Chemical Engineering Science, 1972, Vol. 27, pp. 227-237. Pergamon Press.

Printed in Great Britain

Changing catalyst performance by varying the distribution of

active catalyst within porous supports
F. SHADMAN-YAZDI and E. E. PETERSEN
Department of Chemical Engineering, University of California, Berkeley, California, U.S.A.

(First received 4 January 197 1; accepted 16 April 1971)

Abstract-Uniform and nonuniform distribution of catalyst on porous supports are compared with
regard to: (1) the catalyst selectivity in a series reaction A -+ B + C, (2) the catalyst deactivation
under a self-fouling reaction A -+ B + C. The nonuniform distribution considered is one in which
the catalyst concentration decreases towards the center of pellets. It is found that this distribution
gives a better selectivity to E and decreases the average rate of deactivation by C. In studying the
fouling of catalyst pellets, the accuracy and usefulness of simplified poisoning models were evaluated.

INTRODUCTION ing the steepness of the activity profile. As shown

WHEN a multistep reaction takes place on a in Fig. 1, this function can approximately re-
mixture of several catalysts, the distribution of present a variety of distributions simply by
the products often depends on the arrangement adjusting (Y. A more general profile is one in
of the catalyst functions. Gunn [ I] and Thomas which activity is not necessarily zero at the
and Wood[2] determined the optimum distribu- center. However, this distribution can also be
tion of a physical mixture of two catalysts x and y approximated by Eq. (1) if values of 1yless than
in a reactor which allows maximum production unity are used.
of a desirable product in a multistep reaction.
Rutkin and Petersen [3] found that the selectivity SELECTIVITY IN A SERIES REACTION
of a dual-function catalyst x-y for intermediate A series reaction A + B -+ C is considered
products can be altered by using pellets which in which A is the reactant and B is the desirable
consist of two or three layers of x and y alter- product. Both steps are catalyzed by the
nately. supported catalyst and are assumed to be
The purpose of this study is to evaluate the
changes in the performance of a supported
catalyst with the changes in the distribution of
a catalytically active material on an inert porous
support.
The distribution obtained by impregnation of
a porous bed depends greatly on the nature of
the catalyst, the nature of the support, the
method of impregnation and the composition of
the impregnating solution. The distribution
considered in this paper is one in which the local
activity (measured as the rate constants) is of
the form
K = K”(” (1)

where 5 is a dimensionless distance from the Dimensionless distances, 4

center of the pellet and (Yis a parameter indicat- Fig. 1. Activity distribution in a catalyst pellet, K/K0 = 5”.

227
 F. SHADMAN-YAZDI and E. E. PETERSEN

irreversible and first order. In this part the select- conditions can be written as follows:
ivity will be determined when the reaction occurs
in a porous spherical pellet.
According to Eq. (l), the local rate constants (10)
are
--
’ d 2iki!! - hs2tn4 + hA2{UJI= 0 (11)
KA = KA” 6 a
t2 d5 (’ d5 )
forA + B (2)
0
at<=0 (12)
Kg=Kgo ; u forB + C. (3)
0
J,=1,4=0 4= 1. (13)
The steady-state material balance equations
The solution to Eq. (10) is
for components A and B within the spherical
pellet are
-1’2ZP(2PhA 5 1’2p1
+=6 (14)
Zp(2Ph.d

where ZP is the modified Bessel function of the

firstkindoforderPandP= l/(a+2).
Equation (14) can be written in the form of
a power series of 4:
The boundary conditions are
+ = .I.Io
amPP (1%
de..,_--dCs_. atr=O (6) where
dr dr
(PhA)2m+P
C* = CAO,CB = 0 atr= R. (7) (16)
am = m!(m+P)!Zp(2PhA).

For the boundary conditions at r = R, it is

After substituting Eq. (15) for 9, Eq. (11) can
assumed that the concentration of products
be solved for 4. The solution is
outside the pellet is much smaller than that
inside the pellet. For reactors with high yields,
this assumption cannot be made, However, the $ = 2 C,{m’P (17)
m=o
final results of this analysis concerning the
where
comparison of different types of distribution will
remain qualitatively valid. - hA2a,_, + hB2c,_,
(18)
Introducing the following dimensionless cm= m/P(m/P+l) -
variables
The rates of consumption of A and production
of B are given by

and defining two Thiele parameters (19)

h,=R&$ h,,R& (9) (20)

