0 ratings0% found this document useful (0 votes)
222 viewsThe Method of Volume Averaging PDF
Uploaded by
Eduardo De LeónCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0 ratings0% found this document useful (0 votes)
222 viewsThe Method of Volume Averaging PDF
Uploaded by
Eduardo De LeónCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 235
The Method
of Volume —
Averaging
Stephen Whitaker eit
GEM SOTA AOA
£10 SHOLIVONIATY ONY AIOE, The Method of
Volume Averaging
by
Stephen Whitaker
Department of Chemical Engineering,
University of California at Davis,
California, U.S.A.
KLUWER ACADEMIC PUBLISHERS
DORDRECHT / BOSTON / LONDON ACLLP. Catalogue record for this book is available from the Library of Congress.
ISBN 0-7923-5486-9
Published by Kluwer Academic Publishers,
P.O. Box 17, 3300 AA Dordrecht, The Netherlands.
Sold and distributed in North, Central and South America
by Kluwer Academic Publishers,
10{ Philip Drive, Norwell, MA 02061, U.S.A.
In all other countries, sold and distributed
by Kluwer Academic Publishers,
P.O. Box 322, 3300 AH Dordrecht, The Netherlands.
Printed on acid-free paper
All Rights Reserved
©1999 Kluwer Academic Publishers
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic ot mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner
Printed in the Netherlands. For
Suzanne CONTENTS
Preface xi
Overview xiii
1. Diffusion and Heterogeneous Reaction in Porous Media i
1.1 Governing Equations and Boundary Conditions bocce eee eed
(isl Boundary conditions 9 CGS
1.2 Spatial Smoothing see bees si 6
1.2.1 Spatial Averaging theorem ©... eee ee ee ee MO
1.2.2 Spatial decomposition ©... 0.2... eee ee eee wee 4
1.3 Length-Scale Constraints
1.3.1 Taylor series expansion . .
1.3.2 Order of magnitude estimates... 2...
1.3.3 Area averaged concentration
1.3.4 The filter .
1.4 The Closure Problem . .
1.4.1 Sources
1.4.2 Simplifications . .
1.4.3 The local problem
1.4.4 Closure variables...
1.4.5 Closed form .
VoisymmericUniGclle
1.6 Comparison Between Theory and Experiment .
1.6.1 Experimental studies
1.6.2 Isotropic systems . . . .
1.6.3 Maxwell’s closure problem . .
1.6.4 Role of the filter
1.6.5 Anisotropic systems
boo ccc 54
1.7 Conclusions . .
1.8 Problems... . . . se ce _ =
vii viii Contents
2. Transient Heat Conduction in Two-Phase Systems 2,
2.1 Governing Equations and Boundary Conditions... . . : 2. .-T2
2.2 Spatial Smoothing . . ook
2.2.1 The filter oe 7B
2.2.2 Two-equation model . . . 719
2.3 Local Thermal Equilibrium 6... ees cee eee ee 22. 79
2.3.1 Estimate of ()®-(Ty)° vse ee eee fe 62
232 Comparison with experimen . 89
QAClome
2.41 Local thermal equilibrium ...........-
2.4.2 Local closure problem»... 61.2% 100
2.4.3 Closure variables 2... 2 ee ee - 102
2.4.4 Arbitrary unit cells. E103
2.4.5 Symmetric unit cells. .. 105
2.4.6 Closed form... .. . 106
2.5 Comparison Between Theory and Experiment... 0. 2-2-2 s see eee 107
2.5.1 Maxwell’s closure problem... ... - CC
2.6 Conclusions ........ boot se 20 cccocote .il4
See rstsrsrr——C at Ae (1.2-2)
B.C.2 Cay = F(t), a he (1.2-3)
Le. cay = He), att=0 (1.2-4)
in which @, is the mixture diffusivity defined by Eq. 1.1-15. In this formulation, we
have used %, to represent the entrances and exits of the -phase at the boundary of the
macroscopic region illustrated in Figure 1.2, and we have used 2%. to represent the
averaging
volume, ¥
Figure 1.2, Macroscopic region and averaging volume
entire interfacial area within that region. It should be clear that, in general, the point
concentration will not be known at of, and Eq. 1.2-3 serves primarily as a reminder of
what we do not know about ¢,y at o%, rather than what we do know. The same can be
said about the initial condition for c4, since we will usually know something about the 8 Chapter 1
average concentration at t = 0, but we will rarely know anything about the details of the
point concentration.
If the information represented by Eqs. 1.2-3 and 1.2-4 were available to us, and if
we knew the position of the -« interface, we could solve the boundary value problem in
order to determine cq, asa function of position and time (see Problem 1-4). We could
then calculate the rate of reaction of species A per unit volume of the porous medium and
make use of this quantity in the design of a catalytic reactor. Even if we could solve this
boundary value problem exactly, we would not do so since the solution would yield
more information than we need. For design purposes, it is sufficient to determine the
average concentration and the average rate of reaction associated with the averaging
volume % illustrated in Figure 1.2. The details of this averaging volume are presented
in Figure 1.3 where we have used the position vector ry to locate any point in the
‘ephase. The position vector x locates the centroid of the averaging volume which may
lie in either the y-phase or the x-phase. The relative position vector y,, is used to locate
points in the y-phase relative to the centroid of %”.
Figure 1.3. Averaging volume for the 7 system
The process of spatial smoothing begins by associating with every point in space an
averaging volume which is invariant with respect to time and space. For the two-phase
system under consideration we can express the averaging volume as
Y = Vy(x) + Velx) (1.25)
Here V(x) represents the volume of the -phase contained within the averaging volume,
and the volume fraction of the y-phase (usually identified as the porosity) is given by Diffusion and Reaction 9
ef, = V,00/% (1.2-6)
In the method of volume averaging one encounters many measures of concentration and
it is necessary to identify each of them in some precise manner. We begin by defining
the superficial average of cy, according to
7
cal, = $e Jew
Vy (x)
dv a
Ixtyy
This definition clearly indicates that the average concentration is associated with the
centroid of the averaging volume, indicated by the position vector x in Figure 1.3, while
integration over V, (x) is performed in terms of the components of the relative position
vector yy. In general, we will avoid the precise notation used in Eq, 1.2.7 and simply
express the superficial average as
i :
fem) = Ffenav (1.28)
Yy
‘The superficial average concentration is not the preferred dependent variable since it is
not a good representation of the concentration in the phase. For example, if ¢4y were a
constant given by cy, the superficial average concentration would not be equal to cy.
Because of this, the preferred dependent variable is the intrinsic average concentration
which is defined by
L
te = Jenl, wv (1.2-9)
mH) d en
‘The relation between these two concentrations is given by
Keay) = by Cay)” (1.2-10)
where it is understood that both are evaluated at x. If one is not careful about defining
average values, it is possibie to make errors on the order of
Even though (c4,)" is the preferred dependent variable, it is convenient to begin the
process of spatial smoothing with the superficial average. Thus we integrate Eq. 1.2-1
over the domain V,, associated with the point located by x and divide by 2% to obtain
Lf 28ay 1
> dv =
al ar y
%
Since V, is not a function of time, we can use the general transport theorem (Whitaker,
1Vc 4, )aV (12-11)
1981) to express the left hand side of this result as 10 Chapter 1
djl d
- ay Jenen wv Gea (12-12)
Vy)
a 2a
ww
Yoo
Because the superficial average concentration is associated the fixed point x, it is
appropriate to make use of the partial derivative and express Eq. 1.2-11 as
BHCay)
or
Ikty,
(¥-(DVeny)) (1.2-13)
Given that V,, is independent of time, Eq. 1.2-6 indicates that €,, is also independent of
time and we can use Eq, 1.2-10 in Eq. 1.2-13 to obtain
ey)”
ot
(WV (@Veay)) (12-14)
Obviously the left hand side of this result is in its final form, i.c., the accumulation of
species A is expressed in terms of the intrinsic average concentration, It is of some
importance to understand that Eq. 1.2-14 is a superficial average transport equation, i.e.,
the term on the left hand side represents the accumulation of species A per unit volume
of the porous medium and not per unit volume of the fluid phase.
1.2.1 SPATIAL AVERAGING THEOREM
It should be clear that we need to interchange differentiation and integration on the right
hand side of Eq. 1. 2-14 in order to express the diffusive flux in terms of (c4,)'. This
will be accomplished by means of the spatial averaging theorem (see Problems 1-5
through 1-7). Given some quantity wy associated with the y-phase, we can express the
spatial averaging theorem for the 7-x systert as
1
Wy, = Vy) + y Jrewya (1.2-15)
bre
in which A,, represents the area of the 7-« interface contained within”. This theorem
represents a three-dimensional version of the Leibniz rule for interchanging
differentiation and integration, It was derived independently by Marle (1967) in a study
of single-phase flow in porous media, by Anderson and Jackson (1967) in a study of
fluidized beds, by Slattery (1967) in a study of two-phase flow in porous media, and by
Whitaker (1967) in a study of dispersion in porous media, Since 1967 other derivations
have been presented by Whitaker (1969), Drew (1971), Bachmat (1972), Gray and Lee
(1977), Whitaker (1985), Howes and Whitaker (1985), Gray et al. (1993), and Quintard
and Whitaker (1994b). When the macroscopic process is one-dimensional, a special
form of Eq. 1.2-15 is appropriate (see Problem 1-8). The averaging theorem can be Diffusion and Reaction ul
extended to include systems with many phases (see Problem 1-9) and the form for three-
phase systems is essential for the study of drying (see Problem 1-10),
While we will repeatedly use the spatial averaging theorem in the form given by
Eq. 1. 2-15, it is of some value to list a more explicit form of this theorem given by
Wy) = Vey, + x foes (12-16)
Aye (x)
Here we have used the notation illustrated in Figure 1.3 in order to be precise about the
area integral, and in order to complete out understanding of this theorem we also need to
be precise about the definitions of the two gradient operators. To be clear about those
operators, we consider the curve, ry(3), that lies entirely in the y-phase 2s illustrated in
dA
Ixty,
Figure 1.4. The directional derivative of yy associated with that curve is defined by
(Stein and Barcellos, 1992)
[wyir(s+As)] - wylr,
By ti |Meat 49) — Volt OD (2-17)
ds aro ra
and the unit tangent to the curve is expressed as
A= (1.2-18)
On the basis of Eqs. 1.2-17 and 1.2-18 we define the gradient of Wy according to
(Hildebrand, 1976)
re
= —.0 1.2-19,
7 a (1.2-19)
In order to be clear about the gradient of the average, we consider the curve, x(s),
illustrated in Figure 1.5 that locates points in both the y and «-phases. For that curve, the
directional derivative of (y,) is given by
d AS] -
Wy) [| Wy X(s+ 5] ~ (Wy KG) 12-20)
as asv0 ds
with the unit tangent vector taking the form,
dx
ae 1.221
oe ¢ )
On the basis of Eqs. 1.2-20 and 1.2-21 we define the gradient of (Wy) according to
diyy) ax
—+ =v 1.2-22)
ds a VY (1.2-22) Chapter 1
Figure 1.4, Arbitrary curve in the y-phase
Figure 1.5. Asbitrary curve in the porous medium
In the averaging theorem the symbol for the two gradient operators is the same; however,
the significance is different as indicated by Eqs. 1. 2-19 and 1.2-22.
The vector form of the spatial averaging theorem is given by
(Vea) = Velay) + Jon a, dA (1.2-23)
Ane Diffusion and Reaction B
and we can use this result with the right hand side of Eq. 1,2-14 to obtain
1
(DV) = VADVery) + F Jr Peay Ath (1.2-24)
Ane
Substitution of this result into Eq. 2-14 allows us to express the spatially smoothed
transport equation as
Cay)"
oe = VADVexy) + # Jn Deny dA (1.2-25)
Aye
Clearly a second application of the averaging theorem is required to complete our
analysis of the diffusive flux, but before doing that we want to use the boundary
condition given by Eq. 1.2-2 to express Eq. 1.2-25 as
Heng)" ; '
epee 9 ee) (1.2.26)
ones ee
Seaman Cision ee
heterogeneous
ei
This form illustrates a key aspect of the process of spatial smoothing, ic., the interfacial
boundary condition given by Eq. 1.2-2 is joined to the governing equation given by
Eq. 1.2/1. This is a general characteristic of the governing equations for multiphase
transport phenomena, and because of its importance we state this characteristic as:
The form of a spatially smoothed transport equation
depends both on the form of the governing point equation
and on the form of the interfacial boundary conditions.
‘When spatially smoothed transport equations are generated intuitively, this characteristic
is sometimes ignored. Later it will become clear that the interfacial boundary conditions
influence the form of the averaged equation at two levels. The first level is illustrated by
Eq 1.2-26 while the second level occurs in the closure problem described in Sec. 1.4.
Returning to Eq. 1.2-26, we note that the molecular diffusivity changes slowly with
pressure, temperature, and concentration, thus it is appealing to ignore variations of @,
within the averaging volume and express Eq. 1.2-26 as (see Problem 1-11)
a
. Aeay)
ot
=V (ZkVeay)) - H J heat (1.2-27)
Ar
In this study we will also ignore variations in the reaction rate constant within the
averaging volume and express Eq. 1.2-27 as 14 Chapter 1
Cay)"
ar
= V(FVeay)) — ay k(Cay re (1.2.28)
Here ay represents the surface atea per unit volume which is given explicitly by
a, = Ag/¥ (1.2-29)
and the area averaged concentration is defined by
1
(Cay de = = py (1.2-30)
Are
If we accept Eq. 1.2-28, our next step is a second application of the averaging theorem
and this leads to
Hew)"
or
: l
= VD! egy) + F Jowcaet = ak egy (12-381)
Are
Here we find ourselves confronted with four different concentrations; however, one of
these can be eliminated by the use of Eq. 1.2-10 which yields
Cay)"
* 3
V4] ey Mea)” + Keay)" VEy + oe fmacay |] — 2, cay
An
(1.232)
One might assume that the area averaged concentration, (cay ys is essentially equal to
the volume averaged concentration, (c,,)’, and under these circumstances the last term
in Eq. 1.2-32 does not present a problem. However, the presence of the point
concentration in Eq. 1.2-32 does present a problem since c,, is only available to us in
terms of the solution of the boundary value problem given by Eqs. 1.2-1 through 1.2-4.
We attack this difficulty in terms of a spatial decomposition that is analogous to the
temporal decomposition used in the study of turbulent transport phenomena.
1.2.2 SPATIAL DECOMPOSITION
Since the intrinsic average concentration is the quantity of interest, we decompose the
point concentration according to (Gray, 1975)
Cr = (Cac. (1.2-33) Diffusion and Reaction 15
in which Z,y is referred to as the spatial deviation concentration. In some respects one
could think of Eq. 1.2-33 as decomposing the point concentration into what we want and
whatever else remains. This is indeed true; however, the important characteristic of Eq
1.2-33 is that it represents a decomposition of length scales. By this we mean that the
average concentration undergoes significant changes only over the large length-scale, L.
In Sez. 1.4 we will see that the spatial deviation concentration 4, is dominated by the
small length-scale ¢, and it is for this reason that we refer to Eq. 1.2-33 as a
decomposition of length scales. When Eq. 1.2-33 is used in Eq. 1.2-32 we obtain
1
Aaa
e, eat i 8yCcag)” + (ony "Vey + H Jrlenydaa
Ane
ar
= Vv.
(4.2.34)
1 =
+ Fe foptaett || = ayklcay
Are
Here we are confronted with three distinct problems that appear in essentially every
application of the method of volume averaging. For this study of diffusion and reaction
in porous media, these three problems take the form:
1. The presence of the volume averaged concentration,
e4q)", inside the integral over the area Ay.
2. The presence of the area averaged concentration,
(Cay )pe> in the heterogeneous reaction rate term.
3. The presence of the spatial deviation concentration
Eqy in the governing equation for (c4y)’
We will consider the first two problems in the next section and then move on to the
determination of 4, in Sec. 1.4 where we will present the closure problem.
1.3 Length-Scale Constraints
In
king about the area integral of m (cay) in Eq. 1.2-34, we must remember that
under the integral sign (c,,)’ is evaluated at points in the y-phase indicated by the
position vector y, illustrated in Figure 1.3. We make this point clear by writing 16 Chapter I
dA a3-D
key,
[i cng) aa -¥ Jaton)!
A A500
This clearly indicates that Eq, 1.2-34 is a non-local transport equation since the
dependent variable (c,,)" is evaluated at points other than the centroid, x. Non-local
problems have been studied by Cushman (1983), Koch and Brady (1987), Quintard and
Whitaker (1990a-b), Cushman and Ginn (1993), Goyeau ef al. (1997) and many others.
The analysis of non-local phenomena, or evolving heterogeneities (Cushman, 1990), is
extremely complex, thus there is a great deal of motivation for avoiding a non-local
problem whenever possible. From Eq. 1.2-34 we can extract a local theory provided
certain Jength-scale constraints are satisfied, and in order to develop these constraints we
need to represent (c,,)" in terms of a Taylor series.
1.3.1 TAYLOR SERIES EXPANSION
In order to remove the volume averaged concentration from the area integral in
Eq. 1.2-34, we use a Taylor series expansion about the centroid of the averaging volume
to obtain
ay)"
= lea] + yy Vea)" | + Ey Pe] + 3-2)
ney
Substitution of this result into Eq. 1.3-1 leads to
i : a t
a foretenn)"aa 7 He fecal aa a fry Weay"| aa
bn an cs
(1.3-3)
i
+s fortran Pear a +
Ane
Since all terms evaluated at the centroid can be removed from the integrals, we have
]
Fe Jelena . la freee feay)” + oy ment) en
i
: li in
(13-4)
[a
+ [> Jeet VW e gy)? +
Are
The terms in braces represent a series of geometrical integrals that are related to the
structure of the porous medium. In their present form the integrals are difficult to Diffusion and Reaction 7
interpret; however, Quintard and Whitaker (1994b) have developed a set of geometrical
theorems that relate these integrals to the spatial moments of the porous medium. These
are given by (see Problems 1-12 and [-13)
F foneat = -Vily (1.3-5a)
Age
1
F Jrmra = -V(y,) (1.3-5b)
br
1
H rer iv = -Wy,¥y) (1.3-5e)
Ar
(1.3-5d)
in which 1), (y,), (¥y¥y), ete, should be thought of as the zeroeth, first, second, etc.,
superficial spatial moments of the 7-phase contained in the averaging volume. If we
recognize that
ay = flav = & (1.3-6)
We see that Eq, 1.3-Sa leads to the well-known result given by
H Jet = -Vey sn
An
When Eqs. 1.3-7, 1.3-5b, and 1.3-Se are used in Eq. 1.3-3, the volume averaged
transport equation given by Eq. 1.2-34 takes the form
Keay”
Yat
. a:
V4] 2] &,Vlem¥ =VeHy): Keay) —$VY¥,) Whey" = te Ping
\ An
= ay Keay ye
(1.3-8)
Here we can see a major problem resulting from the use of the expansion given by
Eq, 1.3-2. If the higher order terms are not negligible, the volume averaged transport 18 Chapter 1
equation will contain a third order derivative of (c4,)’ and this leads to a
mathematically ill-posed problem.
