Вы находитесь на странице: 1из 59

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

net/publication/253328255

Heat Transfer in Catalytic Packed Bed Reactors

Chapter · January 1989

CITATIONS READS
15 1,540

1 author:

Stephen Whitaker
University of California, Davis
235 PUBLICATIONS   15,049 CITATIONS   

SEE PROFILE

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

Diffusion of Charged Species in Liquids View project

Interpore Interview with Prof. Stephen Whitaker View project

All content following this page was uploaded by Stephen Whitaker on 17 May 2014.

The user has requested enhancement of the downloaded file.


CHAPTER 10

HEAT TRANSFER IN CATALYTIC PACKED BED REACTORS

Stephen Whitaker

Department of Chemical Engineering


University of California
DA VIS, CA California, USA

CONTENTS

INTRODUCTION, 361

ENERGY, 366

JUMP CONDITIONS, 372

HEAT TRANSPORT IN POROUS CATALYSTS, 373


Volume Averaging, 376
One-Equation Model, 379
Closure, 382
Heterogenous Thermal Source, 389
Effective Thermal Conductivity, 390
Micropore-Macropore Systems, 395

FLUID PHASE HEAT TRANSPORT, 399


Volume Averaging, 402
Traditional Model, 403

CONCLUSION, 413

NOTATION, 414

REFERENCES, 416

INTRODUCTION

In this chapter we consider a line of analysis that leads from the axioms of continuum mechanics
for muiticomponent systems to the thermal analysis of packed bed catalytic reactors. We begin
with the axioms for mass, momentum, and energy and from these we obtain an equation for the
temperature in a muiticomponent system. Jump conditions at phase interfaces are constructed that
can be used, along with the governing equation for the temperature, to determine the temperature
field when chemical reactions are taking place.
Since most reactor design problems involve muitiphase systems with interfacial regions that can
only be determined in an average or statistical sense, a detailed knowledge of the temperature field
is not possible. This situation naturally leads to a search for average temperatures and the governing
differential equations for averaged quantities. Use of the method of volume averaging provides a
precise route from the point equation for the temperature to the thermal design equations for a
packed bed catalytic reactor.
After a presentation of the axioms for muiticomponent, reacting systems, we direct our attention
to the heat conduction problem in a porous catalyst pellet. The constraint of local thermal equilib-
rium is imposed in order to obtain a one-equation model for the heat conduction process. Then

361
362 Multiphase Reactor Operations

the form of the temperature dependence for the area-averaged rate of reaction is discussed, and a
general approach to heat conduction in micro porous media is presented. After completing the
analysis of heat conduction in porous catalysts, we move on to study the fluid-phase heat trans-
port process. An equation for the volume-averaged temperature is derived, and the assumptions
that must be made in order to obtain the traditional model are identified. The general closure
problem for the combined fluid and solid-phase heat transport process is then developed, and the
general form of the two-equation model for heat transfer in catalytic packed bed reactors is pre-
sented. The non-traditional convective and conductive transport mechanisms are discussed and the
route to non-isothermal effectiveness factors is outlined in terms of the method of volume averaging.
The design of packed bed catalytic reactors can be thought of as a model chemical engineering
problem, since it involves so much of what we think of as chemical engineering. For example,
in the typical design one encounters: fluid mechanics, heat transfer, mass transfer, chemical kinetics,
process dynamics, and control, adsorption, and thermodynamics. The generality of the problem is
illustrated in Figure 1, and in this chapter we will consider only the thermal aspects of packed bed
catalytic reactor design. While the scope of this presentation might appear to be rather limited
within the broad domain of reactor design, we will follow a route that has wide applications.
In a simplistic sense, our objective in catalytic reactor design is the determination of the rate
of reaction per unit volume. Since reaction rate coefficients are highly temperature dependent,
reactor design requires a knowledge of the temperature and the prediction of that temperature is
the objective of this chapter. The development presented here is meant to compliment a previous
study of transport processes with heterogeneous reaction [1]. That study dealt exclusively with the
mass transfer process that takes place in catalytic packed bed reactors with a special emphasis on
diffusion in porous catalysts. The axioms associated with mass transport in multi-component
systems can be expressed in terms of the species body illustrated in Figure 2 and are given by

dl
~
dt VA(t) PA
dV = l
VA(t)
r dV
A
(1)

(2)

Chemical Transport Phenomena


Kinetics and
Thermodynamics

Chemical Reactor Design


complex flow fields, multiphase systems,
nonlinear kinetics, adsorption and
heterogeneous catalysis, simultaneous
heat and mass transfer

Dynamics
and
Figure 1. Foundations of chemical reactor
Control
design.
Heat Transfer in Catalytic Packed Bed Reactors 363

• • •• •
• •

o Species A
• Species B

Figure 2. Motion of a species body.

Here VA(t) represents the volume of the species A body illustrated in Fig. 2, PA represents the species
mass density, and rA represents the mass rate of production of species A owing to homogeneous
chemical reaction.
The species continuity equation can easily be extracted from Equation 1 and it takes the form

apA
-
at + V- . (PAV-A) = rA (3)

Here YA represents the species velocity and it is this velocity that describes the motion of the species
body illustrated in Figure 1. If we define the total density P and the mass average velocity Y
according to
A=N

PY = L
A=l
PAYA (4)

we can sum Equation 3 over all N components in the system to obtain

ap
-+ V'(pv) =0 (5)
at - -
Both the species continuity equation and the total continuity equation will be of use to us in the
thermal analysis of a packed bed reactor. In thinking about these two continuity equations, we
must remember that PA is the dependent variable in Equation 3 while P plays the same role in
Equation 5. To solve Equation 5 and thus determine P as a function of time and space, we need
to know the mass average velocity y, which will be determined by the laws of mechanics. In the
design of a chemical reactor it is more important to determine PA as a function of time and space,
and to do this we need to know the species velocity YA and the mass rate of production of species
A per unit volume, rA • The former is determined by the laws of mechanics for muIticomponent
364 Multiphase Reactor Operations

systems while the latter is determined by quantum mechanical considerations as suggested in


Figure 1.
The laws of mechanics for multicomponent, reacting systems are stated in terms of the species
body shown in Figure 2 and we list them as

1. Linear Momentum

-dt
1 v dV
d v A(l) PA_A = 1 V A(l)
b dV +
PA_A fAA(I)
t dA
_A
B=N
+ r A(t) "
Jv L.. _AB
r A(t) r A_A
P dV + Jv v* dV (6)
B~l

2. Angular Momentum

-dt
1
d v A(l) -r x PA_A
v dV = 1 v A(l) -r x PA_A
b dV + f r x _A
AA(I) -
t dA

B~N

Jrv A(t) --rx L Jrv A(t) -rx(rAv1)dV


+ PABdV+ (7)
B= 1 - --

3. Diffusive Source of Momentum


A=N B=N
L L
A=18=1
PAB = 0 (8)

4. Chemical Reaction Momentum Source


A=N
L
A=l
r A yl = 0 (9)

The linear momentum equation for a species body is based on the concept that the time rate
of change of linear momentum of the body is equal to the force acting on the body plus the source
of momentum owing to chemical reaction. In Equation 6 we note that PAYA represents the species
A linear momentum, while PAQA and lA represent the body force and surface force respectively that
act upon the species A body. In the absence of electric fields fiAQA can be replaced by fiAI2 where
¥ is the gravitational vector; however, if one is concerned with the motion of ionic species the body
force PAQA takes on a crucial role.
The velocity of species A produced by chemical reaction has been designated by yl with the
thought that it need not be equal to the continuum velocity YA' In Equation 6, the term PAB
represents the force per unit volume that species B exerts on species A and we expect that this
force will be important when species A and B are moving at different velocities, i.e., PAB is intimately
related to the process of diffusion. It is of some interest to note that the source of linear momentum,
r Ayl, is not invariant to a Galilean transformation; however, the time rate of change of momentum
on the left side of Equation 6 also fails to be invariant but combines with the momentum source
owing to chemical reaction in order to produce a result which is invariant to a Galilean trans-
formation.
In writing the angular momentum axiom given by Equation 7 we have assumed that all torques
are the moments of forces [3] so that the effect of intrinsic angular momentum has been neglected.
The influence of intrinsic angular momentum on the mass average angular momentum principle
has been discussed by Aris [4], and it may be of some interest to explore the influence of this effect
on the species angular momentum principle.
Axioms 3 and 4 simply indicate that there can be no source of mass average momentum as a
result of the diffusive force PAB' and that linear momentum is conserved in the course of chemical
reactions.
Heat Transfer in Catalytic Packed Bed Reactors 365

The derivation of the species stress equations of motion parallels that for single component
systems [5] and leads to

(10)

while the angular momentum principle given by Equation 7 yields the expected symmetry con-
dition given by

(II)

It is shown elsewhere [I, 6] that Equation 10 is the origin of the Stefan-Maxwell equations and
thus Fick's law for binary systems.
If we define the following mass average quantities
A=N
P= L:
A=J
PA (12)

A=N
~ = L:
A=l
(PA/P)~A (13)

A=N
I,.., = L: (IA ~ PA!!A!!A)
A=l-
(14)

A=N
Q= L:
A= 1
(PA/P)QA (15)

in which !!A is the mass diffusion velocity given by

(16)

we can sum Equations 10 and 11 to obtain the traditional results given by

P (a~
at + -v· V'v)
-- = pb- + V'-"
.T (17)

T=P
::::: :::::
(18)

Here we have used Equations 8 and 9 to eliminate the diffusive source of momentum resulting
from the force PAD and the chemical reaction momentum source which is coupled to the mass
source term in Equation 3. If one decomposes the total stress tensor according to

(19)

one can express Equation 17 as the viscous stress equations of motion.

av )
P ( ----
at + -v . V'v
-- = ~ V'p
- + pb- + V'-"
.r (20)

(21)

When the viscous stress is given by Newton's law of viscosity, the Navier-Stokes equations result.

y
P
-- ~ V'p
(aat + -v . V'v) - + pb-,...-
= + 1IV"v (22)
366 Multiphase Reactor Operations

Here we have assumed that the viscosity 11 is constant, and this result confirms the idea that the
Navier-Stokes equations represent the governing equations for the mass average velocity.

