Vous êtes sur la page 1sur 8

Chemical Engineering

Science,1961, Vol. 18, pp. 227 to 2.44. PelgamonPressLtd., London.Printed in Great Britain

Intraparticle diffusion and conduction in porous catalysts-II

Complex reactions


University of Minnesota, Minneapolis 14, Minnesota, U.S.A.

(Receiued 10 Au@& 1960)

Abstract-The problem of intraparticle diffusion and conduction is considered for complex

reactions taking place in porous catalytic spheres. It is assumed that Knudsen diffusion prevails
and that heat generated in the reaction is conducted through the porous structure to the catalyst
surface. Equations are written for an arbitrary number of reactions and criteria are developed to
determine the number of conservation equations for complex systems. A particular numerical
example is worked both approximately and exactly by the same technique as used in the previous
paper; the term exactly meaning that partial pressure and temperature proilles may be calculated
numerically as accurately as desired. The approximate method should be satisfactory for almost
all industrial work.
R&mm&-Le probleme de la diffusion et de la conduction a l’interieur d’une particule eat consider6
pour des reactions complexes se prod&ant avec des catalyseurs poreux de forme spherique. On
suppose que la diffusion de Knudsen pr&aut et que la chaleur degagee par la reaction eat conduite
jusqu’a la surface du catalyseur a travers la structure poreuse de celui-ci. Des equations sont
posees pour un nombre arbitraire de reactions et des crit&ia developpea pour determiner le nombre
d’equations de conservation pour des syst&mescomplexes. Un exemple nmnerique particulier eat
CtudiCIi la fois approximativement et exactement par la technique utilisee dans les precedents
articles. Le terme exact montre que la pression partielle et le proiil de temperature peuvent @tre
calcules numeriquement de fapon aussi precise que l’on veut. La m&bode approximative devrait
etre satisfaisante pour la plupart des cas industriels.
Zusammenfassung-Das Problem der intrapartikuliiren Diffusion und Leitung wird filr
komplexe Reaktoinen betrachtet, die in porosen Katalysatorkugeln stattfinden. Dabei wird
angenommen, dass Knudsen-Diffusion vorherrscht und dass die Reaktionswiirme durch die
poriise Struktur zur Katalysatoroberfliiche geleitet wird. Fur eine willkurliche Zabl von Reak-
tionen werden Gleichungen angegeben und Kriterien entwickelt, um die Zahl der Bilanzgleichungen
fiir komplexe Systeme zu bestimmen. Ein nmnerisches Beispiel iat sowohl angeniihert als such
exakt mit der gleichen Methode durchgerechnet, die in der vorhergehenden Arbeit benutzt wurde ;
mit Hilfe des exakten Ausdrucks k&men Partialdruck und Temperaturproill numerisch heliebig
genau berechnet werden. Die Niiherungsmethode sollte fur die meisten industriellen Arbeiten

arbitrary system of chemical reactions will be

IN A previous paper [l] the authors considered considered [S].
the problem of a chemical reaction taking place The chemical species wiII be denoted by Mi
inside a single porous catalyst sphere. It was and the stoichiometric coefficient of the ith species
shown how the temperature and partial pressure in the jth reaction will be aij. Suppose there are
profile could be calculated and how the effective- m reactions
ness factor could be estimated. An approximate
method was developed and it was shown that the iaijMi=O; j=l,2,...,m
approximation made by a suitable linearization
gave results of satisfactory accuracy. It is the where in a particular aii is positive for
purpose of this paper to generalize the methods a product and negative The heat
for a reactant.
of the previous paper to complex reactions. An of reaction for the jth reaction is ~j and fijis
* Present address : The Ohio Oil Company, Littleton, Colorado, U.S.A.

_ -
the rate of formation of the ith species in the jth
reaction in moles per unit of catalytic surface
D dp! ji,
area per unit of time. The quantity fu may be
l dP

a function of all n partial pressures, the temper- dp,

ature and the total pressure. We suppose that 2dP
the set of chemical species {2Mi} contains all of
the species in the reaction system.
It is proposed to show how the partial pressure
and temperature profile may be computed both !!&
exactly and approximately. The same methods ndp- - _
may be used if the conductivity of the particles =
is great enough so that temperature gradients and A is the matrix
may be neglected.
a11 a12 * * - %a

DERIVATIONOF EQUATIONS* A = a,, a22 . . . a,

It is assumed that Knudsen diffusion obtains .

within the particles so that the diffusion coefficient

is given by .

an1 a,, . . . 4,

Consider the matrix AT, the transpose of 2

for the ith chemical species. Ei is independent
of the temperature and depends on the pore
properties as well as the molecular weight.
Assuming that all flow is by molecular diffusion
a mass balance on a single particle is given by

