430 Ind. Eng. Chem. Fundam.

1902, 21, 438-447

Experimental Criterion for the Absence of Artifacts in the

Measurement of Rates of Heterogeneous Catalytic Reactions

Rostam J. Madon' and Mlchel Boudart'

Department of Chemical Engineering, Stanford University, Stanford, California 94305

An experimental criterion has been developed to assess whether measured catalytic activity is independent of the
influence of transport phenomena. A simple theoretical analysis and experimental verification of the criterion have
been carried out. This test may be used when heterogeneous catalysis is carried out in batch, continuous stirred
tank, and tubular plug-flow reactors, and is best suited to be used with supported metal catalysts. In this case,
the test should be performed with two or more samples with different metal loading but of similar metal dispersion.
Absence of physical artifacts is established by the constancy of turnover numbers. I f the reaction is exothermic,
the test should be repeated at another temperature.

Introduction fusivities of fluids, it is difficult to obtain the effective

It has been stated (Boudart, 1972; Carberry, 1966; diffusivity for different catalysts. Satterfield and Saraf
Satterfield, 1970) that catalytic data should be free from (1965) have shown that diffusional characteristics may
all transport influences to obtain kinetic rate expressions change throughout a pellet due to compression of the
and to yield the correct activity or selectivity of the cat- catalyst powder in a pelletizing machine. Intraparticle
alyst. Though the dangers are known, it should be em- criteria include surface concentrations and surface tem-
phasized that scaling down to laboratory studies is fraught peratures. These are usually not known and bulk values
with problems to eliminate concentration and temperature can be substituted only if there are no interphase gradients.
gradients both within catalyst particles (intraparticle Though the heat transfer criteria can be used to check
gradients) and between the external surface of particles for temperature gradients in the presence of diffusion, the
and surrounding fluid (interphase gradients). latter effect could cause the rate and activation energy to
In the past, a number of interesting and useful criteria be low and the criteria to be obeyed. In the absence of
have been developed to check for the influence of heat and diffusion with the true higher values of the rate and ac-
mass transfer on observed catalytic rates. Most of these tivation energy, the criteria could be disobeyed for the
criteria have been explained in a detailed review by Mears same reaction.
(1971a). Several criteria have been assembled in Table I. Finally, the particle size is another important parameter.
The Weisz-Prater criterion (1954) for intraparticle diffu- Often powders with a wide size range are used in laboratory
sion, for which only directly observable quantities are re- reaction studies, and Aris (1957) showed that the mean
quired, has been criticized for lack of generality by Pet- effectiveness factor is given by
ersen (1965a). The Petersen criterion (Petersen, 1965a), B = CBia[
however, does not contain only observables and is hence i
more difficult to use. The more generalized criterion of Aris has warned against using average values for particle
Bischoff (1967), where the integration of the rate expres- size to calculate 7, and in this light the above criteria could
sion is involved, or that of Hudgins (1968)and Stewart and not be used except by using the largest and hence the most
Villadsen (1969), where the differentiation of the rate conservative value for the radius.
expression is necessary, all suffer from the fact that the Besides using the above correlations, experimental tests
rate expression must be known, be it the power law type are often carried out to check for diffusion. In a flow
or the Langmuir-Hinshelwood type law, and if it is the system, if the space velocity is kept constant and then if
latter then numerical values for all equilibrium constants the conversion remains constant as the flow rate is
must be obtained before the criteria can be used. Hudgins changed, the influence of external mass transfer is said to
(1972) derived a criterion for interphase diffusion, but it be negligible. A similar test is used in slurry reactors where
suffers from the same problem as stated above. The the constancy of the reaction rate is checked as the agi-
criteria to check for interphase and intraparticle temper- tation of the slurry is increased. Chambers and Boudart
ature gradients are due to Mears (1971b) and Anderson (1966) have indicated that at the low particle Reynolds
(1963), respectively. numbers encountered in the laboratory the heat and mass
To use the several criteria, some of the following ex- transfer coefficients are quite insensitive to changes in flow
perimental properties are necessary: effective diffusivity rates, and hence the above diagnostic test may fail to in-
in the pores of the catalyst, the heat and mass transfer dicate that the reaction rates are free from interphase
coefficients at the fluid-solid interface, and the thermal concentration gradients. Moreover, this test cannot give
conductivity of the catalyst. A t the low particle Reynolds any information regarding intraparticle diffusion.
numbers used in laboratory packed bed reactors, the ac- If activity of a catalyst is affected by altering pellet size,
curacy of the value of h obtained from known correlations then this activity is said to be influenced by pore diffusion.
is subject to question (Madon, 1975). Though numerous However, if there is no change in activity with pellet size
studies have been carried out for bulk diffusion and dif- one cannot say for sure that there are negligible intra-
particle concentration gradients. The influence of internal
diffusion in the small pores of a bimodal pore distribution
Corporate Research-Science Laboratories, Exxon Research may still be important, and the test may just indicate that
and Engineering Co., Linden, NJ 07036. there is no diffusional influence in the larger set of pores.
0196-4313/82/1021-0438$01.25/0 @ 1982 American Chemical Society
 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982 439