ENERGY
In the previous paragraphs we have briefly considered the axioms of mass and mechanics for
multicomponent, reacting systems, and we are now in a position to consider the axioms for the
energy of such systems. The structure of the axioms for mass and mechanics was that of balance
equations containing source terms and constraints on the source terms. The axioms for energy are
similar in form, and for the species A body shown in Figure 2 we express the axioms as

+ it· v
AA(t) _A _A
dA + Jv
I' A(t) p A_A v dV
b . _A

(23)

2. I
A=N (
r A eA + -1 VAl ) = 0 (24)
A=! 2

(25)

4. eA is a state function (26)

Here we have used eA to represent the internal energy per unit mass for species A, while qA repre-
sents the conductive heat flux vector and q! represents the radiative heat flux vector. One must
think of g! as a measure of the rate of exchange of photons between species A and the surroundings
[7], while the term QAB represents the rate of thermal energy exchange between species Band
species A. For example, if a region in space is subject to a flux of photons that are absorbed only
by species B, the energy of species A will increase owing to the exchange term, QAB. In axiom 1 we
have suggested, in keeping with Equations 6 and 7, that the species A produced by chemical re-
action need not have the same internal or kinetic energy per unit mass as the continuum values
designated by eA and 1vi.
The nature of the first axiom given by Equation 23 is that the rate of increase of internal and
kinetic energy of species A is equal to the rate of supply of non-mechanical energy (the first two
terms on the right side of Equation 23), the rate of supply of mechanical energy (the third, fourth,
and fifth terms on the right side), and the source of internal and kinetic energy owing to chemical
reaction. The second axiom is analogous to Equations 2 and 9 while the third axiom is the thermal
analogue of Equation 8. The fourth axiom is the central theme of thermodynamics and states that
the internal energy is a function of the thermodynamic state of the system.
The differential equation associated with Equation 23 is given by

= -Y·(gA+g!)+
B=N
BJ;! QAB+Y·CIA·YA)+PAQA·YA+
B=N (1)
BJ;! PAB·YA+rA eA+ 2vA'
(27)
Heat Transfer in Catalytic Packed Bed Reactors 367

In our studies of mass we were interested in PA and this, in turn, required knowledge of the species
velocity, YA. In our study of the energy of multicomponent systems, we are primarily interested in
developing an equation for the temperature and we have a secondary interest in an equation for
the enthalpy. In order to develop these equations from Equation 27 it will be useful to eliminate
the mechanical terms by means of the species mechanical energy equation. This is derived by
forming the scalar product of Equation 10 with the species velocity YA in order to obtain

(28)

Considerable manipulation is required to obtain this result and the species continuity equation
given by Equation 3 is used as part of the development. For single component systems this result
reduces to the traditional form of the mechanical energy equation [8] that is so useful in the analysis
of incompressible flow processes. In our case, we wish only to subtract Equation 28 from Equation
27 in order to obtain the species thermal energy equation given by

a B=N
;3t(PAeA) + Y·(PAeAYA) = -Y·(9A+9~)+ B!;t QAB+IA:YYA

+ rA(eA + ~ VA 2
- YA . YA + ~ vi) (29)

This result will be more useful to us when expressed in terms of enthalpy, thus we define the
species enthalpy per unit mass hA according to the relation

(30)

Here PA represents the partial pressure of species A, and when this result is used in Equation 29
we obtain

(31)

About the partial pressure for species A, we should note that it is defined by

PA = PA2 (aea
PA
A
)
S,PB ,Pc, ..
(32)

where s represents the entropy per unit mass. We define the total pressure as

(33)

and one can prove as a theorem that

p=- -
(au)
av S.mA.mo •.
(34)
368 Multiphase Reactor Operations

Here we have used the classic thermostatics nomenclature in which U is the total internal energy,
S is the total entropy and rnA' mB, etc. represent the mass of species A, B, etc. Strictly speaking,
Equation 34 is only applicable for systems at equilibrium whereas Equations 32 and 33 are restricted
only by the constraint of local thermodynamic equilibrium. This means that a single temperature
can be used to characterize the internal energy of all the species, and it is not the species enthalpy
equation given by Equation 31 that we need but instead the total enthalpy equation. This is ob-
tained by summing Equation 31 over all N species leading to

(35)

Here the two total heat flux vectors are defined by


A~N

q =
-
I
A~l-
qA' (36)

In addition, we have used axiom 3 to eliminate the term involving QAB and axiom 2 to eliminate
the term involving e! + tV!2. One of our objectives is the derivation of an equation for the total
enthalpy per unit mass which is identified as h and defined by
A~N

ph = I
A=:; 1
PAhA (37)

To achieve this we need to use the decomposition indicated by Equation 16, and in addition we
will decompose the species stress tensor according to

(38)

in which PA is defined by Equation 32. If we focus our attention on the right side of Equation 35
and use Equations 16 and 38 we can obtain, after some algebraic manipulation, the following result

A~N

+ y. ¥p + I
A=l
(!A:¥!:!A
-
+ !:!A . ¥PA)

(39)

Here we have defined the total viscous stress tensor by

(40)

One can now use the second axiom for mass to obtain

(41)
Heat Transfer in Catalytic Packed Bed Reactors 369

and use of the fourth axiom for the mechanics of multicomponent systems yields
A=N
'\' v*' v_ = 0
L.. r A_A (42)
A=l

With these simplifications we can express the enthalpy equation as

(43)

It is important to note that the last term in this result represents a source of kinetic energy owing
to chemical reaction and it can be expressed as

(44)

From this we can see that the last two terms in Equation 43 represent a diffusive rate of work and
a diffusive kinetic energy source owing to chemical reactions. One can put forth convincing argu-
ments that these terms are negligible in the total enthalpy equation, thus we express Equation 43 as

(45)

The enthalpy of species A per unit volume can also be represented in terms of the molar concen-
tration cA and the partial molar enthalpy HA according to

(46)

This allows us to write Equation 45 as

(47)

which is essentially the form given by Bird, Stewart and Lightfoot [9]. Equations 45 and 47 are
particularly useful relations for the development of jump conditions at phase interfaces where we
are concerned with enthalpy changes associated with adsorption and chemical reaction.
There are a variety of routes that one can follow to obtain an equation for the temperature from
Equation 45. In this development we will use Equation 37 along with the decomposition indicated
by Equation 16 in order to obtain a total enthalpy equation given by

(48)

This is a convenient starting point for the development of a macroscopic enthalpy balance; how-
ever, for our purposes we wish to simplify the left hand side of this result using the total continuity
equation given by Equation 5. This leads to

ata (ph) + Y. (phy) = Pat


ah
+ py . yh (49)
370 Multiphase Reactor Operations

By axiom 4 given earlier as Equation 26 we know that h is a function of the state of the system,
and we can identify the state of the system in terms of the temperature, pressure and the N-l in-
dependent mass fractions. This means that the enthalpy per unit mass can be represented by

(50)

The partial derivative with respect to time takes the form

ilh (ilh) (ilT) (ilh) (il P) A~N-l (ilh) (ilWA) (51)


at = ilT p,w at
+ ilp T,w + at A"f;, ilw A T,p,WB Tt
Here we have used a sUbscript W to indicate that all N - 1 independent mass fractions are held
constant, and we have used a subscript WB to indicate that all independent mass fractions except
W A are held constant. The constant pressure heat capacity for a mixture is defined by

(52)

and the derivative of the enthalpy per unit mass with respect to the total pressure is related to the
coefficient of thermal expansion for a mixture according to

ilh) 1
( -ilp T,w =-(I-TfJ)
P
(53)

Here fJ can be represented as

P
fJ = -r;1 (ililT ) p,w
(54)

Use of these relations in Equation 51 leads to

ilh= c (ilT)
- - 1
+-(I-TfJ) (il- P) + A~N-l
L (~ilh ) (ilWA)
~ (55)
ilt p ilt P ilt A= 1 ilw A T,p,WB ilt

and by analogy we can express the gradient of the enthalpy as

(56)

Substitution of these two results into Equation 49 allows us to express that equation as

il DT Dp A=N-l (ilh) (ilWA )


~(ph) + Y'(phy) = pC p - + (1- TfJ)- + L ~il~ P ~il~+ y' YW A (57)
ilt Dt Dt A= 1 W A T,p,WB t

in which the material derivatives are given by

Dp ilp
-=-+v'Vp (58)
Dt ilt --

A bit of manipulation with the total continuity equation leads to

ilWA
+ v' VWA ) ilPA + V, (PAV A) -
P( ~ ilt - -
= -
ilt - -
V, (PAU A)
- -
(59)
Heat Transfer in Catalytic Packed Bed Reactors 371

and use of the species continuity equation given by Equation 3 allows us to express this as

(60)

Use of this result in Equation 57 and substitution of the latter into Equation 48 yields an equation
for the temperature given by

(61)

One can follow the developments in Slattery [5] to show that

(62)

This result can be used to express the reaction rate term in Equation 61 as

(63)

To the right side of this result we can add and subtract hNrN and then use Equation 2 to obtain

(64)

Because the diffusive mass fluxes are con stained by


A=N
I
A=l
PA!!A = 0 (65)

we can repeat this same procedure to obtain

(66)

Substitution of Equations 64 and 66 into Equation 61 leads to a greatly simplified form given by

(67)

In general, one wants to express the source term owing to chemical reactions in terms of partial
molar enthalpies and molar rates of production per unit volume. Because of this, we express the
last term in Equation 67 as
A=N A=N
I hArA = A=l
A=l
I HARA (68)

in which RA is the molar rate of production of species A owing to chemical reaction. Use of
372 Multiphase Reactor Operations

Equation 68 in Equation 67 allows us to write the temperature equation as

(69)

This result is equivalent to that given by Bird, Stewart and Lightfoot [9]; however, one must be
careful to note the difference in the interpretation of q. In this approach q represents only the heat
flux resulting from conduction, and it is expressed by Fourier's law as -

(70)

for isotropic media.

JUMP CONDITIONS

In this development we will ignore all tangential interfacial transport in the construction of jump
conditions at phase interfaces. This means that surface diffusion will be ignored, and we will consider
only the transport of mass and energy normal to the Y-K interface shown in Figure 3. The interface
moves with a velocity '::Y as indicated in Figure 3, and we would like to develop the jump condition
for some generic transport process described by

ai/J + yo . (,I,U)
~ = yo . n + (71)
at - 'I' - --
(J

"{- phase

Figure 3. Moving phase interface.


Heat Transfer in Catalytic Packed Bed Reactors 373

Here t/I represents some quantity per unit volume and the convective flux, diffusive flux and source
of this quantity are represented by t/lg, nand (J respectively. It is important to keep in mind that
in this presentation t/lu and n have no tangential components. Given this restriction, one can follow
developments in Slatt~ry [5], to arrive at the jump condition given by

(72)

here ~ represents the quantity per unit area and 8 represents the source of this quantity per unit
area.
In most reactor design applications one uses the molar form of the species continuity equation,
thus Equation 3 is expressed as

aCA
-at + V- . (CAVA)
- = RA (73)

and the jump condition is given by

(74)

In this representation cAis the surface concentration (moles/m2) and RA is the heterogeneous molar
rate of production of species A (moles/sm 2). In general, it is convenient to use the molar form of
the enthalpy equation in order to develop the thermal jump condition. Thus we make use of
Equation 47 and neglect interfacial mechanical work terms in order to obtain

Here HA represents the partial molar enthalpy of the adsorbed species, and this quantity will be
important whenever the heat of adsorption is significant.

HEAT TRANSPORT IN POROUS CATALYSTS

In Figure 4 we have illustrated a fluid-solid system in which the (J-phase represents porous
catalyst pellets. Our first objective is to analyze the heat transport process that occurs in the
catalyst pellets and we will then move on to the analysis of the fluid phase shown in Figure 4. Our
treatment of both processes will be based on Equation 69 which we list here as

(76)

For a single reaction, the individual reaction rates can be expressed in terms of the stoichiometric
coefficients vA and the overall reaction rate according to

A= 1,2, ... ,N (77)

This allows us to express the source term in Equation 76 as

(78)
374 Multiphase Reactor Operations

Figure 4. Heat transfer in a fluid-solid system.

Here ~H, represents the heat of reaction that is negative for exothermic reactions and positive
for endothermic reactions.
In our study of heat transport in porous catalysts, we can neglect the effect of homogeneous
reactions, the mechanical energy source terms, and the convective and diffusive fluxes of energy. This
leads to an energy equation of the form

pc -
aT = - v . (q + qR) (79)
P at - - -
The importance of radiant energy transport depends on both the absolute temperature and the path
length for radiant energy transport. At room temperature, radiation is an important mechanism
of energy transport in fiber glass insulation [10, 11J; however, for the typical porous catalyst the
path length for radiant energy transport is small enough so that this mechanism can be ignored
for the typical reactor design problem. This allows us to write our energy transport equation as

aT
Pc p - at = V·
- (kVT)
- (80)

We are interested in developing the volume-averaged form of this equation for the catalyst pellet
illustrated in Figure 5. The macroscopic length scale is identified as L and one should think of
Heat Transfer in Catalytic Packed Bed Reactors 375

Figure 5. Porous catalyst pellet.

this as the effective particle diameter, dp. It is impossible to solve Equation 80 for both the fluid
and solid phases that make up the porous catalyst, and we seek instead to develop an equation
for the average temperature in the region identified by v". The details of this region are shown
in Figure 6 where the solid phase has been identified as the K-phase and the fluid phase as the
y-phase. One should keep in mind that the y-phase and the f3-phase are the same fluid phase;
however, it is convenient to refer to the fluid phase that is inside the porous catalyst as the y-phase.

