Sie sind auf Seite 1von 30

Journal of Porous Media 10(3), 219248 (2007)

Dispersion in Porous Media with and without


Reaction: A Review
N. Rudraiah1,2 and Chiu-On Ng3
1

National Research Institute for Applied Mathematics, 492/G, 7th Cross, 7th Block (west),
Jayanagar, Bangalore 560082, India
2

UGCCentre for Advanced Studies in Fluid Mechanics, Department of Mathematics,


Bangalore University, Bangalore 560001, India

Department of Mechanical Engineering, University of Hong Kong, Pokfulam Road,


Hong Kong, China
E-mail: rudraiahn@hotmail.com
ABSTRACT
The dispersion in fluid-saturated deformable or nondeformable porous media with or without chemical reaction is studied analytically, considering
a series of particular cases selected through different practical problems
using different dispersion models. The required basic equations are obtained using mixture and homogenization theories. Different analytical and
numerical models valid for large time (asymptotic) and for all time (transient) are explained. The formulations of the problems considered here are
so general that they are applicable to deformable and nondeformable particles amenable to both analytical and numerical procedures. The validity
of the results obtained from an analytical method is verified by comparison
with available experimental and cell model results, and good agreement is
found.

219
Received September 7, 2005; Accepted February 13, 2006
c 2007 Begell House, Inc.
Copyright

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

220

Rudraiah and Ng

NOMENCLATURE

Cb
d
h
i, j, k
k
K
Ka
l
L
M
ma
mc
N
N2
P
Pe
Q
~qf

nonlinear drag coefficient


radius of the particles
distance between the plates
unit vectors in x, y, and z directions
permeability of porous media
= 6 a , Stokes drag
linear drag in Darcy term
p
= 1/2 /f , material characteristic length
characteristic length along the flow direction
volumetric rate
mass of the particles
= N h/l, micropolar parameter
number density of the particle
= /2 + , capillary number
pressure

= U h D, Peclet number
= kP, Darcy velocity
fluid velocity

~qs

solid velocity

= t20 s h2

= kt0
= t0 /a

= a t0 f h2

R1
R2
R3
R4

1. INTRODUCTION

Flow through and past porous media has attracted


considerable interest in recent years because of its
importance in science, engineering, and technology
(see Rudraiah et al., 1979; Kaviany, 1999; Nield and
Bejan, 1999; Vafai, 2000). In this article we review
works on dispersion in porous media (DIPM) because
of its importance in many practical applications, as
discussed briefly in Section 1.1.

Re
Rh
t0
u
u

U
u1 , u2
v
v1 , v 2

= uh/f , Reynolds number

= Re 1 N 2
characteristic time
displacement vector
average velocity
characteristic velocity
velocities of fluid
relative velocity as in Eq. (3.5)
velocities of suspended particles

Greek Symbols
,
micropolar viscosities

= y/h

one of the Lammel constants


0
effective viscosity as in Eq. (2.5)
a
apparent or effective viscosity

interaction between the constraints


f
viscosity
. of fluid

=h
k, Porous parameter
f
s

stress for fluid


stress for solid
stress vector
microrotation velocity

1.1 Importance of the Study of Dispersion in


Porous Media (DIPM)

Notable practical problems that require the study of


DIPM are the extraction of energy from geothermal
regions, analyzing thermal energy storage systems,
thermal insulation, solar collectors with porous absorbers, the evaluation of the capability of heat removal from particulate nuclear fuel debris that may
result from a hypothetical accident in a nuclear re-

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

actor, flow of liquids through ion exchange beds,


drug permeation through human skin, chemical reactors for economical separation or purification of
mixtures, removal of volatile organic compounds from
the unsaturated zone of an aquifer, biomechanical applications such as cartilage in synovial joints (see
Sueiu et al., 2003; Ng et al., 2005), formation of
plaques in coronary artery diseases, ultrafiltration and
dialysis in kidneys, inhaling of aerosols through airways, estimation of air and water pollution, and so
on. Much interest has also been evinced in (1) wetting and dewetting of solids in liquids because of
their importance in polymeric adhesive joints between
solids, where wetting of solids by melt before it solidifies is essential; (2) the efficient recovery of crude
oil from the pores of the petroleum reservoir rocks
by displacement with immiscible water (Taber, 1980;
Posner and Gill, 1973), where a relation between the
displacement rate and pressure is desired; (3) pollutant transport resulting from municipal, industrial, and
agricultural wastes through the surface of the Earth,
polluting ground water where control of pollutants
is desired; and (4) the tiny dust particles floating in
the air, known as aerosols and which are gradually
choking people to death, where an understanding of
the spreading of aerosols is required. The main interest in all these cases is the study of dispersion
(i.e., spreading leading to flow-enhanced diffusion) in
porous media.
The literature on hydrodynamic DIPM is very
sparse in spite of its versatile applications in many
branches of science, engineering, and technology.
This DIPM is the macroscopic outcome of the actual
movements of the individual solute particles through
the pores and the various physical and chemical phenomena that take place within the pores. In general,
such movements and phenomena result from (1) the
external forces acting on the liquid; (2) the microscopic intricate geometry of the pore system; (3) the
molecular diffusion caused by solute concentration
gradient; (4) variations in the liquid properties, such
as density and viscosity, which affect the flow pattern;

221

(5) the change in the solute concentration due to the


chemical and physical forces within the liquid phase;
and (6) the interaction between the liquid and solid
phases. In this dispersion, there are two transport phenomena involved, namely convection and molecular
diffusion. In general, this convective mass flow may
be laminar or turbulent. In this review we study DIPM
only in laminar flow through porous media with and
without chemical reaction, taking into account molecular diffusion in addition to dispersion but neglecting
the adsorption effect.
1.2 Objectives of This Review
Our main objectives of writing this review are the
following:
1. To present a mechanism of dispersion of a dynamically neutral material quantity with and without
chemical reaction in a fluid-saturated porous medium
with the motive that the results obtained may be effectively used in some typical cases to understand the
mechanism of mixing and separation processes.
2. To provide an improved model for DIPM that
agrees with the experimental data of Harleman et al.
(1963) and the cell model results of Simpson (1969).
3. To summarize specific dispersion models that are
used in the literature and discuss the validity of these
models.
4. To derive the required equations using both
mixture and homogenization theories.
1.3 Literature Survey
Reviewing the existing literature to obtain a critical
outlook on recent advances is certainly one way of
assessing the area. But this requires considerably more
space and would be best suited for a monograph,
where space is not a constraint. An alternative to this,
where the space is a constraint, is to focus ones
attention only on ones own work. Even our work
covers many aspects of dispersion through and past
porous media, and not all these aspects can be covered in the space limit at our disposal. We therefore

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

222

emphasize only some selected important problems of


dispersion through and past deformable and nondeformable particles in porous media using mixture and
homogenization theories, which form the backbone
for achieving the objectives discussed in Section 1.2.
The work on dispersion in the absence of porous
media is fairly well known (see Rudraiah et al., 1979),
of which Taylor (1953), Aris (1956), and others were
mainly valid for large time after injecting a slug into
the fluid. Later, Gill and Sankarasubramanian (1970)
generalized it to cover transient phenomena, called a
generalized dispersion model.
In a porous medium the irregularity of the streamlines arises from the complicated geometrical structure. Therefore the hydrodynamic DIPM is highly
complicated and microscopic due to the intricate geometry of the pore system. In that case the best one
can do is to formulate the problem by considering
a molecular picture involving forces of attraction and
repulsion using mathematical simplifications like (1)
statistical averages and (2) homogenization theory.

1.3.1 Statistical Averages


The usual statistical averages used in porous media
are spatial, that is, volume averages and ensemble
averages. One of the important mathematical aspects
in deriving the average properties of the fluid through
porous media is the Slattery-Whitaker theorem. This
theorem, discovered simultaneously and independently
by Slattery (1967) and Whitaker (1967), enables one
to express the volume averages of space derivatives
in terms of the space derivatives of volume averages.
For details, one can refer to the books listed in the
References and to the work of Tam (1969), Maloy et
al. (1988), Sahni and Jue (1989), Hulin et al. (1990),
and Rudraiah (2001). These averages are needed because in general, it is impossible to know what is
happening in each of the pores of a porous medium.
Here the best one can expect is knowledge of the
statistical averages of the physical quantities over a
certain representative volume of the system.

Rudraiah and Ng

1.3.2 Homogenization Theory


The average procedures mentioned previously inevitably involve mathematical simplifications with the
result that the porous media are modeled only incompletely (see Mei et al., 1996). Some theories obtained
from these averages bypass the microscale details because they make use of the potential nature of Darcys
law to govern flow through densely packed porous
media based on a scale much greater than the pore
size. Here a mass transport equation is assumed either to govern the mass transport on a similar large
scale or formally justified using the Reynolds averaging procedure together with the closure hypothesis,
usually based on a gradient diffusion model (i.e., kmodel). These deficiencies are overcome using the
homogenization theory developed on the basis of periodic porous media. Mei et al. (1996) used this theory
to study the dispersion phenomena in porous media.
For details, one may refer to the work of Mei (1992)
and Mei et al. (1996). In this review we briefly explain how these averages and homogenization theory
are used to study the DIPM by considering some
practical problems arising in environmental studies,
biomechanics, and chemical engineering. Before dealing with particular examples, we first explain the work
done on DIPM.
Incorporating the above concepts, it appears that
Saffman (1959) was the first to study the dispersion
of a dynamically neutral quantity in a viscous fluid
flowing through a porous medium under the assumption that the medium may be regarded as equivalent to
a statistically isotropic network of straight capillaries.
His discussion was restricted to the case in which
the dispersion is primarily due to macroscopic mixing.
This arises from the randomness of the streamlines
through the network and has an effective diffusivity of the order UL, where U is the mean velocity
and L is the length scale of the capillaries. These
results are valid when D << U L. Later, Saffman
(1960)relaxed his earlier assumption of D << U L
using the random network of capillaries to represent a
porous medium. Using a Lagrangian correlation func-

