You are on page 1of 30

Transport in Porous Media 2 (1987), 327-356 327

© 1987 by D. Reidel Publishing Company.

Diffusion in Anisotropic Porous Media


JIN-HWAN KIM*, J. ALBERTO OCHOA, and STEPHEN WHIT AKER **
Department of Chemical Engineering, University of California at Davis, Davis, CA 95616, U.S.A.

(Received: 2 September 1986; revised: 19 May 1987)

Abstract. An experimental system was constructed in order to measure the two distinct components
of the effective diffusivity tensor in transversely isotropic, unconsolidated porous media. Measure-
ments were made for porous media consisting of glass spheres, mica particles, and disks made from
mylar sheets. Both the particle geometry and the void fraction of the porous media were determined
experimentally, and theoretical calculations for the two components of the effective diffusivity tensor
were carried out. The comparison between theory and experiment clearly indicates that the void
fraction and particle geometry are insufficient to characterize the process of diffusion in anisotropic
porous media.

Key words. Diffusion, anisotropy, diffusivity tensor, volume averaging.

o. Nomenclature
Roman Letters
d YK interfacial area between 'Y- and K-phases for the macroscopic system, m 2
dye area of entrances and exits of the K-phase for the macroscopic system, m 2
AYK interfacial area contained within the averaging volume, m 2
a characteristic length of a particle, m
b average thickness of a particle, m
CA concentration of species A, moles/m 3
.. Co reference concentration of species A, moles/m 3
(CA)Y intrinsic phase average concentration of species A, moles/m3
CA CA - (CAP, spatial deviation concentration of species A, moles/m
3
C (CA) Y / Co, dimensionless concentration of species A
9lJ binary molecular diffusion coefficient, m 2/s
Deff effective diffusivity tensor, m 2/s
Dxx component of the effective diffusivity tensor associated with diffusion
parallel to the bedding plane, m 2/s
Dyy component of the effective diffusivity tensor associated with diffusion
perpendicular to the bedding plane, m2/s
Deff effective diffusivity for isotropic systems, m2/s
f vector field that maps V (CA) Y on to CA, m
h depth of the mixing chamber, m

*Current address: Department of Chemical Engineering, Chonnam National University, Korea.


** Author to whom correspondence should be addressed.
328 JIN-HWAN KIM ET AL.

H EL / h, dimensionless parameter
I unit tensor
Iy characteristic length of the 'Y-phase, m
Ii lattice vectors (i = 1,2,3), m
L characteristic length for (CA)Y, m; and the depth of the sample chamber,
m
length of the sides of a two-dimensional unit cell, m
molar flux vector for species A, moles/m 2 s
phase average of the molar flux vector for species A, moles/m 2s
unit normal vector directed from the 'Y-phase toward the K-phase
ro radius of the averaging volume, m' .
r position vector, m
time, s
t* characteristic time, s; dimensionless time (tDxx / L 2)
"if averaging volume, m 3
x distance, m
X x / L, dimensionless distance

Greek Letters
E Vy/"if, porosity
An nth eigenvalue
'T tortuosity

1. Introduction
Diffusion in porous media is a central issue in the subject of reactor design
(Satterfield, 1970; Jackson, 1977; Luss, 1977) and in a wide variety of mass
transfer operations (Cussler, 1984). The transport of nutrients to the roots of
plants (Tinker, 1970) and the recovery of methane from coal beds (Smith and
Williams, 1984) are processes which are intimately involved with diffusion in
porous media. Under certain circumstances the rate of drying is controlled by
diffusion of water vapor through a porous medium (Whitaker, 1977) and when
drying involves biological materials the structure of the porous medium can be
quite complex (Crapiste et al., 1986). Because of the complexity of the
diffusional processes that take place in porous media, there exists an enormous
body of literature dealing with both the experimental determination and the
theoretical prediction of effective diffusivities. The complexities consist of simul-
taneous bulk, Knudsen, and surface diffusion in bidispersed systems that are
difficult to characterize (Dullien, 1979). In realistic systems the diffusion takes
place in the presence of adsorption, reaction, temperature gradients, and con-
vective transport caused by mole number changes associated with chemical
reaction or imposed pressure gradients. In addition, one must sometimes deal
DIFFUSION IN ANISOTROPIC POROUS MEDIA 329
with structural changes that occur during the diffusion process (Gavalas and Kim,
1981; Bhatia and Perlmutter, 1983).
The theoretical approaches are as diverse as the problems and they include the
'dusty gas' (Mason and Malinauskas, 1983), random sphere models (Strieder and
Aris, 1973), nonuniform capillary tube models (Foster and Butt, 1966), random
capillary tube models (Johnson and Stewart, 1965; Gavalas and Kim, 1981),
general statistical methods (Beran, 1968; Batchelor, 1974), effective medium
theory (Webman, 1982) and multiscale methods (Bensoussan et al., 1978; Chang,
1982). The theory used in this study is based on the method of volume averaging
(Anderson and Jackson, 1967; Marie, 1967; Slattery, 1967; Whitaker, 1967) and
the closure scheme developed by Ryan et al. (1980, 1981). The general subject
of heat and mass transport in porous media has been reviewed by Carbonell and
Whitaker (1984) and the specific matter of diffusion has been considered by
Whitaker (1986) with a special emphasis on micropore-macropore systems and
the combined process of bulk diffusion, Knudsen diffusion and Darcy flow.
In this work we are concerned with the simplest possible diffusion process, i.e.,
bulk diffusion of species A under dilute solution conditions with N B , N c ,
etc. = O(N A ,) or constrained by equi-molar counter diffusion in a binary system.
The pressure and temperature are constant and there is neither adsorption nor
reaction at the gas-solid interface. The system is illustrated in Figure 1 in which
the solid is identified as the K-phase and the gas as the 'Y-phase. The governing

Averaging
Volume V
Fig. 1. A granular porous medium.
330 JIN-HWAN KIM ET AL.

point equation for the concentration of species A is

(1.1 )

and the boundary conditions are


(1.2)

B.C.2 CA = @P(r, t), on dye (1.3)


Here d YK represents the entire interfacial area of the system and dye represents
the area of entrances and exits of the y-phase for the macroscopic system. In
general, the boundary condition at the entrances and exits is known only in terms
of averaged quantities rather than the point concentration given in Equation
(1.3), and this latter boundary condition serves to remind us of what we do not
know rather than what we do know.
The local volume averaged form of Equation (1.1), subject to Equation (1.2),
is given by

€ a(~)Y = V.[€rzlJ (V (CA)Y + ~y to'K DyKCA dA)]. (1.4)

Here € represents the void fraction, Vy is the volume of the y-phase contained in
the averaging volume illustrated in Figure 1, and AyK represents the y-K
interfacial area contained within the averaging volume. In deriving Equation
(1.4), we have neglected variations of rzlJ within the averaging volume, and we
have imposed the length scale constraint (Carbonell and Whitaker, 1984, Sec. 2)

(Lro)2 ~1 (1.5)

in which L represents the characteristic length scale for (CA)Y.