/Y-PhaSe

Figure 6. Averaging volume within a porous catalyst.


376 Multiphase Reactor Operations

At the K-y interface there will be a source or a sink of thermal energy owing to heterogeneous
chemical reactions. In order to identify the conditions at the catalytic surface, we assume that the
K-phase is impermeable and write the mass and energy jump conditions as

(81)

(82)

These relations are obtained directly from Equations 74 and 75 using the convention that Dy. =
-D.,. One can use Equation 81 in Equation 82 and arrange the latter in the form

(83)

Once again we will restrict our discussion to the case of a single reaction so that the individual
rates of heterogeneous reaction can be expressed as

A = 1,2, ... ,N (84)

The heat of reaction is defined in a manner analogous to Equation 78


A=N
IlH, = L
A=l
vAH Ay (85)

and this allows us to express the thermal jump condition as

(86)

In reactor design calculations, it appears that the left side of this result is universally neglected.
Since the catalytic surface can often be treated as quasi-steady from the point of view of mass
transfer [12, 13], it is attractive to impose a comparable constraint for the heat transfer process.
This quasi-steady constraint can be expressed as

(87)

and this allows us to write the thermal jump condition as

(9, - 9.) . D,. = .1.H,R (88)

At this point we are ready to use the method of volume averaging [14-17] to develop equations
for the average temperature in the region V. shown in Figure 5.

Volume Averaging

The detailed problem under investigation is the heat conduction process associated with the
system illustrated in Figure 6 and it can be described as

in the K-phase (89)


Heat Transfer in Catalytic Packed Bed Reactors 377

Boundary Condition 1:

T. = T y, at the K-y interface (90a)

Boundary Condition 2:

at the K-y interface (90b)

(pcp),
aT, =
at y. (k,YT,), in the y-phase (91)

Boundary Condition 3:

T. = G(t), at A.e (92a)

Boundary Condition 4:

Ty=F(t), at Aye (92b)

Here A .. and Aye represent the entrances and exits of the K and y-phases at the surface of the
porous catalyst illustrated in Figure 5. In general, the boundary conditions at the interface between
the catalyst pellet and the surrounding fluid are known only in terms of average temperatures;
however, the boundary conditions given by Equations 92 need to be stated as a matter of com-
pleteness. With the exception of the source term in Equation 90b, this problem is identical to that
studied by Nozad et al. [18J among many others.
Our first objective is to derive equations for the local volume-averaged temperatures in the K and
y-phases. The K-phase superficial average temperature is defined by

(93)

where V. is the volume of the K-phase contained within the averaging volume V•. The volume
fraction of the K-phase is given by

(94)

and the volume fractions for the K and y-phases are naturally constrained by

(95)

To obtain an equation for <T.> we associate an averaging volume with every point in space and
form the integral of Equation 89 to obtain

1 v.
(pcp). v". f (aT.)
at dV = v".1 fv. y. (k.YTK) dV (96)

It is important to note at this point that we have not assumed that (pcp). is constant, but rather
we have neglected variations in this quantity within the averaging volume v".. Since the K-phase
is assumed to be rigid, we can interchange integration and differentiation on the left hand side
of Equation 96 to obtain

a<T.>
(pcp). -at- = <y. k.YT.> (97)
378 Multlphase Reactor Operations

Use of the spatial averaging theorem [19] allows us to interchange integration and differentiation
on the right hand side of Equation 96 or Equation 97. This yields

(98)

in which A., represents the interfacial area contained within the averaging volume.
It is consistent with the assumption made concerning (pcp). to neglect variations of k. within the
averaging volume. This allows us to write

(99)

so that Equation 98 takes the form

(\00)

In general we would like to work with the K-phase intrinsic average temperature which is defined
by

<Ty=~r
V Jv • .
T.dV (101)

and this average temperature is related to the superficial average by

(102)

We can use this representation in Equation 100 to obtain

(103)

At this point we are confronted with two problems: the area integral of !}.,T. and the interfacial
flux term. The former problem can be simplified by the use of Gray's decomposition [20]

(104a)

T, = <T,>' + T, (104b)

in order to obtain

(105)

It has been shown by Carbonell and Whitaker [21] that this relation can be simplified to

(106)
Heat Transfer in Catalytic Packed Bed Reactors 379

provided the macroscopic length scale ilJustrated in Figure 5 is large compared to the radius of
the averaging volume, v.,. The averaging theorem can now be used to develop the geometrical
relation

(107)

so that Equation 106 takes the form

(108)

When this result is substituted into Equation 103 we obtain the governing equation for <T >".
K

(109)

The analysis of the y-phase heat conduction process is identical to that just given and leads to

(110)

If we were interested in developing a two-equation model of heat conduction in porous catalysts,


we would need to utilize the decomposition given by Equation 104 for both TK and Ty in the
flux terms in Equations 109 and 110 and then develop a closure scheme for TK and Ty. This type
of approach has been explored for the process of diffusion in a micropore-macropore model of a
porous catalyst [22]; however, it is not needed here because the constraints associated with local
thermal equilibrium are generally satisfied.

One-Equation Model

The one-equation model is based on the assumption that the energy transport process can be
characterized by a single temperature. This assumption is valid when the system is in a state of
local thermal equilibrium [23-25] and this condition appears to prevail for almost all practical
problems of heat conduction in porous media. It seems intuitively appealing that the spatial average
temperature, given by

(111)

would be the proper single temperature to characterize the heat conduction process. One can
proceed on firm ground by representing the K and y-phase temperatures by

<T >" =
K <T) + T. (112a)

<Ty)y = <T) + Ty (112b)

where T. and T y represent large scale spatial deviations as opposed to the smalJ scale deviations
1'. and 1'1" When Equations 112 are used in Equations 109 and 110 and the latter two equations
380 Multiphase Reactor Operations

are added, we obtain

[E.(pCp). + Ey(pCp)y] o~~> = y. {E.k.[ Y<T> + ~.


ty 11.y1'. dA]
+ E,k,[ Y<T) + ~, fAy. 11,.1', dA ]} + ~ ty 11K'
01'. 01',
. [k.YT. - kyYT ,] dA - {EK(pCp)K at + EJpCp), at
- y. (E.k.yT.) - y. (Eyk,YT,)} (113)

Local thermal equilibrium requires that the following constraints be satisfied

01'. o<T>
-«--
OTy 0<T)
-«-- YTK« Y<T>, YTy« Y<T) (114)
ot ot' ot ot'

and when these conditions prevail, we can discard the last four terms in Equation 113 and use
the boundary condition given by Equation 90b to arrive at

(115)

Here <p> is the spatial average density defined in the manner suggested by Equation 111 and
C p is the mass fraction weighted heat capacity defined by

(116)

In order to be confident about neglecting the last four terms in Equation 114, one needs estimates
of 1'. and Ty so that the magnitude of the terms involving these large scale deviations can be
determined. The route to these estimates is not obvious; however, a place to start is with Equations
111 and 112, which can be used to express 1'. and 1', as

1'. = <T.>" - <T> = E,«T.>' - <Ty>Y) (117a)

Ty = (Ty>y - <T> = E.«TY - <T.>") (117b)

Here it becomes clear that we need an estimate of <T.>' - <T,>' if we are to be able to deter-
mine under what circumstances local thermal equilibrium is valid. The first development of an
estimate of <T.>" - <T,>' was given in a study of drying porous media [24] and was subsequently
extended to packed bed catalytic reactor design [25]. The crux of the estimation scheme is asso-
ciated with the interfacial flux terms in Equations 109 and 110, and for the case under investigation
these estimates take the form

(118a)

(l1Sb)
Heat Transfer in Catalytic Packed Bed Reactors 381

Here t. and t, are the characteristic lengths illustrated in Figure 6, and a., is the interfacial area
per unit volume that can be written explicitly as

(119)

In obtaining these estimates of the interfacial flux we have used <T;)., to represent the area-
averaged temperature at the K-l' interface and this average is expressed as

(120)

Here T, is equal to TK and T, according to the boundary condition given by Equation 90a. We
can eliminate <T;)., from Equations 118a and b by means of the flux boundary condition given
by Equation 90b, and this allows us to estimate the interfacial fluxes given by Equations 118a and b as

(121a)

(121b)

Use of Equations 121a and 117a in Equation 109 allows us to obtain estimates of TK in terms of
the derivatives of <T K>' that appear in Equation 109. These estimates of T. can then be used in
Equation 114 to determine under what circumstances Equation 115 is a valid representation of the
heat transfer process. The analysis is very long and algebraically complex and the results for the
packed bed heat transfer problem have been presented elsewhere [25]. Following that previous
development, we can express the constraints associated with local thermal equilibrium as

(122a)

(122b)

(123a)

(K~ff)
kw
(4-)
L a. w
<1 (123b)

Here we have used the convention that

W= K,Y; f/ = K,l';

and we have used IX to represent the thermal diffusivity according to

(124)

where K~ff is the effective thermal conductivity.


382 Multiphase Reactor Operations

The constraints given by Equation 122 result from the direct comparison of transient and con-
ductive terms in Equation 113. For example, Equation 122a results from the constraints

aT a<T)
- K« - -
aT, a<T)
-«-- (125)
at at' at at
while Equation 122b is derived from comparable constraints concerning the conductive terms. The
constraints given by Equation 123 are obtained by requiring that the large scale deviation terms
in Equation 113 are small compared to the interfacial flux term in that equation. These constraints
involve the term .1.H,R as indicated by the form of Equation 115. It is shown elsewhere [25] how
this heterogeneous thermal source is eliminated from the analysis in order to arrive at Equations
123.
If we use as an approximation

(126)

we can replace the constraints given by Equations 122 and 123 with

(l27a)

(127b)

provided (t.kw/twk.) and K~frlkw do not differ greatly from unity. For porous catalysts t. is usually
extremely small compared to .;;:j and L, thus the condition of local thermal equilibrium is gen-
erally valid. This means that we can return to Equation 115 and direct our attention to the develop-
ment of reliable expressions for T Kand T yo

Closure

When local thermal equilibrium is applicable we can use a single temperature to describe the
heat transfer process, and that single temperature is governed by Equation 115 which we express as

<p)C p a~:) = y. {EKk{Y<T) + ~K ty !!K,TKdA] + E,k,[Y<T) + ~, fAy, !!'KT, dA]}


- aK,<.1.H,R)K' (128)

To obtain a closed form of this result we must be able to represent TK and T, as functions of
the dependent variable. The most reliable source of information concerning these fields rests with
the governing differential equations for TK and f,. These can be derived following the method of
Crapiste et al. [26] which we will illustrate in the following paragraphs.
We begin our development with Equation 89 and use the decomposition given by Equation
l04a to arrive at

(129)

The intrinsic phase average of this result is given by

(130)
Heat Transfer in Catalytic Packed Bed Reactors 383

At this point we use the same idea that led from Equation 105 to Equation 106 and assume that
averaged quantities can be treated as constants within the averaging volume. This means that Equa-
tion 130 takes the form

V. f [01'.
1 v. (pcp). at - y. (k.yT.) dV -] = - [O<T.>"
(pcp). -o-t- - y. (k.y<T.)") ] (131)

and we can use this result in Equation 129 to obtain an integral-differential equation for T. given
by

(132)

While this governing equation for 1'. appears to be more complex than the governing equation for
T., it is susceptible to several simplifications that lead to a relatively simple boundary value problem
for 1'. and also for 1',. To begin with, we can express Equation 127a as

(133)

and note that when this constraint is satisfied the T.-field is quasi-steady. Since we have already im-
posed this constraint in the development of the one-equation model, it is appropriate to impose it
at this point and expression Equation 132 as

-
y. (k.yT.) =
.f
V1 -
v. y. (k.yT.) dV (134)

In some cases one may wish to take into account the spatial variations of the effective thermal con-
ductivity over an entire catalyst pellet. However, the closure problem for T. is a local problem and
in the domain of this local problem it is quite acceptable to impose the constraint