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

tion, Saffman (1960) computed the longitudinal and


lateral effective diffusivities that are valid for all values of U L/D less than some large value. Rudraiah
(1976) studied the dispersion of soluble matter in a
laminar flow through porous media between two parallel plates using a statistical approach and obtained a
modified Taylors dispersion coefficient using the capillary model of Taylor (1953). Rudraiah (1976) also
studied the longitudinal dispersion of solute in a channel bounded by porous layers using a Taylor model
with Beavers and Josephs (1967) slip condition. He
compared his results with those of Fung and Tang
(1975) pertaining to the use of a no-slip condition
and found that the effect of slip is significant only in
the case where the membrane is permeable to solvent
but not to the tracer. However, in the case where
the membrane is permeable to both the solvent and
the tracer, he found that his results coincided with
those of Fung and Tang (1975), implying that the slip
condition is negligible in that case. Rudraiah et al.
(1985) extended the Taylor dispersion model to study
the longitudinal diffusivity D for laminar flow in a
porous medium of rectangular cross section of breadth
a and height b using a Brinkman model. They showed
that D is about 2D0 for large values of the aspect
ratio b/a, where D0 is the longitudinal diffusivity in
the absence of all variations across the channel. In

particular, they showed that D


for large
is.uniform

values of porous parameter = b
k , where k is
the permeability of porous media. When 0, they
found good agreement with the viscous flow results of
Chatwin and Sullivan (1982).
The previous analysis of dispersion through porous
media is concerned with mathematical models based
on a single capillary tube, a bundle of capillaries, an
array of cells, and so on. However, in recent years,
the dispersion through porous media was also studied using a statistical model, where the microscopic
motion was converted into a macroscopic description using average properties, as explained before.
Bear and Bachmat (1967) obtained, using a statistical
approach, insight into the dispersion phenomena in

223

porous media and gave a relationship between the dispersion and medium characteristics. They showed that
the dimensionless diffusion coefficient increases with
an increase in Reynolds number Re and approaches
infinity as Re . Later, Simpson (1969) investigated the DIPM using a cell model and showed that
the dimensionless diffusion coefficient is independent
of Re and approaches asymptotically a constant value
as Re increases. The conclusion of Bear and Bachmat (1967), namely that D increases with Re and
approaches inutility as Re , is in contradiction
with the experimental observation of Harleman et al.
(1963) and with the probabilistic approach of Simpson (1969). Chandrasekara et al. (1980) obtained a
dispersion coefficient using a deterministic approach
following Taylor (1953) that is in agreement with the
experimental observations of Harleman et al. (1963)
and the cell model of Simpson (1969). The work of
Manz et al. (1988), Patterson et al. (1996), and Bejan et al. (2004) is also fundamental to understanding
DIPM.
The works on DIPM discussed previously are based
on the assumption of using Darcy velocity throughout
the channel. However, in some practical problems, the
structure of the porous media is such that the particles are sparsely packed at the walls and densely
packed along the axis, with the result that the flow
has to be governed by the Brinkman (1947) model
rather than by the potential nature of the Darcy model.
Rudraiah and his research students and collaborators
(see Shivakumara and Venkatachalappa, 2004) studied hydrodynamic DIPM using the Brinkman model
with and without the deformation of solid particles.
Such work is fairly sparse in the literature, and their
work is briefly discussed in this review. Before discussing these specific problems, we review briefly in
Section 1.4 different types of models used to study
DIPM.
1.4 Dispersion Models
In the literature the following analytical and numerical
approaches are used to study the dispersion of solute
in a medium without or with porous media.

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

224

1.4.1 Analytical Approaches


1. The Taylor and Aris dispersion models, as explained in Section 1.2, are valid for a large time after
injecting a solute into a medium.
2. The generalized dispersion model, as explained
in Section 1.2, was developed by Gill and Sankarasubramanian (1970) and Barton (1983) to study dispersion in a clear fluid, and Shivakumar et al. (1987)
developed this model to study DIPM valid for all
time.
3. With the Lighthill approach, Lighthill (1969) obtained an analytical solution for the unsteady convective diffusion equation using an asymptotic expansion
after neglecting axial molecular diffusion. Lighthills
approach, however, takes care of only one initial condition on diffusion of the concentration distribution,
and his solution is valid for a time less than about
0.1a2 /D, where a is the radius of the tube and D is
the molecular diffusivity. Later, Chatwin (1971, 1973)
extended Lighthills results following the approach of
Saffman (1971).
4. In the random walk approach the tracer particles
are represented by a large number of marked particles,
whose paths are traced as they move through the fluid.
This approach models the physics of dispersion more
realistically than the standard diffusion equation.
5. The rate of longitudinal dispersion of tracer
particles in a flow depends on the distribution of concentration variation across the flow. These variations
are generated by the combination of a longitudinal concentration gradient and cross-stream velocity
shear. In that case the response is not instantaneous,
where the dispersion coefficient, D(t), is not only
time-dependent, but also, there is a memory of the
concentration distribution at earlier times. These features are accurately reproduced by the delay-diffusion
equation discussed by Smith (1981).
1.4.2 Numerical Approaches
6. Jayaraj and Subramanian (1978), Harden and Shen
(1979), and references therein studied numerically the

Rudraiah and Ng

shear dispersion of solute particles using the conventional finite difference approximation methods.
7. Guymon (1970) and Smith et al. (1973) used
the finite element method to solve the diffusion equation. The work of Smith et al. (1973) is concerned
with using Rayleigh-Ritz and Galerkin procedures. A
proof of solving the problem using the finite element
approach is equivalent to solving the original partial
differential equation depending on a theorem of Euler
and is given by Zienkiewicz (1971).
8. The Barton-Stokes approach (Barton and Stokes,
1986) is based on the combination of the spectral
method of Gottlieb and Orszag (1977). A Fourier
transform is taken in the flow direction, and the
cross-sectional variation is described using finite differences. This process reduces the partial differential
equation for concentration to a system of ordinary
differential equations, which are solved numerically
using finite differences.
9. In Smiths approach, Smith (1981) showed that
for a sudden uniform discharge in a bounded shear
flow, the asymptotic concentration distribution at moderately large times can be well approximated by a
tilted Gaussian. The tilt parameter makes the concentration distribution suitably skewed. This gives an
alternate approach to express the longitudinal concentration distribution using Gaussian distribution from
the classical approach of Taylor (1953) and Gill and
Sankarasubramanian (1970).
10. In the integral transform approach the diffusion
equation for the concentration distribution is solved
using Laplace transform on time and Fourier transform on space in the flow direction. This leads to
an eigenvalue problem, which can be solved either
numerically or analytically.
11. In addition to the above models, there are
contributions by Whitaker (1967) and Brenner (1980)
to DIPM, but the details are omitted here for want of
space.
To achieve the objectives of this review, the plan
of our work is as follows. In Section 2, dealing
with mathematical formulation, we obtain the basic

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

equations based on the mixture theory involving the


motion of solids and fluids in porous media. In Sections 3 and 4 we study DIPM using some examples
selected from practical problems. We use both largetime and all-time models to study dispersion with and
without chemical reaction. In Section 5 we discuss the
dispersion in deformable porous media using mixture
theory. In Section 6 we use homogenization theory to
study DIPM with chemical reaction on the solid-fluid
interface.

225

+ ( q ) = 0
t
the conservation of momentum,

(2.2)

q
+ q q = p + q
t

+ [Xp (0 0a ) + 0a ] 2 q

Cb s
f
q qf
k k q q
k

(2.3)

where
2. MATHEMATICAL FORMULATION
We model the flow through porous media as a continuous binary mixture of particles and fluid phases,
denoted by = s for particles and = f for fluid,
where each point in the mixture (i.e., porous media) is occupied simultaneously by both particles and
fluid. The porous medium is assumed to be homogeneous and isotropic with body force F~ . Both particle
and fluid phases are assumed to be intrinsically incompressible. Bulk compression of the mixture can
arise only by a decrease in the fluid fraction. Kenyon
(1976), Bowen (1980), and Barry et al. (1991) used
mixture theory to derive the basic equations for deformable porous media. In this article we follow their
mixture theory to obtain the required basic equations
for the mixture of particle and fluid with deformable
media. In this case we assume that deformation is
so small that the intrinsic properties of the mixture of solid and fluid remain approximately uniform
throughout the deformation. The density and volume fraction of each phase in the mixture are given
by

=
Iv

(2.1a)

= /d

(2.1b)

where
is density, and
I is the intrinsic density, d

v is the volume of each constituent.


Then, following Barry et al. (1991), the required
basic equations for each phase are the conservation of
mass,

(
Xp =

1 for solid
0 for fluid

( + 2)
j+k
0 = i +

(2.4)

0a =

a
,
f

(2.5)

and the conservation of species

1 s
C
0
+ q C +
+ 0 C + s
t
b t
b
q
2
+ = Dm C
b

(2.6)

where C is the concentration of particles in the


fluid layer, s is the mass of particles absorbed per
unit length of the mixture surface, 0 is the decay
constant equal to log(2/t1/2 ), t1/2 is the half life, q
is the diffusive flux perpendicular to void axis, and
Dm is the molecular diffusivity; other quantities are
defined in the Nomenclature. We make use of these
basic equations to discuss the particular problems in
Section 3.
3. TYPICAL EXAMPLES
In this section we discuss some important problems
dealing with DIPM using Taylors (1953), Ariss
(1956), and Gill and Sankarasubramanians (1970)
models. The theory developed is so general that it
can be directly used to understand the practical problems given in Section 1.1. In Section 1.3 we pointed
out that the conclusions of Bear and Bachmat (1967)

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

226

Rudraiah and Ng

contradict the experimental observations of Harleman


et al. (1963) and the probabilistic results of Simpson
(1969). The aim of this section is to provide a proper
theory, following Chandrasekara et al. (1980), that fits
both the experimental and probabilistic results.

Using Eq. (3.5) and the nondimensional quantities

3.1 Taylors Diffusion Coefficient for


Nondeformable Porous Media

and following Taylor (1953), assuming partial equilibrium, Eq. (3.2) takes the form

We consider a steady laminar flow through nondeformable porous media to understand DIPM. The
physical model consists of laminar flow of a viscous
fluid flowing through a permeable material between
two infinite parallel plates situated at y = h.
Following the analysis of Chandrasekara et al.
(1980), the momentum and concentration equations
are respectively given by

C
t
L
1 =
t =

C0
t
u
x u
t
y
=
=
L
h

C =

2C
h2 v C
=
2
DL

(3.6)

(3.7)

The solution of Eq. (3.7), using Eq. (3.5) and the


conditions

C/ = 0 at = 1

(3.8)

is
2

d u 1
u=P
dy 2
k

(3.1)

C
C
2C
+u
=D 2
t
x
y

(3.2)

where P = (1/) p/x. The solution of Eq. (3.1)


satisfying the conditions u = 0 at y = h is

cosh
u=Q 1
Q = kP
cosh
.
=h
k = y/h

cosh C
h2 Q tanh 2

+C0 (3.9)
C=
DL

2 2 cosh
where C0 is a constant to be determined using the
entry condition.
The volumetric rate at which the solute is transported across a section of the channel of unit breadth
is

(3.3)

From Eq. (3.3) the average velocity is

Z1
u
=

1
2

u d = Q 1
1

tanh

(3.4)

Since we are considering advection across a plane


moving with the mean speed of flow, the fluid velocity
relative to this plane is given by

v=u u
= Q

tanh cosh

cosh

Zh
2h3 20
5
2
50 C
M = Cdy =
+ 02 + 2
DL
6 2
2
h

2 h3 u
2 5 50
0 C
=
2 2
(3.10)
L D0 6 2
2

where 20 = Q tanh / and 0 = coth 1 and 0


is replaced by u
using Eq. (3.4).
Following Taylor (1953), we assume that the variations of C with are small compared with those in
the longitudinal direction, and if Cm is the mean concentration over a section, C/ is indistinguishable
from Cm / so that Eq. (3.10) may be written as

(3.5)

M =

2D Cm

(3.11)

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

227

D = aR + bS

where

h u

20
h
5 50

D =

D20 L
6 22
2 2
2 2
h u

=
F0 ()
D
2

(3.12)

and F0 () represents the terms in the square brackets

of Eq. (3.12), with 1 20 and (h/L).