The spatial deviation concentration is defined by
CA = CA - (CA)Y (1.6)
and a closure scheme is required in order to represent CA in terms of (CA)Y. One
can follow the original development of Ryan et al. (1981) or the more compre~
hensive treatment of Crapiste et al. (1986) to find that the governing equations
and boundary conditions for CA are given by
(1.7)
-D yK · rzlJV CA = DYK . rzlJV (CA)Y, at d YK (1.8)
CA = <G(r, t), at dye. (1.9)
While the volume-averaged diffusional process may be unsteady, the diffusional
process for the spatial deviation concentration, CA, can be treated as quasi-steady
DIFFUSION IN ANISOTROPIC POROUS MEDIA 331
on the basis of the constraint
!Jflt*
-~1
[2 . (1.10)
"
Here I" represents the characteristic length for CA and it can be thought of as
comparable to the pore diameter as illustrated in Figure 1. In Equation (1.10) we
have used t* to represent a characteristic process time and it is the smallness of I"
(relative to L) that makes Equation (1.10) a universally acceptable constraint.
Obviously there is no profit in solving Equation (1.7) over the entire domain of
interest, and instead one seeks solutions over some representative region such as
the region illustrated in Figure 2. To develop a boundary condition for this
region that replaces Equation (1.9), one treats the region shown in Figure 2 as a
unit cell in a spatially periodic porous media (Brenner and Adler, 1987). Under
these circumstances the boundary value problem takes the form
(1.11)
(1.12)
(1.13)

Fig. 2. Representative region of a spatially periodic porous medium.


332 JIN-HWAN KIM ET AL.

in which r represents the position vector and Ii represents the three, nonunique
lattice vectors needed to describe a spatially periodic porous medium. In addition
to satisfying Equations (1.11) through (1.13), we also require that the average of
CA be zero. This is consistent with Equation (1.6) and the length scale constraints
that are discussed in detail by Carbonell and Whitaker (1984).
Since we have already neglected variations of the molecular diffusivity within
the averaging volume in order to obtain Equation (1.4), it is consistent to neglect
these variations within the unit cell in order to obtain Equation (1.11) from
Equation (1.7). It is easy to show (Whitaker, 1986, Sec. 2; Ryan et ai., 1981) that
CA is a linear function of the gradient of the volume averaged concentration, thus
CA is expressed as
CA = I • V (CA)Y (1.14)
where the boundary value problem for the vector I is given by
V21 = 0, (1.15)
0.16)

I(r + Ii) = I(r), i = 1,2,3. (1.17)


In practice, the periodicity condition is used with r = r. where r. is the position
vector locating a point at the entrance or exit of a unit cell. Under these
circumstances Equations (1.15) through (1.17) determine I to within an arbitrary
constant vector. This arbitrariness is removed by requiring that the average of I
be zero. This is constitent with Equation (1.14), the length scale constraints given
by Carbonell and Whitaker (1984), and the fact that the average of CA is zero.
In order to obtain the closed form of Equation (1.4), we use Equation (1.14) in
the area integral in Equation 0.4) leading to

LYK DyKCA dA = [LYK DYKI dA] . V (CA)Y. (1.18)

Here we have treated V (CA)Y as a constant with respect to integration over AYK
on the basis of Equation (1.5). This result allows one to express Equation (1.4) as
a(CA)Y
e -- -= V . (eD eff • V(CA)Y) (1.19)
at
in which the effective diffusivity tensor is defined by

eD eff = er!iJ [1+ ~ f


'Y AYK
DyKI dA]. (1.20)

One can also define an effective diffusivity tensor according to

(1.21)
DIFFUSION IN ANISOTROPIC POROUS MEDIA 333
and this definition appears to dominate in the reactor design literature where the
main interest is in the phase average molar flux (Whitaker, 1986, Sec. 2) given by
(1.22)

For isotropic processes it is common to express Equation (1.21) as


D~ff = E0J /T (1.23)

in which T is the tortuosity. In terms of Equation (1.21) this parameter is given by

(1.24)

provided Deff is an isotropic tensor.


For many processes of interest, and certainly for our experimental studies, EDeff
is constant and Equation (1.19) takes the form

a(CA)1' _ . l'
E - - - ED eff . V V(CA) . (1.25)
at
Under these circumstances, the skew-symmetric part of Deff makes no con-
tribution to the diffusive flux and the comparison between theory and experiment
needs to be carried out with the symmetric part of D eff . Thus, in our work we
make use of Equation (1.25) and define Deff according to

EDeff = Eqj; [1+ ~ f


l' AYK
~ (nl'Kf + fnl'K) dA]. (1.26)

Before going on to a discussion of our experimental studies and theoretical


calculations for anisotropic systems, we need to review prior studies of isotropic
systems and report our experimental results for a packed bed of glass spheres.

1.1. ISOTROPIC SYSTEMS

It is worthwhile to note that there are no homogeneous, isotropic porous media;


however, there exist media in which the diffusion process is isotropic. By isotropic
one means invariant to any coordinate rotation and by homogeneous one means
invariant to any coordinate transformation. In order for a porous medium to be
isotropic and homogeneous, we require that the geometry, i.e., the position of the
fluid-solid interface, be identical for all observers regardless of their location
(homogeneity) or orientation (isotropy). A little thought will indicate that this
situation is impossible. For example, the porous medium illustrated in Figure 3 is
invariant to translations of the type r = r + Ii but it is not invariant to an arbitrary
translation. Thus, the porous medium is not homogeneous. In addition, the porous
medium illustrated in Figure 3 is invariant to 90° rotations about certain points,
but it is not invariant to an arbitrary coordinate rotation about any point. Thus
the porous medium is not isotropic. On the other hand, the system illustrated in
334 JIN-HWAN KIM ET AL.

----l-- ----r-----n--
I_I I
I f--- ~ ----l

I I
I
I I
I
I I
I ,
___ ~ _____ I______ L __
I

I
1.0

I
I I

I I
I
I
I I
-l--_ _ _ _ _.,;:x

- :_1
+-I - ___ -1_
I --

Fig. 3. Two-dimensional square array as a model of a spatially periodic porous medium.