If the phase is uniformly distributed about the centroid, and the radius of the
averaging volume is large compared to the characteristic length scale for the -phase,
«<8 (1.3-9)
it is intuitively appealing to think of (y,) as being zero. In a study of transport in
ordered and disordered porous media, Quintard and Whitaker (1994a-e) considered the
characteristics of the spatial moments in some detail, and they suggested the following
definition of a disordered porous medium,
A porous medium is disordered with
respect to an averaging volume 2”
when Viy,) <1
This definition of disordered is obviously well-suited for the method of volume
averaging; however, it appears to be consistent with the concepts encountered in other
studies of disordered systems (Strieder and Aris, 1973; Torquato, 1986; Shah and
Ottino, 1987; Rubinstein and Torquato, 1989, Dagan, 1989).
Goycau et al, (1997) examined the geometrical characteristics of dendritic systems
that appeared to be ordered and non-homogeneous; nevertheless, they found that
suitable averaging volumes could be found such that V(yy) << and we will make use of
this inequality to impose the restriction
Vly4)-Vleay® << €,W(c4)t (13-10)
Under these circumstances Eq. 1.3-8 simplifies to
Cay"
& ae
1 =
= VID,/ e,Vcay)” — Vy y¥y)! VV Ay)” F Jntadtt
bn
— yk Cay) ye
ee
and we now need to direct our attention to the term involving VV(c4,)" Diffusion and Reaction 19
1.3.2 ORDER OF MAGNITUDE ESTIMATES
In this development, and elsewhere, we need estimates of terms such as V{c4y)¥and
VV(cay)" (See Problems 1-14 through 1-16). We could estimate the first derivative of
the average concentration as
Veeay)* = O[ACcay)" /AL) (1.3-12)
where A(cay)’ represents the change of the average concentration that occurs over a
distance AL. The estimate of the gradient of the average concentration could also be
expressed as
Veeqy)” = (Ae ay)" /Le) (13-13)
in which L, represents a characteristic length associated with the average concentration,
‘The estimate of the second derivative of the average concentration takes the form
Wem)" = Of (Peay) /a1] (13-14)
in which A(V(c4,,)¥) represents the change of the gradient of the average concentration
that takes place over the distance AL. Alternatively, one could follow the type of
nomenclature used in Eq. 1.3-13 in order to estimate the second derivative of the average
concentration as
Wie)’ = O Vem)" )/ Le (13-15)
Here L,, represents a characteristic length associated with the first derivative of (c,y)"
in the same sense that L, represents a characteristic length associated with the average
concentration. Both L, and Lp, will be functions of time for a transient process, and
both may be significantly different than the generic large length-scale, L, that is
illustrated in Figure 1.2, Rather than use Eq. 1.3-15 as our representation of the order of
magnitude of the second derivative, we will simplify the nomenclature to
Way) = O[Veuy)" [Ler] (13-16)
In this form, L,, accounts for both the change in V(c4,)* and the distance over which
this change takes place, and when the gradient of the average concentration is a constant
one must remember that L,, is infinite. The conventions represented by Eqs. 1.3-13 and
1.3-16 will be used throughout this monograph.
It will be left as an exercise for the reader to demonstrate the order of magnitude
estimate (see Problem 1-17)
(y4¥,) = Oley?) (13-17) 20 Chapter 1
which immediately leads to
Viyayy) = O(de yr? /L.) (13-18)
Here we have followed the convention indicated by Eq. 1.3-13 and defined L, by the
estimate
Ve, = O(de,/L,) (13-19)
We now return to Eq. 1.3-11 and note that the term involving V¢¥y¥,)'VV(cay)" can
be estimated as
‘ct
Vey y¥y):VV(cay)” = of Josmen"| (1.3-20)
This provides the important inequality
Wy yy WV leqy)” << AeyV(cay)” (13-21)
when the following length-scale constraint is satisfied.
2
Te (13-22)
Once again we note that neither L, nor Le, is necessarily equal to the generic large
length-scale, L , illustrated in Figure 1.2. One must make a judgment about L, and Lo:
for every problem, and one must not think of the length scales as being unique for any
given problem. For example, in the center of the region illustrated in Figure 1.2, one
might think of L, as being infinite since the porosity is uniform. On the other hand, at
the surface of a catalyst pellet it is appropriate to use L, =r, because the volume
fraction of the y-phase changes significantly over the distance r, in this particular region.
Sometimes one represents Eq. 1.3-22 as
pic DB (13-23)
with the idea that L represents the smallest large length-scale associated with the
problem under consideration. On the basis of either Eqs. 1.3-22 or 1.3-23 we simplify
Eq. 1,3-11 to
[
Acay)" 1 ~
eae "4 £,Vdeqy) + fmt tt = ak Can ye
Ane
or
\
(13-24)
and we are now confronted with only two problems: The presence of the spatial
deviation concentration @,, in the area integral, and the area averaged concentration Diffusion and Reaction a
(Cay)yx in the heterogeneous reaction rate term. We will deal with the area averaged
concentration first and then move on to the determination of Z,, in the next section.
1.3.3 AREA AVERAGED CONCENTRATION
‘We begin our analysis of the area averaged concentration with the decomposition given
by Eq. 1.2-33 which allows us to express (c4y yg aS
(0m
I 2
Cade = xs rey, 7 Ee es, (13-25)
Ae
In the next section, we will show that for all practical problems of diffusion and reaction
in porous media the spatial deviation concentration is small compared to the volume
averaged concentration, ie.,
ee ce (cae (1.3-26)
This means that Eq, 1.3-25 takes the form
1
cape = Fe fico (13-27)
Aye
and we can use a Taylor series expansion of (c,,)" around the centroid of the averaging
volume to express this result as
(Cay) = Cay) + Vy dye" Wheay)! + Ey ¥y dy VW (Cay )” + oe (13-28)
No precise estimates of (yy); are available; however, one can easily generate
arguments in favor of
Wye > 1, closure process is quasi-steady (14-20)
When the macroscopic process is transient, one can argue that t* must be constrained by
Dt
Lele
= O(1), macroscopic process is transient (14-21)
however, both these conditions can often be satisfied because £,, is generally very small
compared to the characteristic lengths associated with (c,,)” and V(c4,)”.
On the basis of Eqs. 14-18 and 1.4-19, we can express the quasi-steady closure
problem as
0 = V(D,VEqy) - €7'Vey-DyWeay)” + ejay k cay)" (14-22)
heen) | ieee | eet
diffusion diffusive source reactive source
BCL ~My DyVEqy — KE gy = My DyV cay)” + kcgy)" at ey. (14-23)
diffusive source Featiive
source
B.C2 Eqy = PC(est), at My (14-24) Diffusion and Reaction 27
Here it becomes apparent that there are volume sources involving %V(c,)¥ and
Cay)", and there are surface sources involving these same two parameters. A little
thought will indicate that the term k{c4y)" in the boundary condition given by
Eq. 1.4-23 represents a sink since it creates a flux of Z,y from the y-phase to the x-phase.
In addition, one can show that this surface sink is exactly balanced by the volume source,
ej'a,k(Cay)", in Eq. 1.4-22. This means that the average reactive source is zero, and
the terms involving k(cq,)” only influence the manner in which Z4, is distributed in the
phase. The diffusive source in Eq. 1.4-23 has the same characteristic, i.c., the average
surface diffusive source is zero; however, it changes sign over a distance £, and it
dominates the manner in which Z,, is distributed in the y-phase.
The order of magnitude of the surface diffusive source per unit volume can be
estimated as
f order of magnitude :
of the surface = ¥ Jk “DV cyy)"|dA
f y Pee A%eay)| (14-25)
| diffusive source dp
= Oa, F%Mer)")
while the estimate of the volume diffusive source per unit volume takes the form
order of magnitude’ '
of the volume . x Jlrs OMe, y"|aa
x J ltr Mer AM cay
diffusive source i, (14-26)
= 0[Ve, ¢ u
O[Ve, BMc4,)")
These two results allow us to estimate the ratio of the volume source to the surface
source according to
volume source
= (1.4-27)
surface source
For the typical porous medium a, ~ £7! and we obtain
£
volume source _ of “+ ne,) ee 1 (14-28)
surface source Ly
and the volume diffusive source in Eq. 1.4-22 can be discarded.
Further simplification of the closure problem can be achieved if we can demonstrate
that Z4y is small compared to (c,,)". Directing our attention to the boundary condition
given by Eq. 1.4-23, we use the estimates 28 Chapter 1
ie DyVEpy = O(DEny/t,) (1.4-29a)
ie DyVeny)” = O(%,AMeay)* /L,) (1.4-29b)
to obtain
BCI of 3 Sea) + Okey) = of Sea") +O(Keay)") (1.430)
‘This can be arranged in the form of an estimate for Z4y
Sy = of LlaLte)aeny” (kta / Kem"
[i + ofeey/%)]’ [E+ O(ee,/%)]
originally given by Ryan et al. (1981). In order to simplify this result we need to express
the parameter k £, /@, in terms of the particle Thiele modulus which leads us to
He gue}
a 7 9°(¢,/L) (1.4-32)
431)
Here 6 represents the particle Thiele modulus which is given explicitly by
o= LfakTD, (14-33)
In this representation the characteristic macroscopic length, L, is usually taken to the
effective particle diameter (Carberry, 1976), and the small length scale 4, is related to
a, by
4, = a) (14-34)
For practical reactor design problems, the particle Thiele modulus is usually less than ten
(g? $100) while 4, is typically many orders of magnitude smaller than L, thus Eq.
1.4-32 indicates that k £ /@ is very small compared to one.
at «<1 (1.4-35)
QD,
Use of this result with Eq. 1.4-31 leads to the following estimate for the spatial deviation
concentration:
t
ey = of($ em" + (5 en (1.4-36) Diffusion and Reaction =
On the basis of Eq. 1.4-35, and because £, << L_, this estimate leads us to conclude that
Ey << Cay)” (14-37)
and we can discard ké,4y relative to k{c4,)" in Eq. 1.4-23. This allows us to express
the boundary value problem for Z4, as
Vin = (2x fem" (14-38)
B.C. DyVEny = My DyVeeay)” + Keegy)", at My (1.4-39)
BC2 Emp = SE), atohe (1.4-40)
While Eq. 1.4-37 represents a reliable estimate of Z4y relative to (c4y)", one must be
careful to remember that the length scales associated with Z4, and (c4y)* are very
different and a similar inequality does not exist for the gradients of Z4y and (c4y)¥. in
the absence of any chemical reaction, Eq. 1.4-39 immediately leads to the estimate
Veg = [May y) (14-41)
This relation between the gradient of Z4y and the gradient of (c4y)" must be kept in
mind whenever simplifications are made in the volume averaged transport equation for
(c4y)" or in the governing equation for Z4y.
1.4.3 THE LOCAL PROBLEM
It should be clear that we have no intention of solving Eqs. 1.4-38 through 1.4-40 over
the entire macroscopic region illustrated in Figure 1.2. Instead, we want to solve for gy
in some representative region and use the computed results in Eq. 1.4-5 to affect the
closure. Such a region is illustrated in Figure 1.6, and if we want to solve for 4y in that
region we must be willing to abandon the boundary condition given by Eq. 1.4-40.
Since the source, (r,t), only influences the @4,-field over a distance on the order of
4y, the loss of the boundary condition at oi%_ is of no consequence. If we treat the
representative region shown in Figure 1.6 as a unit cell in a spatially periodic porous
medium, we can replace the boundary condition imposed at o%, with a spatially
periodic condition on E4y.
If we accept the model of a spatially periodic porous medium in order to solve the
closure problem, we know that the unit normal vector must satisfy the condition 30 Chapter 1
ny (Eth) = Mye(F),
1.2.3 (1.4-42)
Here ¢; represents the three non-unique lattice vectors that are required to describe a
spatially periodic porous medium (Bensoussan, ef al, 1978; Brenner, 1980; Sanchez-
Palencia, 1980). While the geometry of our representative region may be invariant to a
transformation of the type, r>F+4;, the governing equation and boundary condition
given by Eqs. 1.4-38 and 1.4-39 will not be unless we can treat (c4y)? and V(c4,)"as
constants. Since (c4,)’ cannot be a constant unless V(c4,)" is zero, there are some
approximations to be made before we can impose a spatially periodic condition on Z4,.
Figure 1.6, Representative region
In order to identify these approximations, we need to express (c,y)? and Vicay)? in
terms of Taylor series expansions about the centroid of the representative region. These
expansions are given by
= ea] + yy Wea |, + byyyy Peay) | + 443)
(Cay)
Mean)"|,., . Veeay)"|, +Yy Vee 4y)"|, (14-44) Diffusion and Reaction 3
in which x represents the centroid of the representative region shown in Figure 1.6, and
Yy locates points in the y-phase relative to the centroid. We can estimate the higher
order terms in Eq. 1.4-43 and use the resulting representation in Eq. 1 4-38 to obtain
a vk
res = (ae oot flees
Using both Eqs. 1.4-43 and 1.4-44 in the boundary condition given by Eq. 1.4-39 Jeads
to the following form:
= yg Ven = Nye Heay)" | vo =r]
(1.4-45)
B.C. (1.4-46)
“(é Je [ + of es -}am Al A,
Here we have used the estimates given by
7 Veeay)” = often) (.4-47a)
Wey) = often”) (1.4-47b)
1
in which y,, is on the order of r, and the length scales, L, and L,,, are defined by
Eqs. 1.3-13 and 1.3-16. The estimate for y,, is based on the idea that the representative
region will never be larger than the averaging volume. In order to discard the higher
term in the governing equation for ,,, we require the restriction
5 yf ake a,k : |
(iebwr|at) «(aioe
while the boundary condition given by Eq. 1.4-46 can be simplified on the basis of
of fren] << mye Vey)" | (1.4-49a)
XX Ley - 32 Chapter 1
k
a , E
<< (Een L (1.4.49)
These restrictions lead to the following constraints
r << Le << Ly (1.4-50)
and when these constraints are satisfied, the closure problem can be expressed as
Jeu (1.451)
en at Ay (14-52)
Periodicity: Ey (et) = Eq ln), F=12,3 (14-53)
in which A,. represents the interfacial area contained within the representative region.
Here we have discarded the boundary condition at 0%, given by Eq. 1.4-40 and
replaced it with the periodicity condition on Z, (see Problem 1-19). This periodicity
condition is consistent with Eqs. 14-51 and 1.4-52 only if the geometry of the
representative region is spatially periodic and the source terms are either constant or
spatially periodic. By imposing the constraints indicated by Eqs. 1.4-50 we have been
able to treat the reactive sources, (ayk/€yD,)cay)" and (k/D,)(ca,)". as constants
while the diffusive source, My. -V(cay)", is spatially periodic because of the nature of
the unit normal vector ny.
It is important to understand that Eqs. 1.4-51 through 1.4-53 represent an
approximation of the boundary value problem given by Eqs. 1.4-38 through 1.4-40 and
the validity of this approximation is an important issue. It should be clear that the &,~
field will be dominated by the governing differential equation and the flux boundary
condition, both of which contain source terms that generate the E,y-field. Neither the
boundary condition given by Eq. 14-40 nor the periodicity condition given by
Eq. 1.4-53 will have an important influence on the Z4,-field. ‘This is especially true for
the values of @4y evaluated at the ¥-K interface, and it is this value that enters the filter
identified by Eq. 1.3-37. The role of this filter for diffusive transport processes has been
explored in some detail by Ochoa-Tapia et al. (1994) for spatially periodic models and
Maxwell's (1881) model, and their numerical calculations indicate why Maxwell's
model provides good results for effective diffusivities and effective thermal
conductivities even when the constraints inherent in the model are not satisfied. Diffusion and Reaction 33
One should keep in mind that the representative region shown in Figure 1.6 can be
arbitrarily complex, In their study of two-phase flow in heterogeneous porous media,
Ahmadi and Quintard (1996) used a log-normal distribution of permeabilities (Dagan,
1989) to construct large unit cells that were representative of complex, disordered
systems. In a study of dispersion, Noble et al. (1998) made use of unit cells in which 80
octagonal cylinders were randomly distributed in order to determine the influence of
disorder on longitudinal and lateral dispersion coefficients. It is clear from these initial
studies of disordered systems that unit cells can indeed be constructed that are
representative of real, disordered systems.
Itis of some importance to note that Eqs. 1.4-S1 through 1.4-53 determine Z,, only
to within an arbitrary, additive constant. Sometimes this indeterminacy is removed by
imposing the condition
Eq = 0 (14-54)
While this constraint is quite plausible, it is not necessary since any additive constant
associated with Z4, will not pass through the filter represented by Eq. 1.3-37. It is
important to remember that in the development of the local volume averaged equation
given by Eq. 1.3-36, we made repeated use of Eq. 1.3-7 which can be used to write
$ Jrveonstantyta = -(Ve,)(constant) (1.4-55)
fre
However, when the area integral in Eq. 1.3-36 is evaluated on the basis of the closure
problem given by Eqs. [.4-51 through 1.4-53, we find that
: ‘for a spatially
“F f ny(constant\dA = 0, — 4 periodic porous (14-56)
re medium
Under these circumstances, the additive constant associated with ¢,, plays no role in the
evaluation of the area integral in Eq. 1.4-5. However, this is not true for all processes,
and in Chapter 4 we will see that a condition of the type indicated by Eq. 1.4-54 plays a
crucial role in the closure problem.
1.4.4 CLOSURE VARIABLES
Given the linear nature of the boundary value problem for Z4,, we propose a solution in
terms of the sources which we express as (see Problem 1-20)
Bay = by -Veeayd"| + Seay)" + Wy (14-57) 34 Chapter 1
This approach to solving the boundary value problem for Z4, is known as the method of
superposition, and we refer to the vector by and the scalar sy as the closure variables.
We can think of y, as an arbitrary function and this allows us to specify b, and s, in
any way we wish, and we choose to determine these closure variables by means of the
following two boundary value problems:
Problem
Vb, . 0 (1.4-58a)
BC. = Vby = Ms at Arg (1.4-58b)
Periodicity: bth) = br), 7=1,2,3 (14-58)
Problem IT
k
Vis, = -e 14-59
- am (1.4-59a)
B.C. = (1.4-59b)
Periodicity: s(Pt4,) = s(r), (1.4-59¢)
We can now develop the boundary value problem for yy by substituting the
representation given by Eq. 14-57 into the closure problem given by Eqs. 1.4-51
through 1.4-53 and imposing the conditions on by and s, required by Eqs.1.4-58 and
1.4-59. This leads to our third closure problem given by
Problem It
Vy, = 0 (1.4-60a)
BC =a Vy, = 0, at Ag (1.4-60b)
Periodicity: Wyre) = Wy), 1=1,2,3 (1.4-60c)
Certainly y, = constant is one solution to Eqs. 1.4-60 and it will be left as an exercise
to prove that this is the only solution (sce Problem {-21). Since this additive constant
will not pass through the filter in Eq. 1.4-5, as we have indicated by Eq. 1.4-56, the value
of Wy plays no role in the closed form of the volume averaged diffusion equation.