The fact that no material is lost in the process is
expressed by the continuity equation for Cm , namely

(3.13)

Equation (3.13), using Eq. (3.11), becomes


Cm
2 Cm
= D
t
2

From Eq. (3.16) it is evident that as R increases,


D also increases linearly with R since a, b, and
S are constants. But the experimental observations
0
reveal that the magnitude of D increases with an
increase in R initially and then attains a constant
0
value for large R. In other words, D becomes
independent of R for large values of R. This inconsistency between the theory (Bear and Bachmat,
1967) and experiment (Harleman et al., 1963) suggests that Eq. (3.16) does not represent the correct
diffusion coefficient for porous media and calls for an
improvement in the theoretical model to obtain a diffusion coefficient whose behavior with R agrees with
the experimental results. Chandrasekara et al. (1980)
removed the above-mentioned inconsistency and obtained the relations
0

Cm
M
= 2

(3.16)

(3.14)

which is the equation governing the longitudinal dispersion. Equation (3.12), using Q = kP , can also be
written in the form


P 2 h2 1 5 50 20 tanh h
D=

(3.15)
D 4 6 22 22

L
For details, one may refer to the work of Chandrasekara et al. (1980).
The form of diffusion coefficient D given by
Eq. (3.12) is not suitable to compare directly with
the available experimental data, unless D appearing in
Eq. (3.12) is modified suitably to take into account the
effects of both molecular diffusion and the intrinsic
properties of the porous media. An attempt in this
direction, using an entirely different approach, was
made by Bear and Bachmat (1967), but their results
are still not in agreement with the cell model results of Simpson (1969) and the experimental data of
Harleman et al. (1963). Here we try to improve their
results. Before presenting the improved model, it is
advantageous to discuss briefly the model of Bear and
Bachmat (1967). In their model the expression for the
diffusion coefficient, after simplification (see details in
Chandrasekara et al. (1980)) is

D =

F () =

h
h2
L R d2

a d` R +

u
d
bS

F ()

5 0 (5 + 0 )

6
22

R=

u
d
v

S=

D
v

(3.17)

If the characteristic length of the channel L is equal to

h3 d2 (which is consistent with Taylors model) and


and the characteristic pore length is equal to the
`,
mean grain diameter d, then Eq. (3.17) has the form
D
R
=
F ()
(3.18)
u
d
aR + bS
From Eq. (3.18) it is clear that the dimensionless
diffusion coefficient approaches a constant value as
R since a, b, and S are all constants. It is noted
that F () in Eq. (3.18) also attains a constant value
of 0.333 as .
Since the aim of this work is also to compare our
results with the cell model results of Simpson (1969),
in what follows, a brief outline of the derivation of an
expression for D /
ud from the probabilistic approach
is given following Chandrasekara et al. (1980).
Following the analysis of Chandrasekara et al.
(1980), we get the nondimensional axial diffusion
coefficient as

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

228

D
1 R
i
= h
u
d
1
4 R + 48
2 S

(3.19)

In Eq. (3.19), 1 and 2 are dimensionless constants


(as a and b in Eq. (3.18)) that are not known uniquely.
Hence to study the behavior of D /
ud with respect to
R, proper values of 1 and 2 are chosen in order to
fit the experimental data.
Equations (3.18) and (3.19) represent the expressions for the nondimensional diffusion coefficients
derived from two different approaches from the Taylors and probabilistic models. It is interesting to
note that both the expressions are of the same form.
In Eq. (3.18) the constant a is associated with the
transverse dispersivity and has a maximum theoretical
value of 0.7 (Bear and Bachmat, 1967). But experimental observations indicate a much lower value for
a around 0.1. Hence we choose a = 0.35 and b =
30, which are associated with the effect of molecular diffusion. Using the above values for a and b
in Eq. (3.18), the expression for the axial diffusion
coefficient becomes
2D
2R
=
F ()
(3.20)
u
d
0.35R + 30S
Equation (3.19) contains two unknown constants 1
and 2 : 1 is associated with the hydrodynamic beta
length h , which is about the same as twice the short
distance memory of the tracer particles. The short
distance memory is expected to be of the order of
one to two grain diameters, and hence 1 will have
a value of 3 or 4. The expected value of 2 is more
difficult to estimate, and therefore a suitable value for
2 is chosen to give a better fit to the experimental
observations.
Following Simpson (1969), we choose 1 and 2 =
3.6. Using these values for 1 and 2 in Eq. (3.19),
we have
1.8R
2D
=
(3.21)
u
d
R + 48S
The experimental observations of Harleman et al.
(1963) together with the results of the present investigation and the results of the cell model are

Rudraiah and Ng

presented in Figs. 1 and 2 for a medium consisting


of crushed quartz sand particles of mean diameter
0.092 cm and 0.14 cm with S = 1.05 103 . It
is seen that the experimental results agree fairly well
with the theoretical results. However, it is of interest to note that the results of the Simpson (1969)
model are independent of grain diameter explicitly
and do not indicate whether the diffusion coefficient
increases or decreases with grain size. But the results
of the present work depend on the grain diameter
explicitly through . An increase in the grain size
increases the permeability k and hence decreases ,
which in turn reduces the value of the diffusion coefficient (Fig. 1). This behavior is in conformity with
the experimental results of Harleman et al. (1963).
Further, it is evident from Fig. 1 that the diffusion
coefficient given by Eq. (3.23) attains a constant value
as R increases, which is also in agreement with the
experimental observations. Finally, we conclude that
the model presented in this review for longitudinal
dispersion flow through porous media based on Taylors analysis agrees with the experimental data and
the cell model results. Furthermore, the present model
gives information on the dependence of the diffusion
coefficient on particle size and in this respect can be
considered a better model than the one derived from
the probabilistic approach (Simpson, 1969).

3.2 Aris Modification


In this section we discuss, following Rudraiah (1976),
modification to Taylors model by using moment approach and show that the total diffusion coefficient
is the sum of the molecular diffusion coefficient and
Taylor diffusion coefficient.
Following Aris (1956) and Rudraiah (1976), we
consider the point moment of the distribution of the
solute through porous media at time t as
Z
pi (, , ) d

Cp (, ) =

(3.22)

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

Figure 1. Nondimensional diffusion coefficient 2D /


ud versus Reynolds number Re

Figure 2. Nondimensional diffusion coefficient 2D /


ud versus Reynolds number Re

229

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

230

where and are defined in Eq. (3.6) and =

Dt h2 . Taking the average of Cp over a cross section


S with area of the channel, we have
1
mp () = Cp =

Z
Cp ds

(3.23)

Rudraiah and Ng
2

C0 = 12 A0 + An (cos n) en

(3.28)

Similarly, we can find C1 (for details, see Rudraiah


(1976)). From these (see Rudraiah, 1976), with
, we find that the center of gravity of the solute
is distributed across the channel according to the
formula

Let u = u
[1 + V ()], where u
is given by
Eq. (3.4), and

V () =

1
cosh
1
0
sinh

(3.24)

Multiplying Eq. (3.2) by and integrating with respect to from to , we obtain


Cp 2 Cp
=
+ p (p 1)Cp2 + pU0 VCp1 (3.25)

2
where U0 = u
h/D when h = L.
The boundary and initial conditions are

C1

U0 A0 2
cosh
constant

(3.29)
0
2
sinh

Putting p = 2 in Eq. (3.25) and recalling that m0 =1,


and following the analysis of Rudraiah (1976), we
obtain, as ,

2U02
5
20
dm2
50
=2 2 + 2 + 2
d
0
6 2
2

where the constant A0 appearing in Eq. (3.29) is taken


as unity.
Now, the rate of change of variance V () of the
distribution of the solute about the moving origin is
proportional to
dm2
lim
= lim
t dt
t

Cp
= 0 at = 1

(3.26a)

Cp (, 0) = Gp ()

(3.26b)

Averaging Eq. (3.4) over the cross section S of unit


breadth, after using Eq. (3.2) and the boundary conditions of Eqs. (3.26a) and (3.26b), we get

Z1
mp
U0 p
= p (p 1)mp2 +
V()Cp1 d (3.27)

2
1

This shows that for p = 0, m0 / = 0, so that


m0 is a constant, which can be taken as unity. This
merely expresses the fact that the total quantity of the
solute remains constant.
Following Rudraiah (1976), with p = 0 and p = 1
in Eq. (3.25) and solving the resulting equation (for
details, see Rudraiah (1976)), we get the solution

(3.30)

D dm2
h d

This, using U0 = u
h/D and Eq. (3.30), becomes

dV
u
2 h2 5
20
50
D+

2
dt
D20 6 22
2
Hence, using Eq. (3.12), we find that
dV
D + D
(3.31)
dt
Now, the effective diffusion coefficient D0 is of the
form
D0
= 1+
D

h
u
D0

5
2
50
02 2
6 2
2

(3.32)

The effective diffusion coefficient given by


Eq. (3.32) may be looked on as the sum of the
molecular diffusion coefficient D and the effective
Taylor diffusion coefficient D . Hence the restrictions
imposed, following Taylor, in Section 3.1 are now
removed.

