You are on page 1of 6

Chemical Engineering Journal 94 (2003) 107112

Approximation of the effectiveness factor in catalytic pellets

S.D. Keegan a,b , N.J. Mariani a,b , S.P. Bressa a,b , G.D. Mazza a,b , G.F. Barreto a,b,
a Departamento de Ingeniera Qumica, Facultad de Ingeniera, Universidad Nacional de La Plata, La Plata, Argentina
b Centro de Investigacin y Desarrollo en Procesos Catalticos (CINDECA), CONICETUniversidad Nacional de La Plata,
Calle 47 No. 257, CC 59, CP B1900AJK, La Plata, Argentina
Received 31 May 2002; accepted 14 December 2002

The 1D model proposed by Burghardt and Kubaczka [Chem. Eng. Proc. 35 (1996) 65] to approximate the behavior of 3D catalytic
pellets has been recently found able to provide accurate results for evaluating effective reaction rates when its parameter is suitable
adjusted [Chem. Eng. Res. Des., submitted for publication]. This parameter represents the contraction of the cross-section available for
diffusion. A formulation coupling a first-order Galerkin approximation with a truncated asymptotic expansion is proposed here to evaluate
the effectiveness factor of single reactions in the range of interest 1/5 < < 5 [Chem. Eng. Res. Des., submitted for publication]. The
formulation provides a 3% level of precision for essentially all normal kinetics of practical interest and a large range of abnormal kinetics.
In particular, this conclusion includes reaction rates approaching a zero-order reaction, for which large deviations arise from the use of
previous approximations proposed in the literature. On the other hand, the extent of abnormal kinetics being accurately approximated is
significantly enlarged.
2003 Elsevier Science B.V. All rights reserved.
Keywords: Effectiveness factor; One-dimensional model; Diffusion-reaction

1. Introduction and literature survey 1/5 < < 5. In this range, Mariani et al. [2] found that
the GC model is precise up to 0.7% for linear kinetics and
Burghardt and Kubaczka [1] proposed a 1D geometric they estimated that the deviations will not rise above 2% for
model, hereafter called GC model, to approximate the be- non-linear kinetics outside the range of steady state multi-
havior of 3D catalytic pellets. The GC model consists in a plicity. Therefore, the GC model seems to be an accurate
body allowing mass flux on a single dimensionless coordi- mean to avoid 3D evaluations for effective reaction rates.
nate z, 0 z 1, and presenting a cross-section of variable For actual application, the GC model will be most valu-
area proportional to (z ). According to the value of the shape able in the case of multiple reactions, as it avoids the heavy
power , the model reduces to a slab ( = 0), an infinitely computational burden of having to solve a set of conserva-
long circular cylinder ( = 1) or a sphere ( = 2). tion equations in 3D. Nonetheless, we will restrain ourselves
To approximate the behavior of a generic 3D pellet in this contribution to a single reaction, with the purpose
shape, the external surface area and volume of the model of presenting an approximation to evaluate the effectiveness
body are assumed to equal those of the actual pellet, Sp and factor without resorting to the numerical solution of the con-
Vp , and according to Mariani et al. [2], is evaluated by servation equation.
matching the behavior of the actual pellet at low reaction Numerical codes and computer facilities have been de-
rates. Some simple particle shapes can lead to high values veloped up to such an extent that the evaluation of effective
of , as a circular cylinder with a height to diameter ra- reaction rates is far from being a numerically challenging
tio 0.85 ( = 3.25) or a cube ( = 4.3). Values < 0 problem, provided that a few isolate calculations should
are not likely for catalytic pellets, but the important case be made. However, it should be realized that relatively
of monolith reactors with catalytic washcoat on channels novel catalytic reactors, as reverse-flow reactors and re-
shows this feature [3]. The range of interest can be set as active distillation processes, or more sophisticated models
for traditional reactors, including CFD simulations, can de-
Corresponding author. Tel.: +54-221-4210711; mand thousands, or higher orders, of spatial and temporal
fax: +54-221-4254277. discretization points, in which the effective reaction rates
E-mail address: (G.F. Barreto). should be evaluated. Besides, the overall solution will most