Figure 3 can be used to solve the closure problem given by Equations (1.15)
through (1.17). When this is done (Ryan, 1984) one finds
Dxy = Dyx = 0, (1.27)
thus, the system is transversely isotropic with respect to the diffusion process. In
Equation (1.27) we have used D xx , D xy , etc. to represent the components of Dell
and we will make use of this convention throughout our discussion. A three-
dimensional array of cubes gives rise to a completely isotropic effective diffusivity
tensor; however, the medium itself is not isotropic but only invariant to certain
translations and rotations. Ryan (1984) has also solved the closure problem for
the system shown in Figure 4, and the values of Dxx and Dyy differ by less than
1% while the off -diagonal components are zero. This means that the process is
essentially isotropic even though the diffusion path in the x-direction is quite
'tortuous' relative to that in the y-direction.
There are a number of unconsolidated porous media that appear to be isotropic
with respect to the diffusion process. This is based on the observation that these
systems provide good agreement with the theoretical values predicted for the
models illustrated in Figures 3 and 4, and with the theoretical values predicted
for a model of randomly overlapping spheres of either uniform or nonuniform
size. We use the word 'appear' here because, aside from our own studies, no
experimental measurements have been carried out which could detect the
presence of anisotropy. In this discussion we have emphasized that there are no
DIFFUSION IN ANISOTROPIC POROUS MEDIA 335

I I
)-----1

I
Fig. 4. Two-dimensional body centered array as a model of a spatially periodic porous medium.

isotropic porous media, but that there may be porous media which are isotropic
with respect to certain transport processes. For convenience we will refer to the
latter as isotropic systems.
Certainly the simplest isotropic system to study experimentally is the packed
bed of glass spheres, and for this system we have results from Hoogschagen
(1955), Currie (1960), and our own work. These results are presented in Table I
with the convention that the y-direction is parallel to the gravity vector. Given
that the beds are packed in an anisotropic force field, it is not impossible that an
anisotropic medium would be generated. However, the results for Dxx and Dyy
are clearly equal, to within experimental error, and we conclude that a packed
bed of uniform spheres is isotropic with respect to the diffusion process. Our
experimental studies were carried out with spheres having a diameter of 3.95 mm
while both Currie and Hoogschagen worked with a range of diameters and
obtained results for mixtures. These experimental results, along with others
obtained by Currie are compared with theoretical results in Figure 5. The
calculations of Ryan (1984) are obviously in excellent agreement with the
experimental data, and this suggests that simple, two-dimensional models can be
used to predict the transport characteristics of isotropic systems. The curve
shown for Weissberg (1963) is an upperbound determined by means of a varia-
tional method and is given by
(1.28)
336 JIN-HWAN KIM ET AL.

Table I. Effective diffusivities for packed beds of spheres

e Material This work Currie Hoogschagen


(diameter) eDyy/ffi eDyy/ffi
eDxx/ffi eDyy/ffi

0.183 mixture 0.101


0.255 mixture 0.150
0.350 mixture 0.230
0.356 mixture 0.220
0.376 5-6mm 0.252
0.377 3.95mm 0.238 0.243
0.383 0.38mm 0.263
0.385 3.95mm 0.246 0.245
(0.248)a (0.254)
0.386 3.95mm 0.258 0.259
0.394 3.95mm 0.257 0.258
(0.260) (0.263)
0.405 0.75-0.80 mm 0.271
0.430 1.0-1.25 mm 0.302

aResults in parentheses were obtained with the use of a magnetic stirrer to insure
uniform concentration in the mixing chambers.

In the region of most interest this result is in excellent agreement with the
experimental data, and as an upperbound it is appropriate that it always predicts
values that are greater than those given by Ryan (1984). The first attempt at a
theoretical prediction of the effective diffusivity is due to Maxwell (1881) who

1.0

This work
• spheres
0.8 Currie
o mixture of spheres
o sand
'" carborundum
.. sodium chloride
Hoogschagen
0.6
() mixture of spheres
'l carborundum

0.4 Maxwell

Wakao and Smith


0.2

0.0
0.0 0.2 0.4 0.6 0.8 10
E

Fig. 5. Effective diffusivities for isotropic systems.


DIFFUSION IN ANISOTROPIC POROUS MEDIA 337
analyzed a dilute suspension of spheres to obtain the result
EDefdr!lJ = E[l +!(1- E)]-I. (1.29)
Although this result was originally considered to be restricted to values of E close
to unity, Hashin and Shtrikman (1962) have shown that it is an upperbound for
any value of E and the comparison shown in Figure 5 certainly confirms this. The
micropore-macropore model of Wakao and Smith (1962) can be used to predict
macropore effective diffusivities and for the constraints associated with Equation
(1.1) their expression reduces to
(1.30)
From Figure 5 we see that this seriously underestimates the effective diffusivity.
From a practical point of view, a reasonable empirical representation of the
experimental data is given by
(1.31)
however, for larger values of E the curve due to Ryan is recommended and
tabulated values are given in the appendix. A crucial aspect of the theoretical
and experimental results presented in Figure 5 is that the effective diffusivity for
isotropic systems is independent of the details of the geometry, i.e., knowledge of
the void fraction is sufficient to predict Deft, the single distinct component of the
effective diffusivity tensor.

1.2. ANISOTROPIC SYSTEMS

While systems that are isotropic with respect to diffusion are readily available in
the laboratory, most natural systems and many processed materials are anisotro-
pic to some degree. The process of extrusion may be the cause of anisotropy in
catalyst pellets (Satterfield, 1970, Sec. 1.8.2) and the experimental results of
Currie (1960) suggest that any kaolin-based catalyst will exhibit anisotropic
behavior. Natural minerals such as clay and sandstone tend to be anisotropic
(Musk at, 1949, Sec. 6.7) as do almost all cellular systems. Wood is a classic
example of an anisotropic organic material and this has important consequences
for the drying process (Spolek and Plumb, 1980). Although numerous anisotropic
systems of practical importance exist, it would appear that none have been
studied experimentally with the objective of determining more than one com-
ponent of the effective diffusivity tensor. In the next section we describe an
experimental system that can be used to determine two components of the
effective diffusivity tensor in transversely isotropic, unconsolidated porous media.

2. Experimental Studies
The experimental apparatus used in this work is essentially identical to that used
by Currie (1960) except for the modifications required for measuring two
338 JIN-HWAN KIM ET AL.

Reference Valve
Chamber (Closed)

Mixing
Chamber

Screen

SECTION A-A SECTION 8-8


Position I Position II

Fig. 6. Experimental equipment for determining two components of the effective diffusivity tensor.

components of the effective diffusivity tensor. A schematic drawing is shown in


Figure 6 which provides the end view of the sample chamber and the two mixing
chambers while the drawing presented in Figure 7 provides an oblique view of
the equipment showing only the fixed sample chamber and the moveable mixing
chambers. The sample chamber had the dimensions of 7.18 x 7.18 x 5.08 cm with
the smaller dimension being perpendicular to the view shown in Figure 6. The
sample chamber was situated in a block of plexiglass that was supported in a
manner that allowed for a 90° rotation, but was otherwise rigid. The mixing
chambers were mounted on a large piece of Plexiglass which could slide along
the fixed block of Plexiglass in which the sample chamber was located. In this
manner, the mixing chambers were individually brought into contact with the
sample chamber in order to generate orthogonal diffusional processes within the
sample chamber. The two sides of the sample chamber that were in contact with
tpe mixing chambers were supported by 150 mesh stainless steel screen that
provided a negligible resistance to diffusion for the process times of interest.
Since the processs operated at constant temperature and pressure, the one-
dimensional diffusion process exhibited equal molar counter diffusion and could
therefore be analyzed in terms of Equation (1.25).
The movable mixing chambers had dimensions of 7.18 x 5.08 x 2.44 cm and,
thus, matched exactly the open surface of the sample chamber. By initiating an
experiment with the less dense gas in the sample chamber, natural convection
currents provided a uniform concentration in the mixing chamber. This was
DIFFUSION IN ANISOTROPIC POROUS MEDIA 339
Mixing Chamber Must be
Moved to Position II for
Measurement of Dyy