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

3.3 Hydrodynamic Dispersion of Pollutant


(Regarded as Porous Media) with
Homogenous and Heterogeneous Reactions
The concept of longitudinal dispersion discussed in
Sections 3.1 and 3.2 can be extended to the system
having homogeneous and heterogeneous reactions. In
this case, in calculating the effective diffusion coefficient, we have to be careful in considering the reaction at the rigid boundaries. We note that in chemical
industries, the axial mixing is usually an undesirable
feature since the degradation of potential due to axial
mixing causes lower levels of conversion in chemical reactors and reduced rates of interface mixing in
separation processes. Transverse mixing, on the other
hand, is often desirable since good transverse mixing
will reduce undesirable transverse temperature gradients. Therefore the study of longitudinal dispersion of
a solute in a porous medium with pollutants flowing
through a rectangular reactor with chemical reaction,
discussed in this section, is motivated by its application in mixing processes. Here we approximate the
fluid with pollutant as fluid-saturated porous media.

3.3.1 Taylors Diffusion Coefficient with


Homogeneous Reaction
The physical model and the assumptions made in
Section 3.1 are also true here, except that we have
the chemical reaction. In this case we assume that
the chemical reaction is first order and that it occurs
under such conditions that the gas film resistance is
negligible. This means that the concentration profile
is expected to vary slowly and uniformly and that
the reaction term is K1 C mol cm3 s1 , which
represents the volume rate of disappearance of the
solute due to chemical reaction. Here K1 represents
the first-order reaction rate constant.
The velocity and average velocity are exactly the
same as in Eqs. (3.3) and (3.4). The equation for
concentration, instead of Eq. (3.2), is

231

C
C
2C
(3.33)
+ (u u
)
= D 2 K1 C
t
x1
y
For the homogeneous reaction, the boundary conditions are given by Eq. (3.8).
Following the analysis of Section 3.1, Eq. (3.33)
becomes

2C
h2 K1
cosh
2
2

C
=
Q

0
1
1
2
cosh
D
2
tanh
x1
h Q C
=
=
(3.34)
Q0 =
DL

L
The solution of this, satisfying the above boundary
conditions, is

2
cosh 1

2
1 (1 1 )
sinh 1

cosh
1

(1 21 ) sinh 21

C = Q0

(3.35)

When both the velocity and the concentration fields


are known, we can determine the dispersion coefficient by calculating the volumetric rate as
2
C p
2h7
M = 2 2
DL3 (1 21 )

1
coth 1
2 2+

1
1 (1 21 ) 1 (2 21 )

1 2
1

+
(3.36)
24
3 sinh 21
Following the procedure explained in Section 3.1
and using the fact that no material is lost in the
process expressed by the continuity equation for C,
given by Eq. (3.12), we obtain an effective dispersion
coefficient D1 in the form

D1 =

h6

p
x

2 2 D

F1 (1 , )

(3.37)

where

2 1
1
3
2 21 24
sinh 21

coth 1
1
2 2+

1
1 (2 21 ) 2 (1 21 )

F1 (1 , ) =

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

232

Equation (3.37), using Eq. (3.4), can be rewritten as


u
h
D1
= Pe2 F2 (1 , ) Pe =
D
D
4
F2 (1 , ) =
2 F1 (1 , )
(1 )

(3.38)

Values of F2 (1 , ) given by Eq. (3.38) are computed for different values of dimensionless permeability parameter and dimensionless reaction rate
parameter 1 and are shown in Table 1. It can be seen
that the effective dispersion coefficient decreases with
increases in both and 1 . Equation (3.3) shows that
as grows, the velocity profiles continue to flatten
out and tend to plug flow as . This flattening
effect is the cause for the decrease in D1 with an
increase in . The decrease in D1 with an increase in
1 is natural on physical grounds since an increase in
1 leads to an increase in the number of molecules
of solute undergoing chemical reaction, resulting in a
decrease in Taylors dispersion coefficient. When
0, following the analysis of Rudraiah (1976), we find
that

D1

2
6 p
2 h x
=
945 2 2 D

(3.39)

which agrees with the result of Wooding (1960).


3.3.2 Taylor Diffusion with Heterogeneous Reaction
The diffusion in a porous channel with a first-order
irreversible chemical reaction taking place both in the
bulk of the fluid as well as at the walls is discussed in
this section.
In this case the diffusion equation is the same as
Eq. (3.34) with the heterogeneous boundary conditions

Rudraiah and Ng

where is the heterogeneous reaction rate parameter corresponding to catalytic reaction at the walls.
The solution of Eq. (3.34) satisfying the boundary
condition Eq. (3.40) is

h2 P C
21 cosh
C() =
+
pL2 D 2 2 (2 21 )cosh
2 2

1 + 2 21 +21 cosh

(3.41)
2 (2 21 )
where = 1 sinh 1 + cosh 1 .
Following the procedure as in Section 3.3.1, we
find that
D2
= Pe2 G (1 , , )
D

(3.42)

where D2 is the effective dispersion coefficient and

G(1 , , ) =

4
(1 )

(2

21)

2 1
24

(3.43)

1
1
+
3 sinh 21
1 2

21 2 + 2 21 + 21

1
coth 1
1
2 2
+
sinh1
1
1(2 21 ) 2 (2 21 )

Values of G (1 , , ) are computed for different


values of and for a fixed value of the reaction rate
constant 1 corresponding to the bulk reaction and are
tabulated in Table 2. From this it is clear that as in
the case of homogeneous reaction in the bulk of the
fluid, the increase in the wall catalytic parameter and
porous parameter causes a decrease in the effective
Taylor dispersion coefficient, which is more sensitive
to porous parameter than to wall catalytic parameter
.
3.3.3 Aris Modification to Taylors Approach

C
+ C = 0

C
C = 0

at = 1
at = 1

(3.40)

The analysis of Sections 3.3.1 and 3.3.2 are true only


when the conditions of Section 3.1.1 are valid, which
is the Taylors approximation, and can be relaxed following the Aris (1956) analysis, as in Section 3.2.

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

233

Table 1
Values of F2 (, )

= 0.4
= 0.4
= 1.2
= 1.6
= 2.0
= 2.4
= 2.8
= 3.2
= 3.6
= 4.0

= 20
0.72035485
0.69463623
0.65638278
0.61064020
0.56210659
0.51413141
0.46884689
0.42730419
0.38984253
0.35638306

10
103
103
103
103
103
103
103
103
103

= 40
0.19305787
0.18665859
0.17713057
0.16572515
0.15358554
0.14155871
0.13016877
0.11967950
0.11017910
0.10165243

= 60
0.87737479
0.84914667
0.80710623
0.75676183
0.70314705
0.64999285
0.59960995
0.55316375
0.51104706
0.47319756

10
103
103
103
103
103
103
103
103
103

10
104
104
104
104
104
104
104
104
104

= 80
0.49897345 104
0.48137464 104
0.45964289 104
0.43145858 104
0.40143686 104
0.37166467 104
0.34339900 104
0.31740040 104
0.29378150 104
0.2725443 6 104

Table 2
Values of G (1 , , )

2.0
4.0
6.0
8.0
10.0

= 10, 1 = 1.0
0.15976940 102
0.15506565 102
0.15314452 102
0.15210047 102
0.15144448 102

= 50, 1 = 1.0
0.83232600 104
0.80538381 104
0.79437999 104
0.78839983 104
0.78464245 104

Since the analysis of Section 3.2 is quite general,


we can apply it also to the present problem. By
taking integral moments of Eq. (3.33) and following
the analysis of Section 3.2, we can show the diffusion coefficients for homogeneous and heterogeneous
reactions, respectively, to be of the form

Dt = D 1 + Pe2 F2 (1 , )

(3.44a)

D2 = D 1 + Pe2 G1 (1 , )

(3.44b)

Comparing Eqs. (3.44a) and (3.44b) with Eqs. (3.37)


and (3.38), respectively, we find that the results of
Sections 3.3.1 and 3.3.2 are good approximations for
large Pe . We note that Eqs. (3.44a) and (3.44b) can
also be derived following the analysis of Rudraiah et
al. (1979) and Erdogan and Chatwin (1967) in the
form

Dt = D 1+Pe2 F2 (1, ) , i.e., Dt = D + D (3.45)


agreeing with Eq. (3.44a). Similarly, we can prove the
relation Eq. (3.44b). The relation Eq. (3.45) implies
that the total diffusion coefficient may be looked on as
the sum of the molecular diffusion coefficient D and
the effective Taylor diffusion coefficient D1 , and thus
it is asymptotically valid for all Peclet numbers.
In Sections 3.3.1 and 3.3.2 the effect of heterogeneous and homogeneous reactions on the longitudinal
dispersion of a solute subject to molecular diffusion
is investigated. Following Taylor, it is shown that the
combined effect of molecular diffusion, advection, and
chemical reaction on the distribution of concentration
of a cloud of solute is ultimately to make it spread
out symmetrically about a point with a mean velocity
u
. The mean concentration Cm satisfies the Ficks
law of diffusion with a longitudinal diffusivity given
by Eq. (3.37). It is found that D1 decreases with

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

234

increases in both and the reaction rate constant 1 .


The physical reason for this is the flattening of the
velocity profile due to increase in . The decrease in
D with increase in 1 is natural on physical grounds
since an increase in 1 leads to an increasing number
of molecules of solute undergoing chemical reaction.

4. GENERALIZED DISPERSION MODEL


In Sections 3.13.3 we have discussed DIPM considering some typical examples usually arising in
environmental and chemical industrial situations using Taylor (1953) and Aris (1956) models valid for
large time. The aim of this section is to explain the
use of a generalized dispersion model valid for all
time, as explained in Section 1.4, considering some
typical examples of biomedical engineering, namely
coronary artery diseases (CAD) and synovial joints.
In CAD the red blood cells, white blood cells, and
platelets; cholesterol; and other fat substances will
be suspended in the blood of arteries bounded by
endothelium, the boundaries of the arteries. These
suspended particles may execute microcirculation in
dispersing through the endothelium. Similarly, with
synovial joints, hyaluronic acid (HA), glycoprotein,
and other macromolecular components disperse from
synovial fluid to cartilage. HA and other components
may also execute microcirculation while dispersing
from synovial fluid to cartilage. The endothelium in
arteries and cartilage in synovial joints are layers of
porous cells that may or may not deform. In degenerative situations (see Ng et al., 2005) they may
be regarded as nondeformable cells forming nondeformable porous layers, while in natural situations,
these cells deform, forming deformable porous layers.
In this section we consider a nondeformable porous
layer with blood or synovial fluid modeled either to
execute microcirculation or be regarded as suspension
of a dusty fluid model. In the subsequent Section 5 we
consider a deformable porous layer. In Section 4.1 we
consider a micropolar fluid model, and in Section 4.2
we deal with a dusty fluid model.

