differ in molecular composition, particularly with respect The Voorhies (1945) relationship between catalyst coke and

to olefins, from virgin gas oils. From the linearity of the data catalyst residence time can be written as,
for the virgin gas oils, the following relationships can be de-
C =
atcb
rived:
where C =
catalyst coke, tc catalyst residence time, and a
=

and b are constants. The coefficient o is to a significant extent

a measure of the coking tendency of a charge stock. It is
reasonable to expect some correlations exist between a and the
catalyst decay constants. Figures 7 and 8 are, respectively,
plots of a vs. a and n vs. a. Both of the catalyst decay con-
From Figures 2 and 3, the values of a0 and ba are 25 and 0.42>
—

stants increase with increasing a, which agrees with the trend

respectively, and the corresponding ones for cq and &i are 20 shown on Figure 4.
and —0.46 for the range of data shown. These values were
calculated by least-squares regression of the data for virgin Summary and Conclusions
gas oils. Relationships of K<i to aromatic/naphthene ratio did
The kinetic rate constants for the catalytic cracking of gas
not correlate as well as the other rate constants. However,
oils and subsequent gasoline formation can be correlated with
Ki does not change drastically with composition (Nace et al., the aromatic/naphthene ratio for a wide range of virgin
1971), and it typically varies only from 1.5 to 2.5 at 900°F.
charge stocks. The catalyst decay constant also correlates
The relationships between a, K0, and Ki vs. the ratio of
directly with the aromatic/naphthene ratio. Finally, gas oils
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

aromatics/naphthenes as represented in Figures 1-3 appear to whose composition tends to produce more coke also give loxver
be independent of the level of the paraffin concentrations in
the charge stocks. The data in these figures cover reasonably cracking and gasoline formation rates and higher catalyst
wide ranges of paraffin concentrations. decay rates.
Since catalyst decay in catalytic cracking is considered to
Literature Cited
be intimately associated with coke formation, it is interesting
Downloaded via CSIC on September 18, 2024 at 09:12:00 (UTC).

to examine the relationship between the catalyst decay con- Akhmetshina, . N., Levinter, . E., Tanatarou, M. A., Mócha-
lo v, Y. D., Neftepererab. Neftekhim. (Moscow), 1969 (8), 10.
stant, a. and catalyst coke. Figures 4-6 are plots of a, K0, Crocoll, J. F., Jaquay, R. D., Petro/Chem Eng., C-24, November
and Ki vs. catalyst coke and show a increases and K0 and Ki 1960.
Fitzgerald, . E., Moirano, J. L., Morgan, H., Cirillo, V. A., App.
decrease with increased catalyst coke. Spectrosc., 24, 106 (1970).
The relationship of the coke deposited to the decay constant Nace, D. M., Voltz, S. E., Weekman, Jr., V. W., Ind. Eng. Chem.
Process Des. Develop., 10, 530 (1971).
is to be expected since coke is the major contributor to catalyst
Reif, . E., Kress, R. F., Smith, J. S., Petrol. Refiner, 40 (5),
decay. The decay constant, a, however, reflects all causes of 237 (1961).
decay such as the adsorption of aromatics and nitrogen com- Service, W. J., Petro/Chem. Eng., C-32, November 1960.
Shnaider, G. S., Mukhin, I. I., Chueva, . A., Kogan, Y. S.,
pounds as well as that coke which remains after stripping. Khim. Tekhnol. Topi. Mosel, 1969 (1), 10.
From Figure 5 and 6, it is clear that those charge stocks which Voorhies, A., Ind Eng. Chem., 37, 318 (1945).
produce more coke also have a decreased cracking and gasoline Weekman, V. W., Ind. Eng. Chem. Process Des. Develop., 7, 90
(1968).
formation tendency. Interestingly, the slopes of K0 and Ki Weekman, V. W., ibid., 8, 385 (1969).
with coke are approximately equal indicating the initial Weekman, V. W., Nace, D. M., AIChE J., 16, 397 (1970).
selectivity (Ki/Kf) is largely independent of coke make. White, P. J., Preprint No. 24-68, API Division of Refining, 33rd
Mid-Year Meeting, Philadelphia, Pa., May 1968.
Average values of the nonstrippable catalyst coke in experi-
ments with catalyst residence times of 5.0 minutes were used Received for review October 29, 1970
in these plots (Nace etal., 1971). Accepted May 3, 1971

Testsfor Transport Limitations in Experimental

Catalytic Reactors
David E. Mears
Union Oil Co. of California, Union Research Center, Brea, Calif. 92621

In experimental studies of heterogeneously catalyzed in an experimental reactor reflect only chemical events,
reactions, one of the first objectives should be to determine gradients must be virtually eliminated from three domains:
whether the investigation is concerned with catalytic kinetics Intraparticle within individual catalyst particles
or with interactions between the kinetics and transport Interphase between the external surface of the particles
phenomena. The intrusion of temperature and concentration and fluid adjacent to them
gradients can lead to severe deviations in a catalyst’s per- Interparticle between the local fluid regions or catalyst
formance, in many cases completely disguising the true particles
kinetics of the reaction. To ensure that kinetic data obtained The latter domain equally w7ell could be called intrareactor

Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971 541
since it applies to gradients within the reactor as a whole. where <RS and (R'cs are, respectively, the rate expression and
While the intraparticle domain has been extensively analyzed its first derivative with respect to reactant concentration
for several decades, the latter two domains only recently evaluated at the external surface. A numerical value of 0.75
have received much attention. appearing on the right side of the equation w'as rounded to 1
Criteria are now available to estimate whether transport in this derivation. This criterion can be extended to negative-
effects in any domain are significantly altering the experi- order reactions by taking the absolute value of the derivative.
mental results. In deriving such criteria, perturbation analysis For simple power-law kinetics (CRS kCsn), the perturba-
=

has proved particularly useful for detecting when deviations tion criterion becomes:
of the rate start to become appreciable. The method involves
expanding the reaction rate expression in Taylor series in (RV 1_
K (3)
concentration, temperature, or both. Criteria involving CsDe \n
various dimensionless Damkohler numbers (Damkohler, where is the reaction order. A nearly identical result was
n
1936) are then obtained which will hold deviations to an
recently obtained by Stewart and Villadsen (1969) as an
acceptable tolerance. This paper compares perturbation illustration of their orthogonal collocation procedure, which
criteria with those derived by other approaches, and evaluates is also applicable to other geometries and kinetics. Equations
means of minimizing transport limitations in experimental 2 and 3 fail for zero-order kinetics because the concentration
reactors. deviations are beyond the scope of the perturbation approach.
For this case, the value of 6 on the right obtained by Weisz
Intraparticle Transport
is recommended.
Intraparticle transport has been analyzed for a wide For reactions involving a change in the number of moles,
variety of reaction kinetics, particle geometries, and thermal the effectiveness factor relationships are shifted as a function
behavior, as well-reviewed elsewhere (Petersen, 1965b; Satter- of the transport mechanism. When the mode of transport
field and Sherwood, 1963). Generally the objective has been is Knudsen diffusion, Otani et al. (1964) showed that a
to calculate an effectiveness factor, , defined as the ratio
pressure gradient develops in the catalyst, but the shift
of the actual rate to that which would occur if the temperature in effectiveness factors is generally negligible. When molecular
and concentration were constant throughout the catalyst diffusion prevails, a decrease in moles enhances the transport
particle. The solutions show' that becomes inversely pro-
while an increase hinders it. Weekman and Gorring (1965)
portional to the characteristic dimension of the particle in showed that the influence of a volume expansion or contrac-
the regime of strong diffusion influence. Unfortunately, tion can be expressed in terms of the volume-change modulus:
direct application of such solutions requires knowledge of the
true kinetic behavior and the intrinsic rate constant. =
(v -

1)Y, (4)
Weisz and Prater (1954) presented a criterion for the
absence of significant diffusion effects with irreversible in w'hich Ys is the mole fraction of reactant at the catalyst
reactions w'hich is independent of the rate constant. It is surface and v is the stoichiometric coefficient in the reaction
assumed that Pick’s first law governs diffusion in the porous A -* vB. Following Kubota et al. (1969), the criterion for an
media, that the effective diffusivity, De, remains independent isothermal particle can be modified to:
of the nature of the reaction, and that the intrinsic catalytic
activity is distributed uniformly throughout. To ensure > (RV 1
(5)
0.95 in an isothermal spherical particle with a first-order CsDe (1 + )
reaction, the criterion requires: In dilute cases (7S 1), the volume-change effect becomes
(R V <
negligible.
1 (1) Anderson (1963) applied the perturbation approach to
CsDe
derive a criterion for the lack of importance of temperature
where (R is the observed reaction rate per unit particle volume,
gradients in catalyst particles. The reaction is assumed to
Cs is the reactant concentration at the external surface of the follow' an Arrhenius temperature dependence. For quasi-
particle, and rv is the radius of the particle. The dimensionless isothermal behavior, the observed rate (R must not differ
group, called Damkohler’s number for diffusive transport, from the rate that w'ould prevail at constant temperature
expresses the ratio of chemical reaction rate to diffusive by more than an acceptable amount, say 5%. The resulting
flux. The merit of the criterion is that it is stated in terms of criterion, in terms of Damkohler’s number involving heat
observables, and a parameter, De, which is independently transport by conduction, is
measurable. It also applies to transport of a gas in liquid-filled
pores if the effective diffusivity is correctly evaluated (Satter- \(R V <
TSR
(6)
field et al., 1968). Weisz (1957) later showed by means of a \ts ~W
linear approximation of the concentration gradient at the
surface that the numerical value on the right is conservatively where | \
is the absolute value of the heat of reaction, X
0.3 for second-order reactions and 6 for zero order. The is the thermal conductivity of the particle, E is the true
corresponding value for first-order reactions is 0.6 by this activation energy, R is the gas constant, and Ts is the absolute
approximation. temperature at the catalyst surface. Again a numerical
When strong inhibition by the reaction product occurs, value of 0.75 on the right has been rounded to 1. This criterion
Austin and Walker (1963) and Petersen (1965a) demon- is valid whether diffusional limitations exist in the particle
strated that the Weisz-Prater criterion can fail. Hudgins or not.
(1968), using a perturbation method, derived a design Kubota and Yamanaka (1969), extending the work of
criterion to ensure > 0.95: Akehata et al. (1961), derived a criterion involving pertur-
bations in both temperature and concentration. The criterion,
(R
——
V ^<-—
(Rs
(2) expressed here in our nomenclature, ensures > 0.95 pro-
CsDe Cs(R'cs vided that:

542 Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971
 Diagnostic tests for detecting heat and mass-transfer effects in experimental catalytic reactors are reviewed
and updated. New perturbation criteria are presented for intraparticle, interphase, and inferparticle trans-
port in both single- and mixed-phase flow in fixed beds. Emphasis is placed on the proper choice of reactor
parameters to minimize deviations from isothermal plug-flow performance.