the differential equations and the boundary The selectivity, S, is defined as

228
 Varying the distribution of active catalyst within porous supports

RI, RB DEACTIVATION BY A SELF-FOULING

S=K=RA-~B (21) REACTION
In this section the effect of the distribution of
-Ds z me, activity on the rate of deactivation of a catalyst
m=1
s= (22)
pellet will be evaluated. For this purpose,
a first-order self-fouling reaction of the type
DA $, ma,+- & 2 mc,
m=1 A + B + C is considered. B is the desirable
product and C is a degraded product which is
Figure 2 is a plot of some of the numerical strongly adsorbed on the surface and deactivates
results obtained for three types of activity the catalyst. For mathematical simplicity, it is
distribution. Diffusivities are assumed to be assumed that the reaction takes place in a flat
equal in these numerical calculations. These slab of porous support impregnated with
results show that for the same values of Thiele catalyst. As will be shown later, the final results
parameters, nonuniform distribution of activity can be applied to spherical pellets, too.
gives a better selectivity to B (S increases as In this case, the rate constant before deactiva-
(Yincreases). tion are
KA = KAo ; a forA + B (23)
0
Parameter = h,
Kg=Kgo F a forB+ C. (24)
- a=0 0
he -_-_ a=2
---_ a=6
--,_____-- ____----- It is assumed that the local activity decreases

50 -___
c ,_--- __-
___-
___---
linearly with the local concentration of poison.
Therefore, the rate constants at any time during
the reaction can be written as follows

K,(x, t) = KAo(;>a( 1 -V) (25)

K&x, t) = KBo($(
x 1- y) (26)
y 1~~__~.~~_~____~__----
where 4 is the amount of C in unit volume of the
slab and q” is the maximum limit of q corres-
ponding to complete deactivation of the catalyst.
It is assumed that C is adsorbed to the same
extent on both catalyst and support. Therefore,
q” is a constant independent of x and t.
The rate of deactivation of the catalyst is
found first by a rigorous analysis and then by
simplified poisoning models.

Method 1. Rigorous analysis

The mass-balance equations for A and B are:
0.11 123456769
’ ’ ’ ’ ’ ’ ’ ’ ’

Thiels parameter, h,
DA%- K,“($( 1 -$C* = 0 (27)

Fig. 2. Comparison of selectivities obtained from three kinds

of activity distribution in catalyst pellets.
DB$$+ K,O($( 1 -$)CA = 0. (28)

229
 F. SHADMAN-YAZDI and E. E. PETERSEN

In these two equations, no terms are included +I= I,+=0 at{= 1 (38)
to account for the conversion of B-C or the
change in q, C, and C, with respect to time. This $=%O atc=O (39)
simplification which is usually called “quasi- a(
steady-state assumption” is justified, because o=o at0 = 0. (40)
the rate of the fouling reaction B + C is much
less than the rate of the main reaction A + B. Adding Eq. (35) to Eq. (36) gives
Therefore, the neglected terms are much smaller
than the other terms in the above equations.
(41)
The conservation equation for C is

qI+c#J= 1. (42)
The boundary conditions for Eqs. (27) (28)
-and (29) are Therefore, Eq. (37) can be written as

C* = c*o,c, = 0 atx=L (30) $-5”(1-w)(l-$) =o. (43)

ac,_aC,
_-= o atx=O
a)u
“n
av
“A
(31)
Using the method described in the appendix,
q=o at c = 0. (32) Eqs. (36) and (43) are simultaneously solved to
find the rate of production of A and the extent
Introducing the following dimensionless of deactivation of catalyst at any time. The
variables average activity of a catalyst pellet is shown by
its effectiveness factor, defined as
I;=;, +=c, @=gg$
CAO’

,~,=CA~KB~~,=zt
q” ’ q”
(33) EA(f) =D*. A0 A0
(44)

and defining the Thiele parameter In dimensionless form this becomes

(34)
&(e) ( >
= G. (45)
Eqs. (27), (28) and (29) become
From Eq. (45), the effectiveness factor is found
as a function of time. The results are plotted as
(35) solid lines in Figs. 4-6.