Mixing Chamber in
Position I for
Measurement of Dxx

Fig. 7. Oblique view of experimental equipment.

verified both by the form of the experimental measurements of concentration


versus time and by the use of a magnetic stirrer during several of the experimen-
tal runs. The results shown in Figure 1 suggest that natural convection in the
mixing chambers was sufficient to provide uniform concentrations.
While natural convection currents were significant in the mixing chambers,
they were insignificant in the sample chamber when the diffusion process was
parallel to the gravitational field. The disparate length scales for the mixing
chambers and the porous medium in the sample chamber would account for
mixing being significant in one region and insignificant in another; however, this
was not the case when the diffusion process was orthogonal to the direction of
gravity. Experiments under these conditions indicated significant natural con-
vection in the porous medium, as erratic concentration changes were observed in
the mixing chamber and the time required for completion of the transport process
was relatively short. Because of this, the experimental apparatus was always
rotated so that the direction of the diffusion process was parallel to the direction
of gravity.
In order to measure the concentration as a function of time, Gow-Mac (type
9225AA) thermistors were mounted in the two mixing chambers and in a third
reference chamber which is not shown in Figure 6. Each chamber contained a
rotating plexiglass valve so that gas mixtures could flow into and out of the
mixing chambers and the sample chamber. For example, when the apparatus was
in Position I, as illustrated in Figure 6, pure nitrogen could be circulated through
340 JIN-HWAN KIM ET AL.

the side mixing chamber while a 20 mole percent mixture of argon and nitrogen
could be circulated through the top mixing chamber. When steady conditions
were obtained, the sliding mixing chamber assembly was moved into Position II
and the argon would diffuse into the porous medium contained in the sample
chamber. In order to prevent leakage and to provide lubrication for the sliding
mixing chamber assembly, vacuum stopcock grease was applied to the flat contact
surfaces between the sliding assembly and the stationary sample chamber assem-
bly. Leakage was small and could be calibrated by filling the entire system with
an argon-nitrogen mixture and observing the change in argon concentration as
air leaked into the system.
In order to calibrate the thermistors, mixtures of argon and nitrogen were
prepared in a flow system using a soap film meter to measure the volumetric flow
rates of the two gases. The thermistor output was a linear function of the argon
mole fraction for mole fractions less than 0.20 and all experiments were carried
out with dilute solutions of argon and pure nitrogen in order to take advantage of
the linear relation between concentration and the thermistor output. The ther-
mistor outputs from the mixing chambers were always measured relative to the
output of a second thermistor in a reference chamber in order to compensate for
drift in the electronic recording equipment.

2.1. MEDIA MATERIAL

Three types of unconsolidated porous media were studied: glass spheres, mica
particles, and mylar disks. Since reliable theoretical and experimental values of
the effective diffusivity for a packed bed of glass spheres are available, this
system was used as a test of the experimental method and the results have been
presented in Table I. Three different mica samples were used and a single type of
mylar disk was employed in order to produce an anisotropic porous medium. The
void fraction of each medium was measured by weighing a known volume of the
porous medium, saturating it with water and re-weighing. In an attempt to
characterize the mica samples, 60 particles were chosen at random from particles
distributed over a microscope slide. A picture (magnified 50 times) of each
particle was taken, and each enlarged particle was carefully cut away from the
photograph. The weight of these 60 pieces of photographic paper was then
compared to the weight of a rectangular piece of photographic paper in order to
determine the average projected area of the particles. This was used, along with
the average thickness, to characterize the geometry of the mica particles. While
somewhat less elegant than the use of an image analyzer, this method was
reasonably simple and quite inexpensive. An average thickness associated with
the mica particles was measured with a micrometer having an accuracy of 0.0001
inches. The geometrical characteristics of the particles used in this study are
shown in Table II. There we have used Dp to represent the diameter of the glass
spheres, the diameter of the mylar disks, and the 'effective diameter' associated
DIFFUSION IN ANISOTROPIC POROUS MEDIA 341
Table II. Geometrical characteristics of porous media particles

Sample Mesh Diameter Projected Effective Effective Average alb


Dp (mm) area diameter length thickness
A (mm2) Dp (mm) a (mm) b (mm)

glass sphere 3.95 12.25 3.95


mica A -14/+24 0.935 1.076 1.17 1.037 0.072 14.4
micaB -241+32 0.598 0.481 0.783 0.694 0.043 16.1
mica C -32/+42 0.423 0.235 0.547 0.485 0.033 14.7
mylar disk 1.778 2.483 1.778 1.576 0.076 20.7

Fig. 8. In situ photographs of mica particles and mylar disks.


342 JIN-HWAN KIM ET AL.

with the sieves that were used to prepare the three mica samples. From the
numbers listed under Dp it is clear that the mica particles were somewhat smaller
than the mylar disks. This is also seen in the photographs (enlarged 30 times) of
mica particles and mylar disks shown in Figure 8. These are photographs of the
sample chamber taken with a microscope after removing the supporting screen.
The process of removing the screen did not appear to alter the structure of the
porous medium, thus the photographs provide a reliable representation of the
structure of the systems under investigation. The lack of a uniform, sharp focus in
these photographs is caused by the small depth of field inherent in the micro-
scope. The projected area of the particles is also listed in Table II along with an
effective diameter, D~, associated with the projected area. In addition, we have
listed the effective length for a square having the same area as the projected area,
along with the average measured thickness of the mica particles and the mylar
disks. The degree of anisotropy of the particles has arbitrarily been characterized
by a/b.

2.2. EXPERIMENTAL PROCEDURE

The first step in the experimental procedure was the packing of the sample
chamber illustrated in Figure 6. The glass spheres were simply poured into the
chamber until it was full, while the mica particles and mylar disks were added in
small amounts and tamped gently in an effort to prepare a highly layered
structure. Tapping or shaking the bed as the sample chamber was being filled
appeared to reduce the degree of anisotropy. The structure of the media is
illustrated by the photographs shown in Figure 8.
After the sample chamber was packed, the apparatus was set in Position I as
illustrated in Figure 6. The sample chamber was saturated with pure nitrogen
through the side mixing chamber, while the isolated top mixing chamber was
saturated with an argon-nitrogen mixture. Although pressure differences in the
system were negligible relative to one atmosphere, the volumetric flow rates to
the two chambers were always maintained equal in order to eliminate any
possible pressure difference.
When steady-state conditions were attained, the rotating values were closed
and the sliding mixing chamber assembly was moved from Position I to Position
II. The thermistor output was recorded as a function of time on a Tegal Scientific
Model 500 strip-chart recorder. The run time depended on the particular
value of the effective diffusivity, but the data were usually collected within three
to four minutes. Three runs were made for each packing and the average value of
the three measured effective diffusivities is reported in this work. In addition to
using the concentration versus time data to compute effective diffusivities, the
final measured concentration could be used to determine the porosity and thus
confirm the gravimetric determination of the porosity. When the three runs were
DIFFUSION IN ANISOTROPIC POROUS MEDIA 343
completed, the equipment was rotated 90° and the procedure was repeated in
order to determine the second component of the effective diffusivity tensor.

