ANALYTICAL EXPRESSIONS FOR EFFECTIVENESS

FACTORS IN NON-ISOTHERMAL SPHERICAL

CATALYSTS
S. K. SCOTT*, T. BODDINGTON and P. GRAY
School of Chemistry, University of Leeds, Leeds LS2 9JT, England

(Received ? March 1983; accepted24 October 1983)

Abstract-In a short paper, two approximate routes are presented to analytical expressions for
effec~ivencss factors 9 in non-isothermal, spherical catalysts in which diffusion is much faster than reaction.
Both aDDrOacheS involve aooroximation of the Arrhenius temcerature-deoendencc of the reaction rate
cons&; The tint uses a m&i~tic function of tbe tempera& and pro&e& via an equation of the
Lane-Emden family. It uses a parameter I which turns out to be the inverse of 1. This is related to the
experimental conditions (ambient temperature, particle size etc.) encapsulated in the dimensionless group
b by a quadratic equation. The second route builds on the familiar exponential approximation and also
leads to parametric solutions q(x) and d(x). These expressions are again no higher than quadratic.
Both stable and unstable stationary-state temperature profiles are covered. Criticality or the
disappearance of stationary-states emerges naturally as the maxima of the expressions a(l) or a(x).
Although approximate, the approaches are remarkably precise-the second route reproduces sub-
critical and cxitieal results from numaical calculation with less than 2/. error.
The treatmentsare readilyextended to cover more complicated boundary conditions for heat-transfer,
for which there are both internal and external (surface) resistances character&d by thermal conductivity
and the surface heat-transfer coetlicient respectively.

1. UUTEODUCIION boundary conditions in between these, where there

~&y&d expressions[l] can he readily obtained for exist both an internal temperature distribution and a
effectiveness faotors of catalytic particles for iso- temperature step at the edge.
thermal systems of slab, cylindrical and spherical Both approaches invoke approximations to the Ar-
geometry and with some approximations for non- rhenius temperature dependence of the reaction rate:
isothermal systems in which diffusion is very fast for the first employs a polynomial form reducing the
slab and cylindrical shape. Hlav%ek[Z] has shown equation to one of the Lane-Emden family; the second
how the tabulated solutions of Chandrasekhar and uses the well-known exponential approximation[6] of
Wares[3] may be used to evaluate the effectiveness Frank-Kamenetskii.
factor for a non-isothermal spherical particle numer- 2. lFlE.Ol&TlCAL BACKGBOUND
ically. Independently, Boddington and Gray[4] have The starting point for spherical particles with dis-
considered the effect of self-heating on reaction rates tributed internal temperatures is the stationary-state
measured in non-isothermal systems and on the errors heat balance equation
in activation energies and reaction orders derived
from such data. They del&d a rate-constant d=T 2 dT QAc
correction-factor f which is the inverse of the corre- (la)
sponding effectiveness factor. Recently we have ob-
tained approximate analytical solutions[~ for the cor- The simplest boundary conditions are
rection factors for spherical geometry, which are of
reniarkable accuracy. Two routes have been followed dTjdr=Oatr=O, T=T.atr=e (lb)
and these may both be used to give simple expressions
for effectiveness factors. These cover both exothetic (Frank-Kamenetskii or DE&let boundary condi-
reactions, for which q is greater than unity, and endo- tions) in dimensionless terms eqn (1) becomes
thermic systems with self-repression and effectiveness
2
factors less than unity. The treatments apply to the
two extreme forms of heat transfer: distributed inter-
d8+2LB+gexp
dp2. ~dp
nal temperatures with no temperature excess at the
surface (Dirichlet boundary conditions) and uniform
internal temperatures with a temperature step at the
particle surface. They also cover the whole range of de/dp=Oatp=OandB=Oatp=l. (2h)

*Author to whom correspondence should be addressed. Here we follow the established convention and denote
Present address: Department of chemistry, Maequarie Unl- the group of terms Q~QAc exp( - E/RTJJ(KRT~/E)
verity, North Ryde, N.S.W. 2113, Australia. by the symbol 6. This is related to the Thiele modulus

1079
1080 s.K. scorretal.