Table I. Various Criteria Used t o Check for Transport Effects in Heterogeneous Catalysis
For Intraparticle Mass Transfer For Interphase Mass Transfer
Weisz and Prater (1954) Hudgins (1972)
R2r R, 0.15 1
DaII = -G 1 =- <-r(cb)
Cs De Cbk, Cb 1
dr(C)/dC C=Cb 1
Weisz (1957) For Intraparticle Heat Transfer
Dan < 6 zero-order reaction Anderson (1963)
90.6 first-order reaction
G0.3 second-order reaction [ - A H [ R2r RgTs - 0.75
Daw = < 0.75---
Petersen (1965a) AT, E T
For Interphase Heat Transfer
Thiele modulus < fi Mears (1971b)
where [ - A H 1 Rr RgTb - 0.15
X= < 0.15- --
ri(C1, C2, T) hTb E yb
ri(5) = For Combined Intraparticle Heat and Mass Transfer
ri(Cis, C2s3 e.., TI
Hudgins (1968) Mears (1971a)
0.75 1 1 (-AH) dr(C,T)
Dan 4 -r(Cs) -- -( )T=T, <1
De dT
cs 1 dr(C)/dC I c=cs
Stewart and Villadsen (1969) or for power law kinetics
0.77 1 1
Dan < -r(Cs) DaII < -
cs I dr(C)/dC I c=cs In-70)
For Combined Interphase and Intraparticle Heat
Bischoff (1967)
and Mass Transport"
Mears (1971a)
R lr 1+ 0 . 3 3 ~ 7
<
CbDe I
n - 7bPb I(1f 0.33nw)
a Note. This overall criterion provides that the NET distortion of the observed rate will be within 5% of the rate under
isothermal, isoconcentration conditions. There is no guarantee of freedom from individual transport processes.

Table 11. The Koros-Nowak Criterion

[rate of reaction]/[ concentration of active material] = constant
1. [rate of reaction (mol t 3 - I g-l of catalyst)]/[weight of active material per unit weight of catalyst]
= r/f, (mol s'l g - l of active material)
= constant
2. [rate of reaction (mol s - * g-l of catalyst)]/[surface area of active material per unit weight of catalyst]
= r/fa (mol s-l cm-2 of active material)
= constant
3. [rate of reaction (mol s - ' g-l of catalyst)]/[surface moles of active material per unit weight of catalyst]
= turnover number
= r/f,
= constant