1385-8947/03/$ see front matter 2003 Elsevier Science B.V. All rights reserved.
108 S.D. Keegan et al. / Chemical Engineering Journal 94 (2003) 107112

often have to be found iteratively. If in addition, operating defined for the application of the FOG approximation and for
conditions or design parameters are being optimized, the > a truncated expansion in terms of 1/ is employed.
number of such evaluations will be augmented by a large
factor. It seems clear that a suitable approximation for
evaluating effective reaction rates will be desirable, even 2. The formulation to approximate
considering nowadays computer speed.
A number of approximations for have been presented The problem to solve for a single reaction is:
in the literature (e.g. [415]). Some of them have specific  
d dCA
purposes, as those of Szukiewicz [11,12] intended to evalu- z z DA (CA ) = L2 rA (CA ) (1)
dz dz
ate transient, in addition to steady state, solutions and those
of Ochoa and co-workers [13,14] intended to provide a fast dCA
evaluation for low and intermediate Thiele modulus (). CA = CAs (z = 1); =0 (z = 0) (2)
They are not suitable to cover the whole range of . For this
where A is the limiting reactive species and L is the diffusion
reason, they will not further discussed. All the quoted con-
length, L = (1 + ),  = Vp /Sp .
tributions were originally intended to deal with any or all of
A dimensionless dependent variable Y is defined as
the discrete values = 0, 1, 2. Except the contribution of
Hong et al. [15] that can only be employed for = 2, the 
1 CA
other approximations can be readily adapted for continuous Y= DA (CA ) dCA ;
values of . However, only some of them can be used for  CAs
the whole range of interest 1/5 < < 5 for a given reac- JA = DA (CA ) dCA (3)
tion rate expression. Each approximation shows some other CAe
distinguishing features, but some common conclusions can
where CAe is the equilibrium concentration. In the case that
be drawn.
DA (CA ) is constant, JA = DA (CAs CAe ) and Y = (CA
For kinetics showing effective reaction orders (ne =
CAe )/(CAs CAe ). Eqs. (1) and (2) become
d ln r/d ln C) higher than around 0.5, most of the approxima-  
tions provide a good estimation of in the applicable range z
= (1 + )2 2 r(Y) (4)
of values. Instead, when 0 ne < 0.5 the error increases dz dz
for all the approximations. In the limit, for the so-called
zero-order isothermal reaction (actually, a step function of Y(1) = 1; Y
(0) = 0 (5)
the reactant concentration) none of them can guarantee er- where r = rA /rAs , and
rors less than 20% in the range 0 < < 5. It is recalled rAs
that low values of ne do not only arise from power-law rate 2 = 2 (6)
expressions, but also for the very important case of some
LHHW rate expression with significant inhibition effects. It is noted that with the given definition of Y, r = 0 at Y = 0.
For abnormal kinetics the approximations of Wedel and The effectiveness factor is the average dimensionless rate
Luss [4]; Gottifredi et al. [6], Gottifredi et al. [8] and Yin and  1
Li [10] can be employed, although the errors rapidly increase = (1 + ) r z dz (7)
as the magnitude of the maximum rate increases. This is
expectable, as these approximations join, by means of a The formulation employed in this work to evaluate is
suitable expression, the limiting behavior of as 0 and presented in Table 1. The two approximations employed will
and, therefore, the position and magnitude of the be described in the following paragraphs.
maximum can hardly be predicted from such information. The first-order Galerkin method is employed in the lower
The aim of this work is to present for the whole range range of . The formulation described in Chapter 2 of [16]
1/5 < < 5 a procedure capable to approximate the is used for FOG. It is based on the approximation
effectiveness factor for essentially all normal kinetics of
Y = 1 + (Y0 1)(1 u) (8)
practical interest and for a wide range of abnormal kinet-
ics. Special emphasis has been put in establishing the pre- where u = z2 and Y0 = Y(0) is the parameter to be deter-
cision level and the limit of application of the proposed mined. The integral of the residuals of Eq. (4) weighted by
formulation. (1u) over the pellet volume is evaluated by a GaussJacobi
From the previously summarized analysis, it was con- (weighting function [(1u)u(1)/2 ]) quadrature. Then, the
cluded that knowing the limiting behavior of at 0 and quadrature points are the zeros {ui } of the Jacobi polyno-
is not enough to deal with strongly abnormal kinet- mials PN (u) defined by
ics. Based on this observation, the first-order Galerkin (FOG)  1
approximation is adopted. It is well known that the FOG ap- [(1 u)u(1)/2 ]uj PN (u) du = 0; j = 0, . . . , N 1
proximation is accurate and computationally efficient [16] 0
from low to intermediate values of . A critical value is (9)
S.D. Keegan et al. / Chemical Engineering Journal 94 (2003) 107112 109

