pore wall, and |s denotes evaluation of the preceding ations are small (see Hassanizadeh and Gray, 1979, for
quantity within the pore but at the solid surface. Because discussion). We continue:
si is unknown, an additional equation is needed to have Average Diffusion Flux:
a determinate system. That extra equation comes from Z
the continuity of chemical potential at the grain surface. i 1
j = f ji dV (7)
This condition leads to an equilibrium relationship, and a V
linear approximation yields: Vf

¯ Average Chemical Reaction Rate:

si = kDi
ci ¯s (4)
Z
¯ 1
where ci ¯s is the solute concentration of fluid at the pore ri = f ri dV (8)
i
V
wall and kD (L) is an equilibrium, pore-scale distribution Vf
coefficient.
The set of Eqs. (1)–(4), together with conditions at the Average Adsorbed Mass Fraction:
outside boundaries of the porous medium and an appro- Z
i 1
priate set of initial conditions, completely specify the so- s = s s si dA (9)
lute transport problem at the pore scale. ρ V
fs A

2.2 Averaging of Pore-Scale Equations where ρs (ML−3 ) denotes the solid mass density.
Note that the average mass density of the sorbed solute
Now we would like to upscale these equations to the is now defined in the form of mass fraction (mass of so-
macro scale. To do so, we need to define an averaging vol- lute per unit mass of grains) si (MM−1 ), as is common in
ume, commonly denoted as the representative elementary solute transport.
volume (REV) (Bachmat and Bear, 1987). This is a well- Averaging of Eq. (1) over V f yields the following
known concept and has been extensively discussed in the macro-scale equation (see, e.g., Whitaker, 1969, 1986;
porous medium literature (see, e.g., Bear, 1988; Bachmat Hassanizadeh, 1986; Gray and Hassanizadeh, 1998):
and Bear, 1986; Hassanizadeh and Gray, 1979). Let us
denote the space occupied by REV with V . Volume V ∂nci ¡ ¢ ¡ ¢
bi
+ ∇ · nci v + ∇ · nJi = nri − U (10)
is in turn composed of two subvolumes: V f occupied by ∂t
the fluid phase and V s occupied by the solid grains. The
boundary of solid grains is denoted with Af s . The su- where v is the average flow velocity and ri is the reaction
perscripts f and s are used to denote the fluid and solid rate; the macro-scale adsorption term Ub i (ML−3 T−1 ) is
phases, respectively. Note that the averaging volume V is defined by:
taken to be invariant in time and space, whereas the sub- Z
volumes V f and V s may vary both in time and space. 1 ¡ i ¢
b i
U = c (v − w) + ji · ndA (11)
We shall integrate the fluid Eq. (1) over V f , and the V
interface condition (3) will be integrated over Af s . First, Af s

we need to define average properties:

where w is the velocity of the solid-fluid interface. Note
Average Mass Concentration: that Ji denotes the macro-scale hydrodynamic dispersion
Z vector and accounts for diffusion as well as the mixing of
i 1
c = f ci dV (5) solutes as a result of pore-scale velocity variations:
V
Vf
i
Ji = ncie
v + nj (12)
Average Flow Velocity:
Z where ev denotes the pore-scale velocity deviations, de-
f 1 fined as:
v = f vdV (6)
V v = v − vf
e (13)
V f