(135)

This allows us to express Equation 134 as

(136)

with a similar result for l'y-


In order to develop the boundary conditions for 1'. and 1', we return to the boundary condition
given by Equation 90a and make use of the decompositions given by Equation 104 and 112 to obtain

Boundary Condition 1:

1'. + <T) + f. = 1', + <T) + f" at the K-y interface (137)

Clearly this reduces to

Boundary Condition 1:

at the K-y interface (138)

and further simplification requires some knowledge of the magnitude of the large scale deviations,
384 Multiphase Reactor Operations

TK and Tr relative to the magnitude of the small scale deviations, i". and f,. Since T, is the difference
between two average temperatures

T, = <T,), - <T) (139)

while f, is the difference between a point temperature and an average temperature

(140)

we should expect that T,« f K. Estimates ofT, and T, are available [24, 25] and they were used to
produce the constraints given by Equations 122 and 123 associated with the one-equation model.
On the basis of the closure problem for fK and f" Nozad et al. [18] have produced estimates off,
and f,. When these estimates are used with the constraints

(141)

one finds a more severe set of constraints than those given by Equations 122 and 123. We list the
constraints associated with Equation 141 as

(142a)

(l42b)

and note that the first of these will always be satisfied at sufficiently long times while the second will
always be satisfied at sufficiently large values of the macroscopic length scale. One should keep in
mind that the constraints given by Equations 122, 123, and 142 are all based on order of magnitude
analysis, and this means that they provide nothing more than indications about when the assump-
tions will succeed or fail. A precise delineation between the one-equation model given by Equation
115 and the two-equation model represented by Equations 109 and 110 requires a comparison of
results from both models for a wide range of parameters. This will undoubtedly be done in the
future, but for the present we must rely on Equations 142 in order to express Equation 138 as

Boundary Condition 1:

at the K-y interface (143)

We are now i'n a position to consider the flux boundary condition given by Equation 91. The repre-
sentations for T, and T, illustrated in Equation 137 are used in Equation 90b to obtain

Boundary Condition 2:

!.!,,' k,yf, =!.!,,' k,yf, + !.!,,' k,Y<T)


+ !.!,K· k'yT, - !.!,K . k'y<T) - !.!,' . k,YT, + ~Hi (144)

In developing the one-equation model given by Equation 115 from Equation 113 we have imposed
constraints of the type

(145)

Because of this, the constraints given by Equations 122 and 123 allow us to neglect the terms in
Equation 144 that involve YT Kand yT,.
Heat Transfer in Catalytic Packed Bed Reactors 385

At this point we can summarize our boundary value problem for l'Kand l' y as

(146)

Boundary Condition 1:

f. = fy, at the K-Y interface (147a)

Boundary Condition 2:

at the K-Y interface (147b)

(148)

Boundary Condition 3:

fK = g(t), at A Ke (149a)

Boundary Condition 4:

fy = fit), at Aye (149b)

The latter two boundary conditions are associated with the boundary conditions given by Equations
92, and while f. and l' yare generally unknown at the surface of the catalyst pellet illustrated in Figure
5 we need to include Equations 149 to have a well-posed boundary value problem. In reality, the
boundary conditions at A Ke and Aye are listed more as a remainder of what we do not know rather
than what we do know about l'Kand l'y' In addition to the governing equations and boundary con-
ditions given by Equations 146 through 149, we note that Equations 104 require that

<f.)" = 0, (150)

These conditions are based on the idea that <T K)" and <T,)Y can be treated as constants within an
averaging volume.
If we ignore the influence of the boundary conditions at A Ke and Aye> we can see that there are only
two terms that are the sources of the fK and fy-fields. Thus if Y<T) and ~H,R are zero, a solution
!.O Equations 146 through 148 is fK = fy = O. Given the potential importance of ~H,R on fK and
T" we need to consider this term carefully. The heat of reaction will be a function of the temperature
at the K-Y interface, and on the basis of Equation 90a we can express this idea as

(lSI)

The functional dependence of the heterogeneous rate of reaction is more complex since it depends
on both temperature and the surface concentration of the reacting species. If the catalytic surface
can be treated as quasi-steady (12), the reaction rate can be expressed in terms of the bulk concen-
trations and we express this idea as

(152)

In terms of the one-equation model given by Equation 128, it is only the area-average of ~H, that
is required. Since the closure problem must lead to a representation of l'Kand 1', in terms of <T)
and perhaps <~H,R)K" we need to decompose ~H,R according to

(153)
386 Multiphase Reactor Operations

At this point we make an important simplification and assume that the deviation, AH,R, can be
neglected relative to the area average. This means that Equation 147a takes the form

at the K-y interface (154)

With this simplification we eliminate the possibility of interaction between the temperature devi-
ations, T. and T" and the concentration deviations, cA and cn, etc. It would seem plausible that the
influence ~f CA , Cn, etc. on T. and T, would be small compared to the influence of <c A )', <cn)', etc.
via <AH,R)KV; however, a detailed analysis remains to be done and we will encounter this problem
again.
If one accepts the use of Equation 154 in place of Equation 147b, the analysis follows the earlier
work of Ryan et al. [27, 28], Nozad et al. [18] and Ochoa et al. [29]. We begin by representing TK and
T, by the following expressions
T. = ~. Y<T) + r<AH,R)., + '" (155a)

T, = f" Y<T) + s<AH,R)., + ( (155b)

in which", and ( are completely arbitrary functions. Because'" and ( are arbitrary, we can specify
the functions ~, £ rand s according to the following boundary value problems.

Problem I

(156)

Boundary Condition 1:

-g =f-' at the K-y interface (157a)


Boundary Condition 2:

at the K-y interface (157b)

(158)

Boundary Condition 3:

at A.e (159a)
Boundary Condition 4:

1= l(t), at A,e (159b)

(160)

Problem II

(161)

Boundary Condition 1:

r = s, at the K-y interface (162a)


Heat Transfer in Catalytic Packed Bed Reactors 387

Boundary Condition 2:

!!y• . k.£r = !!y• . ky£s + 1, at the K-Y interface (162b)

(163)

Boundary Condition 3:

r = r(t), at A.e (164a)

Boundary Condition 4:

s = s(t), at Aye (l64b)

<rY = 0, (s)Y = 0 (165)

When g,1, r, and s are specified in this manner, one can follow the analysis of Nozad et a!. [18] or
Whitaker [30,31] to show that i/J and ~ make negligible contributions to T. and Ty respectively.
It should be clear at this point that the closure problem is nearly as complex as the original prob-
lem stated by Equations 89 through 92. However, it should be intuitively appealing that T. and T,
are not influenced by the boundary conditions imposed at A .. and Aye. The same holds true for the
fields g, f, rand s, and this means that it should be sufficient to solve for these functions in some re-
presen"t~ive region such as that shown in Figure 7. The results can then be used to evaluate T.

Figure 7. Unit cell in a spatially periodic porous medium.


388 Multiphase Reactor Operations

and Ty which are given by

TK = ~. Y<T> + r<~H,R>KY (166a)

Ty =!. Y<T> + s<~H,R>KY (166b)

If we want to solve for g, f, rand s in the region illustrated in Fig. 7, we must be willing to abandon
the boundary conditions given by Equations 159 and 164. This leaves us without any long range
variables in the closure boundary value problems and the region shown in Figure 7 automatically
becomes a unit cell in a spatially periodic porous medium. Under these circumstances the boundary
conditions given by Equations 159 and 164 can be replaced with the following periodicity conditions

~(r +0 = ~(r), r(r +0 = r(r), i = 1,2,3 (167a)

!(r + 0 = [(r), s(r + 0 = s(r), i = 1,2,3 (167b)

Here {, represents the three non-unique lattice vectors needed to construct a spatially periodic porous
medium. Many of the details concerning spatially periodic porous media are given by Brenner [32J.
We now return to Equation 128 and make use of Equations 157, 162, and 166 in order to express
our one-equation model for the temperature as

+ Y . {[kK
~
- ky fv. !!Kyr dA] ~} -
<~H,R>K'
~
aKy<~H,R>KY (168)

Here we have treated Y<T> and <~H,R>KY are constants within the averaging volume, thus these
terms have been removed from the integrals over the interfacial area, Aq
Without solving Problem II for rand s, we can use the boundary conditions to estimate the mag-
nitude of r as

(169)

When this result is used in the next to the last term in Equation 168 we obtain

(170)

Here we have again used w = K, Y and 1'/ = K, Y with the restriction that w # 1'/. The constraint

(171)

is similar in nature to that given by Equation 122b at it allows us to discard the term represented
in Equation 170 relative to the last term in Equation 168. This allows us to express the one-equation
model for the temperature as

(172)
Heat Transfer in Catalytic Packed Bed Reactors 389

Here JS.'ff represents the effective thermal conductivity tensor, which is defined by

(173)

Since ~'ff is a rather weak function of the temperature, one finds that Equation 172 is generally
written as

(174)

Under these circumstances the skew-symmetric part of ~eff makes no contribution to the heat flux
and it is appropriate to represent the thermal conductivity tensor as

(175)

At this point we are confronted with two problems: (1) determination of the effective thermal con-
ductivity for the porous catalyst, and (2) evaluation of the heterogeneous thermal source.

Heterogeneous Thermal Source

The heat of reaction, ,1H" will depend on the temperature at the K - Y interface, and on the basis
of Equation 90a we can express this idea as

(176)

The functional dependence of R will be more complex since it depends on the temperature of the
surface and the surface concentrations of the various reacting species. However, in many cases a
catalytic surface can be treated as quasi-steady [12] and under those circumstances the functional
dependence of can be expressed as

(177)

Without offering any specific constraints, we assume that the average of products that makes up the
heterogeneous thermal source can be expressed as the products of averages. We express this idea as

(178)

Given the temperature dependence of the heat of reaction indicated by Equation 176, one can wonder
how the area average of ,1H, depends on <T>. This question has been explored elsewhere [33] and
that study indicates that <,1H,>" can be expressed as

(179)

in which y is the position vector relative to the centroid of the averaging volume. Whenever the
following constraint is satisfied

(180)

one can simplify equation 179 to

(181)
390 Multiphase Reactor Operations

Similar constraints can be developed for the functional dependence of (R)., on (T), (cA )', (c B )',
etc., and when these constraints are satisfied one can express Equation 178 as

(182)

Here it is understood that ~H, and R depend on the volume-average quantities (T), (cA )', (c B )',
etc., in the same manner as they depend on T., cA , cB , etc., as indicated in Equations 176 and 177.
It would appear that the form indicated by Equation 182 has been universally used by those who
study the nonisothermal catalyst pellet problem; however, the validity of that relation certainly needs
to be examined for cases of practical importance. Use of Equation 182 in Equation 174 leads to the
following form of our one-equation model

(183)

In accepting this simplified form one must keep in mind that R will be an exponential function of
the temperature and the constraint comparable to that given by Equation 180 may be violated in
the presence of high temperature gradients.

Effective Thermal Conductivity

From Equation 175 and closure Problem I (Equations 156-160) we see that the effective thermal
conductivity tensor depends only on k., and k, and the geopetry of the system under consideration.
Even though the heterogeneous thermal source term (~H,R)., appears in the general closure prob-
lem for f. and f" we have been able to show that it gives rise to a source term of negligible impor-
tance in the governing equation for (T). This means that the effective thermal conductivity can be
measured in passive systems (without chemical reactions) and used in the analysis of active systems
(with chemical reactions). In addition this means that Problem II of the closure problem given by
Equations 156 through 165 need not be solved. Problem I of the closure problem can be simplified
by repeating a proof given by Ochoa et al. [29] which allows one to show that the Laplacians of
~ and f are zero. This leads to a closure problem given by

Closure Problem

"12q = 0, in the K-phase (184)

~ = f, at the K-y interface (185a)

at the K-y interface (185b)

"1 2f= 0
- '
in the y-phase (186)

i = 1,2,3 (187)

(D' =0 (188)

The proof given by Ochoa et al. [29] relies upon the spatial periodicity of g and f, but aside from
that restriction it is quite general and has other applications within the method of volume averaging.
There are numerous experimental results for the effective thermal conductivity of unconsolidated
porous media, and these have been considered from the theoretical point of view by Nozad et al.
[18]. The theoretical results are based on the solution of Equations 184-188 and the spatially peri-
odic model of a porous medium illustrated in Figure 8. Unhappily, the parameter cia is adjustable
and without a detailed understanding of the solid mechanics of unconsolidated media we can only
search for a value of cia that provides good agreement between theory and experiment. In Figure
Heat Transfer in Catalytic Packed Bed Reactors 391

Figure 8. Model for particle-


particle contact.