1.4.5 CLOSED FORM
Substitution of Eq. 1.4-57 into Eg. 1.4-5 leads to the closed form of the governing
differential equation for (c4,)". Since (e4,)¥ and Vieqy)"are evaluated at the Diffusion and Reaction =
centroid of the representative unit cell, these terms can be removed from the area integral
in Eq. 1.4-5 and this leads to
Cay)" 1
by = Vie Dd Vecay)” + F Jona -Weeay)"
'
An
(14-61)
1
Dy = O|1+ Ee fobrat (14-62)
The
and we represent the vector associated with the chemical reaction as
yD,
.- oY Jorvesra (1.4-63)
The
Use of these two definitions in Eq. 1.4-61 leads to the following form of the closed
equation
Acqy)"
ys = V-(eyDeg “Weeay)") + V-(wleay)) — aykXeny)” — (14-64)
Since u has the units of velocity and is directly proportional to k, one can think of
uXcay)" as a convective transport term that is generated by the heterogeneous reaction.
Paine et al. (1983) have shown that this contribution to convective transport can be
important when convection itself is important; however, we will demonstrate in the
following paragraphs that this term is negligible for the case of diffusion in porous
catalysts.
At this point in our analysis, we should think about how the interfacial boundary
condition given by Eq. 1.2-2 has influenced the volume averaged transport equation.
For the process of diffusion and reaction, we found that the interfacial boundary
condition entered into the analysis in two ways:
1. In the process of spatial smoothing, the interfacial flux was
incorporated directly into the volume averaged transport
equation. That gave rise to the reaction rate term in Eq. 1.2-26. 36 Chapter 1
2. In the closure problem we found that the heterogeneous
reaction rate term acted as a source for the Z4y-field. That gave
rise to the convective-like term in Eg. 14-64.
Although the second contribution will be negligible in this case, one should always keep
in mind that boundary conditions influence the final form of a volume averaged transport
equation both in the spatial smoothing process and in the closure problem. This is
particularly clear in the case of the chromatographic equation and Forchheimer’s
equation (Whitaker, 1997).
Turning our attention to the convective-like transport term, we can use Eqs. 1.4-59
to develop the estimate given by (see Problem 1-22)
Ds, = O(¢,k) (14-65)
and this can be used with Eq. 1.4-63 to conclude that
u = O(k) (1.4-66)
From this result, and the fact that a, = ¢)', we can see that the last two terms in Eq.
1.4-64 are restricted by
aykkepy)® >> V-(wleay)*) (4-67)
and under these circumstances Eq. 1.4-64 simplifies to
cay)”
a = VeleqDag -Weeay)") — aiyk (Eny)” (14-68)
In this particular case, we see from Eqs. 1.4-62 and 1. 4-58 that the effective diffusivity
is a linear function of the molecular diffusivity and that it depends on the geometry of the
Porous medium through the b,-field. The boundary value problem for the b,-field is
purely geometrical in nature and we will examine solutions of this problem in Sec. 1.6 in
order to compare theory with experiment for the effective diffusivity, In addition to
using the closure problem given by Eqs. 1. 4-58 to determine the effective diffusivity
tensor, one can also use it to prove that Dg is symmetric (see Problem 1-23).
1.5 Symmetric Unit Cells
In this section we develop some symmetry properties for the s, and b,-fields that occur
when symmetric unit cells are used to solve the two closure problems given by
Eqs. 1. 4-58 and 1.4.59. Although the unit cell illustrated in Figure 1.6 is not symmetric,
most calculations are carried out using relatively simple unit cells that are completely
symmetric. Under these circumstances one can prove that u=0, and in the following
paragraphs we present Ryan's (1983) proof. We begin with an explicit representation of
Eqs. 1.4-59 for the x, y,2-coordinate system so that the closure problem takes the form Diffusion and Reaction 37
Problem I
a ot 2
ae eae (1.5-1a)
ax’ ay’ az? eB,
as, ays,
B.C.1 ames t mae = HM. A Ag (1.5-1b)
Periodicity: Sy(r+6) = Se), §=1,2,3 (15-1e)
Here we have used n,,n,, and n, to represent the components of the unit normal vector
nj, in order to simplify the nomenclature.
Since Eqs. 1.5-1 are valid in any coordinate system, we consider the transformation
X,y,Z— X,Y, and express the closure problem in this new coordinate system as
Problem II’
(1.5-2a)
BCI (1.5-2b)
Periodicity: Be+4,) = 5(r), 151,23 (1.5-2c)
In Eqs. 1.5-2 we have used 5, to mean
5, = 509.2) (15-3)
We now consider a particular coordinate transformation given by
oo hr, (15-4)
and for this case we see that Eq. 1.5-2a takes the form
os el =
4,77. -(24) 7
In the boundary condition given by Eq. 1.5-2b, the transformation specified by Bq. 1.5-4
leads to the following relations for the components of the normal vector
RR, = 1,G,%,2) = n(-1y,z) = —n, (4,2) (1.5-6a)
In addition, the symmetry of the unit cefl also leads to
ny = nyGy2) = ny(-x,y,2) = my(%y2) €1.5-6b)
A, = m7) = n,(-%y,2) = n,(%9,2) (1.5-6¢) 38 Chapter 1
‘The derivatives in the boundary condition given by Eq. 1.5-2b can now be expressed as
See eS SS (15-7)
ax y ay az az
and this allows us to write that boundary condition as
a5, a5, a5,
Bl ~E “ay ast = HD, athe (1.5-8)
To be absolutely clear about the form of Problem II’, we summarize these results to
express this part of the closure problem as
Problem I’
2s teste
as (15-98)
ax?" ay? ay? £9,
a5, a5, a5,
BC.1 a “oy . nt = KD, aay (1.5-9b)
Periodicity: H(rt+4) = F(n), 1=1,2,3 15-90)
When this boundary value problem is compared with that given by Eqs. 1.5-1, we
conclude that
5y (15-10)
for the coordinate transformation given by Eq. 1.5-4. We can also write Eq. 1.5-10 as
SyO% 2) = yy.) (15-11)
and from this we see that s, is symmetric in x for a unit cell that is symmetric in x. One
can repeat this proof of Ryan (1983) for cells that are completely symmetric to show that
Sy is completely symmetric. For this case we can prove that the vector u is equal to zero.
Because of the periodic nature of s,, we can express the area integral of s, over the
entrances and exits of any unit cell as
Jesu = 0 (15-12)
Vy
ae
When this condition is used with Eq. 1.4-63, and the divergence theorem is applied in
order to convert the area integral to a volume integral, we obtain the following
expression for the vector u: 39
e 1
u = 6%, qfysav (15-13)
‘y
Ba ley, (15-14
Sg tlval ar ee
2 (15-15)
2 ox 2
This leads to the condition
0 (1.5-16)
and for a completely symmetric unit cell we have
u = 0, completely symmettic unit cell (15-17)
Given that the convective transport term in Eq. 1.4-64 is negligible, as indicated in Eq
1.4-67, this result for completely symmetric unit cells is not particularly important;
however, the characteristics of symmetric unit cells are important and the analysis
leading to Eq. 1.5-17 will prove to be useful in other studies.
Turning our attention to Problem I, we examine the x-component of the vector by
for the transformation given by Eq. 1.5-4. We designate i-b, as , and express the
closure problem for this component as
Problem I
ab, ab, Hb,
sj+it+ife=zo (1.5-18a)
ax? ay? "ae
ab, ab, ab,
BCI -E TGE Tae Tee Ae (1.5-18b)
Periodicity: by(r+l,) = blr), i=1,2,3 (1.5-18c)
In the %,3,2 coordinate system we express this problem as 40 Chapter 1
Problem I’
ob,
= (15-19)
_ ab,
BC - ay (1.5-19b)
ay
Periodicity: by(rt+é)) = b, 5 (1.5-19¢)
Here we have used b, to mean
, (15-20)
i
For the specific transformation given by Eq. 1.5-4, we can repeat the analysis given by
Eqs. 1.5-2 through 1.5-9 to obtain
Problem I’ _ _
ab, ab, . ab,
+it+it=o0 (1.5-21a)
ax? ay? az?
i oo
BC1 = me SE my SSE Totes al A (1.5-21b)
Periodicity: byrte;) = be), i (15-21c)
The key idea here is that the x-component of the unit normal vector on the right hand
side of Eq. 1.5-18b controls the behavior of the b, -field and n, is skew-symmetric in x.
By comparing Eqs. 1.5-18 with Eqs. 1.5-21, we see that
by = - (1.5-22)
and we can be more explicit by expressing this result as
Dy (Hz) = — by (—¥ 52) (1.5-23a)
For a completely symmetric unit cell, the above proof can be repeated for the y and
z-components of the vector b,, and this leads to
By (2) = = by Oo-y.2) (1.5-23b)
By(% YZ) = by (2, YZ) (1.5-23c)
These conditions can be used to reduce the computational domain, and in the original
numerical studies (Ryan, 1983) this reduction was of considerable importance. Diffusion and Reaction 41
1.6 Comparison Between Theory and Experiment
Our objective in this section is to compare the theory that Jed to Eq. 1.4-68 with
experiment; however, since the effective diffusivity is independent of the reaction rate it
is sufficient to consider only the case of passive diffusion in porous media. The
comparison between theory and experiment will be restricted to systems for which
ey and Dy can be treated as constants, thus Eq. 1.4-68 can be simplified to
Cay)"
So = Deg VVC uy” (6-1)
It is of some interest to note that we have retained the porosity in this result rather than
canceling €,, from each side of Eq. 1.6-1. This has been done in an effort to avoid errors
on the order of &,, since numerous errors of this type have made their way into the
literature.
It is important to remember at this point that our formulation of the diffusion
problem was based on the restrictions given by Eqs. I.1-12 and 1.1-13 and on the
definition of the mixture diffusivity given by Eq. 1.1-15. This led to the molar flux
being represented by
_
Veay (1.62)
and one can quickly draw upon the developments in Secs. 1.2 and 1.4 to demonstrate
that the superficial volume average molar flux takes the form
(Nay) = = ByDag -Vecay)” (1.6-3)
In many problems of practical importance, one is only interested in the transport of
species A between a porous medium and the adjoining homogeneous fluid. We have
shown this system in Figure 1.7 in which the «region represents the porous medium and
the n-region represents the homogenous fluid. The molar flux associated with the @-n
interface can be expressed as
‘moles of species A
transferred per unit} =~ gy -&,Dgp Vege) (1.6-4)
time and area
however, this is often written in the form
‘moles of species A
transferred per unit} =~ myq ‘Diy “Vicqy)™ (1.6-5)
time and area
In this case the porosity has been incorporated into the effective diffusivity tensor, and
this is entirely a matter of choice; however, one is less likely to commit an algebraic
error if the porosity is retained as a separate factor as we have done in Eqs. 1.6-1 and 42 Chapter I
1.6-4. The general problem of constructing conditions at the boundary between a porous
medium and a homogeneous fluid has been discussed by Ochoa-Tapia and Whitaker
(199Sa, 1995b, 1997, 1998a, 19986).
Non
ayregion
Figure 1.7. Boundary between a porous medium and a homogeneous fluid
1.6.1 EXPERIMENTAL STUDIES
Experimental studies that are consistent with Eqs. 1.1-12 and 1.1-13 are very difficult to
carry out, especially in the presence of a chemical reaction. On the other hand, one can
perform experiments with passive systems at high molar concentrations that are
consistent with the theory leading to Eq. 1.6-1 and the closure problem that is used to
determine the effective diffusivity defined by Eq. 1.4-62. These experiments are done
with binary systems at constant temperature and pressure, and for these systems the
single Stefan-Maxwell equation takes the form
Nay = ay(Nay+Nay) — ZasVeay (1.6-6)
since the total molar concentration, cy, can be treated as a constant. In experimental
systems that are set up to provide transient, one-dimensional molar fluxes, one can argue
that
(Nay) = -(Na,) (16-7)
If one is willing to assume that the average of spatial deviations is zero, one can make
use of Eq. 1.6-7 in conjunction with Eq. 1.6-6 to obtain the following representation for
the intrinsic volume average molar flux of species A
Way) = (Fay Rey +Ney)) = (Dra¥eny) (1.6-8)
‘On the basis of the result given by Eq. 1.4-37, one can make use of the inequality Diffusion and Reaction 43
(Ee(May Nay) << Way) (1.69)
and this allows us to express the superficial average molar flux as
Way) = —(GasVeay) (1.6-10)
From here one can obtain Eq. 1.6-3 where Zp replaces @%, in the expression for the
effective diffusivity tensor given by Eq. 14-62. Since all of the experiments that we
wish to compare with the theory were obtained under conditions that led to Eq. 1.6-7, the
justification of Eq. 1.6-10 is an important matter.
1.6.2 ISOTROPIC SYSTEMS
It is worthwhile to note that there are no homogeneous, isotropic porous media. By
isotropic one means invariant to any coordinate rotation and by homogeneous one
means invariant to any coordinate translation, and a little thought will indicate that
there are no porous media that could satisfy these geometrical requirements. On the
other hand, there exist porous media for which the diffusion process is isotropic and for
which the effective diffusivity tensor and the porosity can be treated as constants. We
think of these porous media as being homogeneous and isotropic with respect to the
diffusion process.
‘As an example, we note that the representative porous mediwm illustrated in Figure
1.6 is invariant to coordinate translations of the type r= F +f), but it is not invariant to
am arbitrary translation and thus it is not homogeneous. Directing our attention of a
simpler system, we note that the two-dimensional, square array illustrated in Figure 1.8
is only invariant to 90° rotations about certain points, but it is not invariant to arbitrary
coordinate rotations about any point. Thus the porous medium is neither homogeneous
nor isotropic. On the other hand, the system illustrated in Figure 1.8 can be used to
solve the closure problem given by Eqs. 1.4-8, and this has been done by Ryan (1983)
to obtain
D, D,
Dy = Dy = 0 (L.6-11)
Here we have used D,,, D,
1-y+ €tC., to represent the components of Dyy, and Eqs. 1.6-11
indicate that the system illustrated in Figure 1.8 is homogeneous and transversely
isotropic with respect to the diffusion process, Ryan also solved the closure problem for
the system shown in Figure 1.9, and for that array one might expect to find values of D,,
that are larger than the values for D,, since diffusion in the x-direction would follow a
more “tortuous path”. However, the computed values of D,, and D,, differed by less
that 1% while the off-diagonal components were zero. This indicates that for practical
purposes the system shown in Figure 1.9 is also homogeneous and transversely isotropic
with respect to the diffusion process. In a more recent study of diffusion in cellular
systems, Ochoa (1989) repeated the original work of Ryan (1983) using several different
numerical techniques, all of which gave good agreement with Ryan's calculations. 44 Chapter I
iL
Figure 1.8, Two-dimensional square array as a model of a spatially periodic porous medium
Figure 1.9. Two-dimensional array as a model of a spatially periodic porous medium
There are a number of unconsolidated porous media that appear to be isotropic with
respect to the diffusion process and in Figure 1.10 we have compared the experimental
data for these systems with several theories, The curve identified as Ryan et al. (1981)
represents the solution of Eqs. 1.4-58 with the single distinct component of the effective
diffusivity tensor being predicted by Eq. 1.4-62. The theoretical results for the models
illustrated in Figures 1.8 and 1.9 are essentially identical and are in good agreement with
the experimental data shown in Figure 1.10. The curve identified by Quintard represents
a solution for the three-dimensjonal array of spheres illustrated in Figure 1.11. The
results of Quintard (1993) are slightly higher than those of Ryan et al. (1981) and are in
excellent agreement with the experimental data. The curve identified by Weisberg
(1963) in Figure 1.10 is an upper bound determined by means of a variational principle
which yields the following analytical representation Diffusion and Reaction 45
£,Deg [Dy = , Weisberg (1.6-12)
1~ Sine,
This result is also in reasonably good agreement with the experimental data, and as an
upper bound it is appropriate that it always predicts values that are greater than either
those of Ryan ef al. or Quintard. The first attempt at a theoretical prediction of the
effective diffusivity is due to Maxwell (1881) who analyzed a dilute suspension of
spheres to obtain the result
€ Dey [By = » Maxwell (1.6-13)
Although Eq. 1.6-13 was originally considered to be restricted to values of €, close to
one, Hashin and Shtrikman (1962) have shown that it is an upper bound for any value of
,, and the comparison shown in Figure 1.10 verifies their analysis.
1.0
Kimet al,
© spheres
Currie
(© mixture of spheres
Oi sand
A carborundum
sodium chloride
Hoogschagen
@ mixture of spheres
Y carborundum
&
1
Figure 1.10, Comparison between theory and experiment for isotropic systems 46 Chapter I
Figure 1.11. Three-dimensional array of spheres as a model of a spatially periodic porous medium
The micropore-macropore model of Wakao and Smith (1962) can be used to predict
effective diffusivities for macropore systems and for the constraints associated with
Eq, 1.2-1 their expression reduces to
8yDey [Zy = Fy,
Wakao and Smith (1.6-14)
‘This result is illustrated in Figure 1.10 and one can see that it significantly under-predicts
the effective diffusivity for values of the porosity in the range of practical importance.
‘The problem of diffusion and reaction in a macropore-micropore model of a porous
catalyst has been examined in some detail by Whitaker (1983a), and a comparison
between theory and experiment is available in terms of effective diffusivities for packed
beds of porous particles (Whitaker, 1988c).
1.6.3 MAXWELL’S CLOSURE PROBLEM
Equation 1.6-13 is of particular interest when viewed in terms of the closure problem
given by Eqs. 1.4-58. Rather than solve that boundary value problem subject to the
periodicity condition associated with the unit cell shown in Figure 1.11, Chang (1982,
1983) proposed a solution subject to a Dirichlet condition imposed at the spherical
surface illustrated in Figure 1.12. This special form of Eqs. 1.4-58 is given by
Vb, = 9 (1.6-15a)
BC Vby = Oe, T= (1.6-15b)
BC2 by = 0 rey, (1.6-15e)
and in order to develop a precise correspondence with the spatially periodic modet
shown in Figure 1.11, we require that the porosity and the area per unit volume for the Diffusion and Reaction aT
two models be equal (Quintard and Whitaker, 19954). The solution to Eqs. 1.6-15
produces the expression for the effective diffusivity given by Eq. 1.6-13, thus this form
of the closuse problem gives reasonable agreement with experimental data as one can see
from the curve identified by Maxwell in Figure 1.10.