2.3. ANALYSIS OF DATA

The experimental data were analyzed in terms of the one-dimensional form of


Equation (1.25). When the flux is parallel to the x-axis, Equation (1.25) reduces
to

(2.1)

with the obvious variation of this result applying to the case of diffusion in the
y-direction. Here species A refers to the argon in the system, and the initial
condition for the experiment can be expressed as

I.e. (CA)"!' = 0, ° ~x~ L, (2.2a)


(CA)"Y = CO, X = L. (2.2b)
The assumption of uniform concentration in the mixing chamber leads to a flux
boundary condition of the form

B.e.1 h a(CA)"Y = -ED a(CA)"Y x= L (2.3)


at xx ax '
while the impermeable surface at the bottom of the sample chamber yields

B.e.2 a(CA)"Y = 0, x = 0. (2.4)


ax
The solution to this boundary value problem is given by Carslaw and Jaeger
(1959, Sec. 3.13) and it can be expressed as

C =_1_+ nf'
2H exp(-A~t*) cos(An X )
(2.5)
1 + H n=O (A~ + H2 + H) cos(An)
in which the various dimensionless quantities are given by
H= EL/h,
The eigenvalues are determined from the equation
An cot An + H = 0, n = 0, 1, 2 ... (2.6)
and the dimensionless concentration in the mixing chamber is

(2.7)

When A~t* is large compared to one, only the first term of the series is significant
344 JIN-HWAN KIM ET AL.

and the concentration in the mixing chamber can be expressed as

( ) 1
Cx=1=--+ ( 2H ) e-1I.2t*
o. (2.8)
1+ H Aij+ H2+ H
Under these circumstances, a plot of the natural logarithm of C(X = 1) versus t*
yields a straight line of slope - DxxAijl L 2, and from this one obtains experimental
values of Dxx. In this work, Equation (2.8) was used to obtain the first estimate of
Dxx and then a curve fitting procedure was used to determine the final value from
Equation (2.7). The difference between the value of Dxx based on Equation (2.8)
and the final value based on Equation (2.7) was about 10%. For times such that
Aijt*}> 1, either Equation (2.7) or (2.8) reduces to

(2.9)

and this result could be used to determine the void fraction which was within 8%
of the values determined gravimetrically.
The measured values of Dxx and Dyy are listed in Table III for the anisotropic
systems studied in this work. The values are made dimensionless by the binary
molecular diffusivity for the argon-nitrogen system. At one atmosphere and
25°C, the Chapman-Enskog formula (Bird et aI., 1960, Sec. 16.4) yields a value
of '3J equal to 0.194 cm2 Is.

3. Comparison of Theory and Experiment


On the basis of the geometrical information about the mica particles and the
mylar disks given in Table II and the photographs shown in Figure 8, it is clear
that the 'isotropic' models shown in Figures 3 and 4 should be replaced with
something comparable to the structures shown in Figure 9. The disparate length

••_.!------
___,__".Ib y

x:

T
. . . . . . ._ ._ _• • L

._.I-;.;;;;..a;;;;--.1
b

t---La --1
._.1
E = 0.75 La ILb = 1 a! b = 15
Fig.9(a).
DIFFUSION IN ANISOTROPIC POROUS MEDIA 345

t ..
, x
1.
t
b
!

,
!

!
!
--I
! Lb
1-: - a ---f~
__ 1
I--L ---.!
a

f =0.85 L/Lb =1 a I b =15

~ ....... n.. zn .r~~


,
I
:-------------------:
I
... n ·T...... n.. n.nj
-------------~
I
:-------------------: f
: I:b
I

)
, __________________ I
~-----------------~ : Ib
t __ - - - - - - - - - - - - - - - - 1
I I I
:___________________ 1 • :I

I - i.
- . - - - La - - _.....1 1-1....
---

E= 0.75 a / b = 15

..• -

-
-
i~
r
1
.- x
--

-
..
Af
~
Ib
T
h
a

f= 0.50 a / b = 15
Fig. 9. Geometrical models of anisotropic porous media.
346 JIN-HW AN KIM ET AL.

scales associated with these structures cause difficulties with the numerical
solution of Equations (1.15) through (1.17), and a detailed discussion of the
numerics is given in the appendix along with tabulated values of Dxx and D yy •
The spatially periodic porous media illustrated in Figures 3, 4, and 9 can be
described in terms of the void fraction, the particle geometry, and the lattice
vectors. In this presentation we consider only a body-centered model of a
spatially periodic porous medium and we describe the geometry in terms of the
void fraction, the particle geometry and the ratio of the lengths of the sides of a
unit cell, La/ Lb. From the models illustrated in Figure 9, it is apparent that a
wide range of configurations can be developed by changing the value of La / Lb.
While theory and experiment indicate that isotropic systems are insensitive to the
details of the geometry, it seems clear that this is not the case for anisotropic
systems. Our success with isotropic systems led us to believe that anisotropic
media could be modeled in terms of €, a / b and some choice of La / Lb (or the
lattice vectors Ii) that would produce good agreement between the theory and the
two experimentally determined components of Dell. That this is not the case is
clearly demonstrated in Figures 10 and 11 where we have shown calculated and
measured values of €Dxx /1}]) and €D yy /1}]) as a function of the porosity. The value
of a/ b of 15. is in the proper range for our mica particles and mylar disks;
however, we have no geometrical information about the vermiculite and mica
used by Currie (1960). While the presentation of theoretical results in Figures 10
and 11 is not exhausitive, it does indicate that one cannot find a single value of

1.0 , - - - - - - - - - - - - - - - - - -.....


Present Work

••• .i. Mylar Disk


V Mica A
0.8
••• 6. Mica B
o Mica C

0.6
y

t.:x • 'b La
Lb Currie's Work
o Vermiculite
• Mica

a
-= 15
b

0.4

0.2

0.2 0.4 0.6 0.8 1.0


E
Fig. 10. Effective diffusivities normal to the bedding plane.
DIFFUSION IN ANISOTROPIC POROUS MEDIA 347
1.0

0.8
-
•••
• • •.b
• Mylar Disk
V Mica A
Il. Mica B
D Mica C

E"Du
.2l
0.6
y

Ux
-

a
b
= 15
- a
La
Lb

0.4

0.2

0.2 0.4 0.6 0.8 1.0


E

Fig. 11. Effective diffusivities parallel to the bedding plane.