Furthermore, changing pellet sizes could cause distortion terion can be successfully applied when kinetic measure-
in the flow field wkich could affect contacting. ments are to be obtained.
To circumvent the problems stated above, we wish to The Koros-Nowak Criterion
develop an experimental criterion which was first suggested The criterion is based on making rate measurements on
by Koros and Nowak (1967) to check if reaction rates were catalyst samples in which the concentration of the cata-
free from mass transfer effects. It will be shown that this lytically active material has been changed. The reaction
criterion can be used as a test for interphase and intra- rate, in the kinetic regime, is directly proportional to the
particle mass and heat transfer either when they occur concentration of the active material (see Theoretical
separately or when the physical events are coupled during Section). The concentration of catalytically active material
the course of a reaction. I t will be shown that when this may be defined as (a) the weight fraction of the active
test is obeyed there is neither poisoning nor poor con- material in the catalyst, (b) the surface area of the active
tacting of the catalyst due to bypassing and channelling. material per unit weight of the catalyst, or (c) the surface
Also, deviations of plug flow behavior in tubular reactors moles of active material per unit weight of the catalyst.
due to transverse and longitudinal temperature and con- The Koros-Nowak criterion can then be explained by
centration gradients (interparticle gradients), backmixing, noting that in the kinetic regime (a) the ratio of measured
and velocity gradients, will cause the criterion to be dis- reaction rates must equal the ratio of the weight fractions
obeyed In all future discussion this test will be referred of the active material in the different catalyst samples, as
to as the Koros-Nowak (KN) criterion. long as the dispersion of the active material does not
In the following sections, the test will be explained, change from one sample to the next, or (b) the turnover
analyzed, and finally demonstrated to show that the cri- number (defined as moles reacted per second per surface
440 Ind. Eng. Chern. Fundarn., Vol. 21, No. 4, 1982