Figure 1.12, Chang’s unit cell
Since the space occupied by the unit cell associated with Eqs. 1.6-15 does not fill the
space occupied by the porous medium, it is surprising that Maxwell’s closure problem
gives such good agreement with experiment. A detailed study of Eqs. 14-58 and the
special form given by Eqs. 1.6-15 has been carried out by Ochoa-Tapia et al. (1994),
and their work indicates that there are significant differences between the by field
determined by Eqs. 1.4-58 and Eqs. 1.6-15. However, it is the area integral of by My,
that determines the effective diffusivity tensor, and this area integral is not strongly
influenced by either the periodicity condition given by Eq. 1.4-58c or the Dirichlet
condition given by Bq. 1.6-15c.
1.6.4 ROLE OF THE FILTER
At this point we are in a position to learn something about the filtering characte
the area integral in Eq. 1.3-37 or in Eq. 1.4-62. To do so we consider the experimental
data for sand (Currie, 1960) that are illustrated in Figure 1.10. We imagine that
photomicrographs of the sand are available and that from them we can construct a
detailed three-dimensional model of the sand. A two-dimensional rendering of the
experimental system is illustrated in Figure 1.13, along with a theoretical model having
the same porosity as the experimental system. Imagine that the boundary value problem
represented by Eqs. 1.4-58 is solved for the experimental system shown in Figure 1.13
and the results are used in Eq. 1.4-62 to determine the effective diffusivity tensor. We
expect that the computed values for Dy would be in excellent agreement with the
experimental values illustrated in Figure 1.10, and this means that the experimental
system illustrated in Figure 1.13 would provide essentially the same result as the 48 Chapter I
theoretical model shown in that figure. This leads us to the question: “What gets
through the filter for the process of diffusion in an isotropic system?”, and the answer is
most certainly, “Not much more than the porosity.”
Experimental system Theoretical model
Figure 1.13. Experimental system and theoretical model for sand
1.6.5 ANISOTROPIC SYSTEMS
While many systems that are isotropic with respect to diffusion are available in the
laboratory, most natural systems and many processed materials are anisotropic. 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) for kaolin suggest that any
kaolin-based catalyst will exhibit anisotropic behavior. Natural minerals such as clay
and sandstone tend to be anisotropic (Muskat, Sec. 6.7, 1949) as do most cellular
systems (Crapiste et al., 1988a-b). Wood is a classic example of an anisotropic organic
material and this has important consequences for the drying process (Bonneau and
Puiggali, 1993; Turner, 1996). Although there are numerous anisotropic systems of
practical importance, no experimental studies exist prior to those of Kim er al. (1987).
Using a system similar to that employed by Currie (1960), Kim ef al. (1987) measured
the components of the effective diffusivity tensor parallel and perpendicular to the
bedding plane. Three systems composed of mica particles were studied and
measurements were also made for an artificial porous medium composed of mylar disks.
Kim et al. (1987) described the properties of the mica and mylar particles in terms of the
following three parameters
1. The average projected area, A
2. The average thickness, &.
3, Aneffective length, a=VA
The properties of the four systems studied by Kim et al. (1987) are listed in Table 1.1
and the degree of anisotropy can be thought of as characterized by the ratio, a/b.
Given the success of the simple two-dimensional models illustrated in Figures 1.8
and 1.9, it would seem reasonable to begin a study of anisotropic systems with similar
models such as those shown in Figures 1.14, From those figures it becomes immediately
apparent that anisotropic systems are much more complicated than isotropic systems. Diffusion and Reaction 49
TABLE 1.1. Geometrical characteristics of porous media particles
projected area | average thickness ] effective length
material A (mm?) b (mm) a (mm) alb
mica A 1.076 0.072 1.037 144
mica B 0.481 0.043 0.694, 16.1
mica C 0.235 0.033 0.485 147
mylar disks 2.483 0.076 1.576 20.7
In the theoretical analysis of Kim et al. (1987), the particles were all parallel to the
x-axis and the influence of the lattice vectors was accounted for by the ratio of the
lengths of the sides of the unit cell, L,/L,, where L, andL, are illustrated in
Figures 1.14.
a
by = 0.85 Lq/Ly =1 a/b=15
Figure 1.14a. Geometrical model of anisotropic porous media
L,/ly=5 a/b=15
Figure 1.14b, Geometrical model of anisotropic porous media 50 Chapter 1
The expectation of Kim et al. (1987) was that anisotropic porous media could be
modeled in terms of the porosity, €,,, the particle geometry, a/b, and some choice of
L, / L, (or the lattice vectors, ¢,). It was further assumed that a two-dimensional model
would be acceptable and that the mean orientation of the particles was sufficient to
characterize the porous medium, That these assumptions were incorrect is clearly
demonstrated in Figures 1.15 and J.16, There we have shown calculated and measured
8, =050 L,/ly=5 a/b=15
Figure 1.14c. Geomettical model of anisotropic porous media
values of ,D,, /G, and €,D,,/‘D, as a function of the porosity for several values of
L,/ L, and we see no choice of L, / Ly that provides good agreement between theory
and experiment, The value of a/b =15 used in the calculations is in the proper range
for both the mica and the mylar disks studied by Kim et al. (1987); however, there is no
geometrical information available for the vermiculite and the mica used by
Currie (1960)
While the presentation of theoretical results in Figures 1.15 and 1.16 is not
exhaustive, it does indicate that one cannot find a single value of L, / Ly that gives good
agreement between theory and experiment, More extensive calculations are available in
Kim et al. (1987) and in Ochoa (1989), and they all lead to the same conclusion that the
two-dimensional models illustrated in Figure 1.14 are insufficient to capture the
important characteristics of diffusion in anisotropic porous media, One of the interesting
characteristics of the more extensive calculations is that under all circumstances the
theory over-predicts the value of D,,. The reason for this seems clear from the in situ
photographs of the mica and mylar disk systems that were taken by Kim ef al, (1987),
and a sketch of the structure seen in the photographs is presented in Figure 1.17. It
would appear that the average orientation of the particles was parallel to the x-axis or
the bedding plane; however, the particles were arranged in groups that were inclined to
the bedding plane at angles in the range of @=+20°. It seems clear that the orientation
of the particles around the bedding plane is responsible for the experimental values of
€,D,,/', shown in Figure 1.16 being smaller than the theoretical values based on the
models shown in Figure 1.14. In addition, it would appear that the effective value of
a/b is less than the Value listed in Table 1.1 because of the manner in which the
Particles align themselves in groups. Diffusion and Reaction
Kimet al,
‘A mylar disk
V mica A
A micaB
O micac
° 02 04 06 08 1
Figure 1.15. Effective diffusivities normal to the bedding plane
° 02 04 06 08
a
Figure 1.16. Effective diffusivities paraitel to the bedding plane
10
51 52 Chapter 1
Figure 1.17. Observed particle orientation for anisotropic systems
On the basis of the results presented in Figures 1.15 and 1.16, we conclude that
successful geometrical models will be based on knowledge of the porosity, &,, the
particle geometry perhaps described by a/b, the mean orientation represented by (8),
and the mean of the square of the orientation, (87). The key to understanding what
geometrical information passes through the filter represented by Eq. 1.3-37 is the closure
problem given by Eqs. 1.4-58, however, at this time we do not know how to extract this
information directly and for the present we rely on intuition and comparison between
theory and experiment.
‘The first improvement on the calculations of Kim et al. (1987) was made by Séez et
al. (1991) who used a three-dimensional model of parallelepipeds in a planar
configuration, This led to theoretical results which provided non-zero values of
€,Dy, | Zy over a wider range of porosities than the two-dimensional models of Kim et
al. (1987). In addition, Séez et al. (1991) were able to obtain reasonably good
agreement between the theory and the experimental results of Kim et al. for €,D,, / Ty
using L,/L,=8 and a/b=15. Not surprisingly, their calculations for the effective
diffusivity parallel to the bedding plane, €,D,, / J, were uniformly too large since the
particles in their model were all parallel to the bedding plane.
The calculations of Séez et al. (1991) were followed by those of Quintard (1993)
who used a three-dimensional array of spheres to produce the results shown in Figure
1.10, In addition, Quintard (1993) studied the problem of anisotropic systems using a
three-dimensional model of cylindrical particles that were not arranged in a planar
configuration. With this model, Quintard was able to reproduce some of the
experimental data for ¢,D,, /7, shown in Figure 1.16; however, the calculations were
rather limited in nature because of the computational cost. Quintard’s (1993) study
clearly established that the mean orientation, (9), of the particles is not sufficient for the Diffusion and Reaction 53
determination of effective diffusivities parallel to the bedding plane and that one must
know something about how the particles are distributed about the mean.
The most recent study of diffusion in anisotropic systems is the work of Ochoa-
Tapia et al. (1994) who made use of Maxwell's closure problem in conjunction with an
ellipsoidal unit cell. The closure problem has the same general form as Eqs. 1,6-15, but
ellipsoidal coordinates are used to achieve an anisotropic system. Values of the effective
diffusivity orthogonal to the bedding plane are represented by €,.D., /@, in order to be
consistent with the nomenclature of Ochoa-Tapia et al. (1994), and the comparison
between theory and experiment for oblate particles is shown in Figure 1.18 for various
10
Kim etal.
A mylar disk
08 Vo omicaA
A mica B
O micaC
06 Currie
ee © vermiculite
© mica
02
0 02 4 06 08 10
Figure 1.18. Experimental effective diffusivities normal to the bedding plane and theoretical effective
diffusivities normal to the major axis of oblate particles,
values of a/b. The parameter a/b represents the ratio of the major axis to the minor
axis and thus has very nearly the same significance as the parameter listed in Table 1.1.
In this case we see good agreement between theory and the experimental results of Kim
et al, (1987) for a value of a/b ~ 9, while the experimental data of Currie (1960) would
be better fit by a value of a/b ~ 15. The value of a/b ~9 is consistent with the results
listed in Table 1.1 and the idea that the particles tend to align themselves in groups as
suggested in Figure 1.17, The predictions for effective diffusivities parallel to the
bedding plane are similar to those shown in Figure 1.16, i.e., the ellipsoidal unit cell
predicts values of €,D,, /‘, that are too large. The use of Maxwell’s closure problem
with ellipsoidal unit cells has the attractive feature that it produces analytical solutions
with a minimum of effort and the results can be used to determine some properties of
anisotropic systems. However, a general theory identifying the geometrical 54 Chapter 1
characteristics that pass through the filter represented by Eq. 1.3-37 still remains to be
developed, and the relation between an experimental system and a theoretical model,
illustrated in Figure 1.13 for isotropic systems, needs to be altered for anisotropic
systems. The idea that more information passes through the filter for anisotropic systems
ig illustrated in Figure 1.19 where we see that the unit cell used for theoretical
calculations must incorporate more information about the structure of the system.
Experimental system Theoretical model
Figure 1.19. Experimental system and theoretical model for anisotropic porous media
1.7 Conclusions
In this chapter we have developed the closed form of the volume averaged transport
equation for bulk diffusion with heterogeneous chemical reaction in porous media. We
have used this process to set up much of the mathematical apparatus associated with the
method of volume averaging, and we have illustrated the general framework of this
procedure which consists of:
1. Spatial smoothing in which the governing equations and boundary
conditions are joined.
2. Simplification of the spatially smoothed transport equation on the basis
of the disparate length scales that often exist for multiphase systems.
3, Development of a closure problem, first in terms of the spatial deviation
concentration and then in terms of the closure variables.
4. Comparison between theory and experiment in the absence of adjustable
parameters.
For systems that are isotropic with respect to diffusion, all four of these objectives have
been accomplished, while in the case of anisotropic systems we are still confronted with
the problem of adjusting geometrical parameters in order to obtain agreement between
theory and experiment. Diffusion and Reaction 55
1.8 Problems*
1-1. Show that the surface transport equation given by Eq. 1.1-19 can be
treated as quasi-steady when k,t* >> 1.
1-2. The interfacial flux constitutive equation given by Eq. 1.1-6 expresses
the net rate of adsorption as
my Nay = kiCay — KaCay> atthe y-X interface (1)
while the rate of adsorption is represented by
{ rate of
ane = kicay, atthe Y-K interface ey
in which k is a constant. If one thinks of the surface as consisting of adsorption
sites, the rate of adsorption should depend on the number of vacant sites. If the
dependence of the adsorption rate on vacant sites is linear, Eq. 2 should be
written as
{ rate of
} = k(1-@)cay. at the Y-K interface 3)
adsorption
in which 6 represents the fraction of occupied sites. Assume that @ is
proportional to the surface concentration and use the chemical kinetic
constitutive equation given by Eq. 1.1-18 in order to develop the following
heterogeneous reaction rate equation:
ke,
Ras =-—-—2L, atthe y- interface (4)
1+ Keay
This is generally known as the Langmuir-Hinshelwood formulation:
1-3, For the process of pure adsorption in the absence of heterogeneous
reaction and surface transport, Eq, 1.1-19 takes the form
OCs
Sat = mye Nays atthe yx interface a
When the linear interfacial flux constitutive equation given by Eq. 1.1-6 is
applicable, this reduces to
aeas
sit = kicy ~ kyCae> athe Y-x interface Q)
Mt
1s to all problems are available from the author. 56 Chapter 1
Under equilibrium conditions, one can use Eq. 2 to conclude that
chs = Keg ay @)
in which the equilibrium constant is defined by
k
Kay = ot 4)
oe 4)
In order to analyze the adsorption process, Eq. 1 is typically written as
OC ay .
Kg = My Nay» atthe y-x interface (5)
which makes use of Eq. 3 even though the surface is not at equilibrium. The
condition illustrated by Eq. 5 is known as local adsorption equilibrium and it
occurs when the departure from the equilibrium condition is small enough so that
it can be neglected. In order to find the constraint associated with Eq. 5,
decompose the surface concentration according to
fas = Cas + Cas ©
and then show that Eq. 5 is a valid approximation when the following constraint
is satisfied:
kat >> t @)
Here 1° represents the characteristic process time.
4-4 A model of diffusion and heterogeneous reaction in porous media can
be constructed in terms of a single cylindrical pore in which the process is
described by the following boundary value problem
cay 1A( Ay) , Pea
= G)-— ~ “|, inthe y—phi
e rar ar) * ae inthe y—phase a)
BC! Cay = Chy, 220 2)
deny
BC2 OGM = hems r=% 2)
a
BC3 <4 = 0, ga (4)
ae
LC. unspecified ©)
Here we have assumed that the catalytic surface at r =r, is quasi-steady even
though the diffusion process in the pore may be transient (Whitaker, 1986b). sion and Reaction ST
The steady-state form of the boundary value problem for c,, could be solved
by classical means; however, in this problem you are asked to develop the area-
averaged form of Eq. 1 and determine under what circumstances the
concentration at r=r, can be replaced by the area-averaged concentration,
(Cay). This is defined by
(ay) = ais Joxremar ©
rio
thus you are seeking a constraint that will allow you to use the approximation
a Oy
In your analysis of this problem, you should not use the formalism of the method
of volume averaging, but instead you should apply the definition given by Eq. 6
directly to the governing differential equation given by Eq. 1. This will lead to the
form
12(,) cae thaw 8)
| San anrdr =
mre mr,
0
rdar\ ar az?
Since the limits of integration are independent of 1 and z, one can interchange
differentiation and integration to obtain
Taf dear), wy Oem)
=f H- satan + Be )
0
Aca 2.
e
Direct integration and imposition of the approximation indicated by Eq. 7 will lead
to a useful form of the area-averaged transport equation.
After you have developed the area-averaged diffusion-reaction equation,
solve the steady-state problem and determine the effectiveness factor defined by
{ee rate o) react
in the pore
5 Cia (10)
total rate of reaction
in the pore if the j
concentration were c,
3
and indicate how n depends on the Thiele modulus. 58 Chapter 1
“1-5. If the function @(¢) is defined by
nab)
on) = Jrcanen @
nea
the one-dimensional Leibniz rule can be used to obtain
i ™o
ae ce a i
ee foben + flubonT - slay Q)
nEatt)
Derive this result beginning with the definition
do yn[ 8etad 00) 8)
dt at at
and express the integral between a(r+Ar) and b(t+Ar) in terms of three
appropriately chosen integrals. Then use a simple geometrical interpretation of
the integrals between a(t} anda(t+Ar) and between b(t) and b(t+Ar) to
complete the derivation. If the integrand in Eq. 1 depends only on the dummy
variable of integration, n, we have the special case
n=)
= fron @)
near)
and the special form of the Leibniz rule is given by
49 db da
Tbe da
7 «SHOE - flan )
The three-dimensional version of this result is encountered in the next problem in
which time is replaced by the arclength along a line.
“4-6. Derive the spatial averaging theorem using an approach analogous to
the derivation of the Leibniz rule in one dimension (Whitaker, 1985). Begin the
derivation by considering the derivative
dg jn [ 2 aot | ay
As
ds 4500}
in which 9(s) is the integral defined by
9) = Jue Q)
ve) Diffusion and Reaction 59
Here s is the distance along any line imbedded in the y-« system illustrated in
Figure 1-6a, and , is any function associated with the yphase. On the basis of
Figure 1-6a. Averaging volume and a two-phase system
Eq. 2, the derivative given by Eq. 1 takes the form
for = fone
VGA) Wis)
d a
s fuay = Jim 3)
Vs) Ss
The two averaging volumes, 7”(s+ As) and 9s), are illustrated in Figure 1-6b,
and a little thought will indicate that the intersection of V,(s+As) and V,(s) will
contribute nothing to the right hand side of Eq. 3. Once the intersection is
removed, you should make use of the projected area theorem (Stein and
Barcellos, 1992, Sec. 17.1) in the form
AV = th-ny dds (4) 60 Chapter 1
Figure 1-66, Differential displacement of the averaging volume
Here dA represents a differential portion of the surface of the sphere located in
the y-phase, and mn, represents the outwardly directed unit normal vector
illustrated in Figure 1-6a. Use of Eq. 4 will allow you to transform the two volume
integrals to area integrals over the entrances and exits of the y-phase contained
within the averaging volume, 2%. Make use of the fact that the derivative with
respect to s can be expressed as (Stein and Barcellos, 1992, Sec. 14.7)
da
fanny 5;
as G6)
where A is a unit vector tangent to the line illustrated in Figures 1-6a and 1-6b.
Finally, you should use the divergence theorem
[Par = Jrevys + Jove ©
vy Ay Ane
in order to obtain the traditional form of the averaging theorem.