= -p”p,,S8 ;fti;
i = 1, 2, 3, . . . n
From the stoichiometry of the reaction it follows and suppose 2 has rank s. For the sake of con-
that venience, suppose the nonzero determinant of
order s is in the upper left hand corner. If it
.& _ fkj_ J;
for all j is not, the species may be renumbered to make
aa. %
it so. Now if the columns of bT are considered
so that
as vectors it follows that each of the last n-s
d columns may be expressed as a linear combination
dT of the ilrst s columns. Therefore, there is a matrix
i=l,2,3,...n 8’ of the form

The question now arises as to the number of

i 0 o....o ‘yI+l,l.. -
- * %a,1
these equations which are independent. This set 0 1 o....o ys+1,2 Yn, 2
of equations may be written in veetor form as
ET = 10 0 l....O *
;(Prq = -p”p&JT
where < and 1 are vectors,

* Notation will be the same as in the first paper. 0 0 .... ...1 y8+],a Yn, 6

Wraparticle ditfusionand conductionin porous catalysts-11
and BT may be obtained from W* by operations Equation (2), of course, may be written in the
on rows above, as by the Simplex Method of form
linear programming. In short, there is a matrix +BG (4)
operator P such that where

aAr=BT 7jlT= f+D,!!?.
. . D !f!?
* dP 1
The essential result of this development is
that if the rank of the stoichiometric matrix is s
where z* is the matrix operator which operates
there are s independent mass conservation
on the columns of 2. Note that the last n-s
equations, the remaining partial pressures having
columns of BT represent the coefficients in the
been eliminated by equation (3).
expansion of the last n-s columns of AT in terms
Now the heat balance must be considered.
of the first s columns of zT, the first s columns
The same assumptions are made here as in the
being considered as a basis in a vector space.
previous paper so that the energy conservation
If one now writes
equation is formally the same

it is clear that this is a column vector with d =o (5)

elements E Uij~j and hence each of the last
j=l Here again pi is the partial molar enthalpy and
n-s elements in the vector is a linear combination
the heat of transport (- 4 RT) will be neglected.
of the first s elements as
One integration can be made directly to give
m 8 ?a
Z as+p,j fj = .Z Y8+p, p x ati lj9 P= l ton-S (~y)k~~+,i’D~l&~=o
j=l g=1 j=l

or from equation (1)

Define now a vector 2 in which Ri is the ith
.j element, then
P (1 -y)k$+i%‘q=O (6)

This may be integrated once to give A vector AH of the heats of reaction may be de-
fined where the ith element is the heat of reaction
D D ‘P, i=s+lton (2) of the ith reaction. Hence
rti qJ-’
where use has been made of the fact that the
gradients are zero at the centre of the particle. Also the reaction system itself may be written
Since Ei = Di 1/T the latter equation may be
integrated to give ATa=a

where Ti? is the vector of chemical species. The

Ei (PC -P&I = 2
YQ Eq (Pq - Pqs),
operator 2 operating as
i = s + 1 ton (3)

Therefore inside the particle there are s independent

partial pressures since the remaining (n-s) may will give another set of chemical reactions where
be eliminated from the rate expressions, and hence the heat.of reaction vector will be
also there are s equations of the type given by
equation (1). ~~*jj=~H’=B*&j (7)
Chem. En&!. Sci. Vol. 18, No. 4. June, 1061.

where a typical element of &ij’ is AH1 = - 80,000 Cal/g mole

AH, = 60,000 Cal/g mole
AHH,’= R, + Ya+l,T &+I + * - * + ytw Q, ;
mA = 25 g/g mole
m, = 36 g/g mole
and also where the elements AHH,’ are a linear m, = 61 g/g mole
combination of the original AH,. From equations mD = 72 g/g mole
(4), (6) and (7) one may write mE = 50 g/g mole
T, = 550°K
(1- y)/CSg+HrB$=O
p,, = 1.50 atm
p, = 1.50 atm
The transpose may be taken to give
pCS = .OlO atm
pDS = pxS = 0.0 atm
R = 0.5 cm
7 = 2oojl
(1 - y) Ic,g + ijIT m’ = 0 y = 0.40
k8 = 8 x 1O-4 cal /set cm “K
If m’ is assumed to be a constant vector, then DA = 3.892 (lo-‘) g mole/atm cm set
DB = 3.244 (lo-‘) g mole/atm cm set
5 (1 - y) k8 ( Ta’2 - T;‘“) =
D, = 2.492 (lo-‘) g mole/atm cm set
D, = 2.293 (lo-‘) g mole/atm cm set
DE = 2.752 (lo-‘) g mole/atm cm set
or if Di is taken a as constant
fAl = 3.36 (1O-s) exp

and hence the temperature or one of the indepen-

dentpi’s can be eliminated from the rate expression.
Hence of the original n + 1 conservation In this problem the rank of the stiochiometric
equations there are only s conservation equations matrix is two so that there are two conservation
which need integrating. Which equations in a equations. The detailed manipulations will not
particular case will be integrated depends on be given here but on the assumption that AH, IE,,
what is desired. In the example to follow in and the diffusivities are constant the following
which AHj and Dj are assumed temperature relations may be derived.
independent, a partial pressure and the temper-
ature will be used as independent variables.