Journal of Porous Media

Upscaling Transport of Adsorbing Solutes 399

Next, the microscopic boundary condition (3) is av- of definitions (4) and (9) leads to the following macro-
eraged overAf s . The time average theorem and defini- scale relationship:
tion (9) are needed. The result is the following macro- i i
scale differential equation for the averaged adsorbed so- b i = − D0 (1 − n) ρs s̄i + SD0 f (c̄i )
U (19)
i
kD d d
lute concentration:
∂ (1 − n) ρs s̄i bi where S (L−1 ) denotes the solid grain specific area, S =
=U (14)
∂t Af s /V . Substitution of this result into the adsorbed mass
Equations (10) and (14) provide two equations to be balance (14) yields the standard linear kinetic adsorption
solved for c̄i and s̄i , provided that an appropriate rela- equation:
tionship is found for the adsorption rate U b i . This term
∂ (1−n) ρs s̄i b i
represents the exchange of mass between solid grains and = U = nkatt f (c̄i )−(1−n) ρs kdet si (20)
∂t
the fluid phase. Such a mass exchange could be due to
thawing, freezing, dissolution, precipitation, or adsorp- where the kinetic rate coefficients, katt and kdet (T−1 ),
tion. Here we consider the case of adsorption only and are defined by:
assume that adsorbed mass has no effect on the fluid or SD0i
katt = (21a)
solid mass density. Phase change, dissolution, and pre- nd
cipitation are neglected. Di
Because there is no phase change, the normal water kdet = i 0 (21b)
kD d
flux at the grain boundary, (v − w) · n in Eq. (11), will be
identically zero. Then, substitution of Fick’s law (Eq. (2)) The mass balance equation for the solutes now becomes:
in (2) Eq. (11) yields: ∂nc̄i ¡ ¢ ¡ ¢
Z i
+ ∇ · nc̄i v̄ i + ∇. nJ i = nr̄i
bi = − 1 ∂c ∂t (22)
U D0i dA (15)
V ∂n − nkatt f (c̄i ) + (1 − n) ρs kdet s̄i
Af s
which, together with Eq. (20), forms a set of two equa-
where ∂(.)/∂n denotes the derivative in the direction nor- tions to be solved for c̄i and s̄i .
mal to the grain surface. We now assume that ∂ci /∂n can
be approximated as follows:
¯ 2.3 Kinetic versus Equilibrium Effects
∂ci ¯¯ ³ ¯ ¯ ´
i¯ i¯
= c s − c pore /d (16) It is shown here that even if the adsorption process can be
∂n ¯Af s described by a linear isotherm at the pore scale, in gen-
¯
i¯
where c s denotes the¯ solute concentration in the fluid eral, it has a kinetic nature at the macro scale. The ques-
tion is when we can assume an equilibrium isotherm at
at the pore wall and ci ¯pore denotes the solute concentra-
the macro scale. That would be the case, of course, if the
tion within the pore at ¯some distance d. We now employ
kinetic process is very fast. This can be studied best if
Eq. (4) to eliminate ci ¯s from (16) and substitute the re-
we write the kinetic Eq. (20) in an alternative form that is
sult back into (15):
commonly employed:
Z i Z
D i
D ¯ ¡ i ¢
bi = −
U 0
si dA+ 0 ci ¯pore dA (17) Ub i = κ (1 − n) ρs KD f ( c̄i ) − s̄i (23)
i
V kD d Vd
Af s Af s i
where κ and KD are the macro-scale kinetic rate coef-
¯There, is of course, no information on the value
¯ of ficient and distribution coefficient, respectively, and are
ci ¯pore . However, it is plausible to assume that ci ¯pore is defined as:
concentration, c̄i . This means Di
a function of average fluid ¯ κ= i0 (24a)
that we may replace ci ¯pore by f (c̄i ) in Eq. (17) and set: kD d
¯ i
i SkD
ci ¯pore = f (c̄i ) (18) KD = (24b)
(1 − n) ρs
Use of this approximation (Eq. (18)), recognizing the fact From (24a), it is evident that the kinetic rate coefficient is
that ci is a constant within the averaging volume, and use the ratio of the micro-scale diffusion mass flux D0i ∆ci /d