1-7. In order to derive the averaging theorem for the divergence of a vector,
replace y, in Problem 1-6 with the vector a, in order to prove that Diffusion and Reaction 61
(Va,) = Vea,) + 3 [maa a
Ag
Form the double contraction of this result with the unit tensor I in order to obtain
the desired result given by
1
(Way) = Va) +o Jone ays @)
dre
1-8. For multiphase transport problems that are one-dimensional at the
macroscopic level, one would like to use an averaging volume in the shape of a
disk or a slab such as we have illustrated in Figure 1-8. To derive the averaging
theorem associated with this special averaging volume, one can follow the
solution of Problem 1-6 to obtain
a-Y| frre -2 Jurmeat =0 (1)
¥, Ae
however, 4. cannot be cancelled in this result en route to the traditional averaging
theorem. Instead we multiply by 2 and divide by % so that Eq. 1 takes the form
1
N-V(y,) - N $ fv =0 Q)
ke
in which N is the normal tensor defined according to
N= A (3)
This is related to the unit tensor and the projection tensor by
l= N+P (4)
in which P is the projection tensor for the plane that is orthogonal to the unit
vector 2. When the thickness of the slab illustrated in Figure 1-8 tends to zero,
show that Eq. 2 can be used to derive the form given by
1
Wy,) = NV) +L vind, Ae =0 6)
An 62 Chapter I
Figure 1-8. Averaging volume in the shape of a slab or a disk
4-9. Develop an extension of the averaging theorem for the y-phase in the
three-phase system illustrated in Figure 1-9. This can be done by considering
the x-phase shown in Figure 1-6a to be made up of the o-phase and the
B-phase.
averaging
volume, Y
Figure 1-9. Three-phase system
4-10. In the process of drying a wet porous medium (Whitaker, 1998), one
encounters the three-phase system illustrated in Figure 1-9 in which the y-phase
represents humid air and the B-phase represents pure water. The continuity
equation for the B-phase is given by Diffusion and Reaction 63
oj
> + V-(pgvg) = 0 (1)
and the volume averaged form of this equation is an essential feature of the
theory of drying. Since the volume of the B-phase contained within the averaging
volume is a function of time, one needs to make use of the general transport
theorem in order to complete the averaging process. The general form of this
theorem is given by
d OV
= av = |—tav :
a Ju J av + Jvow ndA )
eo Huy a
in which %(t) represents an arbitrary, time-dependent volume and o¥,(r)
identifies the bounding surface of this volume. The speed of displacement of the
bounding surface is given by w-n in which n is the outwardiy directed unit
normal vector. Since %;(t) is arbitrary, it can be set equal to V4(t) in order to
obtain the special form that is needed in the analysis of Eq. 1. A key quantity of
interest in the drying process is the mass rate of evaporation per unit volume that
is defined by
A 1
(i) = Joss = W)-11ydA @)
Ay
In this problem you are asked to develop the volume averaged form of Eq. 1 in
order to illustrate how the mass rate of evaporation, (yn), is related to the time
rate of change of the volume fraction of the B-phase, e,, and the superficial
average velocity, (vg). Begin your analysis with Eq. 1 and consider the density
to be constant to obtain the final form of the volume averaged continuity
equation. If the o-phase is impermeable, the mass transfer boundary condition
that can be extracted from Eq. 1.1-2 is given by
B.C.l Pp(vg—W)-gg = 0, at orf, (4)
and if the porous medium is rigid, we have
VpeMyy = WMgs = 0, atch, 6)
At the B-y interface, Eq. 1.1-2 leads to the following boundary condition
B.C2 Pp(vp—W)-Mpy = Pay(Vay—W)-Mpy + at fy ©
1-11. In the method of volume averaging, one often uses the approximation
(DV eu) = TlVeay) «) 64 Chapter I
even though @%, may vary significantly over the macroscopic region of interest.
In general, the molecular diffusivity will depend on both the concentration and the
temperature; however, in this problem we consider only the functional
dependence of @, on the concentration c,,. The objective of this problem is to
determine the constraint associated with Eq. 1 when @%, on the right hand side is
evaluated at the average concentration, (c,,)’. Begin by representing @ in a
Taylor series expansion about the concentration (c,,)” to show that (Whitaker,
1998, Sec. Vill)
az, mH 7 y a (2)
a. a + He +
Keng)” "Baie, ey Keay)"
Consider only the linear term in Say and make use of the over-estimate
EnyVeay) = O(Fay(Veay)) (3)
to demonstrate that Eq. 1 is valid when the following constraint is imposed:
Fey << 1 (4)
Here we have made use of the estimate for ay given by
By = O(2,Veay)") 6)
on the basis of the analysis presented in Sec. 1.4.
1-12. Use the averaging theorem for the +x system to prove that
1
co Jrvas = -Ve,
An
Begin by letting y, in Eq, 1.2-15 be a constant.
1-13. Use the averaging theorem for the y-« system to prove that
1
Fe mvt = -Vy,)
Aye Diffusion and Reaction 65
Begin by letting y, in Eq. 1.2-15 be equal to the position vector r, that locates
points in the y-phase. Then let r, = x+y, in order to prove the above result
1-14, Given some function, f(x), that has an estimated magnitude of A in
the domain 0
ye = Ve Deg Keay") — vk (Ceay)”) @)
Most of the analysis associated with this problem is available in the text;
however, some effort must be directed toward the quantity (c4,),. that appears
in the superficial average of Eq. 1. Chapter 2
Transient Heat Conduction in Two-Phase
Systems
In this chapter we consider the problem of heat conduction in two-phase systems without
the complications associated with adsorption, chemical reaction, phase change, etc, Our
objectives in this chapter are two-fold:
1. To study a process in which wansport occurs in two phases.
2. To indicate how a one-equation model can be developed in
order to describe transport in a two-phase system.
In the previous chapter we analyzed a process involving transport in one phase
(the y-phase) with the second phase (the «-phase) providing only a boundary condition.
Extending our studies to include transport in two phases is an important step since there
are many important processes involving heat and mass transfer in both two and three-
phase systems, and in many of these systems one can develop a reasonable model of the
transport process using a single transport equation. Knowing how to derive this single
transport equation and knowing when it is a valid representation of the process are the
key objectives of this chapter.
2.1 Governing Equations and Boundary Conditions
‘The system under consideration is illustrated in Figure 2.1 and the governing equations
and boundary conditions are given by
at :
opps = V-(ksV%), in the B- phase 1-1)
BCI Th = T,, atthe B-o interface (2.1-2)
BC2 ~nggkgVTp = — Ngo koVT,, atthe B-o interface (2.13)
aT, A
Poss = V-(kgVT,), in the o- phase (21-4)
B.C3 Ty = Fest), atopy (2.15)
= Heat Conduction 73
BC4 Ty = Gest), a ohey (21-6)
LC. Th = #(@), att=0 (21-7)
Leo Tz = Sr), att=0 (2.1-8)
‘The boundary conditions given by Eqs. 2.1-2 and 2.1-3 indicate that the temperature and
the normal component of the heat flux are continuous at the B-o interface. If chemical
reaction, adsorption, or phase change takes place at the B-o interface, a jump in the heat
volume,
Figure 2.1. Macroscopic system
flux can occur and the heat transfer process becomes coupled to a mass transfer process.
Under these circumstances, the problem becomes considerably more complex.
In our formulation of this transient heat conduction problem, we have used , are defined by the estimates
Vey, = vm (2.2-13) Heat Conduction 77
In arriving at Eq. 2.2-11, and the constraints indicated by Eq. 2.2-12, we have assumed
that the B-o system is disordered (Quintard and Whitaker, 1993a, 1994a-e) and this
eads to the first length-scale constraint in Eq. 2.2-12. The definition of a disordered
system and the development of the length-scale constraints given by Eq. 2.2-12 are given
in Sec. 1.3. The o-phase transport equation is analogous to Eq. 2.2-11 and we list the
result as
i
aT ,)° o, -
&a(Pep), 7 Y [Ke] ee MKa)” + H Jnenfeaa
Aas
(2.2-14)
a
+ # fre kgVT,dA
Ao
Here we have made use of the nomenclature given by
Myo = —Mgy. Agg = Aap 22-15)
‘The first of these results from the fact that ngg represents the unit normal vector pointing
‘from the B-phase toward the o-phase whereas the unit normal vector ngg is pointing
from the 6-phase toward the B-phase. The two interfacial areas represented by Agg and
Agp %¢ identical; however, we have used a nomenclature that is consistent with the unit
normal vectors in an effort to avoid making a sign error in terms of ng, and np.
Equation 2.2-14 is naturally restricted by the analogous form of Eq. 2.2-12 which we list
for completeness as
2
oe ee (2.2.16)
If we were interested in developing separate equations for both the f-phase and the
G-phase, we would represent the interfacial flux in both Eqs. 2.2-11 and 2.2-14 in terms
of average temperatures and spatial deviation temperatures and then go on to develop the
closure problem for 7 and 7. This is often necessary if convective transport in the
B-phase is important; however, for most problems of pure conduction one can use a
single average temperature to describe the heat transfer process. In the next section we
will derive the one-equation model for transient heat conduction, and we will develop
the constraints associated with the principle of local thermal equilibrium,
The essential feature of the principle of local thermal equilibrium is that the equality
(h)’ = (T,)° (22-17)
represents a reasonable approximation. If one accepts this idea, Eqs. 2.2-11 and 2.2-14
can be added to obtain 8 Chapter 2
ary k = k, .
rc, 32 = VJ cephg +eok, WIT) + 5B / noha + Fe foogisaa
Ate Aon
(2.2-18)
in which we have adopted the nomenclature represented by
(PCy = EplPep)p + Eo (Pepa (2.2-19)
In Eq. 2.2-18 the single temperature (T) is referred to as the spatial average
temperature which is defined by
= | tav = e(h)? + e9(%)? (2.2.20)
cS
and when the condition of local thermal equilibrium is valid we have
(Ty = (hy = (EY (2.2.21)
To obtain 2 closed form of Eq, 2.2-18 we must develop representations for 7 and 7,
and this is done in Sec. 2.4, There we will find that the conduction terms on the right
hand side of Eq. 2.2-18 can be expressed as
(egky +Egko VAT) + 2 | matin + #& Jrertaas = Ky VT) (2.2-22)
4a Jo
in which K,y is the effective thermal conductivity tensor. This leads to the closed form
of Eg. 2.2-18 given by
oc, 22 = V-(Ky Vir) (2.2.23)
and the macroscopic boundary conditions for this equation are available in the work of
Ochoa-Tapia and Whitaker (1998a)
2.2.1 THE FILTER
As in our study of diffusion and reaction in a porous catalyst, our volume averaged
equation contains a filter that we identify explicitly as
ky [ spatial
FJ ssehaa +Z frosted = 4deviation (2.2-24)
ie S filter
In our previous study of diffusion and reaction, we found that little more than the
porosity passed through the filter for isotropic systems, while the situation was more Heat Conduction =
complex for anisotropic systems. For the case of heat conduction, we will find that the
porosity is an important parameter, and under some circumstances the particle-particle
contact will play a key role in the determination of the effective thermal conductivity
Before going on to the derivation of Eq. 2.2-23 and the comparison of predicted
values of Kyy with experimental results, we need to explore the idea of local thermal
equilibrium; however, those readers who are willing to accept the condition of local
thermal equilibrium can move directly to Sec. 2.4.
2.2.2: TWO-EQUATION MODEL
When local thermal equilibrium is not valid, one must develop the closed form of
Eqs. 2.2-11 and 2.2-14, and Quintard and Whitaker (1993b) have shown that the two-
equation model for transient heat transfer is given by
amy?
560) pe = V(Kyp HTP + Kyq WT)? has
= a,h((%)° ~(7,)°)
foley eX = V(Kag PUI) + Koo VW)*)
(2.2-26)
= ayh((T,)° ~(Tp)*)
In their theoretical studies Quintard and Whitaker (1993b) were able to prove that the
cross coefficients were equal
Ky = Kop (2.227)
and solutions of the closure problems indicated that these coefficients were generally on
the order of the smallest of either Ky, or Kg. Detailed numerical experiments were
used to test the theory leading to Eqs. 2.2-25 through 2.2-27, and we will draw upon
those results in our discussion of local thermal equilibrium in the next section,
2.3 Local Thermal Equilibrium
The phrase local thermal equilibrium is strictly associated with heat transfer in
multiphase systems, and it should not be confused with the concept of local
thermodynamic equilibrium (Truesdell, 1969), In studies of multiphase transport
phenomena, one encounters one-equation models of heat and mass transfer that are
constructed on the basis of local thermal and mass equilibrium. In this section we will
examine the concept of local thermal equilibrium for transient heat conduction in a two-
phase system and leave a variety of more complex heat and mass transfer problems for
further study, 80 Chapter 2
If one believes, on a purely intuitive basis, that (7p)" and (T,)° are essentially
equal and can be represented by a single temperature, it is not unreasonable to put forth
the assumption (Whitaker, 19882)
ASSUMPTION: (T)® = (T,)° (23-1)
and then explore the consequences. The theoretical consequences have already been
indicated by Eqs. 2.2-18 through 2.2-23. While intuition often encourages the use of
Eq. 2.3-1, a more prudent approach (Whitaker, 1977) is to represent the two intrinsic
average temperatures in terms of the spatial average temperature and two Darcy-scale
spatial deviation temperatures
(Tpy® = (T) + Thy (Ty) = (T) + Ty (2.3.2)
and then identify the conditions for which and 7, are negligible. Itis obvious that the
condition indicated by Eq. 2.3-1 will be satisfied when the spatial deviations are zero,
Le.
Ty = 0, Ty = 0, local thermal equilibrium is valid (2.3-3)
This condition only occurs at thermodynamic equilibrium; however, it is possible that
for some processes fy and 7, will be small enough so that these spatial deviation
temperatures can be neglected. In order to determine what is meant by small enough, we
substitute Eqs. 2.3-2 into Eqs. 2.2-11 and 2.2-14 and add the result to obtain
XT) 7 ky he .
(oye, <2 = V-[(eptp tek VAT) + 2f 7 Fda
ah at,
tplPep Zr + CaPep ase ~ V
(2.3-4)
If the last four terms are negligible (as opposed to being zero as indicated by Eq. 2.3-3),
a valid one-equation model can be extracted from Eq. 2.3-4. By identifying the
conditions for which the last four terms in Eg. 2.3-4 are negligible, the intuition
represented by Eq. 2.3-1 will be replaced with a set of constraints that will indicate the
domain of validity of the one-equation model.
We can simplify the last four terms in Eq. 2.3-4 by using the decompositions given
by Eqs. 2.3-2 to conclude that (see Problem 2-1)
tly = ~€afy = €ptq((T)? -(To)’) 23-5) Heat Conduction 81
and if we are willing to ignore variations of the volume fractions in the last four terms we
can express Eq. 2.3-4 as
aT) k = = =
CG, MP = Vlegkg tegkedV(T) + Ef mates +e frasfoaa
ne
Ase Acp
- { etal (Pep Ip (Pe po] (6H)? (Tad) -v [tals A) ( ca? ay")
(2.3-6)
Here it becomes clear that we must be able to estimate the temperature difference
(T)* - (T° if we want to determine when the condition of local thermal equilibrium is,
valid.
The process of discarding the last four terms in Eq. 2.3-4 deserves some thought
since there is more than one way in which the restrictions (Whitaker, 1988a) can be
arranged. To begin with, it is prudent to discard various terms in an equation relative to
the smallest non-zero term thet is retained. Since the left hand side of Eq. 2.3-4 will be
zero for a steady-state process, it seems wise to center our attention on the conduction
erm on the right hand side of Eq. 2.3-4. It is reasonable to use Eq. 2.2-22 as a
representation of this term, and then base the condition of local thermal equilibrium on
the following restrictions:
RESTRICTIONS:
Tegel epdp (Pepe) A(T" —(Ta)®) << V(Key VK) 23-7
0 V- [ep (hy Ke) ¥((Ts)" ~(Tq)°)] << V-(Kuy -VKr) (2.38)
To obtain useful forms of Eqs. 2.3-7 and 2.3-8, we need to estimate the time and space
derivatives of (7) -(T,)° and (7). The estimates for (Ts)° —(T,)° are given by (see
Sec. 1.3.2)
Tp )° - (Ts)
2 (aya) . of Meat") (23-9)
Roe
(Ty)? -(Te)") = of Sea 400") (23-102)
Ly 82 Chapter 2
A((75)° = (T5)°)
(2.3-10b)
Lails
V7((15)" ~(T)%)
in which 1° is a characteristic process time, Ly is the characteristic length associated
with the changes in (7s) -(T,)®, and La; is the characteristic length associated with
the changes in V((7)*—(T,)®). Since the temperature difference (T})® —(T,)%
represents a deviation, as indicated by Eqs. 2.3-2, we can approximate the change in this
difference with the difference itself. This leads to the estimates,
(7,)" =(T,)°)
a a) \
5)" -(7a)*) = 9 B (2.3-11)
Ty -(r.)%
¥((T)—(Tq)°) = of (P= aN*) (2.3-12a)
Ly
es (my -)°)
V°((1)°-(T.)") = O} a (2.3-12b)
Latha
Turning our attention to the right hand side of Eqs. 2.3-7 and 2.3-8, we estimate the
gradients of the spatial average temperature as
VT) = 2) (2.3-13)
r
A(T)
vvur) = O 2.314)
: (i) on)
Here we think of L;, as the characteristic length associated with changes in the gradient
of the spatial averaged temperature, and if the gradient is constant the form of Eq. 2.3-14
indicates that L,, is infinite. This situation occurs when the volume averaged heat
conduction process is steady and one-dimensional.
When the estimates given by Eqs. 2.3-11 through 2.3-14 are used in Eqs. 2.3-7 and
2.3-8 we obtain the following two restrictions associated with the condition of local
thermal equilibrium:
EpEol (Pep \p — Pepa fErr Lr (ea) a
(2.3-15)
Kye air) | Heat Conduction 83
I (2.3-16)
SE o(kg ky) ((Tp)P=(Ta)”)
K A(T)
eff
In extracting the second of these from Eq. 2.3-8, we have assumed that the length scales
for (Tp)P —(T,)° are the same as those for (7), ie LyLy = Ly,Ly. For many systems
of practical importance, the physical parameters and the length and time-scales will be
such that
FpEo[ (Popp — Pepa Ere
s O(1 (2.3-1
Kye (1) (23-17)
pee Soe 4
fpfoKn=te) _ ott) (2.3-18)
Key
Under these circumstances, the condition of local thermal equilibrium will be dominated
by the quantity «Hy —(Ty)%)/ A(T), and it is our ability to estimate this quantity that
allows us to determine when local thermal equilibrium is valid and when it is not, The
best estimate of (7,)*-(7,)° is based on the governing differential equation for this
quantity, along with the appropriate boundary and initial conditions. An analysis of the
general heat transfer process is given by Whitaker (1991b), and in the following
paragraphs we consider the special case of transient heat conduction.