/(R% (~AH)(R'„V'2 of Equation 9, which also covers diverse kinetics, appears

_

(7) more generally applicable.

\ De X /
A criterion for isothermal operation of a catalyst particle
in which (R'r,i ar.d 6V,S are, respectively, the derivatives of the is now apparent from inspection of Equation 9:
rate expressions with respect to concentration and tempera-
ture, evaluated at the surface. While applicable to both endo- ]7ß| < 0.05 n (13)
thermic. and exothermic reactions, the criterion becomes Petersen (1965b), by means of an asymptotic solution, derived
imaginary for rare exothermic cases in which the second a similar criterion with a value of 0.3 on the right for first-
term in the brackets becomes larger than the first. order reactions.
The criterion can be extended to these cases for which For most cases encountered in practice, Akehata et al.
> 1 by squaring both sides. For power-law kinetics (except
7?
(1961) note that the term in Equation 7 accounting for the
zero order), = 1 ± 0.05 if:
intraparticle temperature effect amounts to less than 10%
nffbV of the term for diffusion. Consequently the effect of tempera-
(-AIJ) <&£,,- JEj
'"'

< ( } ture distribution within a catalyst particle is usually much

\ts
^

r i). rts i less important than the internal diffusion problem. Transport
or in terms of the observed rate and dimensionless parameters: limitations occurring outside the catalyst particle must
now be considered.
<5irv2 1
< (9)
C,De |n
-

yß Interphase Transport
in which: For vapor-phase systems, the greater part of the resistance
to heat transfer is often in the boundary layer or “film”
around the particle, rather than within (Carberry, 1966).
7 (10)
This happens because the effective thermal conductivity
of the solid is usually much larger than the conductivity
expresses the sensitivity of the reaction rate to temperature in the gas phase. In the case of a highly exothermic reaction,
and:
the particle temperature can become considerably greater
(-AII)DeC. than that of the bulk stream.
ß =
(11) A criterion for detecting the onset of a heat transport
XT,
limitation in the film has been derived by the author (Mears,
represents the maximum temperature difference that can 1971b) with the perturbation approach. The heat transfer
exist in the particle relative to the surface temperature, resistance of the film is assumed to be lumped at the surface.
(T —

Ts)/Ts. Typical values are 10-40 for the former, If the observed rate is to deviate by less than 5%, the cri-
and —0.1-0.1 for the latter parameter (McGreavy and terion requires:
Cresswell, 1969). The last criterion can 'also readily be
derived by the orthogonal collocation procedure of Stewart AH Sir
<0.15 RJ\
—

and XTlladsen (1969). !x (14)

E
The special case for which ?i 7/31 is close to or equal to
—

where Tb is the temperature of the bulk fluid, and h is the

zero requires further consideration. When this occurs, the
heat transfer coefficient given by correlations (Gliddon and
heat effect compensates for the diffusion effect so that the
rate stays nearly constant until the concentration within Cranfield, 1970; Littman et al., 1969). This dimensionless
the catalyst has dropped to less than 80% of the surface
criterion is valid whether transport limitations exist in the
value. Such deviations exceed the range of applicability particle or not. It show7s that interphase heat transport
of the perturbation approach, so this peculiar case is better limitations can be expected when reaction heats and rates
are high, and flow rates low (small h).
described by the asymptotic analysis of Petersen (1965b).
The latter method gives a value of 13 for the right side of Note the similarity in form to Equation 6, with h replacing
Equation 9 when n and yß both equal 1.0. X/rP. Comparison reveals that the interphase heat transfer
resistance becomes limiting before the intraparticle resistance
It is interesting to compare the criterion of Equation 9
as long as:
with the more widely known one of Weisz and Hicks (1962)
for exothermic first-order reactions: Bip =
hdp/X < 10 (15)

(hr,,2 7ß in which the Biot number expresses the ratio of thermal

exp < 1 (12)
cj)e + ß
resistance of the particle to that of the film. Condition 15
is usually met in laboratory reactors because of the low
This relationship was derived by analogy to the isothermal flow rates.
case, with a conservative simplification of the differential For isothermal cases with first-order kinetics, Carberry
equation. The criterion safely predicts positive deviations % (1961) showed that the effect of bulk mass transport across
> 1.05) for exothermic reactions, but fails to predict nega- the film can be assumed negligible relative to surface-reaction
tive deviations for endothermic reactions. Thus the criterion kinetics provided that:

Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971 543
 r\k¡kca <0.1 (16) This equation is applicable to both endothermic and exo-
thermic reactions, predicting the occurrence of positive devia-
in which kc is the gas-particle mass transfer coefficient, and
tions for the latter.
a is the external surface-to-volume ratio of the particle.
This expression is now readily generalized to other reaction
orders by the perturbation approach, allowing a 5% deviation Interparticle Transport
as before. The criterion can be expressed in terms of another Transport effects, occurring both radially and axially,
Damkohler number: within the reactor as a whole are particularly difficult to
evaluate and control. If neglected, radial temperature
=
&rv/Cbkc < 0.15/ra (17)
gradients can lead to reaction rates several thousandfold
where Cb is the bulk fluid concentration. Comparison with greater at the axis than at the wall (Denbigh, 1965). Less
Equation 14 for typical cases shows that heat transfer frequently, bypassing or axial dispersion of mass may seriously
across the film causes excessive deviations long before mass distort the reactor performance.
transfer becomes limiting. The criterion also shows that One common approach to minimizing these effects is to
zero-order reactions are not affected by interphase transport. adopt the differential reactor, in which low conversions are
One case in which severe interphase mass transport limita- taken. But high precision of chemical analysis is required,
tions might be expected is the trickle-flow reactor, in which a and interphase gradients are not always eliminated. The
liquid reactant trickles down the catalyst pellets while a recycle reactor provides a means of operating differentially
reacting gas flows either in the same or opposite direction. while maintaining a finite overall conversion. Flow rates
This type of reactor is extensively used in the petroleum past the catalyst can then be independently varied to elim-
industry for reactions such as hydrotreating and hydro- inate interphase and interparticle gradients. The well-stirred
cracking of heavy oils. heterogeneous catalytic reactor, in which pellets are suspended
Satterfield et al. (1969) examined the transfer of hydrogen in wire cages attached to a rotating shaft, offers another
across the liquid “film” for a first-order hydrogenation reac- alternative. The merits and demerits of these reactors have
tion. In close parallel to the last criterion, they concluded been ably discussed in recent reviews (Carberry, 1969;
that mass transfer from the gas phase to the outer surface Smith, 1968).
of the pellets will not be a significant resistance provided: Integral reactors are sometimes necessary, however, for
both practical and theoretical reasons. For complex reaction
(ñrp/C*kLS <0.15 (18)
networks, it may prove difficult to prepare synthetic feeds
where kLs, the overall liquid side mass transfer coefficient which match the composition of the actual fluid at various
can be estimated by procedures outlined in their paper, stages of conversion. High conversion data are usually prefer-
and C* is the equilibrium concentration of hydrogen in the able in statistical discrimination between rate expressions of
bulk liquid. the Langmuir-Hinshelwood type (Barnard and Mitchell,
An important conclusion was that mass transport through 1968; Mezaki and Happel, 1969). Barnard and Mitchell
the liquid film cannot be a significant limitation unless the (1968) contend that more accurate results can be obtained
catalyst particles themselves are operating at a very low by measuring integral data over the whole of the conversion
effectiveness factor. This results because the effective diffu- vs. reciprocal flow rate graph than in an analysis using only
a part of the graph. Moreover, integral data are more selective
sivity within the catalyst does not usually exceed a value of
about Vs of that in the boundary layer or “film” adjacent because a prospective rate equation is required to satisfy
to the pellet, and the film thickness is always substantially two conditions: effectively, rate vs. contact time and rate
less than the catalyst radius. The same conclusion applies, vs. initial concentration. Differential data have only to fit

of course, to the vapor-phase case discussed previously. the observed variation of rate with initial concentration.
It is now obvious that a criterion can be derived for the But to obtain reliable data from integral reactors, the investi-
combined intraparticle-interphase case by the perturbation gator must take extra precautions to ensure that heat and
mass transfer effects do not interfere.
approach. In effect, one merely replaces Dirichlet boundary
conditions: In the radial direction, this means enhancing both heat
and mass transport. Dilution of the catalyst with inert
T =
Ts, C =
Cs at r =
rp (19) solids is often useful because it effectively decreases the ratio
of catalyst volume to wall area for heat transfer. Caldwell
with mixed boundary conditions:
and Calderbank (1969) described strategies for optimizing
dT reactor performance by varying the dilution ratio with axial
h(T, -

Tb) distance. In experimental reactors, our objective is to elim-

dr s
at r =
rp (20) inate the usual rise of the centerline temperature to a “hot
dC spot.” Their analysis, based on a one-dimensional model,
'e j
dr
kc(Cb
-

cy shows that a zero temperature gradient can be imposed

s

along the reactor length by using a dilution ratio which

Excluding n —

y^b[ close to zero, the resulting criterion for decreases linearly with conversion. Since this distributes the
tj = 1 ± 0.05 becomes: reaction rate uniformly down the bed, the dilution ratio
also decreases linearly with reactor length. This approach
1 + 0.33
also applies fairly well to practical cases of many parallel
CbDe \n
—

76^61 (1 ~\~ 0.33 ) reactions provided that the kinetics can be lumped (Hutchin-
son and Luss, 1970). In an actual three-dimensional reactor,
where yb and ß„ are both evaluated at bulk fluid, rather than
surface conditions. The analogous criterion to Equation 13
radial variations will still occur.
An approximate criterion for evaluating the importance
for isothermal operation is then:
of radial heat transport limitations has been derived by the
[ + 0.3 nybx\ < 0.05 n (22) author (Mears, 1971b). It applies at any cross section in reac-