Volume 13, Number 5, 2010

400 Raoof & Hassanizadeh

to the amount of mass that is to be adsorbed, that is, and (2) an equivalent upscaled 1-D model for the cross-
i
kD ∆ci . Obviously, the faster the diffusion process, the sectionally averaged concentration. These models allow
larger the kinetic coefficient will be (and thus the smaller us to investigate some of the assumptions made in our
the kinetic effect). On the other hand, the larger the upscaling approach and also to verify results of that ap-
amount of mass to be adsorbed, the more important the proach.
kinetic effect will be.
It is interesting to note that Eqs. (24) allow us to write
3.1 Flow and Transport at Pore Scale (Single-
the macro-scale kinetic rate coefficient as a function of the
Tube Model)
macro-scale distribution coefficient; this can be achieved
i
by eliminating kD between the two equations in (24a). Consider a long single tube with a constant circular cross
We obtain: section with radius R0 . We assume fully developed steady
SD0i state laminar flow in the tube (Poiseuille flow) so that the
κ= s i
(25)
ρ (1 − n)d KD velocity distribution is given by (Daugherty and Franzini,
The inverse of the kinetic coefficient, κ , is the char- 1965):
−1
" µ ¶2 #
acteristic timescale of the kinetic process. Now, if this r
v(r) = 2v̄ 1 − (27)
characteristic time is of the same order of magnitude as, R0
or bigger than, the residence time of solutes in pores, d/v
(i.e., for κ−1 ≥ d/v, then the kinetic process will be im- where r is the radial coordinate, v(r) is the local fluid ve-
portant. This condition can be represented by the dimen- locity, v̄ is the average flow velocity, and R0 is the radius
sionless number σD : of the cylinder. In the single-tube model, adsorption oc-
curs only at the wall of the tube (Fig. 1). In this study,
ρs (1 − n)vKD i
changes in the radius of the tube due to the adsorption
σD = i
(26)
SD0 process are neglected.
The mass transport in the tube is given by the following
Thus, if σD ≥ 1, then kinetic effects are important. equation:
Note that if the flow velocity is very small, then the kinetic
µ ³ r ´2 ¶ ∂c · 2
effects become negligible. In the limiting case of no flow, ∂c ∂ c
as is the case in batch experiments, the equilibrium rela- + 2v̄ 1 − = D0
∂t R ∂z ∂z 2
tionship (24) applies with no approximation. Thus batch µ ¶¸ (28)
1 ∂ ∂c
experiments can be used to obtain the macro-scale distri- + r
r ∂r ∂r
bution coefficient.
In the following section, we will perform numeri-
where D0 (L2 T−1 ) is the molecular diffusion coefficient
cal experiments to explore the assumptions leading to
and z is the axial direction along the tube.
Eqs. (24a) and also find an approximate value for d in
In the tube, the wall acts as an adsorbent, and therefore
Eq. (16).
the adsorption relation appears in the boundary condition
of the differential Eq. (28). Adsorption to the wall may
be described by the following equation, which prescribes
3. NUMERICAL UPSCALING OF ADSORBING that the diffusive mass flux to the wall is the only source
SOLUTE TRANSPORT of accumulation at the wall:
¯
Perhaps the simplest step in upscaling is to replace the ∂s ∂c ¯¯
= −D0 (29)
3-D flow and concentration fields within the pore (or a ∂t ∂r ¯s
tube) by 1-D fields, whereby velocity and concentration
are averaged over the pore cross section. As mentioned in where s (ML−2 ) is adsorbed mass per unit area. This is
Section 1, this upscaling has been considered for homo- the same as Eq. (3), where the first Fick’s law is assumed.
geneous reactions as well as dissolution. Here we treat Similar to Section 2, we assume that linear equilibrium
the upscaling of adsorptive solute transport. adsorption holds at the wall, characterized by the distri-
To analyze the scale dependence of the adsorption bution coefficient value kD (L):
process, we have developed two models: (1) a single-
tube model to simulate details of transport within a pore s = kD c|s (30)

Journal of Porous Media

Upscaling Transport of Adsorbing Solutes 401

FIG. 1: Conceptual representation for the single-tube model. The velocity profile is assumed to be parabolic

Next we introduce the following dimensionless variables The standard package Flex-PDE (Flex-PDE, 2005) has
and parameters: been used to numerically solve this set of equations. We
have simulated solute transport for a range of values of
∗ c ∗ r ∗ z ∗ v̄t
c = , r = , z = , t = parameters P ep and κ. The solution of Eqs. (32)–(34)
c0 R0 R0 R0 ∗ ∗ ∗ ∗
(31) results in concentration filed c (z , r , t ) for different
v̄R0 s k D values of Peclet number, P ep , and dimensionless dis-
P ep = , s∗ = , κ=
D0 c0 R0 R0 tribution coefficient, κ. This concentration field and its
where c0 is the concentration at the inlet boundary. cross-sectional average, c̄∗ (z ∗ , t∗ ), may be considered to
So we will end up with pore-scale dimensionless mass be equivalent to “observation data.” We then use these
transport equation: observed data to obtain the relationship for upscaled ad-
sorption parameters.
∗ ∗
· 2 ∗
∂c ∂c 1 ∂ c
+ 2(1 − r∗2 ) ∗ =
∂t∗ ∂z P ep ∂z ∗2
µ ∗
¶¸ (32) 3.2 Flow and Transport at 1-D Tube Scale
1 ∂ ∗ ∂c
+ ∗ ∗ r One can obtain the first level of upscaling by averaging
r ∂r ∂r∗
the pore-scale concentration over the cross section of the
where P ep is the pore-scale Peclet number, which ex- tube. This results in the 1-D average concentration field
presses the ratio between the magnitude of the advective c̄∗ (z ∗ , t∗ ), to which we refer as the 1-D tube scale. Now,
and diffusive transport terms. The dimensionless bound- the question arises what the governing equation should be
ary conditions for our model are as follows: for this average concentration. To answer this question,
we examine a typical breakthrough curve of average con-
at z ∗ = 0, c∗ = 1.0 (33a) centration for an adsorbing solute at a position far farm
∂c∗ the tube inlet (e.g., at z ∗ = 100), as shown in Fig. 2.
at z ∗ = ∞, = 0 (33b) It is evident that the breakthrough curve shows a non-
∂z ∗
ideal behavior with a nonsymmetric derivative. The tail-
∂c∗
at r∗ = 0, = 0 (33c) ing observed here suggests that at this scale, adsorption
∂r∗
should be described as a kinetic process. The governing
∗ ∂c∗ ∂s∗ equations for 1-D transport of kinetically adsorbing solute
at r = 1, = −P ep ∗ (33d)
∂r∗ ∂t are as follows:
Substitution of Eq. (30) in the boundary condition (33d) ∂c∗ ∂c∗ 1 ∂ 2 c∗ ∂s∗
results in: ∗
+ ∗
= ∗2
− ∗ (35)
∂t ∂z P e ∂z ∂t
∂c∗ ∂c∗
= −P e p κ at r∗ = 1 (34)
∂r∗ ∂t∗ ∂s∗ ∗ ∗
∗
= katt c∗ − kdet s∗ (36)
It is evident that P e and κ are the two parameters in ∂t
p
this set of equations that control the transport and reaction where P e = v̄R0 /DL , which involves the average flow
processes within the tube. velocity v̄ and the longitudinal dispersion coefficient DL .