mole of the active material) or the reaction rate referred different kinds of reactors used to carry out chemical re-
to unit surface area of the active material must be invariant actions. In all following discussions f, will be used to
as the concentration of active material is changed. Table denote the concentration of active material. (The dis-
I1 summarizes the K N criterion. Another approach is to cussions, of course, will also be valid for f, or fa.)
plot the logarithm of the reaction rate in mol s-l g-’ catalyst When investigating heterogeneous catalytic reactions in
vs. the logarithm of the surface concentration of active a batch reactor, the system is assumed to be in a stationary
material; a slope of 1 denotes that artifacts are not influ- state, and either reactant consumption or product for-
encing rate measurements. Explanations regarding the mation is observed with respect to residence time. The
deviation of this slope from unity are given in the Theo- KN test may be used in a batch reactor system as follows.
retical Section. Let us take w1grams of a catalyst in which f m = f,, and
Koros and Nowak (1967) have explained how one can note the conversion at a certain residence time t . In a
change the concentration of active material without second experiment, take w 2grams of a catalyst in which
changing the dimensions and pore diffusion properties of f m = f m , SO that
the pellets. Their method consists of mixing catalyst ~2 = wl(fm,/fm,)
particles with particles of an inert powder and then making
pellets from this mixture. The initial particles in the and note the conversion at the same residence time t as
mixture must be much smaller than the final pellets used the first experiment. If the conversions at the same res-
in the reaction. The inert powder must have the same idence time in the two experiments are identical, then the
diffusional characteristics as the catalyst powder with KN criterion is obeyed.
which it is mixed. For both metallic and nonmetallic Rates are directly obtained (Boudart, 1968; Denbigh,
catalysts this test can be used conveniently as follows. If 1966) in the case of continuous stirred tank reactors
we dilute a given catalyst, by the method described above, (CSTR) operating at the stationary state, and hence the
by a certain factor a , then for the K N criterion to be KN criterion is easily used. The CSTR condition states
obeyed the observed rate should change for the diluted that mixing is complete so that all the properties of the
catalyst over the initial undiluted catalyst by the same reaction mixture are the same in all parts of the reactor
factor cy. It should be noted that no measurement of the and in the exit stream. The concept of the CSTR has been
number of surface sites is involved. used, by experimentalists in the laboratory, in the form
There are two cases when the K N test, carried out by of the recycle reactor pioneered by Temkin (1962) and the
diluting a given catalyst as described above, cannot be continuous stirred tank catalytic reactor used by Tajbl et
used. First, when the catalyst has a bimodal pore distri- al. (1966). The recycle reactor consists of a loop which
bution, internal diffusion in the micropores may be im- includes a reactor containing a small amount of catalyst
portant in both the diluted and undiluted catalyst samples, and a recirculation pump. At the stationary state, when
and if the KN criterion is obeyed it may just indicate that the flow rate of chemical components in and out of the loop
there is no diffusional influence in the larger set of pores. is small compared to the recirculation rate, the composition
Secondly, if the particle size of catalysts used in the re- in the loop and exit line is practically identical, thus
action is close to the size of the powder from which the obeying the CSTR condition and allowing the rates to be
“diluted” pellets are made, as is sometimes the case for obtained directly.
catalysts used in liquid phase slurry reactors, then the KN In plug flow reactors, rates cannot be measured directly
test as described above cannot be used. unless the reactor is operated in a differential manner.
Rather, the conversion of a reactant is obtained at a known
In the case of supported metal catalysts, concentrations space velocity which may be defined as the molar flow rate
of active material may he changed by changing the metal of the reactant divided by the weight of the catalyst. We
loading on the inert support but keeping the preparation must modify the way to use the KN criterion from the
and pretreatment procedure identical, and the dispersion method described above to a method compatible with the
of the metal (defined as the ratio of the number of moles observables in the tubular flow system. Let us define a
of surface metal to total metal) close for all the catalyst quantity analogous to space velocity and denote it by p ,
samples. Techniques (Whyte, 1973; Benson and Boudart, the active space velocity
1965; Wanke and Dougharty, 1972; Benson et al., 1973) are
now available to measure surface areas of the active metal p = tiO/wf, (1)
in supported catalysts. The selective chemisorption of If two experiments are performed and f , is changed but
hydrogen or carbon monoxide has been used by Sinfelt p is kept constant in each experiment, then the KN cri-
(1969) on 11 different supported metal catalysts to obtain terion is obeyed if the observed conversion is invariant in
reaction rates per unit area of active material. The ti- both cases (see Theoretical Section). Another way is to
tration of adsorbed oxygen by hydrogen first used to obtain values of conversion at different flow rates and to
measure platinum surface areas by Benson and Boudart plot conversion vs. l / p using a catalyst sample with f, =
(1965) has been recently extended to obtain the surface fm1. If this plot coincides with the plot when we repeat the
areas of rhodium (Wanke and Dougharty, 1972) and pal- experiment using a catalyst sample with f , = f,, where f,,
ladium (Benson et al., 1973). When surface concentration # f m , then the KN criterion is obeyed.
of the active material in different catalyst samples can be Before concluding this section describing the K N cri-
measured, the K N test may be carried out, with caution, terion, it must be stressed that the test does not require
for catalysts with bimodal pore distributions. Hence, if a knowledge o f the number o f sites involved in the reac-
we have two catalysts, 1 and 2, with surface concentrations tion.
of active material f,, and f,, respectively,where f,, = afmg, Theoretical Section
then if the concentration of active material in the micro- A. Intraparticle Mass and Heat Transfer. Three
pores of catalyst 1 is different from that in catalyst 2 by relations on which much of the following discussion will
the same factor cy, then the K N criterion cannot be used be based are
to check for the effect of intraparticle mass transfer in
catalysts with bimodal pore distributions. r = kvdC,) (2)
Let us now see how the KN criterion may be used in the k 0: f m (3)
 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982 441

I
1 n@
1 u 1n@

Figure 1. Typical plots of In 7 vs. In $J for isothermal and noni- Figure 2. A plot of In 7 vs. In $J for a nonisothermal exothermic
sothermal exothermic and endothermic reactions (Weisz and Hicks, reaction (Weisz and Hicks, 1962).
1962).