544 Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971
tors with linearly decreasing dilution, and at the “hot spot” Dilution to maintain isothermal operation is not without
in undiluted or uniformly diluted beds. At these locations, the hazard, because bypassing effects can influence the conversion
axial temperature gradient equals zero so the sensible heat at large dilution ratios. For cases with uniform dilution,
term can be eliminated from the differential energy balance. van den Bleek et al. (1969) developed a stochastic model
Axial conduction can also be neglected for beds with length- quantitatively describing this influence and obtained a
to-particle diameter ratios sufficiently great that plug flow is criterion to determine the allowable degree of dilution.
approached (generally L/dp > 30). Numerical solutions (Car- Their criterion ensures that the bypassing effect will be an
berry and White, 1969; Valstar et al., 1969) of the conser- order-of-magnitude less than , the relative experimental
vation equations suggest that radial mass transport effects are error (percent) in the conversion. Rearranged and expressed
usually negligible in comparison to heat transport; thus the in our nomenclature it becomes:
concentration can be assumed radially constant. These sim-
L b
plifications permitted an analytic solution (Chambre and > 250 (27)
Grossman, 1955) to the radial temperature distribution. dP
When the heat transfer resistance at the wall is negligible where L is here the bed length including diluent. The criterion
(i.e., R0/rp > 100), analysis (Mears, 1971b) for a 5% deviation can be used in calculating the minimum bed length when the
yields the criterion: dilution ratio is fixed by requirements of bed isothermality.
7 (R6ñ02 RTw In contrast to the radial case, mass transfer in the axial
< °'4 (23) direction must be minimized to prevent significant deviations
keTw E
from plug flow and lower conversions. Petersen (1965),
where ke is the effective thermal conductivity of the bed, as applying the asymptotic solution approach to the case of
given in correlations such as Agnew and Potter (1970) and isothermal, first-order reactions, derived an approximate
Yagi and Kunii (1959) for vapor-phase systems and Week- criterion for freedom from axial dispersion effects:
man and Meyers (1965) for trickle-flow, and (ft6 is the average
reaction rate per unit bed volume at a given bed depth. In a =
VkbDa/v < 1 (28)
integral reactors the latter can usually be evaluated from where kbis the rate constant per unit catalyst bulk volume,
the observed rate constant and the concentration at the
Da is the axial eddy diffusivity, and v is the superficial fluid
cross section of interest.
velocity. In agreement with earlier findings (Wehner and
Note that the functional form of this criterion is similar
Wilhelm, 1956), it shows that the dispersion effect is generally
to Equation 6 with the reactor radius replacing the particle
negligible except for cases involving short beds and high
radius and the thermal conductivity of the bed replacing
conversions. But the problem can be much more severe in
that of the particle. Since R„ >>> rp and ke approaches the
trickle flow reactors because of the higher eddy diffusivities.
order of magnitude of at low Reynolds numbers, the
The author (Mears, 1971a), utilizing perturbation solutions
interparticle transport problem is usually much more severe for power-law kinetics (Burghardt and Zaleski, 1968),
than the intraparticle one.
derived a design criterion which will hold the deviation
The local rate per unit bed volume is related to (ft, the
in the required reactor length to less than 5%. Expressed in
average rate per unit particle volume, by the expression: the same manner as the last criterion, it requires for the
(i -

0 first-order case:
(ftfi (ft (24)
(1 + b) « < 0.22 (29)
in which e is the bed void fraction, and b is the ratio of diluent Thus the new criterion is about four and a half times more
to catalyst volume. conservative than the earlier one. This results because
Comparison of Equation 23 with the interphase criterion, asymptotic criteria actually indicate the transition region
Equation 14, shows that the interparticle resistance becomes between the kinetic-controlled and transport-influenced
limiting first unless: regimes (Petersen, 1968), rather than the point at which
transport effects are just starting to become significant.
(1
~

<0 /hdp\ ñ0
To minimize axial dispersion effects, the practical expedient
< °- (25)
(1 + b) \ke } rp available to the experimenter is to lengthen the bed, which
This condition is achieved only for low values of the radial lowers the axial concentration gradient and, hence, the driving
force for dispersion. Consequently the most convenient form
aspect ratio (R0/rp) or high dilution ratio. Thus the magni-
tudes of the heat transport resistances in experimental reactors of the criterion is one giving the minimum reactor length
are generally in the order: interparticle > interphase > intra-
for freedom from significant dispersion effect. Expressed as
the ratio to particle diameter, the new criterion becomes:
particle.
When the heat transfer resistance at the wall is significant, L kbr 20 C0
the interparticle criterion becomes: t- > 20 —