(36) Method 2. Approximation by poisoning models

Rigorous solution to the poisoning equations
(37) becomes very complicated when the reaction is
more involved or when some of the assumptions
made here do not hold. In these cases, using
and the boundary conditions become a model for poison deposition simplifies the
230
 Varying the distribution of active catalyst within porous supports

~ . Center plane and it either exists in the feed or is produced

mostly in the outer region of the pellet.
/ D e a c t i v a t e d section The other common model is "pore-center
(core) poisoning" in which the core is deactivated
first and expands outward. This is particularly
~,, ~ . , . . . . . ......
useful for self-poisoning reactions where the
concentration of intermediate products which
produce poison is large inside the pellet.
Another useful model is "uniform poisoning"
in which the poison is assumed to be distributed
uniformly inside the pellet. This model is an
intermediate between the two extremes of
"shell" and "core" poisoning.
In this part the above simplified models will
I_ L _ be used instead of the rigorous method of the
previous part. The accuracy and usefulness of
-- x o r C = x / L these models will be evaluated in cases where
a comparison can be made with the rigorous
solution.
Fig. 3. Core-poisoning in an infinite slab. The rate of deposition of poison at any point
inside the pellet is
problem considerably and very often is the only
practical method. x '~ q
(46)
One of these models, which is most often used,
is "pore-mouth (shell) poisoning". In this model
it as assumed that there are two distinct regions Depending on o~, two cases can be distin-
in a pellet at any time: the inner core which is guished:
completely active and the outer shell which is
totally deactivated. The deactivated shell C a s e I, ot > 0
gradually progresses towards the center of The changes in KB and CB with respect to x are
the pellet. This model is suitable in cases where in opposite directions. Therefore, R c is roughly
the poison is strongly adsorbed on the surface assumed to be independent of x (uniform

0"8oI
0"9

. 07 ~ l
hA
~ . ~
a =0
Parameter = hA
/ From rigorous method
From core-poisonirK3 m o d e l

~ 0"6

"~- 0.5- ~

0"2 " ~ - - ~ - "5 ~ %" "~ " ~ " ~" "~ ~''~
0"1 - -- ~ ~

I0-I I IO I02
Dimensionless time, 0

Fig. 4. D e a c t i v a t i o n o f a c a t a l y s t p e l l e t by a s e l f - p o i s o n i n g r e a c t i o n , ~ = 0.

231
 F. SHADMAN-YAZDI and E. E. PETERSEN

a =2
Parameter = h,
- From rigorous method
--- Fr+Teuniform- Poisoning

I IO IO’
Dimensionless time. e
Fig. 5. Deactivation of a catalyst pellet by a self-poisoning. reaction, LY= 2.

I IO I02
Dimensionless time. e
Fig. 6. Deactivation of a catalyst pellet by a self-poisoning reaction, (Y= 6.

poisoning model). This means that in Eq. (35) I,!Ican be written as a power series in 5:
o is a function of 0 but not 5. The solution to this
equation is IJJ= 2 am[mlP (49)
m=o
where
(47)
(P6ym-p
am = m!(m-P)!Z_p(2P6). (50)
where
8 = h,dl-o. (48) From Eqs. (45) and (49),
232
 Varying the distribution of active catalyst within porous supports

Solving Eq. (54) for $ gives

E
A (51)
cash (hA<- h,&)
’ = cash (h/j - hA{P) (57)
Since o is independent of location, Eq. (43) can
and
be integrated with respect to 5:
E (e>
A
= tanh(hA-hb) (58)
h.4 *

The mass-balance equation for C is

Equation (53) is numerically integrated by

replacing the summation by a sufficient number which in dimensionless form is
of terms which provides the desired accuracy.
Then substituting for w in Eq. (51), values of E.,,
4-p_ ’ (1-$I d&I.
are found as a function of time. The results for
two values of CYare shown as dotted lines in
de I 5P
(60)

Figs. 5 and 6. Substituting Eq. (57) for $I and evaluating the

integral gives
Case II, a! = 0
In this case, the rate of poison deposition is
maximum at x = 0 and minimum at x = L. This &-
de - (l - {P) -+nhh,(l--5,). (61)
A
means that the central part of the pellet is subject
to relatively fast poisoning. Therefore, core Equation (61) is then solved numerically and
poisoning seems to be a good model. Actually, the solution is used in Eq. (58) to find the
after some deactivation of the inner part of the effectiveness factor as a function of time. The
pellet, the distribution of activity becomes similar results are shown as dotted lines in Fig. 4.
to that of Case I and uniform poisoning model fits
better. However, to avoid the complication of DISCUSSION AND CONCLUSION
using two different models in one case, only This study shows that the performance of
core-poisoning model is applied. a porous catalyst can be changed considerably
As shown in Fig. 3, the slab is divided into two by using a nonuniform distribution of catalyst
sections. The equation of mass balance for A in on a support.
the active section is In a series reaction A + B + C production of
C can be suppressed by distributing the catalyst
%-hAz+=O {p~{s 1 (54) in such a way that activity decreases towards the
center of the pellet. Figure 2 shows the extent
where of the improvement in selectivity resulting from
such distributions.
$=l ate= 1. (55) It should be mentioned that the nonuniform
distribution of catalyst causes some decrease in
Since there is no reaction inside the deactivated
the net rate of consumption of A and production
zone,
of B. This is of course expected, because the
inner parts of the pellets are not used completely.
(g,-,=($Lo
=O* (56) Figure 7 is a comparison between the rates of