Volume 13, Number 5, 2010

402 Raoof & Hassanizadeh

FIG. 2: Breakthrough curve of average concentration (solid line) resulting in non-Gaussian derivative (dashed line);
based on simulations from single-tube model with continuous input

Note that P e is the tube-scale Peclet number; it is differ- studied upscaling of mineral dissolution within a single
ent from the pore-scale P ep , which is based on the diffu- pore. Reactive flow experiments were performed in a
sion coefficient (see Eq. (31)). cylindrical pore, 500 µm in diameter and 4000 µm long,
∗ ∗
Furthermore, katt and kdet are dimensionless adsorp- drilled in a single crystal of calcite. They employed the
tion and desorption rate coefficients, respectively. These single-tube model to simulate the concentration of Ca2+
are related to dimensional coefficients katt and kdet resulting from dissolution of calcite. The kinetic calcite
through the following relationships: dissolution formula (from Chou et al., 1989) was assumed
to hold at the wall of tube. Then, the flux-average concen-
katt R0 kdet R0
∗
katt = ∗
, kdet = (37) tration of Ca2+ , calculated from the single-tube model,
v̄ v̄ was compared to the measured Ca2+ concentration from
∗
These equations contain three parameters: P e, katt , the experiment for different pH values and flow condi-
∗
and kdet . We have evaluated these parameters by fit- tions. They found an excellent agreement between mod-
ting the solution of this set of equations to the average eling results and results of experiment. This, we believe,
breakthrough concentration, c̄∗ (z ∗ , t∗ ), obtained from the is an indication of the applicability of our procedure in
single-tube model for various values of pore-scale Peclet upscaling from pore to tube scale.
number (P ep ) and κ. This procedure results in relation-
∗
ships between the three upscaled parameters (P e, katt , 3.2.1 Upscaled Peclet Number (Pe)
∗
and kdet ) and their corresponding pore-scale parameters
(P ep and κ). We have used the cross-sectional averaged Here we assumed that the dispersion is not affected by
concentrations from the single-tube model to find corre- the adsorption process. Therefore we evaluate the up-
sponding upscaled adsorption parameters. A similar ap- scaled Peclet number (Pe) for the case of a nonadsorb-
proach was used by Li et al. (2008) for upscaling of disso- ing solute (i.e., κ = 0). This is done by fitting the break-
lution processes under steady state flow conditions. They through curve from the single-tube model to the solution

Journal of Porous Media

Upscaling Transport of Adsorbing Solutes 403

of the 1-D transport equation (Eq. (35)) with no adsorp- have been checking the covariance matrix to make sure
∗ ∗
tion (i.e., katt = kdet = 0). Figure 3 shows the result- that parameters are not correlated.
ing graph, where P e is plotted as a function of pore-scale By repeating the procedure for a range of pore-scale
Peclet number (P ep ). In the same figure, the Taylor dis- parameters (1 < P ep < 300, and 0.1 < κ < 20) and
∗ ∗
persion formula (Taylor, 1953; Aris, 2005) for upscaled founding the corresponding upscaled katt and kdet we
Peclet number in a tube is also plotted (Eq. (38)). The ex- could find relations between these sets of parameters. Fig-
∗ ∗
cellent agreement indicates the accuracy of our numerical ures 5 and 6 show the plots of katt and kdet as a function
code in capturing the transport within the tube and also of P ep and κ.
shows that the Taylor assumption is valid for this prob- From relations shown in Figs. 5 and 6, we found the
∗ ∗
lem. We use Eq. (38) to obtain upscaled Peclet number best-fit formulas for katt and kdet :
(P e) in simulation of adsorbing solute in the next section:
∗ 4.0(1 − e−3κ )
katt = (40a)
1 1 P ep P e0.95
p
= + (38)
Pe P ep 48 9.0
∗
kdet = (40b)
∗ ∗
(0.5 + 4.5κ)P e0.95
p
3.2.2 Upscaled Adsorption Parameters (katt and kdet )
Next, we define the upscaled distribution coefficient (KD )
Since we employed the Taylor formula to estimate the up- by:
scaled Peclet number, there are only two upscaled param- ∗ ¡ ¢
∗ ∗ katt
eters, katt and kdet , left to be determined as a function of KD = ∗
= 0.4 1 − e−3κ (0.5 + 4.5κ) (41)
pore-scale parameters: kdet