=
—

In -

(30)
dp Pea Pea Cf
! \(5{^ < 0.4 RTU,/E
in which is the space time and C0 and C¡ are the initial
keTw [1 + 8(?·„/%>)Bito] and effluent concentrations. The axial Peclet number based
in which the wall Biot number is a measure of the ratio of the on particle diameter:
thermal resistance of the bed to that of the wall. The data vdp
of Yagi and Kunii (1959) yield wall Biot numbers of 0.8 to 2 Pe (31)
Da
for Re < 100 and 0.05 < rp/R„ < 0.2, so the denominator
becomes of the order of 2. Thus heat transfer resistance at the is a measure of the ratio of convective to dispersive transport.
wall aggravates the interparticle transport problem in small The advantage of the second equality in Equation 30 is that
laboratory reactors. it puts the criterion in terms of the observed conversion and

Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971 545
the independently obtainable Peclet number. Given in numbers as low as one. However, the interparticle heat trans-
terms of reactor length, the new criterion is 20 times more fer coefficient (Yagi and Kunii, 1959) becomes quite weakly
conservative than the asymptotic criterion. dependent on G at low Reynolds numbers. Thus, as Chambers
For nonfirst-order reactions obeying simple power-law and Boudart noted, this diagnostic test requires very precise
kinetics, the corresponding criterion is: conversion data or an extended range of flow rates.

L 20 n C0
> (32) Design Criteria for Fixed-Bed Reactors
d„ Pe¡"ln Ct In designing a laboratory reactor, transport limitations
where nis the order of the reaction. This general expression can be minimized through proper choice of the parameters.
gives Equation 30 for first-order reactions and zero for zero- Because of the large Arrhenius effect of temperature on the
order reactions, which are not affected by axial dispersion. reaction rate, maintaining isothermal operation is by far the
Note that the axial dispersion problem becomes more severe most critical consideration. The criteria reviewed provide a
with increasing conversion or reaction order, or with de- means for assessing the best ways to achieve this.

creasing Peclet numbers. Reducing the reactor radius, an obvious choice, is partic-
Extensive studies have shown that the Peclet number is ularly important since the interparticle criterion depends on it
generally close to 2 for gases at Reynolds numbers greater both explicitly and implicitly. The latter dependence occurs
than 2 and to about 0.5 for liquids when Reynolds numbers because the interparticle heat transfer coefficient becomes
are less than 10 (Levenspiel and Bischoff, 1963; Miller and effectively proportional to the mass velocity, G, at Re > 100,
King, 1966). In trickle-flow, still smaller values are found and G is inversely proportional to R02 (at constant mass
in the liquid phase at low Reynolds numbers (Hochman and flow). The interparticle heat transfer criterion then becomes
Effron, 1969; van Swaaij et al., 1969). For example, at a effectively proportional to the fourth power of R0 and the
Reynolds number of 4, typical of bench-scale reactor condi- interphase criterion to the 1.2 power at higher Reynolds
tions, these correlations yield Peclet numbers of the order of numbers. Hence decreasing the reactor radius is a vital step
0.1. Consequently, the criterion shows that the minimum in minimizing both transport resistances.
reactor length for freedom from backmixing effects may be an The question then arises as to how far the radial aspect
order-of-magnitude greater in trickle-flow than in vapor- ratio (R0/rp) may be reduced before wall effects become
serious. Because the packing is more open near the wall, a
phase operation..
larger fluid velocity and lower catalyst bulk density occur
Empirical Tests for Transport Limitation there. As a result the conversion is lower near the wall,
While the analytical criteria are useful, estimation of causing an apparent lower activity for the bed.
Kondelik and Boyarinov (1969) calculated the effect of
parameters involved can be difficult at times. For such cases
a well-established empirical criterion is to run the reaction on
radial void fraction distribution on reactor efficiency as a
function of R0/rP. Their model divides the reactor into
progressively smaller particle sizes, Usually produced by
annular channels based on statistical distribution data and
crushing. If the reaction rate per unit particle volume changes
utilizes the Kozenv equation (Carman, 1956) to compute the
with diameter, strong transport influence is indicated.
volumetric flow rates in individual channels. Radial mass
Graphical (Weisz and Prater, 1954) and analytical (Gupta transfer is accounted for by a lumped radial mass transfer
and Douglas, 1967) techniques are available for estimating
coefficient D. In the R0/rp range of 5-10, the calculations
the effectiveness factors in this case. When the reaction rate
show a minimum in the effective bed activity whose depth
remains independent of the particle size over an appreciable
increases with conversion. But the minimum is reduced with
range, the reaction is considered free from intraparticle
improved radial transfer, and the maximum value assumed
gradients on the macroscopic scale.
for D still appears to underestimate the contribution from
An alternate empirical approach, suggested by Koros and
Nowak (1967), is based on the fact that the rate of reaction eddy diffusion. Consequently, the gain in isothermality
in the kinetic regime is proportional to the number of sites through reducing the reactor radius appears to more than
compensate for the slight channeling loss to R0/Rp ratios as
per unit volume (S), but proportional to S to the 0.5 power
low as 4. However, wall effects may be more serious in trickle-
in the internal diffusion regime. The test involves pelletizing
mixtures of finely divided catalyst with an inert powder of flow, so larger ratios should be used for this operation.
Finally, in an existing reactor of larger diameter, dilution
comparable diffusional characteristics. If the fraction of
of the catalyst bed with inert particles can be utilized to
catalyst in the mixed pellet is /, the ratio of rates must also reduce heat generation per unit volume. Linearly decreasing
be equal to f in the kinetic regime. This method avoids
dilution ratios are strongly recommended for integral reactors.
distortions in the flow field which changing pellet sizes may
But in diluting a bed of fixed length, it should be remembered
cause.
that the mass velocity is reduced proportionately at constant
A common empirical test for the possible influence of
external heat and mass transfer limitations consists of space velocity. The interparticle criterion shows that dilution
will be advantageous in minimizing radial gradients only if
checking the effect of flow rate on conversion at constant the reactor is operating at Reynolds numbers sufficiently low
space velocity. If no such effect is found, it is concluded that
that the effective thermal conductivity is relatively insensitive
transfer of heat and mass to the particles does not influence
to the mass velocity (i.e., Re < 100).
the reaction rate. Chambers and Boudart (1968) warned
that this diagnostic test may fail because of its lack of sensi-
Conclusion
tivity at Reynolds numbers of the order of 10. This observa-
tion was based on a reanalysis of the interphase heat transfer The elimination of transport influences in studies of cata-
data of De Acetis and Thodos (1960). More recent data lytic kinetics is a complex problem. The criteria presented
(Gliddon and Cranfield, 1970; Liftman et al., 1969; Petrovic provide tools for evaluating and minimizing the various
and Thodos, 1968) indicate that both h and kc remain propor- transport resistances. Like any diagnostic test, each must be
tional to between the 0.95 and 0.64 powers of G to Reynolds applied with due caution and understanding. If carefully used,