233
 F. SHADMAN-YAZDI and E. E. PETERSEN

a=0
I.0

0.9 a=2

;//

“ii 0 2 3 4 5 6 7 6 9
Thlele porometer, h,

Fig. 7. Comparison of the rates of production of B for three kinds of

activity distribution.

production of B using different activity distribu- Fig. 1, the ratio of the net amount of catalyst
tions. The loss in the rate of production of B used for nonuniform impregnation of a spherical
is small, specially for large values of h,. How- pellet to that used for uniform impregnation is
ever, the rate of production of B per unit mass of equal to 3/5 for (Y= 2 and l/3 for (Y= 6.
catalyst is significantly higher when the catalyst Another advantage of the nonuniform distri-
is distributed nonuniformly (Fig. 8). This is due bution of activity is the decrease in the rate
to a considerable savings in the amount of of deactivation of the catalyst when there is
catalyst used for impregnation. For example, a self-fouling series reaction. This is illustrated
with the type of activity distribution shown in in Fig. 9 where the effectiveness factor is plotted
against time for three kinds of activity distribu-
tion. Even though the initial rate of reaction is
._ less for larger (Y’S,after a relatively short time
.g
::
3-6
hB=4 this difference diminishes, and finally the rate
3.4
corresponding to a larger (Ybecomes higher than
g 3.2

z-. 3.0
that corresponding to a smaller CL
As was mentioned before, the results obtained
for the deactivation of slabs of catalyst can be
applied to spherical pellets, too. This is because
the difference between the effectiveness factors
of a spherical pellet and an infinite slab becomes
unimportant for large values of Thiele parameter.
It can be shown that as hA becomes very large,
Eqs. (35)-(37) become identical to corresponding
equations in spherical geometry. This is shown
for the pellets with uniform activity distribution
0 I2 3 4 5 6 7 6 9 [4]. However, for the type of distribution
Thiele parameter , h4 considered here, this argument is still valid and
Fig. 8. Comparison of the rates of production of B per unit even more accurate. The reason is that in these
mass of catalyst for three kinds of activity distribution. distributions most of the catalyst is in the outer

234
 Varying the distribution of active catalyst within porous supports

10.’ I IO 102

Dimensionless time, 8
Fig. 9. Comparison of deactivation of catalyst pellets having three kinds
of activity distribution.

shell of the pellets; this obviously reduces the impregnation has occurred. Actually, there will
effect of the central core of the pellets and be some decrease in the cost of impregnation
increases the similarity between the slab and because of a considerable savings in the amount
the spherical models. of catalyst used.
As h, and h, decrease, the effect of activity Finally, the usefulness and accuracy of the
distribution on selectivity and poisoning becomes simplified poisoning models were evaluated in
less pronounced. This effect diminishes com- this analysis. As shown in Figs. 4-6, the results
pletely at the limit where both steps in the predicted by these models are close to the
reaction are kinetic controlled (hA = hB = 0). results of the rigorous method.
The simple reaction considered in this analysis
has some industrial significance. A good example
of this degradation reaction is the dehydrogena- NOTATION
tion of hydrocarbons. There, the desirable A initial reactant
intermediate product may react further to form coefficient of the m’th term in a series
a,
degraded products which sometimes deactivate solution
the catalyst surface by coke formation. B intermediate product
Nonuniform distribution of activity does not C undesirable product
have much effect on the selectivity in a parallel concentration of component i inside the
ci
B catalyst pellet
7
reaction A because it changes the rate of concentration of component i outside the
L ’ catalyst pellet
c coefficient of the m’th term in a series
production of B and C almost equally. solution
Obtaining a nonuniform distribution does not Di effective ditfusivity of component i
necessarily require a more difficult and costly EA effectiveness factor defined by Eq. (44)
method. Distributions similar to the one studied hi Thiele parameter, Lm or Rm
here can be obtained by withdrawing the catalyst 1, modified Bessel function of the first kind
pellets from impregnating solution before total and of order n