Equation 2 indicates that the observed rate is dependent

on the effectiveness factor which in turn is a function of
the dimensionless Thiele modulus defined in eq 4.
Equation 3 is characteristic for heterogeneous catalytic
reactions taking place in the kinetic regime. The rate of
a catalytic reaction, defined in moles per unit time, is an
extensive property and is directly proportional to the
amount (Le., weight, surface area, or surface moles) of
active material; e.g., if the amount of active material is
doubled then the rate should be doubled. Before pro- Figure 3. A plot of In 7 vs. In $J for a nonisothermal exothermic
ceeding, we must clarify the above explanation by stating reaction (Weisz and Hicks, 1962).
that the KN criterion is not the same as the well-known
test in which if the weight of a catalyst is doubled, say in intraparticle diffusion and low values of q. When diffu-
a slurry reactor, the reaction rate should be doubled if sional influence exists but is small, the limits for curves
there is no influence of interphase mass transfer. In the A and B are in between those described above: -1 Ip I
KN criterion, the concentration of active material in the 0 and ' I 2Is I1.
catalyst, rather than just the weight of the catalyst, is For the exothermic reaction a similar situation exists for
changed, and in the kinetic regime the reaction rate is part b to c (Figure 2). Between a and b, p is greater than
directly proportional to the concentration of active ma- zero and hence s is greater than 1. The largest (d In q)/(d
terial. In 4) will be at the inflexion point a' where s will also have
Let us consider the curves A, B, and C in Figure 1. The its largest value. At point b, p is zero and hence s is unity,
slope, of any curve, at any point is similar to the zero diffusional influence point a. The ef-
dlnq dlnv fectiveness factor at b is higher than unity, and the rate
p=-=- (-1 Ip Ia) is hence higher than the true isothermal rate. An easy way
dln4 dln$ to resolve this ambiguity is to perform the experimental
where criterion at another temperature which would move the
CY = 0 for A and B; a > 0 for C; q = v point in question either to the left or right of point b. If
Using eq 2, 4,and 6 s is unity at two different temperatures, then one can be
assured that one is not concerned with the rather rare
r = kd%!(c,) occurrence of being at point b. It is interesting to note that
at point c' where q is unity and there is perfect compen-
sation (the rise in the reaction rate due to the increase in
the temperature within the pellet compensates the de-
crease in the measured rate due to the influence of mass
transfer), the value of s will tend to 1/2, indicating correctly
that the exothermic nonisothermal reaction is being in-
fluenced by intraparticle diffusion. As the Arrhenius
number
Y = E/R,Ts (9)
s is the slope obtained on plotting In r against In f ; s will and the heat generation parameter
be referred to as the Koros-Nowak number or slope. P = (-WDeCs/XTs (10)
When
increase, graphs as shown in Figure 3 may arise (Weisz and
-
p-0,s-1
and when
p - - -
-1, s 1/2
The former limit corresponds to 7 1, i.e., no intra-
Hicks, 1962). Here the values of p are greater than zero
m and hence s -
between points P and Q. A t Q and R, (d In q)/(d In 4)
m. Between Q and R, p 0 and s < 1.
It must be noted that if p is a large negative number s may
particle diffusion, whereas the latter limit indicates severe become less than 0.5 and also zero or a negative number.
442 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982

Table 111. Numerical Values for the Case of an q = l/A (13)

Isothermal, Irreversible, First-Order Reaction on a
Spherical Catalyst Pellet where
rl d -p s = 1 i @/2)
0.059 50 0.98 0.51
0.14 20 0.95 0.53
0.27 10 0.90 0.55 and
0.37 7 0.83 0.58
0.48
0.67
5
3
0.75
0.55
0.62
0.73
n
r(CJ = ?(CIS,C28, CSs, ..., (15)
0.81 2 0.35 0.82 The apparent rate constant k may be taken out of the
0.94 1 0.12 0.94 function r ( C ) . Hence
0.98 0.5 0.032 0.98
r(Cs) = kro(Cs)
Between R and S, p > 0 and s > 1. Beyond S the argu- r0(CJ may be a power rate law or a Langmuir-Hinshelwood
ments are similar to the isothermal case. type expression.
For an irreversible, isothermal first-order reaction in a
spherical catalyst pellet the above conclusions may be
obtained analytically from results of others (Satterfield,
1970)