Table 1 coefficients 1 and 2 (Eq. (T7)) are evaluated analytically

Formulation of the proposed approximation (e.g. from Gonzo and Gottifredi [18]), while (Eq. (T8))

First-order Galerkin approximation (used when ) provides continuity with the FOG approximation.
N For a zero-order reaction or normal LHHW kinetics be-
2 1 r(Yi )
G () = 1 i (T1)
3+ 1 ui having alike, B shows a small maximum shortly after
where (=0 in these cases) when > 3. Although not very impor-
Yi = 1 + (Y0 1)(1 ui ) (T2) tant in magnitude, this kind of errors is undesirable. In the
N present range of interest, 3 < 5, we suggest using the
1 Y0 = 21 (1 + )2 R(Y0 ); R(Y0 ) = i=1 i r(Yi ) (T3)
solution for = 3 with a suitable scaled Thiele modulus,
= min[0 , M ]
as follows. Define
2 + 2
S0 () =3
where S =
2 2
2 (1 + /5) 2 + 2
S ()
20 = ; 2M = 9 G (0 ) (T5) =3
(1 + )R(0) (1 + )2
96 64
S0 = , S = (11a)

Asymptotic approximation (used for 3, > )
(1 + )(3 + ) (1 + )2
1 2 Then,
B () = + 2 + 3 (T6)

= (S )=3 (3 < 5) (11b)