La I Lb that gives good agreement between theory and experiment. From the
photographs shown in Figure 8, it appears that groups of particles are packed
together in a manner considerably different than the models illustrated in Figure
9. This is especially clear in the photograph of the mylar disks and it seems
possible that the 'effective' value of alb is different than the value given for an
individual particle in Table III. For this reason, it seemed reasonable to explore a
range of values of alb and these results are shown in Figure 12 for a single value
of La I Lb = 1. Once again we see that there is no single value of alb that gives
good agreement between theory and experiment.
While the use of adjustable parameters to obtain agreement between theory
and experiment is generally of little value, it is of some interest in this case to
adjust alb as a function of E in order to fit the experimental data. This fit is
illustrated by the dashed line in Figure 12 and this provides us with 'experimental
values' of alb as a function of the porosity, E. When this functional dependence
of the geometry on the porosity is used to determine D xx , we obtain the
comparison illustrated in Figure 13 in which we have used values of the
parameter n ranging from n = 1.0 for E~ 1.0 to n = 15 for E~O.O. Here we can
see that even with adjustable values of alb, we cannot fit the experimental data
for Dxx and Dyy using models of the type shown in Figure 9. Similar conclusions
are reached if one chooses La I Lb as the adjustable parameter, and it is clear that
more of the details of the geometry of anisotropic systems must be incorporated
348 JIN- HW AN KIM ET AL.

Table III. Effective diffusivities for anisotropic systems

Packing material e eDxx/qj} eDyy/qj} Dxx/Dyy

mica A 0.612 0.376 0.114 3.30


0.660 0.420 0.112 3.75
0.710 0.475 0.127 3.74
mica B 0.656 0.385 0.114 3.38
0.714 0.428 0.122 3.51
0.729 0.410 0.158 2.59
0.736 0.499 0.167 2.99
mica C 0.676 0.426 0.105 4.06
0.715 0.447 0.125 3.58
0.747 0.509 0.137 3.72
mylar disk 0.549 0.372 0.092 4.04
0.570 0.397 0.098 4.05
0.593 0.419 0.112 3.74
0.620 0.431 0.117 3.68

into the theoretical calculations if a reliable predictive theory is to be developed.


This will be the objective of subsequent studies.
Directing our attention solely to the experimental data, we note that the values
of EDxx /r:!iJ for anisotropic systems are only 10 to 20% lower than Ryan's curve
for isotropic systems shown in Figure 5. This suggests, as a first approximation,

1.0 r----....,.-----r---,.----,----,

Present Work
... Mylar Disk
0.8 "" Mica A
I>. Mica B
o Mica C
Currie's Work
o Vermiculite
0.6 • Mica

EDyy

'J) La
- = 1
0.4 Lb

0.2

0.2 0.4 0.6 08 1.0


E

Fig. 12. Influence of effective particle geometry on the effective diffusivity normal to the bedding
plane.
DIFFUSION IN ANISOTROPIC POROUS MEDIA 349
1.0 I I I I ./
/
/
Present Work /
... Mylar Disk /
-
0.8 l- v Mica A /
h. Mica B /
0 Mica C /
/
0.6 - ;
/ -
fD xx ~ =1 I lfl
Lb I v
T i ...... ;p %
0.4 - ~ =15 n- 1 i ...... V h.
-
b

/
0.2 - I -
/
/'
./
.-
I

0.2 0.4 0.6 0.8 1.0


f

Fig. 13. Influence of effective particle geometry on the effective diffusivity parallel to the bedding
plane.

one could use Ryan's curve to estimate the effective diffusivity parallel to the
layered structure. We express this idea as
(3.1)
The experimental data also indicate that Dyy is about one fourth of D xx , thus we
could estimate the effective diffusivity normal to the layered structure as
(3.2)
In using Equations (3.1) and (3.2), we have in mind that EDeff/r:!iJ is to be taken
from Ryan's curve in Figure 5 which is quite close to the curve identified by
La /Lb = 15 in Figures 10 and 11.

4. Conclusions
An experimental method for measuring components of the effective diffusivity
tensor has been developed, and results have been obtained for transversely
isotropic, unconsolidated porous media. The experimental results are in good
agreement with other studies, and the theoretical calculations indicate that the
effective diffusivity tensor cannot be accurately predicted in terms of only the
void fraction and the particle geometry. The comparison between theory and
experiment clearly indicates that the details of the geometry of anisotropic
systems exert an important influence on effective transport coefficients.
350 JIN-HWAN KIM ET AL.

Appendix: Numerical Calculations


In this appendix we discuss the solution of the closure problem given by
Equations (1.15) through (1.17) and list some of our numerical results. A generic
unit cell is illustrated in Figure 14, and for such a unit cell one can show (Ochoa,
1987) that the periodicity condition given by Equation (1.17) leads to the
following boundary conditions for fx and fy .
Problem for fx

~ymmetry condition: afx=o


ay °
, y = ,± Lb /2 , (AI)

Skew-symmetry condition: fx = 0, x = 0, ± La, (A2)

Problem for fy

Symmetry condition: afy


ax = ° °
,at x = ,± L a, (A3)

Skew-symmetry condition: fy = 0, at y = 0, ± 4 /2. (A4)

With these symmetry and skew-symmetry conditions, the closure problem can be
solved in one quarter of the original unit cell shown in Figure 14. This reduced
domain and its boundary r are shown in Figure 15.
For this geometry the non-zero components of the effective diffusivity tensor

I-01 --a/2
........
~I 1""'...
1- - - - a/2 ------'~~I

fb

Fig. 14. Unit cell of an anisotropic, spatially periodic porous medium.


DIFFUSION IN ANISOTROPIC POROUS MEDIA 351
y

Boundary of
Domain, r

Fig. 15. One quarter of a unit cell.

given by Equation (1.26) can be expressed as


EDxxl'2iJ = e - 2(fx)b I LaLb, (AS)

eDyy l'2iJ = E- 2(fy)alLaLb. (A.6)

Here (fx) indicates the interfacial average of fx at x = a12, and (fy) represents the
interfacial average of fy at y = b 12. Note that in obtaining Equations (A.5) and
(A6) we have used the skew-symmetry of fx and fy about the x-axis and y-axis
respectively.
In addition to the conditions given by Equations (AI) through (AA), we also
require the boundary conditions at the Y-K interface, and for the system
illustrated in Figure 15 these take the form

Problem for fx

afx =-1 x = a!2, (A7)


ax '
afx = +1 x = La - a12, (A.8)
ax '
afx =0 y = b12, (Lb - b)/2. (A9)
ay ,
352 JIN-HWAN KIM ET AL.