DC PA -- 2Dr! PD
We will consider a reaction system

1. A+B+C

2. 2C+D+E

in which the reactions are irreversible and second

order in the forward direction. The reaction is
carried out in a porous pellet with the following I DA AH,
DD m,PA - ‘;; ;-’ T
information available.

Intrapmticle diffusion and conduction in porous catalysts--II

The two conservation equations may be written where

-- I
2 dPA +m(T,pA)
-- =O
d@ r d,.

+ WPAs
p2 (6 - a) sinh TV
y(r) = 4x3- ~%o
?!+%- fAl(M,) +$(,l)] A2 (B - a)
p (T. PA)
ks (1 - Y) + p2 p)As - boo) - fi ( AaT, - aoo) sinh h r
___ (9)
fi2 PP sg h ( 3 - a) sinh h T
w (T, PA) = ~--fa~
A with

The solution of this set of equations follows in h = -$ [- U + d(U” - 4v)]“a

principle the solution for a single equation except
that it is more complicated. Our purpose here /L.= ;; [- U - l/(u2 - 4v)]l’a
is to sketch the solution not to present the
detailed technique. A particular example will be
u= al, + ho
solved rather than a general one but it will be
obvious what changes need be made for other 71= ctl,b,o - bna,o
cases. It will be recalled that a certain linear
A2+ a,, - b,,
approximation was made and that will be done
here. Let
a=X8 + b,, - a,,

a00 + alOpA + al7 /3= tLe

~___ + a11 - bll

and CL” + ho - alo

boa + hopA + b,,T With these functions the profiles may be plotted
be linear approximations to q ( T,pa) and w ( T,pA)
and this will be done.
respectively. Then the solution of the equations As a generalization of the previous paper the
two differential equations are now written

;; + $g + a,OPA + a,,T = Q (d - a,

+ boo+ hoPA + 4, T = 0 d2PA

dr2 + f ‘2 + b,oP, + b,,T = W(r) - boo
_ = 0 = CP? atr=O
dr dr QW = %o +a,@~ +%T -g(T,I)A)
T = T, W (r) = boo + b,oPA + bnT - w (TV PX)
If it is assumed for the moment that Q (r) and
P =PA, >
W(r) are known functions of r, these two equations
should give an approximation to the temperature may be changed to integral equations by the
and partial pressure profiles, say the first approx- method of variation of parameters to give
imation. This solution may be found by direct
methods and it is T(r) = X (r) - Y (r) +
T, (r) = X (r) - Y (r)
+ [a Q (4 - w(e)] G, (r, 4 ~JL- (10)
PA1 tr) = B x (r) - a y (r)

Chem. Engng. Sci. Vol. 18, No. 4. June. 1961.


Q (4 - We)] 6 0.94 de
w (T, pA) = - 8.632 (10’) exp

- ET?



PA b-1= I= (r) - EY (T) + where use has been made of the formula
1 pP S, = 2X/7. In order to study the nature of

[aQ(e)- W(E)]Gp(r,Odc-

the functions q and w calculations were made and
three-dimensional models made which are shown
in Figs. 1 and 2. These models are interesting

s - a
[ PQ (4 - W(c)]GA(T,4 de
but they must be analysed further to find the
region in the T,pA plane of interest. Since
q (T,, pAa) is positive it appears that there will be
a net exothermic effect so that T > 550” K in
lsinhp(1 --)sinhpe, 0 <E <r the particle. Since p, and pc must not be negative
r(/3 -aa)t~sinhp ’ ’ PA 2 0’2493
G,,(r, c) =
lsinh~rsinh~(1 -e), T > 609.2 - 40.55
I r (/3 - a) p sinh p
In addition PA < 1.50. A further boundary line
and where GA(r, E) is the same except h is sub- may be obtained from the condition that if the
stituted for CL.Equations (10) and (11) are solved first reaction goes to completion and the second
by successive substitutions to obtain a sequence one is insignificant, then p, < l-50 and therefore
of profiles which it is hoped will converge in the
T < 632.6 - 40.55 p,
limit to the solution of the original set of differen-
tial equations. That is, the profiles T, (r) and The resulting region of interest is shown in Fig.
pAl (r) are substituted into Q (c) and W (e) to 3. With this information the linear approxima-
obtain new profiles T,(r) and pAa (e) and this tions can be found and in this problem these are
process is iterated to give a sequence of profiles
( T, (‘% PA” k)}.
From these equations the effectiveness factor
may be computed from
E F _ - 3 Da (dpi/dr),,l moles of J&
. . i
Ra pp Sg ; (f -), unit Of time c
j=l cr.2
h and,p could be complex numbers but this alters
the method in detail rather than principle.
With the data presented at the beginning of
this Section it may be shown that
q (T, pA) = 5.60 (lOa) exp