 1 where () =3 is the value of (Table 1) calculated with
1 = [P(1)]0.5 ; 2 = [P()]0.5 d; (T7) = 3 and (S )=3 is the value from the algorithm in
(1 + )1
P() = 2 0 r(Y) dY Table 1 at S and = 3.
It is worth recalling the number of times that r(Y) should
3 G () 2 2
1 (T8) be calculated for the algorithm in Table 1. When ,
G (0 ) demands N calculations of r(Y) and the iterative pro-
cess for G () demands about 3N calculations. For FOG2 ,
Eq. (T3) in Table 1 is the resulting expression that allows this implies eight evaluations in all.
the evaluation of Y0 . {i } is the set of weighting
 factors of When > , 1 and 2 (Eq. (T7)) should be numerically
the quadrature just described, normalized to i = 1. For evaluated, for which 10 calculations r(Y) were employed.
N = 2 (FOG2 ) These should be added to the N calculations for G (0 ) and,
just in the case = M (Eq. (T4)), the 3N calculations
(3 + ) 12(3 + )/(7 + )
u1,2 = ; for G (M ). More details about computational aspects are
9+ discussed by Keegan et al. [17].
u2 (1 + )/(5 + )
1 = ; 2 = 1 1 (10)
u2 u 1
3. Accuracy of the proposed formulation
Once Y0 is evaluated, from Eq. (7) is calculated by
a Radau (end-point u = 1)Jacobi (weighting function
We will consider for r = rA /rAs expressions of the fol-
u(1)/2 ) quadrature. The same set {ui } of quadrature
lowing form,
points applies, and the corresponding weighting factors can
be written in terms of {i }. Recalling that r(1) = 1, the (1 + A)d (1 C)
r = exp
final result is that in Eq. (T1). (1 + AC)d 1 + (1 C)
If N = 1 is chosen, the method becomes the one-point  n 
C Cen (Q/Qe )m
orthogonal collocation (1POC) method described by Villad- . (12)
sen and Michelsen [16]. The general case with N quadrature 1 Cen (Qs /Qe )m
points will be referred as FOGN method. It is anticipated C is the limiting-species dimensionless concentration, C =
that FOG2 will be the recommended choice in most cases, CA /CAs , and Q represents the dimensionless concentration
on the basis of results obtained from numerical experiments. of a product, Q = CQ /CAs , that is assumed to be stoichio-
The solution from FOG, G , is used up to = , metrically related to C according to Q = Qs + 1 C. The
Eq. (T4), where 0 (Eq. (T5)) makes Y0 = 0 in the solu- equilibrium concentration C = Ce is taken as a parameter
tion of Eq. (T3). Note that the evaluation of G (0 ) from and Qe follows from stoichiometry. The diffusivity DA will
Eq. (T1) does not require any iteration, as it corresponds to be taken as a constant, so variables C and Y are related by
Y0 = 0 and thus r(Yi ) r(ui ) (Eq. (T2)). Further remarks C = Y(1 Ce ) + Ce . The reaction orders n, m are parame-
about FOG in general and the definition of in particular ters which should be assigned to complete the definition of
can be found in Keegan et al. [17]. the driving force in Eq. (12).
At > , the proposed approximation B (Eq. (T6) in The inhibition term (1+AC)d involves two further param-
Table 1) is a truncated expansion of in terms of (1/). The eters, the dimensionless adsorption constant A = KA CAs
110 S.D. Keegan et al. / Chemical Engineering Journal 94 (2003) 107112

Table 2
Maximum errors for FOG approximation (FOG2 , unless otherwise stated) and conditions at which they take place
Group Type of kinetic expression and parameter ranges Maximum (%) Conditions

A Irreversible nth order endothermic reactions: n = 03, () = 5, 2.8 = 4; n = 0

= 0.2
B Irreversible LHHW reactions: n = d = 1, 2; A = 030; isothermal 2.4 = 4; n = d = 2; A = 20; endothermic
and endothermic () = 5; = 0.2 cases.
C Reversible reactions: n = m = 0.5; Qs = 0; Ce = 0.5, 0.9; 3.1 = 1; Ce = 0.9; endothermic
isothermal, endothermic () = 5; = 0.2 and exothermic
() = 5; = 0 cases.
D Irreversible exothermic nth order reactions, DMAX = 1: n = 03. 2.8 = 0; n = 2
E Irreversible LHHW reactions (d > n), DMAX = 1: isothermal (n, d, A) = 3.2 = 5; (n, d, A) = (1, 2, 4.3)
(1, 2, 4.3; 0.5, 1, 8) and exothermic (n, d, A) = (0.5, 1, 4.6); () = 0.3;
= 0 cases.
F Irreversible nth order exothermic reactions, DMAX = 2 (FOG3 ): 3.8 = 5; n = 0.5
n = 03
G Irreversible LHHW reactions (d > n), DMAX = 2 (FOG3 ): 3.6 = 5; (n, d, A) = (1, 2, 6.4)
isothermal (n, d, A)=(1, 2, 6.4; 0.5, 1, 13) and exothermic (n, d, A) =
(0.5, 1, 9); () = 0.3; = 0 cases.