∗
According to Eq. (41), upscaled distribution coefficient
katt = f (P ep , κ) KD is independent of Peclet number. It is practically a
(39)
∗
kdet = f (P ep , κ) linear function of pore-scale distribution coefficient κ, as
shown in Fig. 7.
The CXTFIT curve-fitting program (Toride et al., 1995) The linearity of upscaled distribution coefficient KD
was used for the purpose of solving Eqs. (35) and (36) as a function of pore-scale distribution coefficient κ also
and fitting the breakthrough curves. Figure 4 shows an shows the validity of our upscaling process. Both KD and
example of fitting the breakthrough curves and the cor- κ are a measure of the capacity of the porous medium to
responding parameters. Through the fitting process, we absorb the mass, and thus they should be linearly related.

FIG. 3: Graph of P e versus P ep . The points are the result of fitting of the breakthrough curve of average concentra-
tion (with κ = 0) to Eq. (35) to find P e. The solid line shows the function by the Taylor formula of Eq. (38)

Volume 13, Number 5, 2010

404 Raoof & Hassanizadeh

FIG. 4: The resulting breakthrough curves for pore-scale (circles) and 1-D upscale (solid line) models. In this
illustration, the pore-scale parameters are P ep = 35 and κ = 5.0, and the corresponding upscaled parameters by
∗ ∗
fitting are found to be katt = 0.14 and kdet = 0.05. Using Eq. (38), P e will be 1.3

(a) (b)
∗
FIG. 5: The relation between macro-scale kdet as a function of pore-scale κ and P ep

(a) (b)
∗
FIG. 6: The relation between macro-scale katt as a function of pore-scale κ and P ep

Journal of Porous Media

Upscaling Transport of Adsorbing Solutes 405

∗ ∗
FIG. 7: The relation between upscaled distribution coefficient (KD = katt /kdet ) as a function of pore-scale dimen-
sionless distribution coefficient κ

4. DISCUSSION OF RESULTS D0i

kdet = i
(45b)
kD wR0
Equations (40) can be converted to dimensional forms us-
ing Eqs. (31) and (37); this gives: These equations agree quite well with Eqs. (42a), which
³ were obtained through numerical averaging. Indeed, if
kD ´
4.0 1 − e−3 R0 v̄ D00.95
0.05
we acknowledge the fact that for a tube, porosity is unity
katt = (42a) and specific surface S = 2/R0 , Eqs. (45) reduce to the
R01.95 following forms:
avg 2D0
9.0 v̄
0.05
D00.95 katt = (46a)
kdet = ³ ´ (42b) wR02
0.5 + 4.5 kRD0 R01.95 avg D0
kdet = i wR
(46b)
We observe that katt and kdet are only weak functions kD 0

of velocity and that they strongly depend on the geome-
try of the pore space and the diffusion coefficient as well
as the pore-scale distribution coefficient. We could com-
pare Eqs. (42a) with corresponding equations derived us-
ing the averaging method in Section 2.2. For¯the case of a
single tube, it is plausible to assume that ci ¯pore at a po-
sition somewhere between the pore center and pore wall
(denoted by wR0 , where 1.0 > w > 0) will be equal
to the average fluid concentration, c̄i . This means that in
Eq. (19), we may set:
f (c̄) = c̄ and d = wR0 (43)
As a result, Eq. (19) leads to the following macro-scale
kinetic adsorption relationship:
b i = nkatt c̄i − (1 − n) ρs kdet s̄i
U (44)
where the kinetic rate coefficients katt and kdet (T−1 ) are
defined by:
SD0i
katt = (45a)
nwR0
where superscript avg stands for “averaging method.” Re-
call that wR0 denotes the radial position
¯ of a point in the
pore where point concentration ci ¯pore is equal to average
concentration c̄i (see Eq. (43)).
Now, in Eqs. (42a), if we neglect the dependence
on velocity and the exponential term and let (0.5 +
4.5kD /R0 ) equate to 4.5kD /R0 , we obtain the following
approximations:
4Di
katt = 20 (47a)
R0
2D0i
kdet = (47b)
kD R0
We notice that the two sets are the same if w = ¯ 0.5,
which means that we should expect concentration ci ¯pore
appearing in Eq. (16) to be equal to the average concen-
tration c̄i at r = 0.5R0 . This possibility has been verified
by plotting concentration breakthrough curves at r = 0,
r = 0.5R0 , and r = R0 as well as average concentration
¯
c̄i in Fig. 8; the agreement between c̄i and ci ¯r=0.5R is ex-
cellent. This result suggests that Eqs. (46) provide a very