2.3.1 ESTIMATE OF (7%)° —(T,)°
The governing differential equation for (7;)’—(T,)% can only be developed in an
approximate sense, and in order to keep the analysis as simple as possible we ignore
variations in the volume fraction so that the B-phase transport equation given by
Eq. 2.2-11 can be expressed as
ah)?
ar
-1 1
y &p F ©
(cpp = Vf kyl VTE + SE [ooh aa + E Jong kVp dA
Abe Me
(2.3-19)
The analogous form for the 6-phase is obtained from Eq. 2.2-14 according to
OT)
or
o ea = cay
(ep )e Vk V(T,)° + $e [ng tt . 5 Jag: haV Te aA
Aop Aop
(2.3-20) 84 Chapter 2
It is not immediately obvious that subtracting Eq. 2.3-20 from Eq. 2.3-19 will lead us to
a useful expression for the temperature difference (7)*~(T,)°; however, we will
explore this possibility to obtain
amy? aT)”
(Pep)p oe Pepa a
. :
fosatiaa - ‘eke J neg ied
Ate Asp
ood
Se fis kg Vida
An
= Vk yV(T)® - koVTe)® +
ES
Here the last term has been extracted from the interfacial flux terms in Eqs. 2.3-19 and
2.3-20 along with the flux boundary condition given by Eq. 2.1-3 (see Problem 2-2).
We now make use of the representations given by Eqs. 2.3-2 so that the above result
takes the form
ah ar, awn a p
(Pep GP ~ (ep) AE + [Pep)p~C0ep e] = ¥-(KyVH -KeVFe)
kgec!
ke . .
+ V-| (ky kg V(T) + J rset aa . JrosTeaa
Ae fog
Exlg)”
. Ce ove kV, da
fea
(2.3-22)
‘This still represents a relatively unattractive form; however, if we use the expressions
given by Eq, 2.3-5 for 7, and 7, we can arrange Eq. 2.3-22 in the form Heat Conduction 85
[estoep)p+es(0¢p eo] > ((T)®-(72)°) ~ ¥-[Ceoky +4) V(t —5)”)]
kgeg =
Wy Digg 7p dA
Ate
an, of
= ~[0¢:25 -0¢p 9] + Vil pk) VT) +
JrasToaa + Cee as kg VIpda
Acs M30
ca
(2.3-23)
Here we can see the beginnings of a transport equation for (7) -(7,)° with the
accumulation and conduction terms on the left hand side, and the so-called source terms
involving the spatial and temporal derivatives of {7} on the right hand side. In order for
this result to be useful, we require that it contain only terms involving (7 )® -(7,)° and
the derivatives of (7)
We can make use of the work of Quintard and Whitaker (1993b) in order to express
the interfacial flux as
1 :
wa Jrso-tsipaa = -ail(h)? ~(T)°) (2.3-24)
Age
and we can introduce the mixed-mode parameters defined by
Pep )po = Fo PopptEpPona» kp = Eakp+€pka (2.3-25)
so that our transport equation for (7j)° ~(7,)° takes the form
(0p 4m 2 (Tp) (Tad) ~ Vhs 0(CH® ~(7.)°)] + peo ayH((T)® —(75)°)
1
F Joscfaa
bso
XT) ky
=- [65 )5~ ep 9] + Vg ky )VIT) +
“1
S fete. Jropfsaa
Ag J
(2.3-26) 86 Chapter 2
When local thermal equilibrium is valid, we will show in Sec, 2.4 that
hy
,, at the B-o interface (23-27)
and we will also show that the spatial deviation temperatures can be represented by
Ty = by-V(T), 7 = by-ViT) (2.3-28)
As an approximation, we can use these two results in Bq. 2.3-26 to obtain the following
transport equation for (73)° -(T,)”
(Pe, bo Z(H )°~e)°) ~ ¥-[Kpo¥U(t53* ~<7e)")}
accumulation Clee
+ (pq) ayh{(7)* -(T)°} (23-29)
exchange
= [59-06 42D + Vflly-teeGy) V0)
duction source
‘accumulation source
Here the second order tensor, Cp,, is defined by
CAS
c
© (lp he)”
Poacby a (23-30)
Ave
and we refer to Cgg as the phase geometry tensor since it provides important
information concerning the physical properties of the two phases and the manner in
which the two phases are arranged. One should note that the right hand side of
Eq. 2.3-29 is zero when the physical properties of the two phases are equal, and this
suggests the intuitively appealing result that
(Ty (Ty) = 0, when kg=k, and (pep) = (Pepa (2.3-31)
One must keep in mind that this conclusion is based entirely on the governing
differential equation for (7p)" —(74)° and does not take into account the influence of
either the boundary conditions or the initial condition.
In order to use Eq. 2.3-29 to estimate ()* -tay°)/acr) we need estimates of
the derivatives of (7})* -(7,)° and (7). We begin with the left hand side of Eq. 2.3-29 Heat Conduction 87
and estimate the space and time derivatives in the manner indicated by Eqs. 2.3-11 and
2.3-12. This leads to
(1)° =(T,)°}
2 (qy?-a1") 7 |! (2.3-32)
t
BL o
97( 9 —)") = na ee") 2.3.33)
[Erle
in which we have again assumed that Ly,L, ~ LyyLp. Use of these results allows us to
estimate the left hand side of Eq. 2.3-29 as
(eye 2 ((T)® ~Ta)”) ~ ¥ [bao I)" -(Ta)")] + (epee) 'ah((T)® ey”)
[ rc) k . :
- jo 2 o | + 24 bine)
t Lyky Epes
‘accumilation condiction exchange
(2.334)
Here one must keep in mind that the signs (+) of the three estimates are generally
unknown, If the terms representing accumulation, conduction, and exchange are the
same order of magnitude, our overall estimate will not be very reliable since we do not
know whether the terms will add to produce a result that is large or will subtract to
produce & result that is small, When we examine the right hand side of Eq. 2.3-29, we
find a different situation since we can use Eq. 2.2-23 to express those two terms as.
= [lop pepo] + V[lhy~KeYl+Cp6) VD]
Fl(ko~ Kall) P|
accumulation source ee
(2.3-35)
[Coe )p = ep a] V:(Kuy -V(T))
WC, V-[(& ~kg)(l+ C5) VT)]
It is reasonable to ignore variations in K.y and (kg—k,)(1+Cg,) so that this result
takes*the form 88 Chapter 2
- [pp ~(0¢y9/°2 + Vi (kp ~kel(l+ Cp) VET)]
eee eae
cont
‘accumulation source eee
23-36)
(Pep)p ~ ep o]K
: hy kg le Cpq) ~ [e020 = endo Ker : Pet VM,
(ks —ko)(l+ Cp) DC, ?
At this point we express the second derivative of the temperature as
ACT)
WT) = of 2%) (23-37)
Linky
in order to obtain an estimate of the form
aT)
= [lop )p- epee + V [lp -ko)(1+Cj0)-¥07)]
Ee
accumulation source: enema
(2.3-38)
[teen )p-Gena]Ker |/ acy
= Olfky —kg)(1+0,q) - (Owe Pero] Kar Wf ary
oft o(1+Cp0) WC, (ar)
in which Cg, and Kay are suitable norms of Cy and Ky. Substitution of this result,
+ with Eq. 2.3-34, into the governing differential equation for (7p)° —(7)° leads to
ving estimate:
; [teeta _ [este
hy ~ To)" | “0 Ko (p)C, Ka |
Ar) Uh) {te y Be
L + of = + of
exchange 2 Opt
conduction accumulation
(2.3-39)
Here we have defined the mixed-mode thermal diffusivity as
ky
Bo.
Ogg = (23-40)
me Cp po
and the mixed-mode, small length scale is given by Heat Conduction 89
ERE gh;
es ow 2.3-41
Bo ai ¢ )
For simplicity we have replaced Ly, L; with 1”, and one must remember that these two
length scales will depend on time for transient processes. It is of some importance to
note that Ly, will be infinite for all steady, one-dimensional conduction processes, and
for these conditions Eq. 2.3-39 indicates (7j)* —(T,)° will be zero. This means that
local thermal equilibrium is always valid for steady, one-dimensional heat conduction,
For many systems we expect that the following estimates will be satisfactory:
((0¢,)p- Pepe) Key
Ke ow, (ks ~ka}{1+ Cpe)
= Ol) (23-42)
(PC, Kb, Ko 0
Under these circumstances the ratio of length scales, (60/1)? will control the estimate
given by Eq. 2.3-39, and will thus determine when the condition of local thermal
equilibrium is valid.
Earlier we pointed out that information about a field, such as (7)*~(7,)°, was
best obtained by an examination of the governing differential equation, the boundary
conditions, and the initial condition, However, the estimate of (73)° -(7,)° given by
Eq. 2.3-39 is based only on the governing differential equation and thus must be used
with some care. For all its limitations, Eq. 2.3-39 does provide us with an estimate of
(mp) -()°) /acry in terms of the system parameters and this means that it can be
used with Eqs. 2.3-15 and 2.3-16 to produce constraints that identify the domain of
validity of the one-equation model. From the estimate given by Eq. 2.3-39 and the two
restrictions given by Eqs. 2.3-15 and 2.3-16, we can sce that local thermal equilibrium
will always be achieved when any one of the following three conditions occurs:
1. Either €g or €, tends to zero.
2. The difference in the B-phase and o-phase physical properties tends to zero.
3. The square of the ratio of length scales, (fg, /L)*, tends to zero.
2.3.2 COMPARISON WITH EXPERIMENT
Every theory should be compared with experiment, and the estimate given by Eq. 2.3-39
is no exception. A comparison with laboratory experiments would be extremely
difficult; however, a comparison with numerical experiments is possible and this has
been done by Quintard and Whitaker (1993b, 1995a) who studied the spatially periodic
array of cylinders shown in Figure 2.3. The experiment consisted of solving a boundary 90 Chapter 2
unit cell
Figure 2.3. System used for numerical experiment
value problem to produce values of Tj and 7, which were integrated over the B and
o-phases contained in the unit cell illustrated in Figure 2.3, This provided the
“experimental” values of (7,)° and (T,)° as a function of x and ¢ while “theoretical”
values of (7) were obtain from Eq. 2.2-23 and the theory described in Sec. 2.4. The
boundary conditions for the macroscopic system were as follows:
BC.1 Th = To. at x=L,, £20 (2.3-43)
BC2 Th=T, a x=0, t>0 (23-44)
Le, Ty =T, Ts =T), t=0 (2.3-45)
The parameters for one of the numerical experiments performed by Quintard and
Whitaker (1993b, 199Sa) are given in Table 2.1, and those parameters are representative
of a porous medium composed of glass beads and air with a volume fraction of the
TABLE 2.1. Properties of a Two-Phase System. Numerical Experiment
| 4% ky (cp)s (ep)s “Ka [hs
(Wim-K)_ | (Wim-K) (im3-K) (im3-K) (Wim-K)
0.026 05 1202 17x 10° 21
continuous phase given by £5 =0.62. The thermal conductivities in the two phases
differ by about a factor of 20, while the heat capacities differ by about a factor of one
thousand. The effective thermal conductivity is dominated by the conductivity of the
B-phase aad is about twice the value of ky. The numerical results for (7p)° and (7, )°
are presented in terms of dimensionless temperatures defined by Heat Conduction ue
(h)* - 7 oo
©) = — ae Oo)" = oY a (2.3-46)
iT, it
A comparison of these two intrinsic average temperatures can be used to determine
whether the condition of local thermal equilibrium is satisfied for the system described
by the parameters given in Table 2.1. This comparison is shown in Figure 2.4 where we
I
—- (05) exp.
08
2 (Bo) exp.
Z
2 6 {@) theory
&
&
a 04
a
3
E
7
ot
x/Ly
Figure 2.4. Comparison between theory and numerical experiments for the parameters in Table 2.1
have also presented results for the dimensionless spatial average temperature defined by
| ia
oo (23-47)
This temperature was calculated using the one-equation model given by Eq. 2.2-23 and it
represents a theoretical result. The values shown in Figure 2.4 are for two
dimensionless times, t= 0.0038 and t = 0.034, where t is defined by
Key t
ts oth — (2.3-48)
IC, Le
We expect steady state, and therefore local thermal equilibrium, to occur for values of t
on the order of 0,3, thus the temperature profiles shown in Figure 2.4 are far from steady
state. One should keep in mind that the system is originally in local thermal equilibrium
according to the initial condition given by Eq. 2.3-45. At short times there is a departure
from local thermal equilibrium; however, the agreement between the experimental values 92 Chapter 2
of (p)® and (@,)° clearly indicates that local thermal equilibrium is achieved for
dimensionless times greater than or equal to 0.0038.
Not only do the results shown in Figure 2.4 indicate that local thermal equilibrium is
achieved for 120.0038, but they also indicate excellent agreement between the
experimental results and the theoretical vafues. One need only think about using (@g)°
and (@,) to compute an experimental value of the spatial average temperature
ey = € (Os)? + €6(0,)° (2.3-49)
in order to see the excellent agreement between the theory and the numerical
experiments.
To explore the validity of the estimate given by Eq. 2.3-39, we express it in terms of
the dimensionless temperatures according to
(0¢p)p- cp) rp ep a) Key
po eGicsmse
2 2
£
l + ic + of :
| exchange L Opt
Se
conduction accumulation
)
a
04)" ~ a)” = of“
23-50)
Quintard and Whitaker (1995a) have evaluated the right hand side of this result for the
conditions illustrated in Table 2.1, and for the two dimensionless times indicated in
Figure 2.4, they obtained the following estimates:
(5) - (@,)% = Of01), t=00038 (2.3-51a)
(@5)° - @,)° = O(0015), t=00345 (2.3-51b)
This first of these would seem to be quite reasonable while the second would appear to
underestimate the difference between (@p)° and (@,)7. Given the number of estimates
that are required to evaluate the right hand side of Eq. 2.3-50, we conclude that
Eqs. 2.3-51 are in reasonably good agreement with the numerical experiments.
One of the interesting characteristics illustrated by (@g)* and (@,)° is the small-
scale fluctuation associated with che characteristic length of the unit cell, fy. These
Auctuations were first observed by Prat (1989, 1990, 1992) in a set of numerical
experiments dealing with the boundary condition between a porous medium and a
homogeneous fluid, and the matter has been investigated in detail by Quintard and Heat Conduction =
Whitaker (1994a-e). In their study, it was proved that the proper average to be used
with spatially periodic porous media is a weighted average known as the cellular
average. By proper average, we mean an average that is devoid of fluctuations at the
small length-scale, The cellular average can be obtained by averaging the results shown
in Figure 2.4 over a unit cell, and a little thought will indicate that this second average of
(@p)* and (©,)% would produce a result that would be indistinguishable from the
theoretical values represented by (@). In addition to proving that the cellular average
was the proper average to use with spatially periodic porous media, Quintard and
Whitaker (1994a-e) demonstrated that the traditional spatial average is the proper
average for disordered systems.
Perhaps the most unusual aspect of the results presented in Figure 2.4 is the fact that
excellent agreement between theory and experiment has been obtained even when the
traditional length-scale constraints have not been satisfied. For the system illustrated in
Figure 2.3 and the results shown in Figure 2.4, we see that the characteristic lengths, &,,
r,, and L are all essentially the same order of magnitude and not at all disparate as
suggested by the traditional constraints given by (Whitaker, 1969)
fy has been used to
represent L,, L, and that L? will depend on time for transient processes. A reasonable
estimate of this macroscopic length scale for transient processes is given by
LW) = O(Ky t/(p)C,) (2.3-54)
and this suggests that one can always find significant differences between
(7)? and (7,)° if the time is small enough. On the other hand, at short times the
difference between (7})* and (7,)% may be controlled by the initial condition and this
was not taken into account in the development of Eq. 2.3-39.
We close this section by noting that Eq. 2.3-53 can be used with reasonable
confidence to identify conditions for which local thermal equilibrium is valid, and that
Quintard and Whitaker (1995a) have given convenient analytical expressions for a,h 94 Chapter 2
that can used to determine fgq as defined by Eqs. 2.3-41. For regular arrays of cylinders
one can use Chang's unit cell to obtain
ahs 8ne%,
= ——__+—®__ _.,, pylinders (23-55)
kg [eq(R+1)~3}e -41n( Je.)
where R is given by
R = epkp/Eoky (2.3-56)
For regular arrays of spheres one finds
2 2
ails 40ce(a.? +0441}
= + theres (23-57)
ky (1+5«)+0(2+)+(04 +207 +307)(1-«
in which the parameters ot and x are defined by
a = (Ines), = Kelle (23-58)
In both Eg. 2.3-55 and Eq. 2.3-57, we have used fy to represent the characteristic length
for the B-phase and the manne in which fp is related to the parameters describing
Chang's unit cell is described by Quintard and Whitaker (1995a).
2.4 Closure
In order to obtain a closed form of Eq. 2.2-18 we need to develop the boundary value
problem for 7 and 7, and much of our analysis in this section follows that presented in
Chapter 1. We recall the original boundary value problem from Sec. 2.1
(ep dp et = V(kgV%), in the B- phase (24-1)
BCL Ty = Ty, atthe B-o interface (24-2)
B.C.2 —ng,-kgVTy = —Mgg-kgV7,, at the B-o interface (2.4-3)
(Pepe = Vi(kgV%), in the 6 ~phase (24-4)
B.C3 Th = Art), at mp, (24-5)
B.C4 Ty = G(e,t), awh, (2.4-6) Heat Conduction 95,
and note that the spatially smoothed form of Eq. 2.4-1 was given in Sec. 2.2 as
EY _ vlad evens . L
Ep (0p), — = 8) EBVCTY +> | Mpa tadA || + Se | Mpg kpV Idd
or
Ate Abo
(24-7)
Here one must be careful to remember that this result is only valid when the following
two length-scale constraints are satisfied
fe << my re << Ugly (2.4-8)
‘The spatial deviation temperature in Eq. 2.4-7 is defined by
hh = h - (HP (2.4-9)
and in order to develop the governing differential equation for 7 we need to subtract the
intrinsic form of Eq. 2.4-1 from Eq. 2.4-1. This is obtained by dividing Eq. 2.4-7 by eg
and the result can be expressed as
aT a
(pep)p = V-(kg¥(Tp)*) + e5!Ven-kgV(T)*
(24-10)
_ k; a
+ eV, SE [ npofea + © fn kg VIgdA
fe he
Tf we subtract Eq. 2.4-10 from Eq. 2.4-1 and make use of the definition of 7 given by
Eq, 24-9, we obtain the following governing differential equation for the spatial
deviation temperature.
af os
eps = VilkgWis) — epiVep-kyVe%)? —
Cis
aaa condiiction ‘conductive source
accumulation
&5!