9 we have shown a comparison between theory and experiment for cia = 0.01 and a wide variety
of experimental data. Much of the scatter results from the fact that the volume fraction for the experi-
mental data ranges from E'y = 0.31 to E'y = 0.52 while the theoretical calculations are based on E'y =
0.36. The geometrical model shown in Figure 8 represents an isotropic system with regard to the
heat conduction process, and it is the single distinct component of lSorr defined by Equation 175
that is shown in Figure 9. Because of the wide range of experimental studies presented in Figure 9
and the reasonably good agreement with the single theoretical results for cia = 0.01, one is tempted
to think of the theoretical result as a universal curve for unconsolidated porous media. This would
most certainly be a mistake. In an unpublished study by Wu [34], the experimental system of Nozad
[35] was re-used to measure the effective thermal conductivity for the systems consisting of spheres
of urea formaldehyde, stainless steel, bronze, glass, and aluminum with air being the continuous
fluid phase. All experimental values for the effective thermal conductivity were in excellent agreement
with those obtained by Nozad except for the aluminum-air system. The porous medium made from
the 1 mm-diameter aluminum spheres used by Nozad produced an effective thermal conductivity
that was four times larger than that measured by Nozad. An inspection of the spheres indicated that
they had been deformed by use in other experiments performed in our laboratory and therefore
produced a porous medium with a higher effective contact area. When Wu repeated the experiments
with new aluminum spheres, the original results of Nozad were reproduced without difficulty. The
conclusion here should be obvious: The effective thermal conductivity is very sensitive to the nature
of the particle-particle contact, especially for large values of the ratio, k./k y. A dramatic example
of this fact is contained in the theoretical study of Batchelor and O'Brien [36]. They considered a
regular array of spheres that were in contact at single points as illustrated in Figure 10. Their system
is similar to that shown in Figure 8 with the squares replaced by spheres and the contact area tending
toward zero, i.e., cia .... O. Batchelor and O'Brien used the method of matched asymptotic expansions
to solve the point-contact heat conduction problem; however, they obtained only the first term of
the inner expansion. This left them with an adjustable coefficient in their expression for the effective
thermal conductivity, but their analysis did provide the correct form for Kerr at larger values ofk./k y.
At large values ofkK/ky. Nozad et al. [18] were able to use a regular perturbation expansion to show
that Kerr was a linear function of k./ky for a finite contact area; however, for point-contact between
~
N

1000 I 71
3:
• Preston 0.4 - 0.49 §,
o Wilhelm and Johnson 0.4 - 0.52 -6'
;:r
• Krupiczka 0.4 - 0.45 DI
UI
c Vershoor and Schuit 0.38 - 0.42 ~

.. Shu mann and Voss 0.35 - 0.44 ::u


.. Kling 0.38 cIa =0_01 ~
DI
<> Waddams
• Vagi and Kuni
0.4 - 0.49
0.4 ~
100 ., Gorring and Churchill 0.45 o
"C
co Kannuluik and Martin 0.31 - 0.35 ~
C This Work 0.39 - 0.41 !!!.
Keff / ky o·
::::I
Void Fraction = € Y UI

10

ly = 0.36 for Theoretical Calculations

10 100 1,000 10,000 100,000

kK/ ky

Figure 9. Theoretical and experimental values of the effective thermal conductivity.


Heat Transfer in Catalytic Packed Bed Reactors 393

Y- phase

Figure 10. Model for point-contact heat transfer.

particles Batchelor and O'Brien [36J were able to show that Kerr was a logarithmic function of kK/ky.
In Figure 11 we have presented the theoretical result for Batchelor and O'Brien along with the experi-
mental data of Swift [37] and some recent results ofShonnard [38]. Swift's data were obtained using
metal powders with the particle diameter being on the order of 100 microns, while Shonnard's results
were obtained using metal hemispheres having diameters on the order of 5 centimeters. Shonnard's
experiments were carried out with a unit cell, such as the one illustrated in Figure 10, with the inten-
tion of testing the Batchelor and O'Brien theory. The theoretical line in Figure 11 is given by

~err = 4 In(kK/k y) - 11 (189)


y

in which the number eleven is an adjustable parameter chosen by Batchelor and O'Brien. Certainly
the theory is in reasonable good agreement with the experimental data of Shonnard and it predicts
quite well the trend of the data obtained by Swift. It is not clear why the packed metal powders used
by Swift exhibit point-contact heat conduction characteristics, but they certainly do just that.
In addition to experimental values of the effective thermal conductivity that are significantly lower
than those illustrated in Figure 9, there are values that are significantly higher. The data of Hadley
[39J are an example for which compressed metal powder gave rise to values of K dr considerable
greater than those illustrated in Figure 9. The higher values resulted from both a reduced value of
the void fraction lOy, and from an increase in the particle-particle contact area. Oppio [40] has ex-
tended the calculations of Nozad and her results are shown in Figure 12. While Hadley's results
contain information about the void fraction, and Oppio's results contain information about the
particle-particle contact, there is not sufficient information from either study to make a meaningful
comparison between theory and experiment.
It should be clear from the experimental and theoretical results shown in Figures 9, 11, and 12,
that the effective thermal conductivity of two-phase systems is a strong function of the ratio of con-
ductivities, kK/ky, and the topology of the two phases. Obviously, the theoretical prediction of Kerr
394 Multiphase Reactor Operations

100 C------,------~------_r------_r----__,

• ••• •
10
• - Batchelor & O'Brien

10

Figure 11. Comparison of theory and experiment for point-contact heat conduction.

1000 r---------,----------.--------~--------_rr7~r_~~

/
c/a=O 5
100

10

10 100 1,000 10,000 100,000

Figure 12. Influence of the particle-particle contact area in the effective thermal conductivity.
Heat Transfer in Catalytic Packed Bed Reactors 395

requires a knowledge of kK and k, and the geometrical structure of the porous medium under
consideration.
While the problem of unconsolidated porous media illustrates some successes and some unsolved
problems, the matter of porous catalyst pellets remains in an empirical-experimental state. The
geometrical structure is often explored by means of mercury porosimetry and Dullien [41J has
pointed out that this experimental technique provides unreliable information concerning the pore-
size distribution. Regardless of whether pore-size distributions can be accurately measured or not,
we know that there is a distribution of pore-sizes in any porous catalyst pellet and that a significant
portion of the pores will be small enough so that free-molecule flow will be the dominant mechanism
of mass and energy transport. Because of this we need to consider the micropore-macropore analysis
of heat conduction.

Micropore-Macropore Systems

In our initial study of heat conduction and reaction in a K-y system, we assumed that the thermal
conductivity of the y-phase underwent negligible variations within the averaging volume. In a system
having micropores in which free-molecule flow takes place, the thermal conductivity of the y-phase
will depend on the pore diameter, thus the approximation of negligible variations in k, is no longer
valid.
The micropore-macropore system that we wish to analyze is illustrated in Figure 13 and it is similar
to the system shown in Figure 6. However, in this case we are confronted with a variety of pore sizes

y-phase

Figure 13. Micropore-macropore system.


396 Multiphase Reactor Operations

and in Figure 13 we have highlighted one set of pores that we intend to treat as a single phase. To
be specific we consider this to be the j-phase which has an average pore diameter dj . The entire y-
phase is to be represented in terms of M phases with M -1 of these being micropore phases while the
Mth phase is the macropore phase. The number of phases required to accurately model the system
wiJl obviously depend on the pore-size distribution. We assume that each individual micro pore phase
can be described by a density, a heat capacity, and a thermal conductivity so that the original problem
posed by Equation 89-92 can be replaced by'

(190)

Boundary Condition I:

at AK " , Jl = 1,2, ... , M (191a)

Boundary Condition 2:

at A K " , Jl = 1,2, ... , M (191b)

Jl = 1,2, ... , M (192)

Boundary Condition 1';

at A"V' Jl = I, 2, ... , M (193a)


Jl#v

Boundary Condition 2':

at A"V' Jl = I, 2, ... , M (193b)


Jl#v

Here we have used Jl and v to represent the M distinct y-phases with the idea in mind that Jl = M
represents the macropore phase. For a system of this type the volume averaging theorem for the K-
phase takes the form

(194)

and one can repeat the development given by Equations 93-109 to obtain

(195)

• In proposing the boundary condition given by Equation 191a, we are neglecting the temperature jump that
occurs at gas-solid surfaces. Continuity of the temperature is, of course, consistent with the assumption of local
thermal equilibrium and because of the small length scales associated with micropores this assumption should be
satisfactory. On the other hand, this intuitive line of thinking needs to be replaced with a rigorous analysis that
will be done at some later date.
Heat Transfer in Catalytic Packed Bed Reactors 397

The analysis leading to Equation 110 can be repeated for one of the y-phases that we identify as the
jl-phase in order to obtain

(196)

This result can be summed over all the micro pore phases and the single macropore phase to obtain

(197)

When the constraint of local thermal equilibrium is imposed, we can add Equations 195 and 197 to
obtain

(198)

For simplicity one might extend the definitions given by Equations 116, 115, and 128 to write

I'=M

<P )C p = E.(PCp)K + L
/.L=1
E.(PCp)" (199a)

(199b)

so that Equation 198 can be written as

(200)

The closure problem for the micropore-macropore problem follows the development given by Equa-
tions 129-167. Without going through the details, we note thai the spatial deviation temperatures can
be represented as

11 = 1,2, ... , M (201)


398 Multiphase Reactor Operations

and the one-equation model represented by Equation 200 takes the form

(202)

Here we have used the boundary condition given by Equation 191a along with arguments presented
earlier in Equations 137-143 in order to obtain

Boundary Condition I:

at AK.' Jl = 1, 2, ... , M (203)

The arguments leading from Equation 174 to Equation 183 need to be examined in detail for the
micropore-macropore system illustrated in Figure 13. If they were valid for that more complex
system, one could simply make use of Equation 183 with the effective thermal conductivity defined
by

(204)

About the E" we should remember that

(205)

and about the k" we should remember

(206)

The problem of evaluating the k" for Jl = 1,2, ... , M - 1 is not completely resolved; however, some
insight is given by Prakouras et al. [42].
The closure problem that must be solved in order to evaluate the terms in Equation 204 is given
by a variation of Equations 184-188, and without discussion we simply list the result as

Closure Problem for Micropore-Macropore System

in the K-phase (207)

Boundary Condition 1:

at AK " , Jl = 1, 2, ... , M (208a)

Boundary Condition 2:

at AK " , Jl = 1, 2, ... , M (208b)

in the Jl-phase, Jl = 1, 2, ... , M (209)

• Boundary Condition 1':

i. = i" at A"" Jl = 1, 2, ... , M (210a)


Jl# v
Heat Transfer in Catalytic Packed Bed Reactors 399

Boundary Condition 2':

at A"" J1 = 1, 2, ... , M (21Ob)


J1#V

~(r +.0 = ~V, fir + .0 = Ur), i = 1,2,3 (211)


J1 = 1,2, ... , M

<L>" = 0, J1 = 1,2, ... , M (212)

With the solution of this closure problem may appear to be excessively complex, it is comparable
in difficulty to a random capillary model [43] and indeed that particular geometrical model can be
solved within the framework of Equations 207-212. One must keep in mind, however, that random
capillary tube models are generally used to predict mass transport and in that case the detailed struc-
ture of the K-phase is not particularly important. For the case of heat conduction, we know that the
details of the geometrical structure of the K-phase are very important. Thus accurate prediction of
the effective thermal conductivity of porous catalysts pellets requires more knowledge of the topo-
logical details than is currently available.