Volume 13, Number 5, 2010

406
FIG. 8: Comparison between average concentration breakthrough curve and breakthroughs of concentrations at three
different positions in the tube
good approximation for upscaled kinetic adsorption coef-
ficients. The parameters w must be seen as an empirical
factor that will be different for different pore structures.
5. CONCLUSION
In this work, we have shown that even if there is equi-
librium adsorption at the pore wall (or at the grain sur-
face), one may need to employ a kinetic description at the
larger scale. This result was obtained through theoreti-
cal averaging from pore to REV scale as well as through
numerical averaging from pore to the tube scale. Both
approaches result in formulas for macro-scale kinetic ad-
sorption parameters as a function of pore-scale parame-
ters such as Peclet number and distribution coefficient.
Formulas from two approaches agree very well. The up-
scaled adsorption parameters are found to be only weak
functions of velocity; they strongly depend on geometry
of the pore and diffusion coefficient in the solution as well
as the pore-scale distribution coefficient.
The relations for upscaled transport coefficients are ap-
propriate for using in pore-network models where tube-
scale relationships are needed to model adsorbing so-
lute transport in individual pore throats. Using the pore-
network model, we could scale up from the pore scale to
the scale of REV or even the core scale.
REFERENCES
Acharya, R., Zee, S., and Leijnse, A., Transport modeling of
nonlinearly adsorbing solutes in physically heterogeneous
pore networks, Water Resour. Res., vol. 41, W02020, 2005.
Raoof & Hassanizadeh
Aris, R., On the dispersion of a solute in a fluid flowing
through a tube, Proc. R. Soc. London, Ser. A, vol. 235,
pp. 67–77, 2005.
Auriault, J. L. and Lewandowska, J., Diffusion/adsorption/ ad-
vection macrotransport in soils, Eur. J. Mech. A, vol. 15,
pp. 681–704, 1996.
Bachmat, Y. and Bear, J., Macroscopic modelling of transport
phenomena in porous media. 1: The continuum approach,
Transp. Porous Media, vol. 1, pp. 213–240, 1986.
Bachmat, Y. and Bear, J., On the concept and size of a repre-
sentative elementary volume (REV), in Advances in Trans-
port Phenomena in Porous Media, pp. 3–19, Martinus Ni-
jhof, Dordrecht, Netherlands, 1987.
Bear, J., Dynamics of Fluids in Porous Media, Courier Dover,
New York, 1988.
Binning, P. J. and Celia, M. A., Pseudokinetics arising from
the upscaling of geochemical equilibrium, Water Resour.
Res., vol. 44(7), pp. 145–155, 2008.
Bolt, G., Movement of solutes in soil: Principles of ad-
sorption/exchange chromatography, Soil Chem. B, vol. 5,
pp. 285–348, 1979.
Bryant, S. L. and Thompson, K. E., Theory, modeling and ex-
periment in reactive transport in porous media, Curr. Opin.
Colloid Interface Sci., vol. 6, pp. 217–222, 2001.
Cao, J. and Kitanidis, P., Pore-scale dilution of conserva-
tive solutes: An example, Water Resour. Res., vol. 34,
pp. 1941–1949, 1998.
Chou, L., Garrels, R., and Wollast, R., Comparative study of
the kinetics and mechanisms of dissolution of carbonate
minerals, Chem. Geol., vol. 78, pp. 269–282, 1989.
Daugherty, H. and Franzini, J., Steady flow of incompressible
fluids in pipes, in Fluid Mechanics with Engineering Ap-
plications, 564, McGraw-Hill, New York, 1965.
Edward, D. A. and Davis, A., Diffusion and convective disper-
sion through arrays of spheres with surface adsorption, dif-
fusion, and unequal solute partitioning, Chem. Eng. Sci.,
vol. 50, pp. 1441–1454, 1995.
Journal of Porous Media
Upscaling Transport of Adsorbing Solutes
Edwards, D. A., Shapiro, M., and Brenner, H., Dispersion and