546 Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971
the criteria can help the experimenter to achieve a more X =
thermal conductivity of particle, g-cal/sec-cm-°C
rigorously defined experimental system. µ =
viscosity, g/sec-cm
v =
stoichiometric coefficient
Nomenclature p =
density, g/cm3
=
space time inverse space velocity, sec
=

a =
superficial (outside) surface area of catalyst particle x =
Damkohler number for interphase heat transport, =

per unit particle volume, cm2/cm3 (—AH) (Rrp/hTb

b = dilution ratio, cm3 inert solid/cm3 catalyst =
Damkohler number for interphase mass transport, =

Bip = thermal Biot number for catalyst particle, hdp/X =

(Rrp/kcCb
Bi„ = thermal Biot number at the reactor wall, hwd,/k„ =

C* =
equilibrium concentration of hydrogen in liquid, Literature Cited
g-mol/cms
Cb —
concentration of reactant in bulk fluid, g-mol/cm3 Agnew, J. B., Potter, O. E., Trans. Inst. Chem. Eng. (London)
effluent concentration of reactant, g-mol/cm3 48, T15 (1970).
Cf =

Akehata, T., Namkoong, S., Kubota, H., Shindo, M., Can. J.

C0 = initial concentration of reactant, g-mol/cm3 Chem. Eng., 39, 127 (1961).
C, = concentration of reactant at outer surface of particle, Anderson, J. B., Chem. Eng. Sci., 18, 147 (1963).
g-mol/cm3 Austin, L. G., Walker, P. L., Jr., AIChE J., 9, 303 (1963).
dp = diameter of catalyst particle, cm Barnard, J. A., Mitchell, D., J. Catal., 12, 376 (1968).
axial eddy diffusivity, cm2/sec Burghardt, A., Zaleski, T., Chem. Eng. Sci., 23, 575 (1968).
Da =
Caldwell, A. D., Calderbank, P. H., Brit. Chem. Eng., 14, 1199
De = effective diffusivity in a porous catalyst, cm2/sec (1969).
E = intrinsic activation energy for chemical reaction, Carberry, J. J., AIChE J., 7, 351 (1961).
g-cal/g-mol Carberry, J. J., Catal. Rev., 3 (1), 61 (1969).
fraction of active catalyst in mixed pellet Carberry, J. J., Ind. Eng. Chem. 58 (10), 40 (1966).
f =
Carberry, J. J., White, D., ibid., 61, (7), 27 (1969).
G = mass velocity, g/sec-em2 of total or superficial bed Carman, P. C., “Flow of Gases Through Porous Media,” Aca-
cross section demic Press, New York, 1956, p 12.
h —
heat transfer coefficient between gas and particle, Chambers, R. P., Boudart, M., J. Catal., 6, 141 (1968).
Chambre, P. L,, Grossman, L. M., Appl. Sci. Res., A5, 245
g-cal/sec-cm2-°C (1955).
hw = heat transfer coefficient at the reactor wall, g-cal/ Damkohler, G., Z. Elcktrochem., 42, 846 (1936).
sec-cm2' °C De Acetis, J., Thodos, G., Ind. Eng. Chem., 52, 1003 (1960).
AH = heat of chemical reaction, g-cal/g-mol Denbigh, K. G., “Chemical Reactor Theory,” Cambridge Univ.
Press, Cambridge, England, 1965, p. 44.
k = intrinsic rate constant per unit particle volume, sec-1 Gliddon, B. J., Cranfield, R. R., Brit. Chem. Eng., 190, 481 (1970).
for first-order reaction Gupta, V. P., Douglas, W., Can. J. Chem. Eng., 45, 117 (1967).
kb =
apparent rate constant per unit bulk catalyst volume, Hochman, J. M., Effron, E., Ind. Eng. Chem., Fundam., 8, 63
=
k(l —

e)77, see-1 (1969).