7 = #P = ”[
4
coth 4 - i] 7 oc k-lI2 (16)
From eq 2,3, and 16, it can be easily seen that under Severe
intraparticle diffusion conditions for any rate expression
dlnq dlnq s =
-- - 4- Before ending the section on intraparticle diffusion, a
d In 4 d4 short corollary may be easily deduced from the above
2 - 4 coth 4 - 4’ csch2 4 algebra.
P’ (12) In eq 7, if the concentration function g(CJ is described
4 coth 4 - 1 by Cisn,where n is the true reaction order, then
See Table I11 for numerical values. It is interesting to note
that for the firsborder reaction considered, at effectiveness
factors 10.8, the Koros-Nowak number is approximately
equal to q.
Computations are not made so easily for other situations, where m is the observed reaction order. From eq 8
e.g., nonisothermal reactions. Equations given by Weisz n = 1 + (m- l)/s (17)
and Hicks (1962), Carberry (1961), and others have to be
solved and numerical differentiation carried out to give Knowing s from the KN criterion and the observed con-
(d In d / ( d In 4) = P. centration dependence m, one can obtain the true order
It is essential to note that the above analysis for intra- of reaction n, for isothermal operation.
particle diffusion holds not only for reactions obeying B. Interphase Mass Transfer. When the reaction
simple power rate laws but also the more complex Lang- rate is strongly dependent on the interphase transport of
muir-Hinshelwood rate expressions including strong reactants to the active material on a catalyst
product inhibition. Analysis, detailed by Satterfield (1970), N = k,a(Cb - C,) (18)
and Hutchings and Carberry (1966), show that In 7 vs. In
4 plots similar in shape to the ones described above are As k, is independent of the concentration of active mate-
obtained. Interphase heat or mass transfer influences can rial, changing the latter will not change the rate and the
complicate the problem (i.e., when the Nusselt numbers KN number s will be zero.
are low) by being coupled with the intraparticle case. For porous catalysts, when there is a large interphase
Hutchings and Carberry (1966) and Carberry (1961) have mass transfer influence the effectiveness factor is usually
shown that interphase heat transfer influence raises q for low, as only the active material at the mouth of the pores
exothermic cases whereas interphase mass transfer tends participates in the reaction.
to reduce 7. Once again the plots In 7 vs. In 4 are similar rate = qkg(Cs) = k,a[Cb - c,] (19)
to the ones described above. I n short, for all cases with
intraparticle diffusional influence, with or without t h e It can be easily shown for a first-order reaction that
complexity of strong product inhibition, intraparticle heat rate = k’Cb
transfer, and interphase heat or mass transfer, p will
deviate from zero and s will deviate from one. where
It should be noted that the value of s equal to unity also
means that the turnover numbers, obtained for different
values off,, must be identical.
Large Intraparticle Concentration Gradients. Un- Remembering that for a severe internal diffusion problem
der extreme diffusion conditions only a thin layer of a 7 0: fm-lI2 and using eq 3 and 20, we have two bounds when
catalyst particle is useful, and the local curvature and the KN criterion is used
hence the shape of the catalyst particle is of minimal im-
portance. 1. a k , << qk rate = constant Cb and s = 0
Aris (1965), Bischoff (1965), and Petersen (196513) have 2. ak, >> qk rate = constant fm1/2Cband s = y2
demonstrated that the effectiveness factor is asymptoti-
cally h-l for a general case, where A is the normalized Though easily shown for a first-order reaction, the same
Thiele modulus limits apply for more complex rate expressions. Values
 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982

of s between 0 and ‘I2would then indicate a coupled in- In ,f (rurfoce mol g

-1
catalyst)

fluence of the interphase and intraparticle mass transfer.