and inhibition exponent d. The Prater () and Arrhenius

() numbers define the exponential term in Eq. (12) for
non-isothermal cases.
Reversibility, inhibition or thermal effects are not ac-
counted for if A, () and Ce , respectively, are taken as be-
ing zero. In the examples discussed below it will be assumed
that a given effect is not included unless the specific param-
eters are explicitly defined. Note that for the reversibility
effect three parameters should be stated: m, Ce and Qs .
Some restrictions for the values of some parameters are
taken from the outset, in order to keep realistic values.
The reaction orders will be taken within 0 n, m 3.
The specific case identified with n = 0 actually means the
zero-order reaction, Cn = 1 if C > 0; Cn = 0 if C = 0. The
non-isothermal parameters will be constrained by 25
and || 0.2. Fig. 1. Error (%) for irreversible nth order isothermal reactions.
Given the set of parameters defining r(Y) and the value of
, the following relative error has been identified the values of from the proposed approximation are exact
for the specific values = 1 and = 3. For the remaining
() num ()
= max values of , the approximation behaves very well, with a
num ()
maximum value (1.67%) at = 5. In general, keeps below
where is the value from the approximate formulation and 2% for all n and .
num arises from the numerical solution of Eqs. (4) and (5). The errors adding an endothermic effect ( = 5) to
From the way in which the approximate formulation has irreversible nth order reactions were somewhat higher than
been built up, always corresponds to some intermediate for the isothermal case, although not exceeding 3% (Table 2,
value of . The results for a number of cases of practical Group A).
interest will be discussed next. Due to space limitations, most Inhibition effects (LHHW kinetics) with d n do not
of the results are summarized in Table 2. Further details and introduce much variation in precision respect to power law
comments can be found in Keegan et al. [17]. kinetics. A number of examples were tried, either at isother-
mal conditions or with endothermic effects (Group B of re-
3.1. Normal kinetics sults in Table 2).
As far as m > 1, the effect of the reversibility term in
Normal kinetics are here understood as those in which Eq. (12) on the level of accuracy of the proposed approxima-
dr/dY is non-negative in the range 0 Y 1. The use of tion is not significant. However, if m < 1 and Qs = 0, the
FOG2 is assumed, unless the contrary is stated. derivative dr/dY tends to infinity at Y = 1, causing fast vari-
For irreversible, isothermal nth order reactions, the values ations of r close to the pellet surface. The deviations for a
of are displayed in Fig. 1 as a function of . For n = 0, highly limited reaction (Ce = 0.9), keeping n = m = 0.5, at
S.D. Keegan et al. / Chemical Engineering Journal 94 (2003) 107112 111

isothermal, endothermic ( = 5; = 0.2) and exother-

mic ( = 5; 0) conditions, are in general bounded
by 3%, except for some values slightly higher (see Table 2,
Group C). Temperature variations become restrained by the
strong reversibility, even for the large value || = 5. When
lower values of Ce are tried (Ce = 0.5), the errors become
smaller than for Ce = 0.9. Nonetheless, the case Ce = 0.5,
() = 5 is particularly interesting as r(Y) shows a minimum
and a maximum within 0 < Y < 1 (this is actually an exam-
ple of abnormal kinetics). Consequently, the corresponding
curve versus also shows a minimum and a maximum.