Problem for fy

;;=+1, y=b/2, (A.lO)

~=-1, y=Lb -b/2, (A. 11)

afy _ _
- - 0, x - a /2, La - a /2. (A. 12)
ax
Determination of fx and fy requires the solution of Equation (1.15) subject to the
boundary conditions given by Equations (A. 1) through (A.4) and Equations (A.7)
through (A.12).
From Equations (A.5) and (A.6) we see that we need only the values of fx and
fy at the 'Y-K interface, and this fact encourages the use of the boundary element
method (Brebbia and Walker, 1980). This method yields the interfacial value of
the field f (here f represents fx or fy) and the flux of f given by
(A. 13)
while avoiding the calculation of either the field or the flux at interior points.
In the use of the boundary element method, we assumed the interfacial values
of f and q to be constant on each element resulting from the discretization of the
boundary f illustrated in Figure 15. In our calculations, the code given by
Brebbia (1978) was used with only a few changes. The most important of these
was the addition of a subroutine to assign the coordinates and the boundary
condition to each of the elements on the boundary, f. It is worthwhile to note
that this approach to the solution for fx and fy requires that the derivative
boundary conditions given by Equations (A.1) and (A.3) and Equations (A.7)
through (A.12) must take the form
Problem for fx
q=O, y = 0, Lb/2, (A. 14)
q = +1, x = a/2, (A.15)
q = -1, x=La -a/2, (A. 16)
q=O, Y = b/2, (Lb - b)/2. (A.17)
Problem for fy
q=O, x = 0, La, (A. 18)
q=+l, Y = b/2, (A. 19)
q=-l, y=Lb-b/2, (A. 20)
q=O, x = a/2, La - a/2. (A.21)
DIFFUSION IN ANISOTROPIC POROUS MEDIA 353
Table IV. Comparison between finite
difference calculations and boundary
element calculations

E EDxx/0J

BEM FDM

0.36 0.213 0.212


0.51 0.326 0.327
• 0.64 0.451 0.449
0.75 0.583 0.576
0.84 0.712 0.703
0.91 0.828 0.820

The HEM utilized 256 elements while


the FDM utilized a regular grid of 60 x
60 points. La = Lb and a = b.

With these boundary conditions, we have solved Equation (1.15) to produce the
Ix and Iy fields and, in turn, the values of eDxx/gj) and eDyy/gj). Ryan et al. (1981)
have produced values of these quantities for La = Lb and a = b using a finite
difference method (FDM), and their results are compared with our boundary
element method (BEM) calculations in Table IV. Obviously the two methods are
in good agreement for this particular geometrical model.
In addition to our boundary element calculations, we also used the finite
element method (FEM) to solve the closure problem. Square elements and linear
shape functions were used in our finite element calculations, and a comparison
with the boundary element calculations is given in Table V. Here we see

Table V. Comparison between finite element cal-


culations and boundary element calculations

E EDxx/0J EDyy/0J

HEM FEM HEM FEM

0.19 0.114 0.109 0.100 0.102


0.36 0.240 0.235 0.200 0.205
0.64 0.526 0.523 0.403 0.411
0.70 0.600 0.596 0.456 0.462
0.74 0.652 0.651 0.495 0.503
0.76 0.678 0.680 0.515 0.530
0.84 0.788 0.788 0.608 0.622
0.91 0.884 0.886 0.711 0.725

The HEM utilized 256 elements while the FEM


made use of 448 elements. La = 15Lb and a =
15b.
354 JIN-HWAN KIM ET AL.

Table VI. Calculated effective diffusivities using the boundary element method

~ alb LalLb ~Dxxl'ZiJ ~Dyyl'ZiJ ~ alb LaILb ~Dxxl'ZiJ ~Dyy!'ZiJ

0.74 15 1 0.741 0.028 0.70 1.5 1 0.585 0.436


0.80 15 1 0.787 0.116 0.75 1.5 1 0.642 0.506
0.88 15 1 0.870 0.225 0.80 1.5 1 0.704 0.584
0.91 15 1 0.902 0.301 0.85 1.5 1 0.770 0.670
0.36 15 10 0.348 0.022 0.90 1.5 1 0.842 0.766
0.50 15 10 0.458 0.138 0.95 1.5 0.919 0.875
0.64 15 10 0.585 0.272 0.30 3 0.106 0.017
0.70 15 10 0.647 0.338 0.40 3 0.226 0.050
0.76 15 10 0.713 0.410 0.50 3 0.356 0.094
0.84 15 10 0.806 0.524 0.60 3 0.494 0.161
0.91 15 10 0.892 0.650 0.65 3 1 0.562 0.209
0.36 15 5 0.162 0.004 0.70 3 1 0.627 0.270
0.50 15 5 0.382 0.007 0.75 3 1 0.689 0.345
0.64 15 5 0.606 0.027 0.80 3 1 0.748 0.434
0.70 15 5 0.685 0.085 0.85 3 1 0.807 0.541
0.76 15 5 0.742 0.185 0.90 3 0.868 0.667
0.84 15 5 0.822 0.347 0.95 3 1 0.933 0.817
0.91 15 5 0.898 0.529 0.25 5 1 0.248 0.022
0.19 15 15 0.114 0.100 0.30 5 1 0.279 0.036
0.36 15 15 0.240 0.200 0.40 5 0.349 0.059
()'.64 15 15 0.526 0.403 0.50 5 0.438 0.086
0.70 15 15 0.600 0.456 0.60 5 1 0.533 0.124
0.74 15 15 0.652 0.495 0.65 5 1 0.588 0.151
0.76 15 15 0.678 0.515 0.70 5 1 0.645 0.186
0.84 15 15 0.788 0.608 0.75 5 1 0.705 0.236
0.91 15 15 0.884 0.711 0.80 5 0.764 0.309
0.05 1 1 0.070 0.025 0.85 5 0.822 0.414
0.15 0.116 0.080 0.90 5 0.880 0.558
0.25 0.168 0.139 0.95 5 0.939 0.749
0.35 0.227 0.205 0.50 7.5 1 0.499 0.034
0.40 0.260 0.241 0.60 7.5 1 0.570 0.087
0.50 1 0.336 0.321 0.65 7.5 1 0.615 0.111
0.60 1 0.427 0.412 0.70 7.5 0.664 0.139
0.65 1 0.478 0.464 0.75 7.5 0.717 0.175
0.70 1 1 0.535 0.519 0.80 7.5 0.773 0.227
0.75 1 0.596 0.581 0.85 7.5 1 0.829 0.311
0.80 0.664 0.648 0.90 7.5 1 0.886 0.451 i

0.85 0.738 0.722 0.95 7.5 1 0.942 0.673


0.90 1 0.819 0.805 0.76 15 0.754 0.068
0.95 1 0.907 0.897 0.80 15 0.788 0.116
0.5 1.5 0.380 0.197 0.84 15 0.827 0.162
0.6 1.5 0.478 0.310 0.88 15 0.870 0.225
0.65 1.5 0.530 0.371 0.91 15 0.902 0.301

excellent agreement between these two methods for a highly anisotropic medium
(La = 15Lb and a = 15b) whereas the finite difference calculations gave poor
results for this geometry (Ochoa, 1987).
The results presented in the body of this paper were obtained using the
boundary element method because the input data were much simpler to handle
DIFFUSION IN ANISOTROPIC POROUS MEDIA 355

than those for the finite element method. In most cases, the boundary element
method was able to provide convergent results for Dxx and Dyy; however, when
the distance between solid particles becomes small relative to La or Lb the
problem becomes more complex and more elements are needed. Since the values
of Dxx and Dyy are known to be zero in certain cases, we could interpolate
between converged values in order to estimate values of Dxx and Dyy in regions
where it was difficult to obtain calculated values. This is the origin of the dashed
line in Figure 11 for a value of La/Lb = 10.
The full range of calculated values of Dxx and Dyy for 1 ~ La I Lb ~ 15 and
1 ~ alb ~ 15 are given by Ochoa (1987) and in Table VI we have listed the
• values used in the preparation of Figures 10 through 13 .