4, by S = e-{Pr)d where (Pr) is the Prater number. The substitutions I# = (1 + @) and t* = Sa5~pZ
In what follows we shall replace the exponential term transform (7) to one of Bmdens equations[7)
in (2a) corresponding to the Arrhenius temperature
dependence of the reaction rate either by the poly- dZ@/dt2 + (2/t) d4/dt + 4$ = 0 t8a)
nomial form
with
exp{B/(l + ~0)) zz a5(1 + ~8)~ (3)
d+/dr = 0 at t = 0, 4 = 1 at t = (&z~~)/~. (8b)
(typically a = 1 and ~1= l/5)
or by the exponential approximation (6 +O) A particular solution (8) which satisfies the boundary
condition at t = 0 is
exp{0 /( 1 + 60)> 2 exp 8. (4)

The effectiveness factor q is defined by 42 = (1 + PO)2 = $-$ = 312 +Yjskp. (9)

actual reaction rate in non-isothermal stationary-state

= ideal, isothermal reaction rate with all reactant at T,

and is given by the relationship The constant of integration is evaluated from 6 via
the outer boundary condition
q = 3r/l5, (5)

where r = -d@/dp
temperature
at p = 1 is the dimensionless
gradient at the surface.
s=
(3 > C(l-1); +;(l_~,~. (10)

Later we will consider more general boundary The upper root of the second equation yields low
conditions at the surface temperature, stable profiles, the lower root corre-
sponds to high temperature excesses and unstable
de/dp+ge=Oatp=l. (6) solutions. Other compact expressions for various
important quantities in terms of 1 are given in Table
Here p = ~a& is familiar in thermal explosion theory 1. In particular we have for the effectiveness factor
as the Biot number.
q = as/l = 2a5/[l + (1 + 4pa5/36)/7. (11)
3. ANALYTICAL SOLUTIONS VIA QIJINTIC
APPROXlMATIONs: DlRICHLET BOUNDARY
This dependence is shown in Fig. 1.
CONDITIONS
With the approximation (3), eqn (2) becomes 4. ANALYTICALSOLUTIONS VIA SECOND-ORDER
REYERSK#NzDIRICHLET BOUNDARY CONDITIONS
d20/dpz + (2/p) d@/dp + 6a5(1 -t- @)5 = 0, (7a) The exponential approximation (4) yields

and we first consider boundary conditions d20/dp2+(2/p)d0/dp+5exp0=0 (12)

d@/dp=Oatp=O, 0=Oatp=l, (7h) and we again consider the Dirichlet boundary condi-

Table I. Parametric expressions for spherical catalyst with Dirichlet boundary conditionsfrom quintic
approximation
parary?tric critical case F(j)*
lim F(R)
equation F(O) 1-c 1

6 3Q(l-R)/w5 6 = 3/4&L = l5/4 3(1-P.)

CT

v&*-1 t/II uf-1)/b = 2.071 5 (1-R.) /z

eo

r (I-Q.)/r l/Zr = 2j (I-LI/v

11 2/Q al5 = 2 1 + I1 - a)

2
(1 + ua) l/l& + (l-k)021 2/(1 + 02) 1 + (1-i) (1-02)

*n~~~rical values determined with a = 1, II = '15, see Discussion

 Analytical expressionsfor effectivenessfactors in non-isothermal spherical catalysts 1081

given by retaining a second term in (14)

(16)

the second equality following From (15).

Table 2 represents the expressions for various
important quantities in terms of the new parameter
x, as well as values at criticality and LimitingForms for
nearly isothermal systems.
For any given system, the corresponding value of
x is obtained from d as the solution of the quadratic
equation:
0
0 1.0 2.ll 3.0 4.0

6 ; x =6--6 &6(1-0.3S)fl
S= (17)
Fig. 1. Dependence of effbctiveness factor on 6 for Dirichlet (1 +x)* d-1.2 .
boundary conditions from quintic approximation to Ar-
rhenius equation. The upper root corresponds to high internal tem-
perature profiles (unstable solutions), the lower roots
tions (7b). An infinite series solution to eqn (12) in
terms of p* is

$-e(p)=ZlnX@)orexp(8,--B))/2=X@),
(13)

where 0, is the central temperature-excess and X(p)

is given by

The parameter y must satisfy the boundary condition

at p = 1,

A remarkably accurate approximation to the solution

(14) may be obtained as follows. Suppose X, repre-
sents X(p = 1). Then a first approximation to v(X,)
is 6
Fig. 2. Dependence of effectiveness factor on 6 for Dirichlet
yYX,-1=x (15) boundary conditions with exponential approximation to
Arrhenius equation: -, second-order reversion formu-
where x = (X, - 1). The second-order reversion is Iae; ----- numerical computation.