Rudraiah and Ng

4.1 Generalized Dispersion in Micropolar Fluid


Bounded by a Porous Layer
In this section, following Rudraiah et al. (2001), we
study dispersion in the presence of suspended particles executing microcirculation, modeled in a mathematically rigorous fashion using Eringen (1966) field
theory of micropolar fluid. This has led to a very fruitful interaction between microcirculation of the group
of suspended macromolecules and the theory of partial differential equations. These interactions lead to
the dispersion of particles from one region to another.
To take care of microcirculation, we consider an incompressible micropolar fluid, in a free-flow region,
representing fluid driven by a constant pressure gradient in a region bounded below by a permeable layer,
where the flow is completely saturated, and above by
an impermeable region. The nominal surface is at y
= 0, and we assume that the fluid is in the region
between y = 0 and y = h and that the porous layer is
in the region h < y < 0, with < x < .
The nondimensional equations of motion for a
steady, fully developed micropolar fluid in a region
following Rudraiah et al. (2000) are
p
2u

+ 2 + 2N 2
=0
y
y

2
u
2
2
4 1N
2m 2 +
=0
y 2
y
Rh

(4.1)
(4.2)

These equations are solved using the no-slip condition:


u = 0 = 0 at y = 1

(4.3)

The effect of a porous layer on the fluid flowing


into a region is governed by the Saffman (1971) slip
conditions

0
d
du
= u
= at y = 0
dy
dy
k
k

(4.4)

where and 0 are the slip parameters and k is


the permeability of the porous layer. The solutions
of Eqs. (4.1) and (4.2) satisfying the above boundary
conditions are

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

2N 2
(A2 sinh my + A3 cosh my)
m
P Re 2 2A1 y
+
y +
+ A4
(4.5)
2
m2

u (y) =

PRe
A1
y 2 (4.6)
2
m
The constants Ai (i = 2 4) are given in the work of
Rudraiah et al. (2000). The expressions for them are
very lengthy and hence are omitted here, but they are
included in the numerical computation.
If a solute diffuses in a fully developed flow, the
concentration c of the solute satisfies the dimensionless diffusion equation
(y) = A2 coshmy +A3 sinhmy

1 2 2
+ u
= 2 2+ 2

Pe

(, , ) = 0 and

(4.7)

= 0 at = 1

With this method the solution of Eq. (4.7) is written in


a series expression in the form

= m (, ) +

fi (, )

i=1

i m
i

where the dispersion coefficients Ki () are timedependent. Equation (4.9) has to be solved subject
to
Xs
2
Xs
m (0, ) = 0 for || >
2
m (0, ) = 1 for ||

(4.11)
m (, ) = 0

Introducing Eq. (4.10) into Eq. (4.9), rearranging


terms, and following Rudraiah et al. (1991, 2000), we
get K1 () = 0 and
1
G1 sinh 2m + G2 cosh 2m
Pe2

X
2 2
+G3 cosh m+ G4 sinh m+ GS + An en Gn

K2 () =

i=1

The expressions for Gi (i = 1 5) and An are very


lengthy; hence they are not given here, and only their
numerical values are incorporated in the process of
computing K2 , K3 , K4 , and so on, and we found
that Ki () (i > 2) are negligibly small compared to
K2 (). Hence the dispersion model of Eq. (4.10) now
leads to

(4.8)

where m is the average concentration distribution


with respect to . Following the generalized dispersion procedure of Rudraiah et al. (1991, 2000) and
neglecting the details for want of space, and following
the analysis explained in Section 1.3, we get
m
1 2 m
m
+ u
2

Pe 2

X
2 f 1 i m
i + 1 m
fi

+ V fi
+

2
i
i + 1
i=1

X
1 i + 2 m
i + 1 m

fi 2

f
= 0 (4.9)
i
Pe i + 2
i
i=1
We assume that

X
m
i m
=
ki ()

i
i=1

235

(4.10)

m
2 m
= K2 ()

2
whose solution is

"
(
)
(
)#
Xs
Xs
1
2
2 +

= erf
+ erf
2
2 T
2 T

where T =

K2 (y) dy.

Equation (4.10) as such has no physical meaning.


However, the analysis reveals that K1 () = 0 and
Ki () (i > 2) are infinitesimal compared to K2 ().
This approximation now gives a physical meaning to
Eq. (4.10), namely, the diffusion equation. The parameters that influence K2 and m are the micropolar
parameter m, coupling number N 2 , and Darcy number D. Here K2 and m are computed. The behavior
of K2 and m reveals that they increase with the

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

236

increase in N 2 and m. We see that as increases,


m tends asymptotically to zero faster in the case of
micropolar fluid than of Newtonian fluid.
4.2 Generalized Dispersion Model Using Dusty
Viscous Fluid
As stated in Section 4, we discuss the generalized
dispersion model using dusty fluid. Here we consider
region 1 as the free fluid modeled as dusty fluid and
region 2 is a nondeformable porous layer modeled
as a dusty fluidsaturated porous layer. To understand
this, we consider the fluid flow in a region bounded
by a permeable surface, denoted by region 1, and
that in a porous layer, denoted by region 2. For this
physical configuration the basic equations governing
the flow, following Rudraiah et al. (2000), are

Rudraiah and Ng

For Region 2:
1 p
d2 u 2
+ 2 u2 =
d2
x

(4.17)

where
KN
m
y
n
w2 = n +
=
h
P (1 n)
k
2

2 =
w2 =
=

m
K
=

The velocity distribution, using the conditions


du1
= 0 at y = 0 u1 = u2 at y = h
(4.18)
dy
u1
u2

= 0
at y = h u2 = 0 at y = h + h0
y
y
are

For Region 1:

For Region 1:

u1
p
2 u1
=
+ 2 + KN (v1 u1 ) (4.12)
t
x
y

v1
= KN (u1 v1 )
t

(4.13)

For Region 2:

v2
= KN (u2 v2 )
t

(4.15)

The quantities are defined in the Nomenclature. Assuming that ui = ui (Y )ent and that vi = vi (Y )ent
(where i = 1 pertains to region 1 and i = 2 to region
2), and eliminating v1 and v2 between Eqs. (4.12)
(4.15), we get the equations in the nondimensional
form, as given below:
For Region 1:
p
d2 u1
+ w2 u1 =
d2
x

(4.16)

1
F1 cos w 2
w

(4.19)

For Region 2:

u2 =

u2
p
2u2

=
u2 + 0 2 +KN(v2 u2) (4.14)
t
x k
y

p
u1 =
x

1
P
F2 cos +F2 sin 2
(4.20)
x

where F1 , F2 , and F3 are constants having lengthy


expressions and are omitted here for want of space;
however, they are considered in the numerical evaluation. For details, see Rudraiah et al. (1991, 2000).
The average velocities for the two regions are given
by
Z1
u
1 =
0

P
u1 d =
x

sin
1
c3
2

(4.21)

0
1+h

Z
1
sin (1 + h0 ) sin
u
2 = 0
u2 d = c1
h
h0
0

1
cos (1 + h0 ) cos
2 0 (4.22)
c2
h
h

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

The particle velocity vi (y) in both regions can be


obtained from Eqs. (4.13) and (4.15) in the form
ui (y)
vi =
1 n

K0 (1 ) =

(4.23)

n=0

M0 (1 ) =

4.3 Dispersion Model

n An sin n en 1

n=0

(i = 1, 2)

237

(4.27)
An
n

sin n en 1

[Bn cos n bn sin n ]

n=0

A slug input of concentration c0 (x, y, 0) and the local


concentration c1 (x, y, t) of the solute in region 1 and
C2 (x, y, t) of region 2 satisfy the convective diffusion
equation

ci
ci
Di
1 2 ci
2 ci
+ u1
=
+
(4.24)
t
xi
(1 + i ) a2i x2i
y 2
where
i = KNi /, xi = x ai , ai =
(1 + i /1 )/(1 + i ) (i = 1 pertains to region
1 and i = 2 to region 2).
The initial and boundary conditions to be satisfied
are

ai
c0 ,
|x1 |
Xs
2
c1 (x1 , y, 0) =
(4.25a)
ai
0,
|x1 | >
Xs
2

!
. X sin (1 + h) sin
n
n
Bn
e

h
n
n=0
!

cos n (1 + h) cos n 2n 2
(4.28)
Bn
e
n h
2n 2

where

An =

Bn =

2 sin n
sin 2n
1+
2n
2 [sin n (l + h) sin n ] h
sin 2n (l + h) sin 2n
h+
2n

and n and n are the roots of the equations


n tan n = n tan n and Bn = tan n (l + h) Bn ,
respectively.
As , Eqs. (4.27) and (4.28) give the following asymptotic representation:

c2 (x, y, 0) = 0

(4.25b)

c1 (, y, ) = c2

(4.25c)

(, y, ) = 0

(4.25d)

K0 () = 20

(4.26a)

M0 () = 20 h sin 0 tan 0 (l + h)

cos 0
[sin 0 (l + h) sin 0 ]

[cos 0 (l + h) cos 0 ]
(4.29b)

c1 = c2
D1

c1
c2
= D2
at y = h
y
y

(4.26b)

c1
= 0 at y = 0
y

(4.26c)

c2
= 0 at y = 1 + h0
y

(4.26d)

Making these equations dimensionless using the scales


used earlier and following the analysis of Section 4.1,
the exchange coefficients K0 and M0 , convection
coefficients K1 and M1 , and dispersion coefficients
K2 and M2 can be obtained asymptotically as

(4.29a)

The asymptotic representation for convection coefficients following the approach of Gill and Sankarasubramanian (1970) can be obtained as

sin sin ( + 20 )
+
(4.30a)
K1 = C 3

( + 20 )
.

sin(+20)
sin
sin20
+
C3 1
0 +1
( + 20 )

I3 (0, 0)
I2 (0, 0)
+
(4.30b)
M1 =
2 (0, 0)
3 (0, 0)

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

238

The dispersion coefficients are given by

Rudraiah and Ng

For Region 1:

2
X

1
I1 (j, 0)
K2 = 2
2j 20 (4.31a)
P1

(j,
0)
1
j=1

1 X I22 (i, 0) X I32 (j, 0)


M2 = 2
+
(4.31b)
P2
(j, 0) j=1 3 (j, 0)
j=1 2
Here the expressions for i (j, 1) and Ii (j, 1) are
omitted for want of space but are included in the
numerical computation. Equations (4.30a), (4.30b),
(4.31a), and (4.31b) can be truncated after the terms
involving K2 and M2 without causing serious error
since the values of Ki and Mi for i > 2 will be
negligibly small. The resulting equations for mean
concentration in the two regions are

a X1
1
a + X1

erf
1 =
+ erf
e (4.35)
2
2 T
2 T
where
Z1
=

K0 (z) dz = K0 1
0

Z1
X = 1 +

K1 (z) dz = 1 + K1 1
0

Z1
T =

K2 (z) dz = K2 1
0

For Region 2:

For Region 1:
1
1
= K0 (1 ) 1 + K1 (1 )
1
1
2
1
+ K2 (1 )
21

1
X2
2 =
exp 0 0
4T0
2T0
(4.32)

For Region 2:

(4.36)

where
Z2
0 = M0 (Z) dz = M0 2 X0 = 2 + M1 2 T0

2
2
= M0 (2 ) 2 + M1 (2 )
2
2
2
2
+ M2 (2 )
22

= M2 2
(4.33)

The mean concentration has to satisfy the initial and


boundary conditions
(
1 ( 1 , 0) =

1
0

if |1 |
if |1 | >

1
2
1
2

a1
a1

(4.34a)

2 (2 , 0) = 0

(4.34b)

1 (, 1 ) = 2

(4.34c)

( , 1 ) = 0

(4.34d)

The solutions for equations (4.32) and (4.33) subject to conditions (4.34) are found to be

These expressions are numerically computed, and the


results are given below.
In this case the fluid is modeled as dusty viscous
fluid. The parameters governing the exchange coefficients K0 and M0 are the parameter , the drag
parameter , and diffusivity ratio . The computed
results reveal that K0 decreases with time initially
and reaches a constant value for large time, where as
M0 decreases with time. The effect of increases in
and is to decrease M0 , but K0 does not vary with
them. The mean concentration 1 decreases with an
increase in time. However, 2 increases initially and
then decreases for large time for all values of , ,
and .