FLUID PHASE HEAT TRANSPORT

In the previous section we analyzed the heat transfer process in a porous catalyst phase in order
to obtain a local volume-averaged equation for the temperature given by

(213)

The principle of local thermal equilibrium was used in order to obtain a one-equation model for
K-Y system illustrated in Figure 6, and the temperature <T> is the single temperature that describes
the thermal state of the K-y system. We now turn our attention to the Ii-phase illustrated in Figure
4, and in order to simplify the nomenclature for this analysis we are going to express Equation
213 as

(214)

Here we have assumed that the heat transfer process in a porous catalyst is isotropic, and we have
used nomenclature that is consistent with Figure 4 and previous studies of heat transfer in two-phase
systems [21,25,44]' The boundary conditions at the Ii-(J interface can be expressed as

Boundary Condition 1:

at the Ii-(J interface (215)

at the Ii-(J interface (216)

and the thermal energy equation for the Ii-phase can be expressed as [7]

(217)

When thinking about these equations one should remember that the system under consideration
contains one solid phase and one fluid phase. The solid phase is identified in Figure 6 as the K-phase
400 Multiphase Reactor Operations

while the fluid is indicated as the y-phase. The combination of these two phases is referred to as
the a-phase as illustrated in both Figures 4 and 5, and the fluid phase that is exterior to the K-y
system is identified as the {3-phase.
The boundary conditions given by Equations 215 and 216 are based on the idea that the volume-
averaged temperature in the a-phase can be treated as a point temperature in terms of its relation
to the temperature in the {3-phase. This idea has been used in the analogous mass transfer problem
[22], and in order to indicate the nature of the approximation, we express a more precise form of
Equation 215 as

Boundary Condition la:

at A. p on the {3-a interface (2ISa)

Boundary Condition I b:

at A yp on the {3-a interface (21Sb)

These boundary conditions should be interpreted in terms of Figure 14, which illustrates an interface
between a porous catalyst and a surrounding fluid phase. From Equations 104 and 112 we know
that the boundary conditions can be written as

Boundary Condition la:

Tp = (T) + T. + T., (219a)

Portion of AK{3

Figure 14. Interface between a porous catalyst and a continuous fluid phase.
Heat Transfer in Catalytic Packed Bed Reactors 401

Boundary Condition 1b:

Tp = <T) + T, + f" at A,p (219b)

Since the heat transfer process in the porous catalyst pellet is restricted by Equation 141, we can
express the idea of continuity of temperature as

Boundary Condition la:

Tp = <T) +f K,
(220)

Boundary Condition 1b:

Tp = <T) + f" at A,p (221)

The closure problem described earlier led us to the results

(222)

and these can be used to express the boundary condition as

Boundary Condition la:

(223a)

Boundary Condition 1b:

(223b)

Here we have adopted the nomenclature associated with the transition from Equation 213 to Equa-
tion 214. From the closure problem for ¥ and f we know that the order of magnitudes of these
two vector fields are on the order of the length scales t, and tK where t, is illustrated in Figure 6.
We express this idea as

¥, f = Q(tK , t,) (224)

and note that if tK and t, are sufficiently small we should be able to replace Equation 223 with
Equation 215. Obviously the magnitude ofy~T. has something to do with this simplification and
we will comment on the boundary conditions given by Equation 215 and 223 later in the analysis.
The flux condition given by Equation 216 can be reconsidered from the same point of view and
the results that are analogous to Equations 223 are

Boundary Condition 2a:

(225a)

Boundary Condition 2b:

(225b)

Justifying the reduction of these flux expressions to that given by Equation 216 will be more difficult
than the simplifications of Equations 223 to Equation 216; however, we will delay commenting on
this problem until some later time.
402 Multiphase Reactor Operations

Figure 15. Averaging volume for a packed bed.

Volume Averaging

At this point we accept Equations 214-217 as a reasonable representation of the heat transport
process under consideration and direct our attention to the volume average form of the governing
equations. The averaging volume to be used is illustrated in Figure 15, and we designate this av-
eraging volume as "Y.
The average of Equation 214 can be expressed as

(226)

in which E. is the volume fraction of the u-phase. This is given explicitly by

(227)

It is plausible to neglect variations of (pcp). within the averaging volume so that Equation 226
takes the form

(228)

Our treatment of the conductive term in this equation follows Equations 97-109 and leads to

(229)
Heat Transfer in Catalytic Packed Bed Reactors 403

There is an alternate form of Equation 228 that will be useful to us, and it can be obtained simply
by use of the divergence theorem. This leads to

(230)

in which An, represents the area of entrances and exists of the a-phase contained within the averaging
volurr.e shown in Figure 15. This result will be useful in a subsequent discussion of the traditional
model of heat transfer in packed bed catalytic reactors. In addition, it provides an interesting physical
interpretation of the conductive term in Equation 229, since comparison of Equations 229 and 230
reveals that

(231)

In our treatment of the [:I-phase we will assume that the flow is incompressible so that Equation
217 takes the form

(232)

The volume-averaged version of this result is given by

(233)

One can impose the restriction

(234)

and follow the analysis of Carbonell and Whitaker [21J in order to express Equation 233 as

= y. {Epk{Y<Tp)P + ~)AP" !!pnTp dA]} + ~ L," !!pn' kpYT pdA (235)

Here the term (pcp)PY' <~pTp) represents the dispersive transport, and a representation for Tp is
required in order to specify even the form of this term. Before exploring this problem in detail, we
want to indicate what must be done to extract the traditional model of packed bed heat transfer
from the development presented thus far.

Traditional Model

If one is willing to accept the following postulate concerning the temperature deviation in the
[:I-phase

(236)
404 Multiphase Reactor Operations

and the following definition of a film heat transfer coefficient

(237)

the fluid-phase thermal energy equation takes the form

Ep(pCp)p [ ----a:--
8<T)P + <Yp)P. Y<Tp)P ]

= V' [(~~ff + ~D)' Y<Tp)P] - ap.h(Tp)P - (Tp)p.) (238)

Here we have used the incompressible form of the continuity equation

(239)

in order to simplify the volume-averaged convective transport term, and we have used the definitions

(240)

where 15D should be referred to as the fluid-phase thermal dispersion tensor.


We h-a ve referred to Equation 236 as a postulate since it is easy to show that the representation
given by Equation 236 is incomplete. About Equation 237 we should note that (Tp)pa represents
an area-average defined by

(241)

and on the basis of the boundary condition given by Equation 215 we have

(242)

At this point we could return to the more precise version of Equation 215 that is given by Equation
233, and note that the area-average of Tp is actually given by

(243)

Given the constraints imposed on ~ and! by Equation 160, it seems plausible that the last two terms
in Equation 243 can be neglected. The precise motivation for this would be stated by the restriction

(244)

It is important to recognize that while Equation 237 can stand as a definition of the film heat
transfer coefficient, it forces a commitment to a symmetric representation of the heat conduction
process in the G-phase. We summarize this traditional model as
Heat Transfer in Catalytic Packed Bed Reactors 405

Fluid-Phase Transport

a<Tp)p
Ep(pCp)p [ -a-t-
p
+ <Y.p) . Y<T p)
pJ
= y. [(~~ff + ~D)' Y<Tp)PJ - ap.h«Tp)P - <Tp)p.) (245)

Boundary Conditions

Boundary Condition I:

(246)

Boundary Condition 2:

(247)

Porous Catalyst Phase

(248)

In this approach one must model the porous catalyst phase as spherical pellets, infinitely long
cylindrical pellets, or flat slabs of porous catalyst that are of infinite extent. With this type of model
one is forced to neglect particle-to-particle transport and the nature of this restriction can be made
more precise by referring to Equation 230. The neglect of particle-to-particle heat conduction is
comparable to

(249)

and this allows us to combine Equations 216, 230 and 237 to obtain

(250)

If the heat transfer process in the porous catalyst phase can be treated as quasi-steady, i.e.

(251)

we can use Equation 250 in Equation 245 to obtain

Fluid-Phase Transport

(252)

where

~;ff = ~~ff + ~D (253)


406 Multiphase Reactor Operations

This representation is found often in texts on reactor design; however, the nature of the restrictions
is not always made clear. In order to be more precise than the representation of the heat transfer
problem given by Equations 245-248, we need to avoid the ad hoc expression for Tp given by Equa-
tion 236 and this requires the development of a closure scheme.

Closure

In order to obtain a reliable expression for Tp we make use of the decompositions


(254a)

(254b)

in Equation 217 and arrange the result as

(255)

At this point we follow the method ofCrapiste et al. [26] which is illustrated by Equations 129-132.
With Equation 255, this approach leads to

(pcp)p [ 8T - + Yp· Y<Tp)P J-


atp + Yp· yTp -
y. (kpYTp)

= ~p Iv. {(PCp){8~p + Yp· YTp + yp· Y<Tp)PJ - y. (kpYTp)} dV (256)

and from Equation 214 we can follow the same approach to obtain a governing equation for T•.
This is given by

(257)

Here we have used the decomposition

(258)

in Equation 214 in order to arrive at Equation 257, and when Equations 254a and 258 are used
in the boundary conditions given by Equations 215 and 216 we obtain

Boundary Condition 1:

(259)

Boundary Condition 2:

(260)

At this point it should be clear that Tp depends on Y<Tp)P because of the volume source term,
YP . Y<Tpi, in Equation 255 and because of the surface source term !!p •. kpY<Tp)P in Equation 260.
However, it should also be clear that Tp depends on other quantities and that the functional

Heat Transfer in Catalytic Packed Bed Reactors 407

dependence given by Equation 236 may be a serious over-simplification. Under many circumstances
one can treat the closure problem as quasi-steady on the basis of the constraints

(261)

where Ip and I. are the characteristic lengths illustrated in Figure 4 and t* is a characteristic process
time. Since the closure problem is a local problem, one can neglect the variations of kp and k. in
the solution for T. and Tp. Because of this, and because of the constraints given by Equation 261,
we can express Equations 256 and 257 as

(pcp)P(Yp· YTp + yp· V(Tp)P) - k pV 2 Tp

= ~P Ivp [(PCp)p(yp. YTp + yp. y(Tp)P - kpV 2 T p)] dV (262)

(263)

If we are willing to solve the closure problem for a unit cell, such as the one illustrated in Figure
7, then Tp and T. are spatially periodic, i.e.

T.(r + I;) = T.(r), i = 1,2,3 (264)

Under these circumstances one can follow the proof of Ochoa et al. [29] to show that Equations
262 and 263 take on a simplified form leading to the following closure problem.

(265)

Boundary Condition 1:

(266)

Boundary Condition 2:

!1p •. kp YTp = !1p •. k.yT. + !1p •. k.y(T.>" - !1p •. kpY(Tp)P, (267)

0= k.V2T. + 4>. (268)

(269)

In order to complete our statement of the closure problem, we need to express 4>. in some useful
form. By definition we have

(270)

and in a previous study by the author [33] it has been shown that

(271)

provided the spatial deviation defined by Equation 258 can be represented as

(272)

In Equation 272 ~, is the position vector relative to the centroid of "Y•.