~ Fy J Po keV Tpaa
baa
interfacial flux
(24-11) 96 Chapter 2
This result is analogous to the governing equation for the spatial deviation concentration
given by Eq. 1.4-8; however, in this case we must deal directly with the interfacial flux
and we begin that process by using Eq. 2.2-8 in order to express the interfacial flux as
1 1 =
F Jota kaVIA = - Teg hy VeRI! + & Joo trata (4-12)
iy fe
When this result is used in Eq. 2.4-11 we obtain the following governing differential
equation for the spatial deviation temperature in the B-phase:
ats a) - evil tae er z
ep poe = VolkpVhy) ~ e5'V-1 55 | mpatada | - fogs hy VI04
fa Ave
(2.4-13)
The third term in this result represents @ non-local contribution since it involves values
of 7} associated with points other than the centroid of the averaging volume. Our
arguments for neglecting the non-local term are identical to those presented in Sec. 1.4,
and we make use of that development in order to express Eq. 2.4-13 as
ot S e5 -
(ey )p a = VilkgVig) - cz Jase kV Fd (2.4-14)
— condVetion is
“eoumlation
interfacial flux.
Even when volume averaged heat conduction process is unsteady, the closure problem
will generally be quasi-steady. This occurs because the following constraints
Og t* aot”
B—>>1, SS
6 3
>>1, closure process is quasi - steady (2.4-15)
are usually satisfied. Under these circumstances we can express the closure problem
associated with Eqs. 2.4-1 through 2.4-6 as
Vi(kgVi) = 7 Je kg VIgdA (2.4-16a)
hen
B.C th = T+ (Gy -(G)’, nom, (2.4-16b)
Evi a ve B
BC2 Mp0 “ko VTp — (2.4-16c)
= gg “kg V(Tq)? , at chy Heat Conduction 7
4
Vik VE) = 5 fing keVEyda (2.4-16d)
Aas
B.C3 Th = flrt), atom, (2.4-16e)
B.C4 T= grt), ato%, (2.4-16f)
Equations 2.4-16 represent the starting point for the nwo-equation model closure problem
(Quintard and Whitaker, 1993b, Sec. III) that leads to the two-equation model
represented by Eqs. 2.2-25 and 2.2-26. If we ignore the contributions of the non-
homogeneous terms associated with B.C.3 and B.C.4, we see that there are thtee source
terms, (7,)° -(T,)°, ViTp)®, and Vi75)° that are the generators of the 7 and 7,-
fields. If we were willing to impose local thermal equilibrium, i.e.,
(hy = (T)° = (T) 2.417)
these three sources would be replaced by the single source, V(7). However, we should
remember that Eq. 2.4-17 is the mathematical consequence of the restrictions given by
Eqs. 2.3-7 and 2.3-8. Just because something is small enough to be neglected in the
macroscopic equation does not necessarily mean that it is smal? enough to be neglected
in the closure problem.
2.4.1 LOCAL THERMAL EQUILIBRIUM
To explore local thermal equilibrium, as it applies to the closure problem, we need to
examine the interfacial boundary conditions which we write as
= FB ~ ((P-()°),. atom, (24-18)
ie
| ang kg VI, = = Mpg ko VTq + gg «(ky — ke) V(T)
(2.4-19)
+ pq “kpa (Tp)? —(Ta)?) sat ohh
Here kg, represents the mix-mode thermal conductivity defined by
kyo = €akp + Egkg (2.4-20)
In order for local thermal equilibrium to be valid at the closure level, we require that the
following two restrictions be satisfied
(hy (Tq)? << Ty. Fy Q.4-21)
Nyy “kpaV((Tp)° —(Ta)®) << gg (kp —ke VI) (2.4-22) 98 Chapter 2
The second of these is easy to put into a useful form, and we need only apply the
estimates given in Sec. 2.3 to express Eq, 2.4-22 as
(%)* -(7)" k=
MB) To)" i 2.423)
A(T)
This result can be used with Eq. 2.3-53 to obtain
oe ky
a (2.4-24)
L) Ka
and for many problems of practical interest this constraint is easily satisfied.
In order to deal with the restriction given by Eq. 2.4-21 we need estimates of the
spatial deviation temperatures. When the constraint indicated by Eq. 2.4-24 is satisfied,
we can simplify the flux boundary condition given by Eq. 2.4-19 to the following form
B.C.2 Nyy kgVIy + Mog kyVTy = Mpg “(ky — kg V(T) (2.4-25)
Here we have placed both spatial deviation fluxes on the left hand side and retained the
source term on the right hand side. One should think of the source as being distributed
between the two phases.
In order to develop a constraint associated with the restriction given by Eq. 2.4-21,
we need estimates of 7 and 7,. When local thermal equilibrium is valid in the closure
problem, Eq. 2.4-18 will provide
th = 7%) (2.4-26)
and we want to distribute the source on the right hand side of Eq. 2.4-25 in a manner that
is consistent with Eq. 2.4-26. To accomplish this, we distribute the fluxes according to
Nga kgVI, = @[Mpq “(ka - kg )¥(7)] (2.4-27)
Nop keV Ty = U=@)[ngg (ky - ky)VT)] (2.4.28)
in which the parameter @ is restricted by 0S @ <1. From these two results we obtain the
following estimates for the spatial deviation temperatures
(2.4.29)
}« T)
-k
‘<1 n] (2.4-30)
3 Heat Conduction 99
If we now impose the condition indicated by Eq. 2.4-26, we find that @ is given by
= ofr it/| te 4) (2.4-31)
and our estimates of 73 and 7, can be expressed as
=
7 : ik
i = 00, = of(ie+74) (kg ky) VT) (2.4-32)
Itis convenient to represent this estimate in the form
ae blk —ko)
Brot *p o
ii, = opt 2 wr 2.4-33
oo [sistas | i
in order to indicate the intuitively appealing result that both 7 and %, are zero under the
following circumstances
1. Bither €g or €, is zero.
2. The thermal conductivities are equal, ky = kg.
3. The spatial average temperature is constant, V(T) = 0.
One might also like to have an estimate requiring that 7 and 7 be zero when
(pep)p = (Pep)gi however, that condition is circumvented by the quasi-steady nature of
the closure problem given by Eqs. 2.4-16.
Use of Eq. 2.4-33 in the restriction given by Eq. 2.4-21 leads to the inequality
73)? — (T5)° fg lathe —k,
AT To" _fba kp he) (2.4-34)
ACT) Ty (lakp + oko)
and with the aid of Eq. 2.4-24 we can express this result as
oe Lp lolky —k
S60) ce fptolkp= ko) (2.4-35)
L Lr (lokp + lpke)
In order to clarify the meaning of this constraint, we recall Eqs. 2.3-37 through 2.3-39
which indicate that L? is given by
= Imly (2.436)
and we use this result in Eq. 2.4-35 to obtain 100 Chapter 2
L60 bplakp ke
fa ge _tplotkp ke)
(2.4.37)
Ln Poa (Cakp + opke)
For many practical applications, the right hand side of this inequality will be on the order
of one, thus the constraint will be satisfied whenever fg, is smal] compared to Ly. We
should remember that this latter macroscopic length-scale is defined by the order of
magnitude estimate
VuT) = O[V(T)/Ly] (2.4-38)
thus L,, is infinite for steady, one-dimensional heat conduction processes.
We are now in a position to make use of the simplifications provided by Eqs. 2.4-21
and 2.4-22 so that the closure problem given by Eqs. 2.4-16 can be expressed as
= eq! -
Vi(kpVij) = E fins ky Vig (2.4-39a)
Ape
B.C. Th = T, atohe (2.4-39b)
BC2 — —Myg kg VI, =—Mpg “ko VTy + Mp9 (kp—kaVIT), at eA%y — (2.4-390)
Vi(kViq) 7 frase Aaa (2.4-39d)
Aas
B.C3 Th = f(rt), atom (2.4-39e)
B.C4 FT, = gBltyt), ato, (2.4-39f)
Obviously we have no intention of solving for 7 and 7, in the macroscopic region
illustrated in Figure 2.1, but instead we want to determine these two spatial deviation
temperatures in some representative region such as we have illustrated in Figure 2.4.
2.4.2 LOCAL CLOSURE PROBLEM
In order to discard B.C.3 and B.C.4 and develop a local closure problem, we need to
adopt the model of a spatially periodic system and replace Eqs. 2.4-39 with
<1
a ion x
kV, = Se | mao kpVigdA (2.4-40a)
be
BC.l Ty = To, at Agg (2.4-40b)
BO2 — -ngq keV
po “keV Tg + pq “(kp —ka)V(T)), 5 at Aggy (2.4-40c) Heat Conduction 101
a1
kW = on nog kg VTqdA (2.4-40d)
Aap
2,3 (2.4-40e)
Periodicity: hir+e) = Re), Rer+e) = RO,
Here it is important to understand that the representative region shown in Figure 2.4 can
be arbitrarily complex (Ahmadi and Quintard, 1996) and thus contain all the important,
Figure 2.4. Representative region of the B-o system
geometrical characteristics of the B-o system. Under these ci
boundary conditions given by Eq. 2.4-40c do not play an important role in the
determination of the effective thermal conductivity tensor. In our formulation of the
local closure problem, we have ignored variations of ky and k, within the unit cell and
we have evaluated the gradient of the spatial average temperature at the centroid. This
latter simplification is discussed in Sec. 1.4 and is based on the length-scale constraint
n << Ly (24-41)
Within the framework of a local closure problem, the area integrals representing the
interfacial flux are related by (see Problem 2-3)
i = 1 =
Jo KVIpdA = a Jrrop #7 (2.4-42) 102 Chapter 2
and this means that the boundary value problem given by Eqs. 2.4-40 contains a single
undetermined integral. If this integral were specified, Eqs. 2.4-40 could be used to
determine % and 7 only to within a single arbitrary constant, Both the integral
associated with the interfacial flux, and the arbitrary constant associated with 7 and 7,
can be determined by imposing the following conditions on the average values of the
spatial deviation temperatur:
Average: (hy = 0, (%)7 = 0 (2.4-43)
In subsequent paragraphs we will indicate how these conditions can be used to solve the
closure problem, and why they are unnecessary for the case of symmetric unit cells.
2.4.3 CLOSURE VARIABLES
Given that the single non-homogeneous term in the local closure problem is proportional
to V(T)|, we express 7 and 7 as
ty by V(T) + Wg (2.4-44a)
2 b, VT) + Wo (2.4-44b)
in which bg and bg are referred to as the closure variables (see Problem 2-4) and it is
understood that V(7) is evaluated at the centroid. We specify these two vector fields
according to the following boundary value problem
kyV?bp = © fe kgVbpdA (2.4-45a)
Abe
B.C.1 bg = bys at Agy (2.4-45b)
BC.2 -ngg-kgVbp = —Mgg-kgVbg + Mpglkg—ky), at Agg (2.4-45¢)
keV'bg = SE fgg haVbgdA (2.4-45d)
Ags.
Periodicity: bg (r+é;) = bg(r), bg(t+é;) = b(n), 1=1,2,3. (2.4-45e)
and leave it as an exercise for the reader (see Problem 2-5) to demonstrate that
Wp = Wo =constant. This constant will have no influence on the closed form of the
macroscopic equation since it will not pass through the filters represented by the area
integrals in Eq. 2.2-18. One can prove that by=b,=constant when ky =k (see Heat Conduction 103
Problem 2-6), thus the source, ngg(ks kg), in Eq. 2.4-45¢ is entirely responsible for
generating contributions to 7, and 7; that will pass through the filter.
It is important to understand that the integrals in Eqs. 2.4-45a and 2.4-45d are
related by
1 1
om J pg -kpVbp dd = - 57 |map-keVbgdA = € (2.4-46)
fs hes
thus there is a single undetermined constant vector in the boundary value problem given
by Eqs. 2.4-45. If this constant were known, one could use Eqs. 2.4-45 to determine bp
and b, only to within another arbitrary constant vector. If the averages of 7 and T,
are set equal to zero, this leads to the following condition on the averages of the mapping
vectors.
Averages’ dbp )® = 0, (b,)” = 0 (2.4-47)
These two conditions can be used to specify both the arbitrary constant associated with
by and b, and the undetermined constant represented by Eq. 2.4-46, and in the
following paragraphs we indicate how this can be done.
2.4.4 ARBITRARY UNIT CELLS
In order to develop a general solution for the closure problem given by Eqs. 2.4-45, we
make use of Eqs. 2.4-46 and 2.4-47 to express the closure problem as
kgV7bp = epic (2.4-48a)
BCl bp = by, at Age (2.4-48b)
BC2 — = ngy-kgVbg = —Mgg*kgVbg + Mgg(kp—kg), at Ag, — (2.4-48e)
k,Vb, = -€gle (2.4-48d)
Periodicity: by (r+#,) = bg(r), by (t+é)) = bg(e), =1,2,3 (24-480)
Averages: (bp)® = 0, (bg)? = 0 (24-488)
‘We now decompose the closure variables according to
bg = bp + By-e (2.4-49a)
bs = bo + By-e (2.4-49b)
where the new closure variables are defined by two boundary value problems. The first
of these is given by: 104 Chapter 2
Problem 1
knV?b = 0 (2.4-50a)
B.C.L by = bg. at Ag, (2.4-50b)
BC2 = Mpg-kgVbp = —Mgg-kgVby + Myg(ky—kg), at Age (2.4-50c)
k,V’b, = 0 (2.4-50d)
Periodicity: bp(r+é,) = bg(r), bo(r+é;) = by~), 1=1,2,3 (2.4-50e)
Average: (b3)° = 0 (2.4-50f)
Here we see that Eqs. 2.4-50a through 2.4-50e determine bj and by, to within a single
arbitrary constant vector, and it is Eq. 2.4-50F that specifies this constant. The boundary
value problem that determines Bp and B, is given by
Problem IT
ksVB, = 65'l (2.4-51a)
B.C.l By = By, at Agy (2.4-51b)
BC2 = ng, :kgVBy = —Mgg°kgVBg, at Agg (2.4-51c)
k.V?B, = | (2.4514)
Periodicity: By(r+@;) = Bg(r), Bg(r+4,) = By(r), 1=1,2,3 (2.4-5le)
Average: (By)? = 0 (2.4-51f)
Here we see that Eqs. 2.4-5la through 2.4-51e determine By and B, to within a single
arbitrary constant tensor, and it is Eq. 2.4-51f that specifies this tensor. On the basis of
Eqs. 2.4-50f and 2.4-51f, we see that the constraint on the average given by the first of
Eqs. 2.4-48f is automatically satisfied. The second of Eqs. 2.4-48f can be used to
determine the constant vector in Eqs. 2.4-48a and 2.4-48d according to
© = —((B,)*)"-(b5)? (2.4-52)
Since the non-homogencous terms in the boundary value problem for By and B, are
both constants proportional to the unit tensor, one can prove that these two tensors can
be determined by the solution of a scalar-valued boundary value problem (see Problem
2-7). For symmetric unit cells, we can prove that the constant vector c is zero and the
solution of the closure problem is simplified. Heat Conduction 105,
2.4.5 SYMMETRIC UNIT CELLS
In order to prove that the undetermined constant represented by Eq. 2.4-46 is zero for
symmetric unit cells, one begins by using the periodicity condition and divergence
theorem for bg and b,, to obtain
ae
(24-54)
Next, one follows the type of analysis given in Sec. 1.5 to prove ( see Problem 2-8) that
the components of bp and b, are skew-symmetric for a symmetric unit cell. Since V? is
a symmetric operator, and the components of bg and b, are skew-symmetric functions,
the volume integrals of V°bp and Vb, must be zero for any unit cell that is completely
symmetric. This means that Eqs. 2.4-53 and 2.4-54 can be expressed as
1 1 :
$ Jove kgVbgdA = x fav bpdV = 0 (2.4-55)
Ato %
: 1
F Jos ko DotA S $f tevv.av = 0 (2.4-56)
ha i,
and under these circumstances the closure problem represented by Eqs. 2.4-45 simplifies
to the form originally presented by Nozad et al. (1985a). This is given by
Vb, = 0 (2.4-57a)
BCL bg = bg, at Ag (2.4-57b)
BC2 =ngg:Vbp = —Mgg'KVbg + myg(I~K), at Agg (2.4-57c)
vb, = 0 (2.4-57d)
Periodicity: bg(r+4,) = bg(r), bo(r+4,) = bg(e), (2.4-57e)
in which the parameter « is defined by
K = ky/ky (2.4-58)
When xk =0 our closure problem for heat conduction naturally reduces to the closure
problem for diffusion given by Eqs. 1.4-58. For non-symmetric unit cells, we have no 106 Chapter 2
proof that the integrals in Eqs. 2.4-45a and 2.4-45d are zero, nor do we have a proof that
this undetermined constant is negligible; however, comparison with experiment suggests
that the effects of non-symmetric unit cells are not particularly important.
2.4.6 CLOSED FORM
In oder to obtain the closed form of the macroscopic equation, we recall
Eq. 2.2-18
7
yc, XD fy J r ke J 7
CAE = V-| epkey +eoky WIT) + <2 | npoFaa + “2 [agg Fda
Pep (pkg +eoka)VT) + Z| pol a | Pop
he As
(24-59)
along with the representations for 7 and 7, that took the form
F = by-ViT) + Wp (2.4.60)
T, = by VT) + Wo (2.461)
When used with a spatially periodic model, the area integrals in Eg, 2.4-59 have the
characteristic that
1 1
7 foracta =O Jresaa =0 (2.4-62)
Ass Aag
and this means that if Wg andy, are constants (see Problem 2-5), they cannot pass
through these filters, Under these circumstances we can substitute Eqs. 2.4-60 and
2.4-61 into Eq. 2.4-59 to obtain the following closed form
ar :
oc, 3 = V(Kiy -V(7)) (2.4-63)
Here the effective thermal conductivity tensor is defined by
(kg —&
Kg = (php teoks) 1 + ose frsabyas (2.4-64)
Boo
and in the next section we compare theoretical calculations of Ky for isotropic systems
with several sets of experimental data, One can prove that the thermal conductivity
tensor determined by Eqs. 2.4-64 and 2.4-48 is symmetric, i.e.,
K Ki,
2 (2.4.65)
eff
and the proof will be left as an exercise for the reader (see Problem 2-9). Heat Conduction 107
2.5 Comparison Between Theory and Experiment
In this section we wish to compare experimental values of the effective thermal
conductivity with theoretical values determined by the solution of Eqs, 2.4-57 and the
representation of Ky given by Eq. 2.4-64. Alll the experimental studies were based on
the assumption that Ky was a constant, thus Eq. 2.4-63 was employed in the form
a
¢, 2 = Ky VOT) (2,5-1)
For comparison with experimental results, it is convenient to work with a dimensionless
form of this result given by
_
ee (es +€0x)I + (2.5-2)
in which K = kg /kp as indicated earlier by Eq. 2.4-58.