mass transfer coefficient between gas and particle, Hudgins, R. R., Chem. Eng. Sci., 23, 93 (1968).
ke =
Hutchinson, P., Luss, D., Chem. Eng. J., 1, 129 (1970).
cm/sec Kondelik, P., Boyarinov, A. I., Collect. Czech. Chem. Commun.,
ke —
effective thermal conductivity across reactor bed, 34, 3852 (1969).
g-cal/sec-cm °C - Koros, R. M., Nowak, E. J., Chem. Eng. Sci., 22, 470 (1967).
kcs = overall mass transfer coefficient through liquid film, Kubota, H., Yamanaka, Y., J. Chem. Eng. (Japan), 2, 238 (1969).
Kubota, H., Yamanaka, Y., Dalla Lana, I. G., ibid., 2, 71 (l969).
cm/sec Levenspiel, O., Bischoff, K., Advan. Chem. Eng., 4, 95 (1963).
L —
reactor bed length, cm Littman, H., Barile, R., Pulsifer, A., Ind. Eng. Chem., Fundam.
L„ = reactor bed length with undiluted catalyst, cm 7, 554 (1969).
McGreavy, C., Cresswell, D. L., Can. J. Chem. Eng., 47, 583
=
integer exponent in power law rate expression (1969).
Pe„ = axial Peclet number, dpv/Da =
Mears, D. E., Chem. Eng. Sci., in press (1971a).
rp =
particle radius, cm Mears, D. E., J. Catal, 20, 127 (1971b).
R =
gas constant, g-cal/g-mol °K -
Mezaki, R., Happel, J., Catal. Rev., 3, 241 (1969).
Re Miller, S., King, C. J., AIChE J., 12, 767 (1966).
=
Reynolds number, vpdp/µ=

Otani, S., Wakao, N., Smith, J. M., ibid., 10, 130 (1964).
Ro = radius of tubular reactor, cm Petersen, E. E., Chem. Eng. Sci., 20, 587 (1965a).
(R = observed reaction rate per unit particle volume, Petersen, E. E., ibid., 23, 94 (1968).
g-mol/sec-cm3 Petersen, E. E., “Chemical Reaction Analysis,” Prentice-Hall,
reaction rate per unit bed volume, g-mol/sec cm3 Englewood Cliffs, N. J., 1965b, pp. 76, 79, 198.
(J\b
Petrovic, L. J., Thodos, G., Ind. Eng. Chem., Fundam., 7, 274
— -

(A'c =
derivative of reaction rate expression with respect to (1968).
concentration, sec-1 Satterfield, C. N., Sherwood, T. K., “The Role of Diffusion in
fft'i = derivative of reaction rate expression with respect to Catalysis,” Addison-Wesley, Reading, Mass., 1963.
0 Satterfield, C. N., Ma, Y. H., Sherwood, T. K., Inst. Chem.
temperature, g-mol/sec cm3 K · ·

Eng. (London), Symp. Ser., 28, 22 (1968).

5 =
active sites per unit particle volume, cm-3 Satterfield, C. N., Pelossof, A. A., Sherwood, T. K., AIChE J.,
S.V. =
space velocity, v/Lu, sec-1
=
15,227(1969).
t =
time, sec Smith, J. M., Chem. Eng. Progr., 64, 78 (1968).
T =
absolute temperature, °K Stewart, W. E., Villadsen, J. V., AIChE J., 15, 28 (1969).
Vals tar, J. M., Bik, J. D., van den Berg, P. J., Trans. Inst. Chem.
v =
superficial velocity, cm/sec Eng. (London), 47, CE 136 (1969).
Y s
= mole fraction of reactant at catalyst surface van den Bleek, C. M., van der Wiele, K., van den Berg, P. J.,
z =
distance in axial direction, cm Chem. Eng. Sci., 24, 681 (1969).
van Swaaij, W., Charpentier, J. C., and Villermaux, J., Chem.
Greek Letters Eng. Sci. 24, 1083 (1969).
Weekman, V. W., Jr., Gorring, R. L., J. Catal, 4, 260 (1965).
dimensionless axial dispersion number, Weekman, V. W., Jr., Meyers, J. H., AIChE J., 11, 13 (1965).
a =
x/kbDa/v =
Wehner, J. F., Wilhelm, R. H., Chem. Eng. Sci., 6, 89 (1956).
ß =
dimensionless maximum temperature rise, ( AH) = —

Weisz, P. B., Z. Physik Chem., NeueFolge, 11,1 (1957).

DJC./XT. Weisz, P. B., Hicks, J. S., Chem. Eng. Sci., 17, 265 (1962).
7 =
dimensionless activation energy, E/RTS = Weisz, P. B., Prater, C. D., Advan. Catal, 6, 143 (1954).
6 =
relative experimental error in conversion Yagi, S., Kunii, O., AIChE J., 6, 97 (1959).
e = bed void fraction
=
effectiveness factor Received for review November 2, 1970
=
volume change modulus (1 =
¡>)YS
—

Accepted March 22, 1971

Ind. Eng. Chem. Process Des. Develop., Vol. 10, No. 4, 1971 547