Table 2. Parametric expressions for spherical catalyst with Dirichlet boundary conditions from second-
order reversion

parametricequation F(X) Critical case Ft5/4)

lim F(x)
x+0

8 12 x (l++~/~l+X)2 10/3 12 x

e 2 In (1+x1 4 In t3/2) = 1.622

0 2/x

r 4 x (1+)/(1+x) (1+.&d 2 4 x

rl (1+x)/ (l+i x) g/5 1 +

:x

2 1t -
Bo- e(o) 2 In ~5tl*~xw&x)P 41 2 In I5[l l +?I* - 4) 2xp2
1082 S. K. %n-r er &.

to low extents of internal self-heating (stable solu- Table 3. Parametric expressions for generalized boundary
tions). For endothermic solutions only the positive conditions from quintic approximation
root is physically realistic: solutions are unique. The -2
explicit equation for 9 in terms of 6 is somewhat 6 (3/Z&) 9, [I1 + USs) - f.1

clumsy, but the dependence is displayed in Fig. 2,

along with the numerical solution for the exponential e Iti - 1)/v
0
approximation.
r (l/b) II - p. (1 + iQl(1 + ties1
5. GENERALIZED BOUNDARY CONDITIONG:
QUINTIC APPROXIMATION
II a5 (1 + llesj3/a
For the generalized boundary condition (6) with
arbitrary values of the Biot number, both the heat
2
balance equation with the quintic approximation @a) (1 + uej l/IL + [(l + wear2 - 9.1 P21
and the particular solution (9) still hold. Because of
the different form of the boundary condition, how-
ever, the various interrelationships between the inte- From these formulae we may construct and illus-
gration constant I and the quantities 6, 0, &,, r and trate the dependence of q on 6 for various Biot
q are altered. Most significantly there is another numbers as shown in Fig. 3.
parameter which determines the behaviour, B. Writ-
ing 0 = 0, # 0 at p = 1, 6, I and 8, are related by 6. GENERALIZED BOUNDARY CONDlTIONs:
SECONSORDEB REVERSION
b = 46 I[( 1 + /.l&)-2 - 11 (18) The infinite series solution (13) and its second-
order reversion (16) still apply when the more general
where Sz = 3/(4a%) is used for the classical critical boundary condition (6) is used. Now, however, we
value at infinite Biot number. have that
The generalized expressions for the quantities listed
above are given in Table 3 in terms of I and 0,. The e=e,=r/fi (20)
values of I and 0, appropriate to a system of specified
initial conditi0nsi.e. specified S and /? are obtained at p = 1, the second equality arising horn the
by solving eqn (18) and definition of r, the temperature gradient at the
surface. in Section 4, the sum of the infinite series
l = [(l + Pee*)- BPW(l + NV (19) x, = 1 + x was defined in terms of the central
temperature-excess and that at p = 1, hence (13) gives
which arises from differentiating eqn (9) and using the
boundary condition at p = 1. e,-e,=zh(l +x); exp e, = (1 + x)~ exp(r//3).
(21)

The factor exp 0, = exp(r/P) plays a very important

role, occurring in many of the generalized expres-
sions. The parametric forms for the various quantities
S, t?,, r and 9 in terms of x and fi are displayed in
Table 4. The equation for the temperature profile
8(p) is the same as that given in Table 2 for the
Frank-Kamenetskii or Dirichlet boundary condi-

Table 4. Parametric expressions for generalized boundary

conditions from second-order reversion

12 x (1+x)
6 exp {- r/B3
(1+x)2

2 In (1+x1 l r/B
e*

4 x (1+++
r
(1+x) 11+J X)