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

5. DISPERSION IN DEFORMABLE POROUS


MEDIA
As explained in Section 3, we study the dispersion
phenomena in deformable porous media using the
Taylor (1953) model.
For simplicity we consider a two-dimensional geometry, shown in Fig. 3. This geometry consists of
fully developed flow through a symmetrical channel
filled with fluid-saturated porous media with boundaries at y = 0 and y = h. A pressure gradient
p/x = G (t) producing an axially directed flow is
applied. Owing to an assumption of infinite channel,
we assume no x-dependence in any of the terms, except the pressure. Since we assume fully developed
flow, we consider for simplicity unidirectional flow
with us as the x-component of displacement of solid
particles in porous media and uf as the x-component
of velocity of fluid. Then the momentum equations,
for the mixture of deformable solid and fluid assuming linear drag (i.e., Cb = 0) in Eq. (2.3) and making
them dimensionless using
y =

ut =

y
h

t =

h2 G0

uf

t
t0

us =
G =

h2 G0

us

G
G0

where h, G0 , and t0 are the characteristic length,


pressure gradient, and time, respectively, and for simplicity, neglecting the asterisks ( ), are
2 s

2us
2us
u
s
f
=
R

R
G

R
R
u
(5.1)
1
1
2
3
t2
y 2
t

uf
2uf
us
f
f
= R3 2 R4 G + R2 R3
u (5.2)
t
y
t

Figure 3. Physical model

239

where Ri (i = 1 4) are dimensionless numbers


defined in the Nomenclature. The boundary conditions
are
uf = us = 0 at y = 0 1 for t > 0

(5.3)

If C is the concentration of deformable particles in


the fluid and diffuses in the fully developed flow given
by Eqs. (5.1) and (5.2), then C satisfies the equation
C
C
+ u
= Dm
t
x

2C
2C
+
x2
y 2

(5.4)

We are interested here in studying the dispersion


mechanism using the Taylor-Aris approach.
5.1 Taylor Dispersion
Following the assumptions made in Section 3.1 and
using us = u and uf = v in Eqs. (5.1) and (5.2),
and assuming steady flow (i.e., /t = 0) and solving
using the no-slip conditions u = v = 0 at y = 0 and
y = 1, we get
h
s
f
y (1 y) + 4
sinh y (5.5)
2
sinh
i
2
y (1 y) sinh + sinh (1 y)
sinh +
2

u=

f
sinh y sinh
2
sinh

+ sinh (1 y)

v=

(5.6)

where
=

R2 R3
Kh2
=
R1
a

2 =

R2
Kh2
=
=
R4
a

If solid particles diffuse in the fully developed


flow, the concentration C of the media, following the
assumptions made in Section 3 after Taylor (1953),
satisfies Eq. (3.2) with u replaced by u where u =
u for solid and u = v for fluid. We can consider
two cases: One is advection of concentration by solid
particles, and the other is advection of concentration
by fluid.

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

240

5.1.1 Advection by Fluid

Rudraiah and Ng

where

In this case, Eq. (3.2) takes the form


C
C
2C
+v
= Dm 2
t
x
y

(5.7)

This equation is made dimensionless using the nondimensional quantities given in Eq. (3.6), using
Z1
v = vdy =
o

2(cosh1)
f

sinh
(5.8)
2 sinh

and neglecting asterisks ( ) for simplicity; using the


Taylor (1953) quasi-steady approximation, we get
2C
h2 W f C
=
2
y
DL

Dro =

h2 v2
F0 ()
D

1 sinh cosh
1
F0 () = 2
2
0
3

(cosh 1)
2

24
+

34

0 =

2 (cosh 1)
2
sinh

(5.9)

where
f
W f = v v = 2
sinh

2 (cosh 1)
sinh y + sinh (1 y)

We solve this equation for the two cases; case


1 corresponds to the situation where the layer is
bounded by insulating types of boundary conditions,
and case 2 corresponds to the layer bounded below by
an insulating type and above by a conducting type of
boundary condition.

which is the equation governing the longitudinal dispersion of solids in the fluid; Dr0 is the dispersion
coefficient.
F0 () is computed for different values of , and
the results are tabulated in Table 3. From Table 3 it
is clear that the function F0 () decreases with an
increase in and approaches zero for large values
of . Therefore the effect of is to control the
dispersion of solid particles in the porous media.
5.1.1.2. Case 2: Upper Boundary Conducting and
Lower Boundary Insulating
In this case the boundary conditions are

5.1.1.1 Case 1: Both Boundaries Insulating


The solution of Eq. (5.9), satisfying Eq. (3.8), is

h2 f
sinh y sinh (1 y)
C=
+
DL2 sinh
2
2

2
y (cosh1) y(cosh1) C

+
+C0 (5.10)

where C0 is a constant to be determined using the


entry condition.
Defining the volumetric rate M as in Eq. (3.10) and
following the same procedure in obtaining Eq. (3.15),
we get
Cm
2 Cm
= D r0
t
2

(5.11)

C
= 0 at y = 0
y

C = 1 at y = 1

(5.12)

The solution of Eq. (5.9), satisfying the conditions of


Eq. (5.12), is

h2 f
sinhy sinh(1y)
C = 1+
+
(5.13)
2
DL sinh 2
2

y 2 (cosh 1) y (cosh 1) sinh C

Note that the first four terms in the square brackets


in Eq. (5.13) are the same as the terms in the square
brackets of the solution (5.10). These two solutions
(5.10) and (5.13) differ only in the last term in the

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

241

Table 3
Variation of F0 () with

100
101
102

F0 ()
1.04 102
2.45 103
1.64 109

square brackets of Eq. (5.13). With this observation


the expression for M takes the form

M =

2h h2 v2
C
F1 ()
L D

(5.14)

We note that the contribution of the last term in


the square brackets in (5.13) to M is zero so that
F1 () = F0 (). Hence we get the same dispersion
coefficient given in Eq. (5.11).
5.1.2 Dispersion of Solid Particles by Advection
In this case we note that according to the momentum
balance equation, the displacement u of solid particles is a function of velocity v of fluid. Furthermore,
the parameters R1 and R2 involve the characteristic time t20 and t0 , respectively, and hence we can
use the displacement u in advection of concentration with a suitable dimension. Then applying the
procedure explained in Section 3.1 in dealing with
Eq. (3.2), considering velocity u for solid particles as
in Eq. (5.5), and computing average velocity u
and
using the approximations made in deriving Eq. (5.9),
we get
2C
h2 W s C
=
2
DL

(5.15)

where W s = u u
/t0 . The expression for W s is very
lengthy and hence is omitted here, but it is included in
solving C.
Here also we discuss the following two cases.
5.1.2.1. Case 1: Both Boundaries Insulating
The solution of Eq. (5.15), satisfying the condition of
Eq. (3.8), is

f
h2 s 3 4 C h2
2y y y
+
C=
4
DL 24
DL sinh

sinh y sinh (1 y) y 2 (cosh 1)


+

2
2

2 3
y(cosh
1)

2y y 4 y 2 sinh +
24

+ Co
(5.16)

As before, C0 has to be determined using an entry


condition.
Following the procedure in obtaining M as explained (Eq. (3.10)), we get
M = 5 103

h3 s2 C

+ G1 + G2 + G3 (5.17)
DL

The expressions for Gi (i = 1 3) are very lengthy


and hence are omitted here, but they are included
in the numerical evaluation. The results reveal that a
suitable value of (> 1) controls the dispersion of
solid particles in the fluid.
6. HOMOGENIZATION THEORY
We first discuss here homogenization theory to derive
the required basic equations, which are then used to
study DIPM with chemical reaction.
6.1 A Brief Description of Homogenization Theory
The homogenization theory is a multiple-scale method
of averaging that can be used to derive phenomenological equations on the basis of micromechanics in
a general and rigorous manner, without any closure
hypothesis. The core assumptions of the theory are
that (1) there exist two or more vastly different length
scales and (2) the submacrostructures are periodic. For

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

242

a two-scale medium the ratio of microscale ` to the


macroscale L is very small, that is, `/L 1,
which can be used as an ordering parameter, enabling
an asymptotic analysis to be carried out. The method
of multiple-scale (spatially and temporally) expansion
can be employed systematically to yield approximate
governing equations at successive orders, together
with proper boundary conditions, including periodicity. These equations define the canonical boundary
value problem in a unit cell. Once solved, their averages over the cell will then yield the macroscale
equations. Despite the idealization, the resultant effective equations can often contain information that
lead to a better insight into the mechanisms of the
problem. See Mei et al. (1996) for a review of some
of the applications of this method.
6.2 DIPM with Chemical Reactions

C
+ (uC) = D2 C in f
t

Figure 4. Schematic of a unit cell for the microstructure


of a porous medium

u=0

Let us now apply the homogenization technique to


DIPM, where the transport is subject to two chemical
reactions, one reversible and the other irreversible, on
the solid-fluid interface. We first define a unit cell
for the microstructure of the porous medium, as is
schematically shown in Fig. 4, in which f is the
pore fluid, s is the solid grain, is the solidfluid interface, and n is the unit outward normal to
the boundary of f . Note that for simplicity, we
have shown only one solid grain in the unit cell. In
fact, the present theory is applicable to any kind of
microstructure of the medium, as long as it is periodic
over the unit cell = f + s .
Consider the mass transport of a reactive chemical
species in the pore fluid. The mass concentration
C (x, t) of the solute phase (i.e., mass of species
dissolved per volume of pore fluid) is assumed to be
governed everywhere by