3.2. Abnormal kinetics

It is very well known that strong exothermic or inhibi-

tion effects producing dr/dY < 0 lead to values > 1 and
eventually to multiple solutions. The maximum value of
(dr/dY) occurring within 0 Y 1, denoted as DMAX ,
has been found a suitable criterion to establish up to what
Fig. 2. Comparison of errors (%) from Wedel and Luss (WL) [4] and
extent the approximation can be used with adequate level Yin and Li (YL) [10] expressions and FOG2 , for LHHW reactions with
of precision [17]. d > n and = 2.
Exothermic and irreversible nth order kinetics and values
of the product () such that DMAX = 1 were tested leading
to always below 3% (Table 2, Group D). Smaller values the errors are still moderate, around 8.0% (d = 2) and 3.7%
of () produce smaller values of , on average, and always (d = 1), and perhaps acceptable for some purposes. The
less than 3%. exact maximal is 1.62 for (n, d, A) = (1, 2, 10), while
Two cases of isothermal and irreversible kinetics FOG2 predicts = 1.66.
(DMAX = 1) with significant inhibition effects n < d and Also shown in Fig. 2 are the results from Wedel and Luss
one case combining exothermic and inhibition effects were [4] and Yin and Li [10] approximations, specially developed
also analised. The parameter A was again taken such that for = 2. The errors from both expressions rapidly increase
DMAX = 1 results. The error remains below 3%, except as A increases, although for this particular value of , Yin and
for a few cases showing slightly higher values (see Table 2, Li [10] approximation behaves better. The Gottifredi et al.
Group E). [6,8] approximations have been also tried for the conditions
The examples shown up to this point allow to conclude in Fig. 2. They present similar results as those from Wedel
that FOG2 is able to approximate normal and abnormal ki- and Luss [4] approximation in the upper range of A, but the
netics showing values of DMAX 1 within a precision of latter is more suitable at small A.
around 3%.
On the other hand, rate expressions showing values of
DMAX increasingly larger than one lead to errors consistently 4. Conclusions
larger than 3%. It has been tried if higher values of N could
enlarge the value of DMAX , while maintaining the precision A formulation coupling a first-order Galerkin approxima-
around 3%. The approximation FOG3 can be judged as able tion with a truncated asymptotic expansion is proposed to
to maintain the desired level of precision up to DMAX = 2, evaluate the effectiveness factor of single reactions in the
if some tolerance is allowed (up to 4%, see Groups F, G in range 1/5 < < 5 of the GC geometrical model (Table 1).
Table 2). Higher values of N do not succeed to extend the The formulation introduces an empirical criterion to shift
value of DMAX above 2, showing that the FOG approxima- from the FOG to the asymptotic approximation, expressed
tion (Eq. (8)) becomes limiting, rather than the precision of by Eq. (T4) in Table 1. In the range 3 < < 5, the formu-
the quadrature to evaluate the integrated residuals. lation is not directly applied, but the results from = 3 are
It is interesting to disclose the behaviour of the approx- employed along with a scaled Thiele modulus, Eq. (11a).
imation for abnormal kinetics beyond the analysed values The FOG2 option, employing two quadrature points, is
of DMAX . The effect of A on for the already consid- suggested for general purposes. FOG2 provides a 3% level
ered LHHW isothermal cases (n, d) = (1, 2) and (n, d) = of precision for essentially all normal kinetics of practical
(0.5, 1) can be appreciated in Fig. 2, using FOG2 and = 2 interest and abnormal kinetics restrained by a maximum neg-
(sphere). At the largest A in Fig. 2, DMAX equals 4.48 and ative slope (dr/dY) = DMAX = 1. In this way, FOG2 al-
3.71 for d = 2 and d = 1, respectively. In spite of these val- lows very precise predictions for reaction rates approaching
ues largely exceeding the safe limit DMAX = 1 for FOG2 , a zero-order reaction, for which large deviations arise from
112 S.D. Keegan et al. / Chemical Engineering Journal 94 (2003) 107112