(1+x)
rl ---i-- exp{- r/61
Fig. 3. Dependence of effectiveness factor on 6 under
t1+3 X)
general boundary conditions for various values of the Biot
number fi, with quirk approximation to rate law.
 Analytical expressions for effectiveness factors in non-isothermal spherical catalysts
6 give the characteristic S-shaped curve. This is because
Fig. 4. Dependence of effectiveness factor on 6 under
general boundary conditions for various vzdues of the Biot
nmnher fl from second-order reversion.
tions, although the critical profile is different from
that given in Table 2 because the critical value of x
is altered by the finite Biot number-see Discussion.
The dependence of tbe effectiveness factor on d for
various values of the Biot number is shown in Fig. 4.
7. DISCUSSION
Both approaches developed here, q&tic approxi-
mation and second-order reversion, are approxi-
mations to the Arrhenius temperature-dependence of
the reaction rate. The status of these and of other
approximate forms have been discussedtl, 81 else-
where. The second-order reversion builds upon
Frank-Kamenetskiis exponential approximation
(E -0) and because of its familiarity tbis is sometimes
regarded as the correct formulation of the problem.
In general, however, it is no more valid than other
possible choices, and the polynomial chosen here is in
fact a better fit to the full Arrheuius equation for a
value of E = l/20.
Tbe results derived from the two treatments arise
naturally in parametric forms. These expressions in
terms of I or x are neat and simple, and they afford
further insights into relationships between different
quantities that are not so obvious from the numerical
solutions. Both eqns (IO) and (17) for S(i) and 6(x)
with Dirichlet boundary conditions respectively are
quadratic equations with natural maxima. For the
polynomial rate-law the maximum value for 6(f)
occurs when I = l/2 and 1(1 - I) = l/4. This is the
critical value of 6 = 3/(4aJ~). For values of d < S,,
there are two roots of (10) for 1, one in the range
1 < I c l/2 corresponding to stable solutions, the
other 0 < I < l/2 leading to unstable profiles. Thus, in
general, for a given set of initial conditions there will
be either two solutions for the effectiveness factor or
1083
none. This behaviour follows that observed in cylin-
drical and slab particles.
For the second-order reversion, the maximum in
6(x) occurs for x = 514 when a,, = 10/3. This is in
remarkably precise agreement with the numerical
result for the exponential approximation a,, = 3.322.
Again subcritical systems may show two roots x for
a given 6. However, for d = 615 the quadratic equa-
tion becomes singular (the second root x + tends to
infinity). The unstable branch of the q-8 curve termi-
nates abruptly here, at the value 4 = 5. For lower
values of 6 there is only one positive solution for x
and hence a unique effectiveness factor. The results
presented here do not show the spiralling displayed
by the numerical solutions, which are singular about
6 = 2. (The eye of the spiral corresponds[l] to q = 3.)
The latter arises from the steepness of the exponential
approximation and has been discussed[9] in terms of
the S - $ solutions by B. F. Gray. We also do not
observe any turning round of the unstable branch to
we have assumed that the rate of diffusion is much
faster than the reaction rate so that the concentration
of the reactant throughout the catalyst can be taken
as constant and equal to the concentration outside
the particle. At high reaction rates this may well
break down. The reaction becomes diffusion con-
trolled, and our model cannot cope with this. Thus
we would not expect our expressions to apply to
systems with high extents of self-heating as typified
by the more extreme unstable states. At such high
temperatures both the quintic and exponential ap-
proximations cease to be good representations of the
Arrhenius dependence which shows a distinct
flattening out, with the term exp(O/(l + ~0)) tending
to the finite value exp(l/a) as 0 -+ m. The two approx-
imations become unbounded as 0 becomes infinitely
large.
In the vicinity of the stable branch, up to and
including the turning point, our expressions are re-
markably accurate. The second-order reversion solu-
tions reproduce the numerical results of HlavS%k[2]
and of Chandrasekhar and Wares [3] to less than 1%
error over the whole of this range. Direct comparison
of the quintic formulae with these classical values is
less appropriate as it is a different approximation to
the Arrhenius term.
Under general boundary conditions, with arbitrary
Biot number, both the expressions S(f) and I(x) are
altered slightly. For the quintic approximation the
product 1(1 - 1) is replaced by f[(l + l(e,)-- 11,
where 8, also depends on I. Clearly the maximum of
this term no longer correspotids to I,, = 1J2 and its
location is now a function of the heat transfer
characteristics /3. The dependence of i, on the Biot
number is shown in Fig. 5. As jl is decreased, 1,
increases corresponding to lower values of 6, 0, and
r at criticality. At the same time the critical value of
the effectiveness factor increases. This is because the
extra resistance to heat transfer at the particle surface
leads to greater extents of self-heating away from the
 S. K. Sccrrr et al.