Nozad et al, (1985a) have solved the closure problem given by Eqs. 2.4-57 for the
simple, two-dimensional unit cell shown in Figure 2.5. This produces an isotropic
Figure 2.5. Unit cell in a two-dimensional, spatially periodic porous medium
effective thermal conductivity tensor, thus only a single component of the vector fields
bg and b, is required in order to determine Kjy. The computed values of Ky /kg are
shown in Figure 2.6 along with results determined by Perrins er al. (1979) who used an
extension of the method of Rayleigh (1892). Similar results have been obtained by
Sahraoui and Kaviany (1993) using direct numerical simulation for heat conduction in a
unit cell, and one can prove (see Problem 2-10) that the results obtained in this manner
are identical to those given by the solution of Eqs. 2.4-57. Figure 2.6 illustrates results 108 Chapter 2
for a wide range of values of K = kg /kp and three different geometries: a square array of
squares, a square array of cylinders, and a hexagonal array of cylinders. One can see a
slight influence of geometry, but for all practical purposes the dimensionless effective
thermal conductivity depends only on €y and K. Once again we should think of the area
integral in Eq. 2.5-2 as a filter, and then ask the question: What characteristics of the
boundary value problem given by Eqs. 2.4-57 pass through the filter? For the case of
non-touching particles, only the volume fraction and the ratio of thermal conductivities
Pass through the filter.
oo T T T ee
Perrins et al.
© (square array of cylinders)
© (hexagonal array of cylinders)
Nozad er al. ep = 019
1 1
1 10 10° 10° 108
«= ky/kp
Figure 2.6. Effective thermal conductivity for two-dimensional arrays of non-touching particles,
One can use the closure problem giver by Eqs. 2.4-57 and an asymptotic analysis
for small values of K~ to demonstrate that Ky /ky is independent of « for large values
of , and the intuitive explanation is quite straightforward. When ky for the non-
touching particles is much, much larger than &g for the continuous phase, the
temperature in each particle becomes uniform, Further increases in ky have no
influence on the temperature distribution in the G-phase and the effective thermal
conductivity is controlled entirely by ky and the geometry of the two-phase system. For
the geometries represented in Figure 2.6, the geometrical effects are almost entirely
accounted for by eg.
2.5.1 MAXWELL’S CLOSURE PROBLEM
For a two-phase system made up on non-touching particles, Chang's (1982, 1983) unit
cell shown in Figure 2.7 represents an attractive model. As we mentioned in Sec. 1.6, Heat Conduction 109
Figure 2.7. Chang's unit cell for the B-c system
Chang’ unit cell leads to Maxwell’s model with the special form of the closure problem
given by
BC bp = (2.5-3a)
Vbp = 0, a 0, that was studied by Batchelor and O’Brien (1977). They
considered a regular array of spheres and developed a solution for large values of
x= kg/ky in terms of an inner expansion. The result took the form
Ky
= 4in(x) ~ 11 (2.5-6)
i
in which the number eleven represents an adjustable parameter that can only be
determined precisely by matching the inner expansion with an outer expansion. The
value of the adjustable parameter was chosen on the basis of experimental data with
values of « ranging from 30 to 3000, The analysis of Batchelor and O'Brien is
compared with that of Nozad et al. in Figure 2.11 where the good contact results
represent an expansion in terms of «~'. This leads to a linear dependence of Kg kg on
x instead of the logarithmic dependence for the point contact condition.
The theory of Batchelor and O'Brien (1977) has been compared with experiment by
Shonnard and Whitaker (1989) and the results, along with those of Swift (1966), are
shown in Figure 2.12. The experiment performed by Shonnard and Whitaker consisted
of bringing two spheres into point contact in a unit cell, thus the experiment was
designed explicitly to test the theory of Batchelor and O'Brien. Errors resulted from the
fact that it was difficult 10 maintain a uniform temperature gradient and it was difficult to Heat Conduction
10°
Cr
10 point contact +
z 10 10? 10° 10 108
K = ke/ky
Figure 2.11. Effective thermal conductivity for good contact, point contact, and no contact
113
completely eliminate natural convection since the diameter of the spheres was on the
order of 5 cm. Nevertheless, the experimental results of Shonnard and Whitaker tend (0
confirm the theory of Batchelor and O’Brien. ‘The results of Swift were obtained with
packed metal powders for which the particle diameter was on the order of hundreds of
micrometers, thus no specific effort was made to create a point-contact porous medium.
100
© Shonnard & Whitaker
4 Swift
40 — Batchelor & O’Brien
Key
ky
10
04
10 10° 10? 108 10°
K = ko/kp
Figure 2.12. Comparison of theory and experiment for point-contact heat conduction 14 Chapter 2
We would expect Swift’s results to exhibit the type of behavior illustrated in Figure 2.9
rather than the point contact behavior shown in Figure 2.12, but it is clear that they do
not.
2.6 Conclusions
In this chapter we have examined the process of transient heat conduction in two-phase
systems. We have developed constraints associated with the principle of local thermal
equilibrium, and these constraints are in reasonable agreement with numerical
experiments. When the condition of local thermal equilibrium is valid, the heat transfer
process can be represented in terms of a one-equation model. The closure problem
associated with the one-equation model was developed and values of the effective
thermal conductivity were compared with experimental values for non-consolidated
porous media. Good agreement between theory and experiment was obtain provided
that the particle-particle contact was taken into account in terms of an adjustable
parameter.
2.7. Problems”
2-1. Show how the Darcy-scale temperature deviations defined by Eqs. 2.3-2
can be used to obtain the form indicated by Eqs. 2.3-5.
2-2. Indicate how the last term in Eq. 2.3-21 is obtained from Eq. 2.3-19,
Eq. 2.3-20, and the interfacial flux boundary condition given by Eq. 2.1-3.
2-3, Show how the flux boundary condition given by Eqs. 2.4-40c can be
used to develop the integral condition represented by Eq. 2.4-42.
2-4. Given the following steady, one-dimensional heat conduction problem
B.C. Ty = 0, x=0 @
ie)
BC2 QB)
BC3 oo)
(5)
* Solutions to all problems are available from the author. Heat Conduction 115
B.c4 To = 0, x= lptla (63)
develop a solution in terms of the closure variables the appear in the
representations
Th = Jp2%s + Bp2%o. Ty = FolsIPp + (HI, oO)
Take g and, to be constants and show that the solution in terms of the
closure variables gives the same result as the direct solution that is given by
249 256
Th = Hoye? -[La,0(242) Ooo _ 8
b = 2P5* ; Tro) * mate) |* @)
56h Q, fp tbe | O5ép 2,
re) S00) Cee) | ences eee a
2, \1+9 ky It+@ ro | Ug ie
in which the dimensionless parameter 9 is defined by
9 = Lpko/lokp (10)
2-5. Develop the boundary value problem for ys and y., which are defined
by Eqs. 2.4-44, and identify the circumstances for which you can demonstrate
that these two variables are equal to a constant. Follow the approach suggested
in Problem 1-21.
2-6. Prove that bg =b, =constant when the thermal conductivity of the
B-phase is equal to that of the o-phase. Restrict you analysis to symmetric unit
cells and note that the proof is similar to that used in Problem 2-5.
2-7. Given the boundary value problem
kgV'B, = el ay
B.C. B, = B,, at Ag Q)
BC.2 = yg kyVBy = —Mgg-kgVB, . at Age @)
k,V°B, = -e;'1 @)
Periodicity: By(r+/;) = By(r), Bo(r+é,) = By(r), i=1,2,3 (5)
Averages: (B® = 0 6)
prove that By and B,, can be expressed as 116 Chapter 2
B, = Bl (72)
B, = Bl (7b)
and derive the boundary value problem for By and B,.
2-8. Consider the closure problem given by Eqs. 2.4-48, and show that it
reduces to the closure problem given by Eqs. 2.4-57 for the case of completely
symmetric unit cells. Consider only the x-components of the vectors bg and b,
given by
by = iby, by = iby 0)
and a unit cell that is symmetric in x. Develop the closure problem for by and b,
and consider a coordinate transformation of the type
ny. > BT Q)
in which ¥=-x, ¥=y, Z=z. The constants in Eqs. 2.4-48a and 2.4-48d are
invariant to the coordinate transformation indicated by Eq. 2. Show that
bg (49.2) = —bg(-2,y.2)5 bg(%y,2) = —by(-*y,2) GB)
and that, by analogy, bg andb, are completely skew-symmetric for a unit cell
that is completely symmetric. Indicate why this leads to the result
° Upon co
c= + Jrve kgVbgdA = x fev bydV = 0 @
he 4,
_ -$ Jo kgVbgdA = -Z J bov*bav =0 oy
Aas Yo
thus proving that Eqs. 2.4-48 simplify to Eqs. 2.4-57 for completely symmetric
unit cells.
2-9. Use Eqs. 2.4-48 in order to prove that the thermal conductivity tensor is
symmetric, Ky =Kiz. Notice that the dyadic product of bp with the
boundary condition given by Eq. 2.4-48c will lead to a boundary condition
containing the term (kg-kg)nggbg, and it is this term which appears in the
expression for the thermal conductivity tensor given by Eq. 2.4-64.
2-10. Sahraoui and Kaviany (1993) determined the effective thermal
conductivity for the system illustrated in Figure 2.5 by first solving the following
boundary value problem associated with the unit cell ilustrated in Figure 2-10. Heat Conduction ales
2 2
eh | eh
= ie 5
O = SP + Zr inthe B-phase a)
BC. Th = T,, atx=0 2)
BC2 Th = Ty, atthe B-o interface B)
B.C3 —nygkgVTy = — Ngo koVTg» at the B-o interface (4)
BC4 Tho= 1, at x=k 6)
ah
B.C5&6 = 0 y=OL 6)
ay
PT . ty
+ —£, inthe o- phase a
ax? ay? :
|. L -|
Figure 2-10. Unit cell used for the determination of the effective thermal conductivity
Given the temperature field in the B-phase, the average heat flux was computed
according to
eb
1p, a
@dleo = ~T j a dy (8)
0
and this was used to determine an effective thermal conductivity on the basis of
= 7
Gono = Ker 9) 118 Chapter 2
The boundary conditions given by Eqs. 2 and 5 are intuitive; however, they can
be proved to be the correct representations for a spatially periodic system on the
basis of the closure problem given by Eqs. 2.4-57. To do so, one begins with
Ty = {T) + Ty = {T) + bg-V(T) (10)
and then considers the special case in which
very = 1X2 a
ox
The fact that i-by is skew-symmetric for a symmetric unit cell can then be used
to develop the boundary conditions given by Eqs. 2 and 5.
2-11. Consider a B-o system in which the o-phase is a rigid solid and the
B-phase is a stagnant fluid. If a uniform, homogeneous thermal source exists in
the o-phase, the heat conduction process can be described by
(ep pe = V-(kjV7j), in the B- phase a
BCL Ty = Ty, atthe B-o interface Q)
BC2 Hyg kyVIy = —Mg,°K,VT, ,at the B-o interface )
(069 = Vi(kgVT;) + ®,, in the o-phase (4)
Consider , to be a constant and develop the one-equation model for heat
conduction. Carefully identify the constraints that must be satisfied and develop
the closure problem that can be used to predict the effective thermal conductivity
tensor.
2-12. Develop the constraints associated with the principle of local thermal
equilibrium for the heat conduction process described in Problem 2-11.
2-13. When chemical reactions take place in porous catalysts, the heat
conduction process can be described by
oT,
(ep yar = V-(kyVT,), in the y—phase ()
BC1 T, = T., atthe y—K interface (2)
B.C2 —my ky VT, = —ny kc VI, + Qe, atthe y—K interface )
You might also like
- Instant download Process Heat Transfer Principles Applications and Rules of Thumb 1st Edition Thomas Lestina pdf all chapter100% (3)Instant download Process Heat Transfer Principles Applications and Rules of Thumb 1st Edition Thomas Lestina pdf all chapter55 pages
- April 18 Homework Solutions: Mechanical Engineering 375 Heat TransferNo ratings yetApril 18 Homework Solutions: Mechanical Engineering 375 Heat Transfer10 pages
- Fundamentals of Heat and Mass Transfer 7th Edition - Bergman, Lavine, Incropera, DeWitt (1) - p0015No ratings yetFundamentals of Heat and Mass Transfer 7th Edition - Bergman, Lavine, Incropera, DeWitt (1) - p00151 page
- Introduction To Convective Heat Transfer Analysis Chapter 8 PDFNo ratings yetIntroduction To Convective Heat Transfer Analysis Chapter 8 PDF84 pages
- Interdisciplinary Applied Mathematics: L. Kadanoff J .E. Marsden L. Sirovich Wiggins F. John (Deceased)No ratings yetInterdisciplinary Applied Mathematics: L. Kadanoff J .E. Marsden L. Sirovich Wiggins F. John (Deceased)289 pages
- Some Unsteady Unidirectional Ows of A Non-Newtonian Uid: T. Hayat, S. Asghar, A.M. SiddiquiNo ratings yetSome Unsteady Unidirectional Ows of A Non-Newtonian Uid: T. Hayat, S. Asghar, A.M. Siddiqui10 pages
- Tests For Transport Limitations in Experimental Catalytic Reactors - DV Mears, 1971No ratings yetTests For Transport Limitations in Experimental Catalytic Reactors - DV Mears, 19717 pages
- A First Course in The Mathematical Foundations of ThermodynamicsNo ratings yetA First Course in The Mathematical Foundations of Thermodynamics77 pages
- The Behaviour of Porous Catalyst Particles in View of Internal Mass and Heat Diffusion EffectsNo ratings yetThe Behaviour of Porous Catalyst Particles in View of Internal Mass and Heat Diffusion Effects8 pages
- 9 Akbari-Ganji's Method: Advanced Numerical and Semi-Analytical Methods For Differential Equations100% (1)9 Akbari-Ganji's Method: Advanced Numerical and Semi-Analytical Methods For Differential Equations8 pages
- ISVR Advanced Course in Aeroacoustics 2011No ratings yetISVR Advanced Course in Aeroacoustics 20112 pages
- Aerodynamics of Wings at Low Reynolds Numbers - Boundary Layer Separation and ReattachmentNo ratings yetAerodynamics of Wings at Low Reynolds Numbers - Boundary Layer Separation and Reattachment130 pages
- How Dilute Are Dilute Solutions in Extensional Flows?: C. Clasen, J. P. Plog, and W.-M. KulickeNo ratings yetHow Dilute Are Dilute Solutions in Extensional Flows?: C. Clasen, J. P. Plog, and W.-M. Kulicke33 pages
- Introduction To Langmuir Probes - Hiden Analytical Tech Info Sheet531No ratings yetIntroduction To Langmuir Probes - Hiden Analytical Tech Info Sheet5318 pages
- PH 8221 - Experimental Physics-Ii: Vacuum TechniquesNo ratings yetPH 8221 - Experimental Physics-Ii: Vacuum Techniques56 pages
- Functions of Matrices Nicholas J. Higham 2024 Scribd Download100% (1)Functions of Matrices Nicholas J. Higham 2024 Scribd Download71 pages
- (Ebook) Airplane Stability and Control: A History of the Technologies That Made Aviation Possible - Second edition by Malcolm J. Abzug, E. Eugene Larrabee ISBN 9780511607141, 9780521021289, 9780521809924, 0511607148, 0521021286, 0521809924 - Own the complete ebook with all chapters in PDF format100% (2)(Ebook) Airplane Stability and Control: A History of the Technologies That Made Aviation Possible - Second edition by Malcolm J. Abzug, E. Eugene Larrabee ISBN 9780511607141, 9780521021289, 9780521809924, 0511607148, 0521021286, 0521809924 - Own the complete ebook with all chapters in PDF format59 pages
- Full Download Theory and Computation in Hydrodynamic Stability 2nd Edition W. O. Criminale PDF100% (3)Full Download Theory and Computation in Hydrodynamic Stability 2nd Edition W. O. Criminale PDF52 pages
- 1603171724-chapt1-3-experimental-physics-notesNo ratings yet1603171724-chapt1-3-experimental-physics-notes21 pages
- U in The Positive X: CL203, Autumn 2009No ratings yetU in The Positive X: CL203, Autumn 200916 pages
- EN Process Performance: Effect of Agitation, Loading and Stabilizer LevelNo ratings yetEN Process Performance: Effect of Agitation, Loading and Stabilizer Level23 pages
- (AMCP 706-283) - Engineering Design Handbook - Ballistic Missile Series - Aerodynamics - U.S. Army Materiel CommandNo ratings yet(AMCP 706-283) - Engineering Design Handbook - Ballistic Missile Series - Aerodynamics - U.S. Army Materiel Command98 pages
- 1960 Copper Dynamics of Liquids in Moving ContainersNo ratings yet1960 Copper Dynamics of Liquids in Moving Containers5 pages
- Molecular Dynamics Modeling of Hypersonic Gas-Phase and Gas-Surface Reactions by T. E. Schwartzentruber, P. Norman, and P. ValentiniNo ratings yetMolecular Dynamics Modeling of Hypersonic Gas-Phase and Gas-Surface Reactions by T. E. Schwartzentruber, P. Norman, and P. Valentini8 pages
- Mathematical Proceedings of the Cambridge Philosophical Society Volume 69 issue 3 1971 [doi 10.1017_s0305004100046880] Packham_ B. A._ Shall_ R. -- Stratified laminar flow of two immiscible fluids.pdfNo ratings yetMathematical Proceedings of the Cambridge Philosophical Society Volume 69 issue 3 1971 [doi 10.1017_s0305004100046880] Packham_ B. A._ Shall_ R. -- Stratified laminar flow of two immiscible fluids.pdf6 pages
- The Structure of Turbulent Shea - A. A. R. TownsendNo ratings yetThe Structure of Turbulent Shea - A. A. R. Townsend416 pages
- Development of Finite Difference Solver For Heat Transfer Analysis in FinsNo ratings yetDevelopment of Finite Difference Solver For Heat Transfer Analysis in Fins24 pages
- Classical and Statisitcal ThermodinamicsNo ratings yetClassical and Statisitcal Thermodinamics111 pages
- Chul Park Viscous Shock Layer Calculation of Stagnation-Region Heating Environment Viscous Shock Layer Calculation of Stagnation-Region Heating EnvironmentNo ratings yetChul Park Viscous Shock Layer Calculation of Stagnation-Region Heating Environment Viscous Shock Layer Calculation of Stagnation-Region Heating Environment7 pages
- Full Download Convective Heat and Mass Transfer Second Edition Seyed Mostafa Ghiaasiaan PDF DOCX100% (3)Full Download Convective Heat and Mass Transfer Second Edition Seyed Mostafa Ghiaasiaan PDF DOCX65 pages
- Ghia V., High-Resolutions For Incompressible Flow Using The Navier-Stokes Equations and A Multi-Grid MethodNo ratings yetGhia V., High-Resolutions For Incompressible Flow Using The Navier-Stokes Equations and A Multi-Grid Method25 pages
- Flow in Porous Media I: A Theoretical Derivation of Darcy's LawNo ratings yetFlow in Porous Media I: A Theoretical Derivation of Darcy's Law23 pages
- Diffusion and Dispersion in Porous Media (S.Whitaker, 1967)No ratings yetDiffusion and Dispersion in Porous Media (S.Whitaker, 1967)8 pages