the use of former approximations. On the other hand, the kinetics, geometry and activity distribution, Chem. Eng. Sci. 36
extent of abnormal kinetics being accurately approximated (1981) 713719.
is significantly enlarged. [7] H.W. Haynes Jr., An explicit approximation for the effectiveness
factor in porous heterogeneous catalysts, Chem. Eng. Sci. 41 (1986)
If some errors between 3 and 4% can be accepted, the 412415.
use the FOG3 approximation allows dealing with kinetics [8] J.C. Gottifredi, E.E. Gonzo, O.D. Quiroga, Approximate effectiveness
showing up to DMAX = 2. factor calculations, in: Whitaker, Cassano (Eds.), Concepts and
Design of Chemical Reactors, Gordon and Breach, New York, 1986.
[9] R.J. Wijngaarden, A. Kronberg, K.R. Westwerterp, Industrial cata-
lysis, Optimizing Catalysts and Processes, WileyVCH, Weinheim,
Acknowledgements 1998.
[10] Q. Yin, S. Li, Rational approximation of the overall effectiveness
The authors wish to thank the financial support of the factor for the gasliquidsolid phase catalytic reaction, Ind. Eng.
following Argentine institutions: ANPCyT-SECyT (PICT Chem. Res. 34 (1995) 37713776.
[11] M.K. Szukiewicz, New approximate model for diffusion and reaction
no. 00227 and no. 06297), CONICET (PIP96 no. 4791) and
in a porous catalyst, AIChE J. 46 (2000) 661665.
UNLP (PID no. 11/I058). G.D.M., N.J.M. and G.F.B. are [12] M.K. Szukiewicz, An approximate model for diffusion and reaction
research members of the CONICET, S.D.K. is a fellow of in a porous pellet, Chem. Eng. Sci. 57 (2002) 14511457.
the CONICET and S.P.B. is member of the CIC. [13] J.O. Marroqun La Rosa, T.V. Garca, J.A. Ochoa Tapia, Approximate
isothermal global effectiveness factor, in: Proceedings of the
15th International Congress of Chemical and Process Engineering
References (CHISA 2002), Paper no. 0801, Congress CD, ISBN 80-86059-33-2,
[14] J.O. Marroqun, T. Viveros, J.A. Ochoa, Evaluation of isothermal
[1] A. Burghardt, A. Kubaczka, Generalization of the effectiveness factor
effectiveness factor for nonlinear kinetics using approximate method,
for any shape of a catalyst pellet, Chem. Eng. Proc. 35 (1996) 6574.
Ind. Eng. Chem. Res. 37 (1998) 37803781.
[2] N.J. Mariani, O.M. Martnez, G.F. Barreto, A one-dimensional
[15] J. Hong, W.C. Hecker, T.H. Fletcher, Improving the accuracy of
equivalent model to evaluate overall reaction rates in catalytic pellets,
predicting effectiveness factor for mth order and Langmuir rate
Chem. Eng. Res. Des., submitted for publication.
equations in spherical coordinates, Energy and Fuels 14 (2000) 663
[3] D. Papadias, L. Edsberg, P. Bjrnbom, Simplified method of effe-
ctiveness factor calculations for irregular geometries of washcoats:
[16] J. Villadsen, M.L. Michelsen, Solution of Differential Equation
a general case in a 3D concentration field, Catal. Today 60 (2000)
Models by Polynomial Approximation, Prentice-Hall, Englewood
Cliffs, NJ, 1978.
[4] S. Wedel, D. Luss, A rational approximation of the effectiveness
[17] S.D. Keegan, N.J. Mariani, S.P. Bressa, G.D. Mazza, G.F. Barreto,
factor, Chem. Eng. Commun. 11 (1980) 245259.
Approximation of the effectiveness factor in catalytic pellets
[5] J.C. Gottifredi, E.E. Gonzo, O.D. Quiroga, Isothermal effectiveness
(electronic version), 2002.
factor. I. Analytical expression for a single reaction with arbitrary
[18] E.E. Gonzo, J.C. Gottifredi, Rational approximation of effectiveness
kinetics. Slab geometry, Chem. Eng. Sci. 36 (1981) 705711.
factor and general diagnostic criteria for heat and mass transport
[6] J.C. Gottifredi, E.E. Gonzo, O.D. Quiroga, Isothermal effectiveness
limitations, Catal. Rev. Sci. Eng. 25 (1983) 119140.
factor. II. Analytical expression for a single reaction with arbitrary