modulus or Frank-Kamenetskiis 6 in catalyst par-

ticles of spherical geometry can he obtained by two
approximate routes. One approach approximates the
Arrhenius temperature-dependence with a quintic
form, the second uses the exponential approximation
aud the reversion of its infinite series solution. The
results apply to all boundary conditions for heat
transfer, from Dirichlet conditions to uniform inter-
nal temperatures, and to both exothermic and endo-
thermic processes. Both routes have good precision,
errors for stable solutions are rarely greater than 1 or
2%. The conditions for criticality correspond to the
merging of the stable and unstable branches, and
arise as the maxima of simple quadratic expressions.

Acknowledgement-We are grateful to British Gas for a

Research Fellowship for S.K.S.

*.
0 10 102 103 NOTATION
catalyst radius
Biot number P Arrheuius pre-exponential factor
Fig. 5. Dependence on &t number of critical values for 1 reactant concentration inside particle, assumed
(qnintic approximation) and x (second-order reversion).
Note the change of scale on the abscissa from linear to constant and equal to concentration outside
logarithmic at Bit numberp = 10. catalyst
diffusion coefficient
activation energy
centre, and it is in the outer region of the particle
that parameter of quintic solution
the majority of the volume lies. reaction order
The dependence of 5 on x for the second-order = coQD /KT,, Prater number
reversion expression is complicated by the extra = - AH, reaction exothermicity
factor exp[ - r (x)//I]. This again shifts the maximum distance from centre of sphere
in 6 and the value of x to which this corresponds. The Universal Gas constant
variation of x,, with b is also given in Fig. 5. = (Sa5p)2p dimensionless distance
The polynomial approximation apparently offers thermodynamic (absolute) temperature
two further parameters 0: and P, whose values must = 6 e&/12, parameter of infinite series solution,
be assigned before the expressions can be used to eqn (14)
obtain numerical results. There are various criteria = parameter of second-order reversion
that may be used to gnide these choices, but the most coustant
natural requirements are that this approximate form =X%/PC, Biot number
should have the same value and slope as the Ar- = - d0/dp at p = 1, temperature gradient at
rhenius dependence at the ambient temperature of the surface
particle (i.e. when 0 = 0). Thus we specify = u,,~QAc~exp(- E/RTJ(RT,Z/E)tc, Frank-
Kamenetskiis reduced version rate at T.
c?(l + j&s)5= exp{0/(1+ &)} at B = 0 = RTJE, dimensionless ambient temperature
effectiveness factor
= (T - T,)/RTz/E), dimensionless tem-
perature excess
d{a(l + rtW5)= dew{@/U + 63) at e = o
thermal conductivity
de d0 constant
= r /a~, dimensionless distance
These lead to the values a = 1 and p = l/5. This surface heat-transfer coefficient
choice for K leads to the striking result which can be =(I + pe)
seen in Table 1, that the parameter I is simply the
inverse of the effectiveness factor, and the other
REFERENCES
expressions then clearly relate 6, 0, and r to rf.
PI Aris R., TheMathematicalTheoryof Difubn d Reac-
tion in Pemzeoble Cataiyzts. Clare.ndon Press, Oxford
8. CONCLUSIONS 1975.
Compact, analytical expressions for the de- PI Hlav6Eck V., Marck M. and KubiEek M., CON. Czech.
pendence of the effectiveness factor on the Thiele Chem. commtm. Engl. E&L 1968 33 718.
 Analytical expressions for effectiveness factors in non-isothermal spherical catalysts 1085

[3] Chandrasekhar S. and Wares G. W., Astrophys. J. 1949
109 551.
[4] Boddington T. and Gray P., Proc. R. Sot. Lond. 1970
A32ll71.
[q Boddinpon T., Gray P. and Scott S. K., Proc. R. Sot.
Lund. 1981 A378 27.
[6] Frank-Kamenetskii D. A., Zlr. jz. Khan. 1939 13 738.
[7] Chandrasekhar S., Introduction to the Study of Stellnr
Structure. University of Chicago Press 1939.
[8] Boddington T., Gray P. and Scott S. K., Trans. Faraday
Sot. II 1982 78 1721.
[9] Gray B. F., Cornbust. F/me 1969 13 50.