(6.1)

where u is the pore fluid velocity and D is the


molecular diffusivity of the solute in the fluid. The
fluid is incompressible, and therefore

Rudraiah and Ng

(6.2)

We suppose that on the solid-fluid interface, the


species may undergo two chemical processes of firstorder kinetics: One is to cause it to decay irreversibly
at a rate of , and the other is to cause it to exchange
reversibly between a solute phase in the fluid and
a solid phase sorbed on the solid at a rate of k.
To avoid excessive loss of the chemical species from
the system within an effective time scale of mass
transport, we assume that the irreversible reaction rate
is one order of magnitude slower than the reversible
one: /k = O (). If we denote the concentration of
the sorbed phase (mass per surface area of ) by Cs ,
the boundary condition can be specified as follows:

DC nC =

Cs
= k (C Cs ) on (6.3)
t

where is the equilibrium partition ratio of the


chemical into the two phases; that is, when chemical
equilibrium is attained, the sorbed phase concentration
is equal to times the fluid phase concentration:
= (Cs /C)equilibrium . Also, we have inserted in
Eq. (6.3) to signify that the decay reaction is a
slow process. For the convenience of the deduction
presented subsequently, we further assume that the

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

porous medium is homogeneous on the macroscale,


that is, that all the material and chemical properties
are constants and invariant with the macrocoordinates.
We emphasize, however, that this assumption is not a
necessary one and is introduced here only for the sake
of simplicity.
Since is very small, we look for the solution of
Eq. (6.1) in terms of the regular perturbation technique. In this technique the unknowns u, C, and Cs
are expressed in the form

(u, C, Cs) = u(0) , C (0) , Cs(0) + u(1) , C (1) , Cs(1)

+ 2 u(2) , C (2) , Cs(2) +


(6.4)

O(2 ):
(0)

Cs
D C (2) + 0 C (1) n C (1) =
t2
(1)

Cs
+
= k C (2) Cs(2) on
(6.9)
t1

Here we define a volume average over the unit cell


as follows. For any function f ,

hf i

(6.5)

=
+ 2
t
t1
t2

(6.6)

x0i

= xi (i = 1, 2, 3) are the macrocoordinates


where

and (t1 , t2 ) = , 2 t are the effective time scales


by convection and dispersion, respectively. On substituting these expansions into the transport Eq. (6.1)
and the boundary condition Eq. (6.3), and collecting
terms of equal power of , we get equations that are
very lengthy, and hence they are pointed out, but their
solutions are given satisfying the boundary conditions

1
||

ZZZ
f dV

|f |
=
||

DC (0) n = k C (0) Cs(0)

on

(6.7)

O():

(0)
Cs

D C (1) + 0 C (0) n C (0) =


t1

(1)
(1)
= k C Cs
on
(6.8)

(6.11)

Likewise, we may define an area average over the


solid-fluid interface as
1
f
||

ZZ
f dS

(6.12)

At O(1), the perturbation equations imply that the


(0)
leading order concentrations C (0) and Cs are independent of the microcoordinates:

C (0) = C (0)(x0i , t1 , t2 ) Cs(0) = Cs(0)(x0i , t1 , t2 ) (6.13)


and they are in local equilibrium:

Cs(0) = C (0)

O(1):

(6.10)

In particular, the porosity is given by

h1i =
= + 0

243

(6.14)

At O() the volume averaging of perturbation equation gives, after using the Gauss theorem and the
boundary condition Eq. (6.8), the leading order effective transport equation:
(0)
~u
C (0)
(0)
+
0 C (0) +
C =0
t1
R
R

(6.15)

where = || / || is the solid-fluid interface area


per volume of the porous medium and R = +
is the retardation factor for the mass transport in the

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

244

medium. At the leading order, the transport is controlled only by convection and the two reactions; dispersion is not effective on this short time scale. While
the irreversible reaction is to diminish the species
through the sink term in Eq. (6.15), the reversible
reaction is to retard the rates of all transport processes
through the factor R, resulting from local equilibrium
between the two phases of the species.
On eliminating the unsteady term between the perturbation equation and Eq. (6.15), we express the
higher-order concentrations as follows:

C (1) = N 0 C (0) + M C (0)

(6.16)

Cs(1) = Ns 0 C (0) + Ms C (0)

(6.17)

where N, Ns , M , and Ms are functions of the microstructure and are determined by solving the following boundary value problems. The problem for N
is
u(0) N D2 N = u
(0) in f

Ms = M | +

Rudraiah and Ng


k R

(6.23)

The functions N and M can be determined only up to


an arbitrary constant. For uniqueness we may further
require them to satisfy

hNi = hM i = 0

(6.24)

by which C (1) = 0.
At O(2 ), taking the volume average of perturbation equation, on using the Gauss theorem and the
boundary condition Eq. (6.9) followed by the substitution of Eqs. (6.16) and (6.17), we may get after some
algebra the higher-order effective transport equation

C (0)
u0
0 (0)
+
0 C (0) +
C
t2
R
R

!
0
DI + D

02 C (0)
=
R

(6.25)

(6.18)
where

with the boundary condition

D N
I n =

(0)

u
R

= k (N + Ns ) on

(6.19)

where u
(0) = u(0) u(0) /R and
I is the isotropic
tensor. The problem for M is

(0)

M D M =
in f
R
2

(6.20)

= k (M + Ms ) on (6.21)
R

From Eqs. (6.19) and (6.21), Ns and Ms are found in


the form

u(0)
Ns = N|
k R

is the O() correction to the convection velocity,

0 =

with the boundary condition

DM + =

D
E D
E
2
u0 = u(1) + M u(0) D hM i +
Ns
R

D (0) E

N
(6.26)
Ms u

(6.22)

Ms + M

(6.27)

0
is the O() correction to the sink term, and D
is the

dispersion coefficient tensor given by

D
E
D (0) E
0
(0)
N
D
=
u
N

D
hNi

R
E
E
D
D

(6.28)
+
u(0) u(0)
kR2
Expressed in index form, the combined diffusion and
dispersion coefficient tensor can be written in the
following symmetric form:

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

Dij
= DI + D
R1

E D N
1 D (0)
Nj
i
(0)
Dij + 2 ui Nj + uj Ni 2 x + x
j
i
=
E
D
E
D
ED
E
D

(0)
(0)
(0)
(0)
j + u
i +

ui
N
N
ui
uj
j
2

2R
kR

D
E
(0)
(0)
(0)
where u
i = ui ui
/R and hNi i = 0 has
been used. One may simplify the expression above by
noting that

D
E
(0)
(0)
(0)Ni
(0)Nj
u
i Nj + u
j Ni = 21 uk
Nj +uk
Ni
xk
xk

Nj 2 Ni + Ni 2 Nj
2

D Ni
Nj
Ni Nj
=D

+
xk xk
2 xj
xi
D
E
D (0) E
(0)
i
+
ui
Nj + uj
N
(6.30)

2R

1
2

where use has been made of Eqs. (6.2), (6.18), (6.19),


and the Gauss theorem. Substituting Eq. (6.30) into
Eq. (6.29), one can easily obtain a compact expression
for the diffusion-dispersion coefficient:

Dij

D
=
R

bi bj
xk xk

D (0) E D (0) E
ui
uj
(6.31)
kR3

where bi = Ni xi . A few remarks regarding the


dispersion coefficient can be made as follows.
1 In principle, this coefficient can be computed numerically for any given periodic microstructure.
This can be a nontrivial task as the computational
domain is typically a three-dimensional complex
geometry.

245

1
R

(6.29)

4 When the species is inert or nonreactive (i.e.,


= = 0), Eq. (6.31) reduces to the one previously obtained by Brenner (1980) using Brownian motion theory and by Mei (1992) using the
homogenization theory.
5 At this order, the dispersion coefficient is affected only by the reversible phase exchange
process, but not by the irreversible decay process.
The latter will indeed affect the dispersion when
the next higher order transport is considered. The
deduction is rather algebraically involved and is
not presented here.
6 The reversible reaction affects the dispersion
through phase partitioning and phase exchange
kinetics. If the phase exchange is sufficiently kinetic (i.e., small k or phase equilibrium not quite
readily achieved), the dispersion coefficient in the
presence of even a very small degree of sorption
can be dramatically different, quantitatively or
qualitatively, from the inert case. A strongly kinetic phase exchange can increase the dispersion
coefficient by one order of magnitude or so. The
opposite is true when the phase exchange kinetics
is weak and the sorption partition is large, for
which the dispersion coefficient will be weighted
down by the retardation factor.

2 It is clear from Eq. (6.31) that the coefficient is


not only symmetric, but also always positive.

ACKNOWLEDGMENTS

3 Although we have derived Eq. (6.31) based on


the assumption that the porous medium is homogeneous and the physical and chemical properties
are constants, one may show that with the use
of a spatial averaging theorem (Mei et al., 1996),
exactly the same expression for the dispersion
coefficient can be deduced, even without this
assumption.

This work was started when one of us (N.R.) was


visiting Hong Kong University. The financial support
by the Research Grants Council of the Hong Kong
Special Administrative Region, China, through project
HKU 7192/04E is gratefully acknowledged. The work
of N.R. was also supported by DST under research
project AS: 237/04, and their support is gratefully
acknowledged.