408 Multiphase Reactor Operations

Since Equations 265-268 indicate a more complex functional dependence for 't we must think of
Equation 271 as an approximation. With this statement as a precaution, we follow the development
given by Equations 179-181 and impose the restriction

(273)

so that Equation 270 takes the form

(274)

Use of Equation 258 allows us to write this result as

(275)

and a Taylor series expansion leads to

(276)

Here all derivatives are evaluated at the reference temperature, <T.)·. One can limit the analysis by

(277)

so that Equation 276 takes the linearized form which we write as

(278)

This representation for 1.


would be satisfactory if ¢. were a function of only the temperature T.;
however, ¢. will depend on the temperature and the concentration of the reacting species as we
indicated in Equation 177. If we limit our discussion to the classic case of a first order, irreversible
reaction, the functional dependence of ¢o can be expressed as

(279)

Here we have used Co to represent the intrinsic phase average concentration of the reacting species,
and we could be more explicit by writing
reacting species (280)

It is not too difficult to show that the more general representation of 1>0 is given by

(281)

so that Equation 268 takes the form

0= k
/j
\721' +
IT
(D¢o)
DT(J
l' + (i!¢o)c
(J DCa {f
(282)

Referring to Equations 265-268, with the latter taking the form given by Equation 282, we can
see that the closure problem represents a situation in which 1'p, 1'., and Co must all be determined
Heat Transfer in Catalytic Packed Bed Reactors 409

simultaneously. A little thought will indicate that C. will be coupled to the concentration deviation
in the ,B-phase, thus the closure problem is quite complex for the general case.
In order to illustrate the nature of the closure problem and the structure of the volume-averaged
transport equations, we will avoid the coupled heat and mass transfer closure problem by imposing
the restriction

(a<p.)
ac.
c.« (a<p.) T.
aTa
(283)

This is a severe limitation since just the opposite is true for many practical reactor design problems,
and the need for a more thorough analysis should be obvious.
The decoupled closure problem can now be summarized as

(284)

Boundary Condition 1:

(285)

Boundary Condition 2:

(286)

(287)

At this point one can follow the closure analysis of Ochoa et al. [29], Whitaker [31], or the earlier
work of Zanotti and Carbonell [45], and express Tp and T. as

Tp = Ep' Y(Tp)P + Gp' Y<T.)a + Rp(Ta)" - (Tp)P) (288a)

Ta = Ea' Y<Tp)/1 + Ga' Y<Ta)a + B.a(Ta)a - (Tp)P) (288b)

The new functions Ea> Gp, etc., are determined by a set of boundary value problems that can be
expressed as

Problem I

(289a)

Boundary Condition 1:

(289b)

Boundary Condition 2:

(289c)

(289d)

i = 1,2,3 (28ge)

(E.)" = 0 (289f)
410 Multiphase Reactor Operations

Problem II

(290a)

Boundary Condition I:

(290b)

Boundary Condition 2:

(290c)

_
0- k.Y'
2
G. +(arP·)G
aT. • (290d)

i = 1,2,3 (290e)

<G.>· = 0 (290f)

Problem III

(29Ia)

Boundary Condition I:

Rp = R. + I, (29Ib)

Boundary Condition 2:

(29Ic)

(29Id)

i = 1,2,3 (29Ie)

(29lf)

Although only a portion of Problem I has been explored in detail [46J, some results for flow in
capillary tubes have been produced by Zanatti and Carbonell [45J for the case of heat transfer in
two-phase systems.
A general theory of heat transfer in catalytic packed bed reactors is obtained by substitution of
Equations 288 into Equations 229 and 235; however, before doing so we need to use the decom-
positions for T p and T. in order to express the local volume averaged equations as

Solid Phase

(292)
Heat Transfer in Catalytic Packed Bed Reactors 411

Fluid Phase

Ep(pCpp)p [ -it-
8<T)P
+ <yp)P . Y<Tp)P ]

= y. {Epk{ y<Tp)P + ~p to !!p.Tp dAJ - (PCp)p(ypTp)}

(293)

Here we see "convective-like" terms in both Equations 292 and 293 that result from the interfacial
flux terms according to

~ to !!P.· kpY<Tp)P dA = {~to !!p. dA}' kpY<Tp)P


= -YEp'kpY<Tp)P (294)

In general, these terms will be small except near the wall of a packed bed, non-adiabatic reactor; how-
ever, the length scale constraints upon which Equations 292 and 293 are based, tend to fail near the
wall, thus it may be misleading to retain these terms when a more thorough analysis would generate
other comparable terms in the wall region. We will leave this problem unanswered for the present,
and turn out attention to the use of Equations 288 in Equation 292 and 293.
When Equation 288b is used in the local volume-averaged transport equation for the porous solid
temperature, we obtain

8<T.)· p
E.(pCp). -8-t- + !l•• ' Y<T.)· + !l.p' Y<Tp)

= y. (~'ff' Y<T.)·) + E.<</1.)· - ap.h.p«T.>" - <Tp)P) + y. (~.p' Y<Tp)P) (295)

In this result, the first term on the left side and the first three terms on the right side are similar in
form to what one might develop on an intuitive basis. However, the remaining terms are all the
result of rigorous analysis and cannot be ignored. The coefficients that appear in Equation 295 are
defined by

!l•• = k.[ yEa - ~ t. !!.p . (!R. + YG.) dAJ (296a)

u.p = k.[ ~ t. !!.p . (!R. - YE.) dAJ (296b)

~~ff = E.k. [! + ~ t. !!.pG. dAJ (296c)

(296d)

(296e)
412 Multiphase Reactor Operations

In writing Equation 295 we have ignored variations in the coefficients defined by Equations 296 in
order to gain some degree of simplification. Although the general closure problem leading to the
determination of the coefficients in Equations 296 has never been solved, we have some experience
with various parts of the closure problem [18,21,22,27,28,29,35,45,46]' On the basis of that
experience, we can make the following crude estimates

gp = Q(g.) = Q(t.) Rp = Q(R.) = Q(I) (297)

In addition, we can estimate gradients as

(298)

where the {3-phase length scale can be estimated

(j _ {t., Re < 1
(299)
p- t.IJRe, Re > 1

In Equations 298 we have used l/I" to represent some quantity associated with the u-phase, while
l/I p has a similar meaning for the {3-phase. On the basis of the results given by Equations 297 we
can estimate the coefficients in Equations 296 as

!!•• = Q(k.ap.) (300a)

!l.p = Q(k.ap.) (300b)

J,$.~ff = Q(E.k.) (300c)

J,$..P = Q(E.k.) (300e)

h. p = Q(k./I.) (300d)

This suggests that none of the terms in Equation 295 can be discarded; however, one should re-
member that estimates are just that. If one models a porous medium as a bundle of capillary tubes
[45], it is not too difficult to show that

!l •• = J,$..P = 0, bundle of capillary tubes (30la)

Because of this, there is motivation for discarding the non-traditional or the non-intuitive terms
in Equation 295, but this should be avoided until detailed calculations have been completed. In
fact, we now have sufficient information to suggest that the term involving J5.. P should not be
neglected. Once again we note that for a bundle of capillary tubes [21], one can-arrive at

bundle of capillary tubes (302)

while for more general porous media there is a great deal of theoretical and experimental evidence
indicating that

(303)

Given the similarity between the boundary value problems for E. and g., the estimates represented
by Equations 300c and 300d seem quite plausible, and this suggests that the term in Equation 295
involving J,$..P should not be discarded on an intuitive basis.
Heat Transfer in Catalytic Packed Bed Reactors 413

We now turn our attention to the fluid phase and make use of Equation 288a in Equation 293
in order to obtain

+ <y'p) P . Y<Tp) pJ + ~pp. Y<Tp ) P + ~p•. Y<T.) a


t3 <T p)P
Ep(pCp)p [ -t3~t~

= y. (~;ff· Y<Tp)P) + ap.hp.«T.)' - <Tp)P) + y. (~P.· y<T.Y) (304)

Once again we encounter some non-traditional convective and conductive transport mechanisms.
The five coefficients in Equation 304 that need to be determined by a solution of the closure problem
are given by

(305a)

up.
-
=
kp
--
V
i np.· (IR.
Apa - '" v
+ VG
- -
(pcp)p
p) dA + -~V
fv, -v.R
-v
p dV (305b)

(305c)

(305d)

(305e)

One can repeat the order of magnitude analysis that led to Equations 300 in order to obtain the
following representations

~pp = Q(kpa p.) + Q[(pcp)p<y.p)P] (306a)

~p. = Q(kpapa) + Q[(pcp)p<y.p)P] (306b)

~;ff = Q(Epkp) + Q[EP(PCp)pt.<y.p)P] (306c)

~J. = Q(Epkp) + Q[EP(PCp)pt.<y'p)P] (306d)

k pit., Re < 1
{ (306e)
kp.JRe/t., Re > 1

Once again our order of magnitude estimates suggest that none of the non-traditional terms in
Equation 304 can be discarded unless a detailed solution of the closure problem indicates that they
are indeed negligible.

CONCLUSION

In this chapter we have illustrated a rigorous approach to heat transfer and reaction in porous
catalysts that leads to a traditional result given by Equation 183. The closure problem associated
with Equation 183 was de-coupled from the mass transport closure problem by means of the sim-
plification indicated in Equation 154. The thermal closure problem that led to Equation 183 allows
one to determine, on a strictly theoretical basis, the effective thermal conductivity for the K-y system.
In our study of heat transfer in the fluid phase of packed bed catalytic reactors, we have illustrated
414 Multiphase Reactor Operations

how the method of volume averaging can be used to arrive at the traditional model given by
Equations 245-248. In doing so, we have used an overly-simplistic representation for Tp (compare
Equation 236 with Equation 288) and an expression for the interfacial heat flux (see Equation 237)
that perhaps does not adequately represent the physics of the process under consideration. For
example, use of Equation 288a leads to

~ t.a !!p.' kp'yTp dA = ap.hp.«T.)" - <Tpl)

+ k{~ t.a !!p.' YEp dA - YEp]' Y<Tpl


+ k{~ fA.a !!p.' YG. dA]' Y<T.>" (307)

rather than the traditional representation given by Equation 237. The advantage of the traditional
model rests largely with the ability to calculate non-isothermal effectiveness factors by means of
the simultaneous solution of Equation 248 and the associated diffusion-reaction transport equation
[1]. In the method of volume averaging the coupling between heat and mass transport takes place
at two levels:

1. Between the volume averaged transport equations.


2. Between the transport equations for the spatial deviations that must be determined to complete
the closure.

In this development we have ignored the coupling at the spatial deviation level. This is apparent
in the simplification that takes place between Equations 177 and 183 with the latter being the
accepted form of the "point equation" in the porous catalyst. In the second level of averaging, we
have ignored the coupling in the closure problem by the imposition of the restriction given by
Equation 283. Under these circumstances, one has lost the opportunity of calculating a non-
isothermal effectiveness factor, thus the closure problem obviously needs to be extended. In the
traditional model the difficulty is removed by operating in a mixed mode: a volume averaged
transport equation is used for the fluid phase while a "point equation" is used for the solid phase.
The validity of this approach needs to be carefully considered and further work is obviously in order.

Acknowledgement

This work was supported by NSF Grant CBT 8611610.