Acknowledgement
This work was supported by NSF Grant CPE 83-08461 (J-H.K., J.A.O., and
S.W.). In addition, the support of Chonnam National University (J-H.K.) and the
Consejo Nacional de Ciencia y Tecnologia from the Mexican Government
(J.A.O.) is gratefully acknowledged.

References
Anderson, T. B. and Jackson, R, 1967, A fluid mechanical description of fluidized beds, Ind. Eng.
Chem. Fundam. 6, 527-533.
Batchelor, G. K., 1974, Transport properties of two-phase materials with random structure, Ann.
Rev. Fluid Mech. 6, 227-255.
Bensoussan, A., Lions, J-L., and Papanicolaou, G., 1978, Asymptotic Analysis for Periodic Structures,
North-Holland, New York.
Bhatia, S. K. and Perlmutter, D. D., 1983, Unified treatment of structural effects in fluid-solid
reactions, AIChE J. 29, 281-289.
- Beran, M. J., 1968, Statistical Continuum Theories, Interscience, New York.
Brebbia, C. A., 1978, The Boundary Element Method for Engineers, Wiley, New York.
Brebbia, C. A. and Walker, R., 1980, Boundary Element Techniques in Engineering, Newnes-
Butterworth, London.
Brenner, H. and Adler, P., 1987, Transport Processes in Spatially Periodic Porous Media, Hemisphere,
New York.
Carbonell, R G. and Whitaker, S., 1984, Heat and mass transport in porous media, in J. Bear and M.
Y. Corapcioglu (eds.), Mechanics of Fluids in Porous Media, Martinus Nijhoff, Dordrecht.
Carslaw, H. S. and Jaeger, J. C., 1959, Conduction of Heat in Solids, OUP, Oxford.
Chang, H-C., 1982, Multiscale analysis of effective transport in periodic heterogenous media, Chem.
Engng. Commun. 15, 83-91.
Crapiste, G. H., Rotstein, E., and Whitaker, S., 1986, A general closure scheme for the method of
volume averaging, Chem. Engng. Sci. 41, 227-235.
Currie, J. A., 1960, Gaseous diffusion in porous media. Part I - a non-steady state method, Brit. J.
Appl. Phys. n, 314-324.
Cussler, E. L., 1984, DiffUSion and Mass Transfer in Fluid Systems, Cambridge University Press.
Dullien, F. A. L., 1979, Porous Media: Transport and Pore Structure, Academic Press, New York.
Foster, R N. and Butt, J. B., 1966, A computational model for the structure of porous materials
employed in catalysis, AIChE J. 12, 180-185.
356 JIN-HWAN KIM ET AL.

Gavalas, G. R. and Kim, S., 1981, Periodic capillary models of diffusion in porous solids, Chern.
Engng. Sci. 36, 1111-1122.
Hashim, Z. and Shtrikman, S., 1962, A variational approach to the theory of effective magnetic
permeability of multiphase materials, 1. Appl. Phys. 33, 3125-3131.
Hoogschagen, J., 1955, Diffusion in porous catalysts and adsorbents, Ind. Eng. Chern. 47, 906-913.
Jackson, R., 1977, Transport in Porous Catalysts, Elsevier, New York.
Johnson, M. F. and Stewart, W. E., 1965, Pore structure and gaseous diffusion in solid catalysts, 1.
Catalysis 4, 248-252.
MarIe, C. M., 1967, Ecoulements monophasiques en milieu poreux, Rev. Inst. Francais du Petrole
22(10), 1471-1509.
Mason, E. A. and Malinaukas, A P., 1983, Gas Transport in Porous Media: The Dusty Gas Model,
Elsevier, New York.
Maxwell, J. C., 1881, Treatise on Electricity and Magnetism, Vol. I, 2nd edn., Clarendon Press,
Oxford. ..
Muskat, M., 1949, Physical Principles of Oil Production, McGraw-Hill, New York.
Ochoa, J. A, 1987, PhD thesis, Department of Chemical Engineering, University of California at
Davis.
Ryan, D., Carbonell, R. G., and Whitaker, S., 1980, Effective diffusivities under reactive conditions,
Chern. Engng. Sci. 35, 10-15.
Ryan, D., Carbonell, R. G., and Whitaker, S., 1981, A theory of diffusion and reaction in porous
media, in P. Sfroeve and W. J. Ward (eds.), AIChE Symposium Series, No. 202, Vol. 77, pp. 46-62.
Ryan, D., 1984, The theoretical determination of effective diffusivities for reactive, spatially periodic,
porous media, MS thesis, Department of Chemical Engineering, University of California at Davis.
Satterfield, C. M., 1970, Mass Transfer in Heterogeneous Catalysis, MIT Press, Cambridge, Mass.
Slattery, J. C., 1967, Flow of viscoelastic fluids through porous media, AIChE 1.13, 1066-1071.
Smith, D. M. and Williams, F. L., 1984, Diffusional effects in the recovery of methane from coal beds,
SPE 1. 24, 529-535.
Spolek, G. A and Plumb, O. A, 1980, A numerical model of heat and mass transport in wood during
drying, in A S. Mujumdar (ed.) Drying '80, Vol 2, Hemisphere, New York.
Strieder, W. and Aris, R., 1973, Variational Methods Applied to Problems of Diffusion and Reaction,
Springer-Verlag, New York.
Tinker, P. B., 1970, Some problems in the diffusion of ions in soils, in J. G. Gregory (ed.) Sorption and
Transport Processes in Soils, S.c.1. Monograph No. 37, Society of Chemical Industry, London.
Wakao, N. and Smith, 1. M., 1962, Diffusion in catalyst pellets, Chern. Engng. Sci. 17, 825-834.
Webman, I., 1982, Macroscopic properties of disordered media, in Lecture Notes in Physics 154,
297-303.
Weissberg, H. L., 1963, Effective diffusion coefficients in porous media, 1. App/. Phys. 34,
2636-2639. f
Whitaker, S., 1967, Diffusion and dispersion in porous media, AIChE 1. 13,420-427.
Whitaker, S., 1977, Simultaneous heat, mass and momentum transfer in porous media: a theory of
drying, Advances in Heat Transfer 13, 119-203. •
Whitaker, S., 1986, Transport processes with heterogeneous chemical reaction, in A. E. Cassano and
S. Whitaker (eds.), Concepts and Design of Chemical Reactors. Gordon and Breach, New York.

You might also like