PA (1.1998 PA - 0.2997) -
540 560 560.
Temperature. T. OK

Fm. 3. Graph showingthe region of interest in the

(2.604 PA + 0.06421T - 39.12)a T,pA plane for the sample problem.

; ._
Intraparticle diffusion and conduction in porous catalyetgr1

U, -/- alspA + u,,T = 8463 + 638.7~~ - 14.37T

PA (r) = 0.2500 + 0.0331 ‘lnh
b, + b,, pA + b,,T = 4676 - 18.70 PA
-a l+‘(~)G~(r,e)&
u = - 33.07 s
v = 263.8 0

h= 4.325 The calculations were carried out as suggested

/L = 3.791 earlier and the partial pressure and temperature
a = - 0.006784 profiles are shown in Figs. 4 and 5. What is called
/? = o*ooo (bu = 0) the final solutions on these curves is a result of
and several successive substitutions in the integral
sinh /LT sinh hr
T(r) = 600.0 + 6.064 - - 4.873 - +
r r

+ [aK (4 - WI] G,, b, 4 A +


+ W(E) G,, (r, c) dc


FIG. 5. Temperature profiles within particle showing

effect of linear approximation.
- Final Solution
- - - - pc First approximation.

equations. The essential point here, however, is

not the number of calculations made but rather
the excellence of the results with the linear
\ approximation, equations (8) and (9). Effective-
0.2 \
ness factors are as follows :
p1, Flnol Solution \
------_ PC. flnol Solution
Fi?UZl Approximate
OO 0.2 0.4 0.6 0.8 1.0
solution solution
Dimensionless Radius, I -
FIG. 4. Particle pressure pro5lea in particle for 0.592 0.533
reactant A and product C (Exact solution). 0.412 0.364
- pA Final solution 603 602
- - - - pc Final solution

Chem.Engng.Sci. Vol. la,No. 4. June,19131.


These calculations were made on a desk calcu- the equations and a technique was described
lator with the radius of the particle divided into which could be used to obtain temperature and
twenty equal parts. The integrals were then partial pressure profiles as rigorously as desired.
computed using a simple trapezoidal rule. Some The approximation method may break down in
attempts were made to determine the effect of some cases but it is recalled that for simple
interval size but no conclusions were drawn. reactions it worked remarkably well.
Attempts were also made to solve the problem on It must be stated that the methods used in
an analogue computer although these results this paper for numerical computations while
were only partially successful. The equations satisfactory for two and possibly three reactions
tend to be unstable when proceeding from the will probably not be suitable for larger numbers of
surface inwards. Some success was achieved reactions. Analogue computer solutions for three
integrating the equations on an analogue computer or more reactions may be less inconvenient but
starting at the particle centre and moving out- this remains to be shown. It has been suggested
wards. However, two values must be guessed, that grid methods, that is the transformation to
the partial pressure and temperature at the centre finite difference approximations for the deriva-
of the particle, and this can be inconveniently tions, be tried but there are considerable diffi-
tedious. culties with this because of the nonlinearity of
the finite difference equations and lack of in-
SUMMARYAND CRITIQUE formation on convergence of the resulting iterative
process. It would seem that some simpler
This paper is a generalization to complex
approximation scheme is needed.
systems of reactions of the previous paper on
Acknowledgements-This work was supported by grants-
simple reactions. It was shown that the rank of
in-aid from the Procter & Gamble Company, the Eastman
the stoichiometric matrix gives the number of Kodak Company, the Standard Oil Foundation (Indiana),
conservation equations which need integrating. the Minneapolis-Honeywell Regulator Company, and by
An approximate method was devised for solving support from the National Science Foundation.

[1] SCHILSON R. E. and AMUNDSONN. R. Chem. Engng. 815. (In press).
[2] SCHIL~ON R. E., Ph.D. Thesis, University of Minnesota, December 195’7.


Vous aimerez peut-être aussi