reaction in two-dimensional model porous media, Phys.
Fluids A, vol. 5, pp. 837–848, 1993.
Flex-PDE, A flexible solution system for partial differential
equations, 1996–2005, http://www.flexpde.com, PDE So-
lutions Inc., Sunol, CA, 2005.
Gramling, C. M., Harvey, C. F., and Meigs, L. C., Reactive
transport in porous media: A comparison of model predic-
tion with laboratory visualization, Environ. Sci. Technol.,
vol. 36, pp. 2508–2514, 2002.
Gray, W. G. and Hassanizadeh, S. M., Macroscale continuum
mechanics for multiphase porous-media flow including
phases, interfaces, common lines and common points, Adv.
Water Resour., vol. 21, pp. 261–281, 1998.
Guo, G. and Thompson, K. E., Experimental analysis of lo-
cal mass transfer in packed beds, Chem. Eng. Sci., vol. 56,
pp. 121–132, 2001.
Hassanizadeh, S., Derivation of basic equations of mass trans-
port in porous media: 1. Macroscopic balance laws, Adv.
Water Resour., vol. 9, pp. 196–206, 1986.
Hassanizadeh, S. and Gray, W., General conservation equa-
tions for multi-phase systems: 1. Averaging procedure,
Adv. Water Resour., vol. 2, pp. 131–144, 1979.
Kapoor, V., Jafvert, C. T., and Lyn, D. A., Experimental study
of a bimolecular reaction in Poiseuille flow, Water Resour.
Res., vol. 34, pp. 1997–2004, 1998.
Kechagia, P. E., Tsimpanogiannis, I. N., Yortsos, Y. C., and
Lichtner, P. C., On the upscaling of reaction-transport pro-
cesses in porous media with fast or finite kinetics, Chem.
Eng. Sci., vol. 57, pp. 2565–2577, 2002.
Knutson, C. E., Werth, C. J., and Valocchi, A. J., Pore-scale
modeling of dissolution from variably distributed non-
aqueous phase liquid blobs, Water Resour. Res., vol. 37,
pp. 2951–2963, 2001.
Li, L., Peters, C. A., and Celia, M. A., Upscaling geochemi-
cal reaction rates using pore-scale network modeling, Adv.
Water Res., vol. 29, pp. 1351–1370, 2006.
Li, L., Steefel, C. I., and Yang, L., Scale dependence of min-
eral dissolution rates within single pores and fractures,
Geochimica et Cosmochimimica Acta, Elsevier, vol. 72(2),
pp. 360–377, 2008.
Maher, K., DePaolo, D. J., and Lin, J. C. F., Rates of silicate
dissolution in deep-sea sediment: In situ measurement us-
ing 234U/238U of pore fluids, Geochim. Cosmochim. Acta,
vol. 68, pp. 4629–4648, 2004.
Mauri, R., Dispersion, convection, and reaction in porous me-
dia, Phys. Fluids A, vol. 3, pp. 743–756, 1991.
Meile, C. and Tuncay, K., Scale dependence of reaction rates
in porous media, Adv. Water Resour., vol. 29, pp. 62–71,
2006.
Mo, Z. and Friedly, J., Local reaction and diffusion in porous
media transport models, Water Resour. Res., vol. 36,
pp. 431–438, 2000.
Mojaradi, R. and Sahimi, M., Diffusion-controlled reactions in
disordered porous media: 2. Nonuniform distribution of re-
actants, Chem. Eng. Sci., vol. 43, pp. 2995–3004, 1988.
Murdoch, A. and Hassanizadeh, S., Macroscale balance rela-
tions for bulk, interfacial and common line systems in
multiphase flows through porous media on the basis of
molecular considerations, Int. J. Multiphase Flow, vol. 28,
Volume 13, Number 5, 2010
407
pp. 1091–1123, 2002.
Pagitsas, M., Nadim, A., and Brenner, H., Macrotransport pro-
cesses in the presence of bulk and surface chemical reac-
tions, J. Chem. Phys., vol. 85, pp. 4038–4044, 1986.
Quintard, M. and Whitaker, S., Aerosol filtration: An analy-
sis using the method of volume averaging, J. Aerosol Sci.,
vol. 26, pp. 1227–1255, 1995.
Raje, D. S. and Kapoor, V., Experimental study of bimolecu-
lar reaction kinetics in porous media, Environ. Sci.;echnol.,
vol. 34, pp. 1234–1239, 2000.
Rashidi, M., Peurrung, L., Tompson, A., and Kulp, T., Ex-