NOTATION

A. w area of the I1-W interface contained CA surface concentration of species A,


within an averaging volume, m 2 kgmoles/m2
A. e macroscopic area of entrances and <cA)W intrinsic phase average concentration
exits for the l1-phase, m 2 of species A in the w-phase,
a. w A.w/V or A.w/v", interfacial area per kgmoles/m 3
unit volume, m - 1 cp constant pressure heat capacity,
12A species A body force per unit mass, kcal/kgK
N/kg Cp mass fraction weighted constant pres-
12 total body force per unit mass, N/kg sure heat capacity, kcal/kgK
cA concentration of species A, eA species A internal energy per unit
kgmoles/m 3 mass, kcal/kg
Heat Transfer in Catalytic Packed Bed Reactors 415

internal energy of species A generated species A heat flux vector, kcal/m K


or consumed by chemical reaction, species A radiant energy flux vector,
kcal/kg kcal/m K
~ector field that maps Y(T,)' onto total heat flux vector, kcal/m K
T"m total radiant energy flux vector,
~ector field that maps Y(Tp)P onto
kcal/m K
T p, m
total heat flux vector for the w-phase,
~ector field that maps Y(T.)· onto
kcal/m K
T., m rate of thermal energy exchange
~ector field that maps Y(TKY onto
between species B and species A,
T., m
kcal/m 3s
~ector field that maps Y(Tpl onto
mass rate of production of species A
T p, m
per unit volume, kg/m3s
G. ~ector field that maps Y<T.)· onto
molar rate of production of species A
T., m
per unit volume, kg-mole/m 3 s
species A partial mass enthalpy,
molar rate of production of species A
kcal/kg
per unit area, kg-mole/m 2 s
h total enthalpy per unit mass, kcal/kg;
R overall homogeneous reaction rate,
film heat transfer coefficient in the
kg-mole/m 3s
traditional model, kcal/m'K
u-phase film heat transfer coefficient
R overall heterogeneous reaction rate,
kg-mole/m's
for the volume-averaged thermal
overall heterogeneous reaction rate in
energy equations (hap # h p.),
the Jl-subphase of the )'-phase,
kcal/m'K
kg-mole/m 2 s
p-phase film heat transfer coefficient
Re Reynolds number
for the volume-averaged thermal en-
!: position vector, m
ergy equations (h p• # h. p). kcal/m'K
entropy per unit mass, kcal/kg K
species A partial molar enthalpy, time, s
kcal/kg-mole temperature in the w-phase, K
adsorbed species A partial molar en- intrinsic phase average temperature
thalpy, kcal/kg-mole
in the w-phase, K
unit tensor T", - (T",)"', local spatial deviation
thermal conductivity of the I]-phase,
temperature, K
kcal/m K spatial average temperature, K
effective thermal conductivity tensor
(T",)W - (T), large-scale spatial de-
for the w-phase, kcal/m K
thermal dispersion tensor for the p- viation temperature, K
~D species A stress vector, N/m 2
phase, kcal/m K
sum of the effective thermal conduc- species A stress tensor, N/m'
tivity tensor and the thermal disper- total stress tensor, N/m'
sion tensor, kcal/m K YA - Y, species A mass diffusion ve-
L macroscopic characteristic length locity, m/s
scale, m species A velocity, m/s
microscopic characteristic length mass average velocity, m/s
scale for the w-phase, m velocity of species A generated or
i = 1, 2, 3 lattice vectors for a spa- consumed by chemical reaction, m/s
tially periodic porous media, m mass average velocity in the p-phase,
unit normal vector pointing from the m/s
w-phase to the I]-phase intrinsic phase average velocity in the
outwardly directed unit normal vec- p-phase, m/s
tor at the entrances and exits of the YP - (Yp)P, local spatial deviation ve-
w-phase locity, m/s
partial pressure of species A, N/m2 v averaging volume for the p-u system,
total pressure, N/m ' m3
force per unit volume exerted by averaging volume for the K-), system,
species B on species A m3
416 Mulliphase Reactor Operations

vw volume of the w-phase contained y! velocity of a moving interface, m/s


within either Vor V., m 3 :i position vector relative to the cen-
VA(t) volume of a species A body, m 3 troid of the averaging volume, m.

Greek Letters

ex k/pc p , thermal diffusivity, m 2/s PA species A mass density, kg/m 3


ex,ff K~ff/ <p )C p , thermal diffusivity for the P total mass density, kg/m 3
porous catalyst phase, m 2 /s <p) spatial average mass density, kg/m 3
AH, molar heat of reaction, kcal/kg-mole ~A species A viscous stress tensor, N/m2
E. volume fraction of the IJ-phase given , total viscous stress tensor, N/m2
by V./V or V.IV. <P." thermal source in the porous catalyst
K k./k y , ratio of conductivities phase, kcal/m 3 s
vA stoichiometric coefficient for species A WA PA/P, species A mass fraction

REFERENCES

1. Whitaker, S., "Transport Processes with Heterogeneous Reaction," in Concepts and Design of
Chemical Reactors, edited by S. Whitaker and A. E. Cassano, Gordon and Breach, New York
(1986).
2. Truesdell, C., and Toupin, R., "The Classical Field Theories," in Handbuch der Physik, Vol. III,
Part 1, edited by S. Fliigge, Springer-Verlag, New York (1960).
3. Truesdell, C., Essays in the History of Mechanics, Springer-Verlag, New York (1968).
4. Aris, R., Vectors, Tensors and the Basic Equations of Fluid Mechanics, Prentice-Hall, Inc.,
Englewood Cliffs, New Jersey (1962).
5. Slattery, 1. c., Momentum, Energy and Mass Transfer in Continua, R. E. Krieger Pub. Co.,
Malabar, Florida (1981).
6. Whitaker, S., "Mass Transport in Porous Catalyst Pellets," Transport in Porous Media, 2, 269-
299 (1987).
7. Whitaker, S., Fundamental Principles of Heat Transfer, R. E. Krieger Pub. Co., Malabar, Florida
(1983).
8. Whitaker, S., Introduction to Fluid Mechanics, R. E. Krieger Pub. Co., Malabar, Florida (1981).
9. Bird, R. B., Stewart, W. E., and Lightfoot, E. N., Transport Phenomena, John Wiley and Sons,
Inc., New York (1960).
10. Glandt, E. D., "Simultaneous Conduction and Radiation in Porous and Composite Materials:
Effective Thermal Conductivity," Ind. Eng. Chern. Fundam., 22, 276-282 (1983).
11. Whitaker, S., "Radiant Energy Transport in Porous Media," Ind. Eng. Chern. Fundam., 19,
210-218 (1980).
12. Carbonell, R. G., and Whitaker, S., "Adsorption and Reaction at a Catalytic Surface: The Quasi-
Steady Condition," Chern. Eng. Sci., 39,1319-1321 (1984).
13. Whitaker, S., "Transient Diffusion, Adsorption and Reaction in Porous Catalysts: The Reaction
Controlled, Quasi-Steady Catalytic Surface," Chern. Eng. Sci., 41, 3015-3022 (1986).
14. Anderson, T. B., and Jackson, R., "A Fluid Mechanical Description of Fluidized Beds," Ind. Eng.
Chern. Fundam., 6, 527-538 (1967).
15. Marle, C. M., "Ecoulements Monophasiques en Milieu Poreux," Rev. Inst. Francais du Petrole,
22,1471-1509 (1967).
16. Slattery, J. C., "Flow ofViscoe1astic Fluids Through Porous Media," AIChE Journal, 13, 1066-
1071 (1967).
17. Whitaker, S., "Diffusion and Dispersion in Porous Media," AIChE Journal, 13,420-427 (1967).
18. Nozad, 1., Carbonell, R. G., and Whitaker, S., "Heat Conduction in Multiphase Systems I:
Theory and Experiment for Two-Phase Systems," Chern. Eng. Sci., 40, 843-855 (1985).
19. Whitaker, S., "A Simple Geometrical Derivation of the Spatial Averaging Theorem," Chern.
Engr. Ed., Winter, 18-21 and 50-52 (1985).

Heat Transfer in Catalytic Packed Bed Reactors 417

20. Gray, W. G., "A Derivation of the Equations for Multiphase Transport," Chem. Eng. Sci., 30,
229-233 (1975).
21. Carbonell, R. G., and Whitaker, S., "Heat and Mass Transfer in Porous Media," pages 123-198
in Fundamentals of Transport in Porous Media, edited by 1. Bear and M. Y. Corapcioglu,
Martinus Nijhoff Publishers, Dordrecht, The Netherlands (1984).
22. Whitaker, S., "Diffusion and Reaction in a Micropore-Macropore Model of a Porous Medium,"
Latin Amer. J. Chem. Engr. Appl. Chem., 13, 143-183 (1983).
23. Whitaker, S., "Simultaneous Heat, Mass and Momentum Transfer in Porous Media: A Theory
of Drying," pages 119-203 in Advances in Heat Transfer, Vol. 13, Academic Press, New York
(1977).
24. Whitaker, S., "Heat and Mass Transfer in Granular Porous Media," pages 23-61 in Advances
in Drying, Vol. 1. Hemisphere Publishing Corp., New York (1980).
25. Whitaker, S., "Local Thermal Equilibrium: An Application to Packed Bed Catalytic Reactor
Design," Chem. Eng. Sci., 41, 2029-2039 (1986).
26. Crapiste, G. H., Rotstein, E., and Whitaker, S., "A General Closure Scheme for the Method
of Volume Averaging," Chem. Eng. Sci., 41, 227-235 (1986).
27. Ryan, D., Carbonell, R. G., and Whitaker, S., "Effective Diffusivities for Catalyst Pellets under
Reactive Conditions," Chem. Eng. Sci., 35, 10-16 (1980).
28. Ryan, D., Carbonell, R. G., and Whitaker, S., "A Theory of Diffusion and Reaction in Porous
Media," AIChE Symposium Series, #202, Vol. 71, 46-62 (1981).
29. Ochoa,1. A., Stroeve, P., and Whitaker, S., "Diffusion and Reaction in Cellular Media," Chem.
Eng. Sci., 41, 2999-3013 (1986).
30. Whitaker, S., "Flow in Porous Media I: A Theoretical Derivation of Darcy's Law," Transport
in Porous Media, 1, 3-25 (1986).
31. Whitaker, S., "Flow in Porous Media II: The Governing Equations for Immiscible, Two-Phase
Flow," Transport in Porous Media, 1, 105-125 (1986).
32. Brenner, H., "Dispersion Resulting From Flow Through Spatially Periodic Porous Media,"
Trans. Roy. Soc. (London), 297, 81-133 (1980).
33. Whitaker, S., "The Role ofthe Volume-Averaged Temperature in the Analysis of Nonisothermal,
Multiphase Transport Phenomena," Chern. Engr. Comm.
34. Wu, L. C., Unpublished Experimental Results, Department of Chemical Engineering," University
of California at Davis (1984).
35. Nozad, I., Ph.D. thesis, Department of Chemical Engineering, University of California at Davis,
(1983).
36. Batchelor, G. K., and O'Brien, R. W., Proc. Roy. Soc. (London), A355, 313-328 (1977).
37. Swift, D. L., "The Thermal Conductivity of Spherical Metal Powders Including the Effect of an
Oxide Coating," Int. J. Heat Mass Transfer, 9, 1061-1073 (1966).
38. Shonnard, D. R., "Experimental Determination of Unit Cell Effective Thermal Conductivities
for Point Contact," M.S. Thesis, Department of Chemical Engineering, University of California
at Davis (1985).
39. Hadley, G. R., "Thermal Conductivity of Packed Metal Powders," Int. J. Heat Mass Transfer,
29,909-920 (1986).
40. Oppio, 1., personal communication (1984).
41. Dullien, F. A. L., Porous Media: Fluid Transport and Pore Structure, Academic Press, New York
(1979).
42. Prakowras, A. G., Vachow, R. I., Crane, R. A., and Khader, M. S., "Thermal Conductivity of
Heterogeneous Mixtures," Int. J. Heat Mass Transfer, 21,1157-1166 (1978).
43. Gavalas, G. R., "A Random Capillary Model with Application to Char Gasification at Chemi-
cally Controlled Rates," AIChE J. 26, 577-598 (1980).
44. Whitaker, S., "Heat Conduction in Porous Media with Homogeneous and Heterogeneous
Thermal Sources," Proceedings Euromech, 194, 39-44, Nancy, France (1985).
45. Zanotti, F., and Carbonell, R. G., "Development of Transport Equations for Multiphase Sys-
tems III: Application to Heat Transfer in Packed Beds," Chem. Eng. Sci., 39, 299-311 (1984).
46. Eidsath, A., Carbonell, R. G., Whitaker, S., and Herrmann, L. R., "Dispersion in Pulsed Systems
III: Comparison between Theory and Experiments for Packed Beds," Chem. Eng. Sci., 38,
1803-1816 (1983).

View publication stats

Вам также может понравиться