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

246

REFERENCES

Aris, R., On the Dispersion of a Solute in a Fluid Flowing


through a Tube, Proc. R. Soc. London, Ser. A, vol. 235,
pp. 6777, 1956.
Barry, S. I., Parker, K. H., and Aldis, G. K., Fluid Flow
over a Thin Deformable Porous Layer, Z. Angewandte
Math. Phys., vol. 42, pp. 633648, 1991.
Barton, N. G., On the Method of Moments for Solute
Dispersion, J. Fluid Mech., vol. 126, pp. 205218,
1983.
Barton, N. G., and Stokes, A. N., A Computational
Method for Sheared Dispersion in Parallel Flows, in
Computational Techniques and Applications, CTAC85,
J. Noye and R. May (eds.), pp. 345355, Elsevier, New
York, 1986.
Bear, J., and Bachmat, Y., A Generalized Theory on Hydrodynamic Dispersion in Porous Media, IASH Symp.
Artif. Recharge Manage. Aquifers, pp. 716, 1967.
Beavers, G. S., and Joseph, D. D., Boundary Conditions at
a Naturally Permeable Wall, J. Fluid Mech., vol. 30,
pp. 197207, 1967.
Bejan, A., et al., Porous and Complex Flow Structures in
Modern Technologies, Springer, New York, 2004.
Bowen, R. M., Incompressible Porous Media Models by
Use of the Theory of Mixtures, Int. J. Eng. Sci., vol.
18, pp. 11291148, 1980.
Brenner, H., Dispersion Resulting from Flow through Spatially Periodic Porous Media, Philos. Trans. R. Soc.
London, Ser. A, vol. 297, pp. 81133, 1980.
Brinkman, H. C., A Calculation of the Viscous Force Exerted by a Flowing Fluid on a Dense Swarm of
Particles, Appl. Sci. Res., vol. A1, pp. 2734, 1947.
Chandrasekara, B. C., Rudraiah, N., and Nagaraj, S. T.,
Velocity and Dispersion in Porous-Media, Int. J. Eng.
Sci., vol. 18, pp. 921929, 1980.
Chatwin, P. C., On the Interpretation of Some Longitudinal Dispersion Experiments, J. Fluid Mech., vol. 48,
pp. 689702, 1971.
Chatwin, P. C., Longitudinal Dispersion of Dye Whose
Concentration Varies Harmonically with Time, J. Fluid
Mech., vol. 58, pp. 657667, 1973.
Erdogan, M. E., and Sullivan, P. J., The Effects of Aspect Ratio on Longitudinal Diffusivity in Rectangular
Channels, J. Fluid Mech., vol. 120, pp. 347358, 1982.
Erdogan, M. E., and Chatwin, P. C., The Effects of Cur-

Rudraiah and Ng

vature and Buoyancy on the Laminar Dispersion of


Solute in a Horizontal Tube, J. Fluid Mech., vol. 29,
pp. 465484, 1967.
Eringen, A. C., Theory of Micropolar Fluids, J. Math.
Mech., vol. 16, pp. 118, 1966.
Fung, Y. C., and Tang, H. T., Longitudinal Dispersion of
Tracer Particles in Blood Flowing in a Pulmonary
Alveolar Sheet, J. Appl. Mech., vol. 42, pp. 536540,
1975.
Gill, W. N., and Sankarasubramanian, R., Exact Analysis
of Unsteady Convective Diffusion, Proc. R. Soc. London, Ser. A, vol. 316, pp. 341350, 1970.
Gottlieb, D., and Orszag, S. A., Numerical Analysis of
Spectral Methods, SIAM, Philadelphia, 1977.
Guymon, G. L., A Finite Element Solution of OneDimensional DiffusionConvection Equation, Water
Resour. Res., vol. 6, pp. 204210, 1970.
Harden, T. O., and Shen, H. T., Numerical Simulation of
Mixing in Natural Rivers, J. Hydraul. Div. Am. Soc.
Civ. Eng., vol. 105, pp. 393408, 1979.
Harleman, D. R. F., Melhorn, P. F., and Rumer, R. R.,
Dispersion-Permeability Correlation in Porous Media,
J. Hydraul. Div. Am. Soc. Civ. Eng., vol. 89, pp. 6785,
1963.
Hulin, J. P., Cazahat, A. M., Guyon, F., and Carmona, F.,
Hydrodynamics of Dispersed Media, Elsevier, New
York, 1990.
Jayaraj, K., and Subramanian, R. S., On Relaxation Phenomena in Field-Flow Fractionation, Sep. Sci. Technol.,
vol. 13, pp. 791817, 1978.
Kaviany, M., Principles of Heat Transfer in Porous Media,
Springer, New York, 1999.
Kenyon, D. E., The Theory of an Incompressible SolidFluid Mixture, Arch. Rat. Mech. Anal., vol. 62, pp.
131147, 1976.
Lighthill, M. J., Motion in Narrow Capillaries from the
Standpoint of Lubrication Theory, in Circulatory and
Respiratory Mass Transport, G. E. N. Wolstenholme
and J. Knight (eds.), pp. 85104, J & A, Churchill,
1969.
Maloy, K. J., et al., Fractal Structure and Hydrodynamic
Dispersion in Porous Medial, Phys. Rev. Lett., vol. 26,
pp. 22952298, 1988.
Manz, B., Gladden, L. F., and Warren, P. B., Flow and
Dispersion in Porous Media: Lattice-Boltzmann and
NMR Studies, AICE J., vol. 45, pp. 18451854, 1988.

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

Dispersion in Porous Media

Mei, C. C., Method of Homogenization Applied to Dispersion in Porous Media, Transp. Porous Media, vol. 9,
pp. 261274, 1992.
Mei, C. C., Auriault, J. L., and Ng, C. O., Some Applications of the Homogenization Theory, Adv. Appl. Mech.,
vol. 32, pp. 277348, 1996.
Ng, C. O., Rudraiah, N., Nagaraj, C., and Nagaraj, H. N.,
Electrohydrodynamic Dispersion of Macro Molecular
Components in Nanostructured Biological Bearing, J.
Energy Heat Mass Transfer, vol. 27, pp. 3964, 2005.
Nield, D. A., and Bejan, A., Convection in Porous Media,
Springer, New York, 1999.
Patterson, A., Dionfrio, A., Allain, C., and Hulin, J. P.,
Tracer Dispersion in Polymer Solution Flowing through
a Double Porosity Porous Medium, J. Phys. II, vol. 6,
pp. 16391654, 1996.
Posner, M., and Gill, W. N., Miscible Displacement with
Combined Free and Forced ConvectionLaminar Flow
in a Vertical Tube, AIChE J., vol. 19, pp. 151158,
1973.
Rudraiah, N., Dispersion of Soluble Matter in Laminar
Flow through Porous Media between Two Parallel
Plates, Bangalore Univ. J., vol. 2, pp. 110, 1976.
Rudraiah, N., Proceedings of the First DST-SERC School
on Mathematical Modelling of Atmospheric Pollution,
Sapna Book House, Bangalore, India, 2001.
Rudraiah, N., Chandrasekara, B. C., Veerabhadraiah, R.,
and Nagaraj, S. T., Some Flow Problems in Porous
Media, PGSAM Series, vol. 2, Bangalore Univ., India,
1979.
Rudraiah, N., Pal, D., and Vortmeyer, D., The Effect of
Aspect Ratio on Longitudinal Diffusivity in a Porous
Medium of Rectangular Cross-Section, Int. Commun.
Heat Mass Transfer, vol. 12, pp. 313322, 1985.
Rudraiah, N., Kasiviswanathan, S. R., and Kaloni, P. N.,
Generalized Dispersion in a Synovial Fluid of Human
Joints, Biorheology, vol. 28, pp. 207219, 1991.
Rudraiah, N., Siddeshwar, P. G., and Ranganath, T. R.,
Anatomy and Biomechanics of Synovial Joints, Part
III, Hydrodynamic Dispersion, in Frontier in Biomech,
Supplemented with Yoga Concepts, N. Rudraiah et al.
(ed.), pp. 137, Sapna Book House, Bangalore, India,
2000.
Saffman, P. G., A Theory of Dispersion in a Porous
Medium, J. Fluid Mech., vol. 6, pp. 321349, 1959.
Saffman, P. G., Dispersion Due to Molecular Diffusion

247

and Macroscopic Mixing in Flow through a Network of


Capillaries, J. Fluid Mech., vol. 7, pp. 194208, 1960.
Saffman, P. G., On the Boundary Condition at the Surface
of a Porous Medium, Stud. Appl. Math., vol. 50, pp.
93101, 1971.
Sahni, M., and Jue, V. L., Diffusion of Large Molecules in
Porous Media, Phys. Rev. Lett., vol. 62, pp. 629632,
1989.
Shivakumar, P. N., Rudraiah, N., Pal, D., and Siddheshwar, P. G., Closed form Solution for Unsteady Convective Diffusion in a Fluid Saturated Sparsely Packed
Porous Medium, Int. Commun. Heat Mass Transfer,
vol. 14, pp. 137145, 1987.
Shivakumara, I. S., and Venkatachalappa, M., Flow
through Porous and Nonporous Media, in Advances
in Fluid Mechanics: A Series on Collected Works of N.
Rudraiah, vol. 2, pp. 4756, McGraw-Hill, New York,
2004.
Simpson, E. S., Velocity and the Longitudinal Dispersion
Coefficient in Flow through Porous Media, in Flow
through Porous Media, R. J. M. De Wiest (ed.), pp.
201214, Academic, New York, 1969.
Slattery, J. C., Flow of Viscoelastic Fluids through Porous
Media, AIChE J., vol. 13, pp. 10661071, 1967.
Smith, I. M., Faraday, R. V., and OConnor, B. A.,
Rayleigh-Ritz and Galerkin Finite Elements for
Diffusion-Convection Problems, Water Resour. Res.,
vol. 9, pp. 593606, 1973.
Smith, R., The Early Stages of Contaminant Dispersion in
Shear Flows, J. Fluid Mech., vol. 111, pp. 107122,
1981.
Sueiu, A. N., Iwatsubo, T., and Matsuda, M., Theoretical
Investigation of an Artificial Joint with Micro-PocketCovered Component and Biphatic Cartilage on the
Opposite Articulating Surface, ASME J. Biomech. Eng.,
vol. 125, pp. 425433, 2003.
Taber, J. J., Research on Enhanced Oil RecoveryPast,
Present and Future, Pure Appl. Chem., vol. 52, pp.
13231347, 1980.
Tam, C. K. W., The drag on a Cloud of Spherical Particles in Low Reynolds Number Flow, J. Fluid Mech.,
vol. 38, pp. 537546, 1969.
Taylor, G. I., Dispersion of Soluble Matter in Solvent
Flowing through a Tube, Proc. R. Soc. London, Ser. A,
vol. 219, pp. 186203, 1953.
Vafai, K., Handbook of Porous Media, Marcel Dekker,

Begell House Inc., http://begellhouse.com Downloaded 2007-5-29 from IP 147.8.93.243 by Dr. Chiu-On Ng (cong)

248

New York, 2000.


Whitaker, S., Diffusion and Dispersion in Porous Media,
AIChE J., vol. 13, pp. 420427, 1967.
Wooding, R. A., Instability of a Viscous Liquid of Vari-

Rudraiah and Ng

able Density in a Vertical Hele-Shaw Cell, J. Fluid


Mech., vol. 7, pp. 501515, 1960.
Zienkiewicz, O. C., The Finite Element Method in Engineering Science, McGraw-Hill, New York, 1971.

Das könnte Ihnen auch gefallen