perimental analysis of pore-scale flow and transport in
porous media, Adv. Water Resour., vol. 19, pp. 163–180,
1996.
Robinson, B. A. and Viswanathan, H. S., Application of the
theory of micromixing to groundwater reactive transport
models, Water Resour. Res., vol. 39(11), p. 1313, 2003.
Ryan, D., Carbonell, R., and Whitaker, S., Effective diffusiv-
ities for catalyst pellets under reactive conditions, Chem.
Eng. Sci., vol. 35, pp. 10–16, 1980.
Sahimi, M., Diffusion-controlled reactions in disordered
porous media–I. Uniform distribution of reactants, Chem.
Eng. Sci., Elsevier, vol. 43(11), pp. 2981–2993, 1988.
Sahimi, M. and Jue, V. L., Diffusion of large molecules in
porous media, Phys. Rev. Lett., vol. 62, pp. 629–632,
1989.
Serrano, S. E., Propagation of nonlinear reactive contaminants
in porous media, Water Resour. Res., vol. 39(8), p. 1228,
2003.
Shapiro, M. and Adler, P. M., Coupled transport of multi-
component solutes in porous media, J. Eng. Math., vol. 31,
pp. 357–378, 1997.
Shapiro, M. and Brenner, H., Taylor dispersion of chemically
reactive species: Irreversible first-order reactions in bulk
and on boundaries, Chem. Eng. Sci., vol. 41, pp. 1417–
1433, 1986.
Shapiro, M. and Brenner, H., Chemically reactive generalized
Taylor dispersion phenomena, AIChE J., vol. 33, pp. 1155–
1167, 1987.
Shapiro, M. and Brenner, H., Dispersion of a chemically re-
active solute in a spatially periodic model of a porous
medium, Chem. Eng. Sci., vol. 43, pp. 551–571, 1988.
Szecsody, J. E., Zachara, J. M., Chilakapati, A., Jardine, P. M.,
and Ferrency, A. S., Importance of flow and particle-scale
heterogeneity on CoII/IIIEDTA reactive transport, J. Hy-
drol., vol. 209, pp. 112–136, 1998.
Taylor, G., Dispersion of soluble matter in solvent flowing
slowly through a tube, Proc. R. Soc. London, Ser. A,
vol. 219, pp. 186–203, 1953.
Toride, N., Leij, F., and van Genuchten, M. T., The CXTFIT
code for estimating transport parameters from laboratory
or field tracer experiments, version 2.0, Res. Rep 137, 121,
Riverside, California, 1995.
van der Zee, S. E., Analytical traveling wave solutions for
transport with nonlinear and nonequilibrium adsorption,
Water Resour. Res., vol. 26, pp. 2563–2578, 1990.
van Duijn, C. J., Mikelic, A., Pop, I. S., and Rosier, C., Effec-
tive dispersion equations for reactive flows with dominant
Peclet and Damkohler numbers, Adv. Chem. Eng., vol. 34,
pp. 1–45, 2008.
408
View publication stats
Weber, Jr., W., McGinley, P., and Katz, L., Sorption phenom-
ena in subsurface systems: Concepts, models and effects
on contaminant fate and transport, Water Res., vol. 25,
pp. 499–528, 1991.
Whitaker, S., Advances in theory of fluid motion in porous me-
dia, Ind. Eng. Chem., vol. 61, pp. 14–28, 1969.
Whitaker, S., Flow in porous media: I. A theoretical derivation
of Darcy’s law, Transp. Porous Media, vol. 1, pp. 3–25,
1986.
White, A. F. and Brantley, S. L., The effect of time on the
weathering of silicate minerals: Why do weathering rates
differ in the laboratory and field? Chem. Geol., vol. 202,
Raoof & Hassanizadeh
pp. 479–506, 2003.
Wood, B. D., Quintard, M., and Whitaker, S., Jump conditions
at non-uniform boundaries: The catalytic surface, Chem.
Eng. Sci., vol. 55, pp. 5231–5245, 2000.
Wood, B. D., Quintard, M., and Whitaker, S., Estimation of
adsorption rate coefficients based on the Smoluchowski
equation, Chem. Eng. Sci., Oxford; New York: Pergamon
Press, vol. 59(10), pp. 1905–1921, 2004.
Wood, B. D., Radakovich, K., and Golfier, F., Effective reac-
tion at a fluid-solid interface: Applications to biotrans-
formation in porous media, Adv. Water Resour., vol. 30,
pp. 1630–1647, 2007.
Journal of Porous Media