235
 F. SHADMAN-YAZDI and E. E. PETERSEN

K volumetric rate constant; KA for A + B and xp distance from center plane to the boundary
KsforB + C of deactivated zone, shown in Fig. 3
KO maximum value of rate constant
L (thickness of slab)/2, shown in Fig. 3
112 integer parameter in series solutions Greek symbols
1 a parameter showing the distribution of
P
a+2 activit in Eq. (1)
4 amount of poison (C) per unit volume of 6 h, ti l-w
porous pellet
5 xorL
cl0 maximum q for complete deactivation L R
r radial distance from the center of a spherical 32
5P
pellet L
R radius of a spherical pellet 8 CAOKBO~
Ri rate of reaction of component i per unit 4O
volume of porous pellets &lCS
Ri rate of reaction of i per unit mass of catalyst +’ DACAo
S selectivity, RslRc
t time J,$
X distance from center plane to any point in
the slab 03

REFERENCES
[l] GUNN D. J., Chem. Engng Sci. 1967 22 963.
[2] THOMAS W. J. and WOOD R. H., Chem. Engng Sci. 1967 22 1607.
[3] RUTKIN D. R. and PETERSEN E. E., To be published.
[4] PETERSEN E. E., Chemical ReactionAnalysis. Prentice-Hall 1965.

APPENDIX where
Equations (35) and (43) can be written as
mi,j+l = wi,j+ (Ae)(l-&,~)g(5~.j, W(J). (66)
T=$f(l.o) (62) For numerical calculations Eqs. (64) and (65) with corres-
ponding boundary conditions (Eqs. (38)-(40)) are solved
$ = (1 -+)s(5* w). (63) simultaneously.
Equation (64) with given “initial slope” and “final value”
is a boundary value problem. One typical solution approach
A closed-end integration formula corresponding to Eq. is to assume an initial value for JI, solve the equation as an
(62) is initial value problem, correct the initial assumption and
repeat this trial and error until the final value calculated is
Jli+l.j_2$i,j+JIi+L,J= ~[~,+,.,f(I;i+l.j,
Wi+l.i) sufficiently accurate. In the present problem, however,
there is a simpler method which eliminates the need for trial
+ l”Jli,j(Si,j, 0i.j) +k-1.lf(<i-l.j~ Wi-l,j)l (a)
and error. Since Eq. (62) is homogeneous, its general solution
where i and j correspond to distance and time increments has the form C& where C is independent of 1 and $,, is any
respectively. particular solution. Suppose the particular solution obtained
According to Heun’s first method, a predictor-corrector in the first trial is JI1 &d the cokesponding final value is I/+~.
formula is constructed for Eq. (63): The fmal value specified in the boundary conditions is unitv.
There, the desirable particular solutidn is obtained if all
the values corresponding to the solution JI1 are multiplied by
~i,j+*-~i,j=~l(l-~r.j+l)g(5i.j+1~~,,j+,) the ratio I/I/I,,. This scaling does not change the initial slope
+ (1 -h.j)g(li,~~~i,j)l (6-5) which is zero.

RCsumk-Les auteurs cornparent la distribution uniforme et non-uniforme du catalyseur sur des sup-
ports poreux en consid&ant:
1 -la si5lectivitk du catalyseur dans une r&action en s&e A -+ B -+ C,
2 -la d&activation du catalyseur dans une r&action d’auto contamination de celui-ci A+ B+ C.

236
 Varying the distribution of active catalyst within porous supports

Les auteurs considbrent une distribution non-uniforme dans laquelle la concentration de catalyseur
diminue en allant vers le centre des grains. 11s remarquent que cette distribution donne une meilleure
selectivite pour la formation de B et qu’elle r6duit la vitesse moyenne de d&activation du catalyseur
causte par C. La precision et l’utilitt des modeles simplifies ont 6tC evahrees pour l’btude de la
d&activation des grains de catalyseur.

Zusammenfassung- Die gleichformige und ungleichformige Verteilung von Katalysatoren auf poriisen
Tragern werden verglichen in Bezug auf: (1) die Selektivitft des Katalysators in eier Folgereaktion
A+ B+ C, (2) die Deaktivierung des Katalysators in einer selbstvergiftenden Reaktion A+ B+ C.
Die betrachtet ungleichfiirmige Verteilung ist eine bei welcher die Katalysatorkonzentration gegen die
Mitte der Pille hin abnimmt. Es wird festgestellt, dass diese Verteilungeine bessere Selektivitat gegen
B ergibt und die durchschnittliche Deaktivierungsrate durch C vermindert. Bei der Untersuchung der
Vergiftung von Katalysatorpillen werden die Genauigkeit und die Brauchbarkeit vereinfachteter
Vergiftungsmodelle kritisch beurteilt.

237