C. Interphase Heat Transfer. Mears (1973) has
shown that interphase heat transfer effects become sig-
nificant before the corresponding mass transfer effects
unless /3 is very small in which case, of course, problems
due to heat transfer will not arise. In the following dis-
cussion, it is assumed that there is no diffusional influence
and C, is equal to c b . Though the arguments are for
exothermic reactions, endothermic reactions may be
treated similarly.
At steady state, the heat released by the reaction is equal
to that transferred to the bulk fluid.
(-mr= ha(T, - Tb) (21)
Let us change the concentration of active material on the
catalyst from f,, to f,, where f,, > f,,, keeping all other
conditions the same
( - W r i = ha(T,, - Tb) (22)
( - m r 2 = ha(T,, - Tb) (23) I Coupled Interphore Transfer I
As f,, > f,,, r2 > rl and hence T,, > Tal. T,, > T,, > T b
indicates interphase heat transfer influence. Figure 4. Plots of In f, vs. In r , to indicate diagramatically the
Let us use the KN criterion, N1 = N2,to check for in- Koros-Nowak criterion for interphase heat transfer (exothermic re-
terphase heat transport influence. Using the Arrhenius action) and its coupling with interphase mass transfer.
form for the rate constant and separating f, from the
preexponential term, we get As explained before, when there is only isothermal in-
terphase mass transfer influence s = 0. These individual
fm,A’ exp(-E/R,T,,)g(C,,T,,)
- influences of external heat and mass transfer can be cou-
fm,
pled as indicated diagramatically in Figure 4 in order that
f m , A ’ exp (-E/RgTa,)g(Ca, Ts,) a value between s = 0 and s > 1 would be obtained. In
the rare case when the coupling gives s = 1,experiments
fm, done at a new bulk temperature would shift the value of
s from 1 if transport influences are important. For en-
Divide by exp(-E/RTb) dothermic reactions as f,, > f,, and external temperature
gradients are important, N 2 < N1and s < 1.
Mears (1973) has recently shown the relative importance
of transport effects, and the order was interphase heat
transport > intraparticle mass transport > interphase mass
transport > intraparticle heat transport. It is useful to
keep these generalities in mind when using the KN cri-

-
This equality holds if either (case 1): T,,
Tb or (case 2): T,, Tar.
-
Tb and T,, - terion. Figure 5 indicates usual values of N and s for the
KN criterion.
D. Deviations from Plug Flow Behavior in Tubu-
Case 1 is easily explained, as when the surface tem- lar Reactors (Interparticle Gradients). Plug flow in
peratures tend to the bulk temperature there is a minimal tubular reactors may be defined as a state of flow when
gradient and hence negligible heat transfer influence. the flow rate and all fluid properties over any cross section
Case 2. (a) When Tss > Tal > Tb, N1 # N2, there is
-
interphase heat transfer influence. (b) T,, > Tal Tb, Nl
normal to the fluid flow are identical and the residence
time of all volume elements, entering the reactor, is the
# N2. This occurs when by increasing the rate we may
enter the heat transfer influence regime. (c) T,, -
Tb. This is inconsistent with eq 22 and 23 because when
Tal>
same. Deviations from plug flow are due to the influence
of transverse and longitudinal temperature and concen-
tration gradients (interparticle gradients), velocity gra-

- -
sides cannot remain identical. Therefore, when T?, Tal
we must have T,, Tal Tb. Hence, the KN criterion,
-
their left-hand sides are different as r2 > rl, the right-hand dients, and backmixing in the tubular reactor. Though
longitudinal concentration gradients are always present
in plug flow reactors, these gradients may influence the
N1 = N 2 , can be used to establish the nonexistence of transport of reagents by causing them to reach the reactor
interphase temperature gradients. outlet more rapidly than they would by bulk flow alone.
It is interesting to rationalize the value of s for inter- Denbigh (1966) has indicated that transverse tempera-
phase heat transfer without diffusion when rate expres- ture gradients most often affect and invalidate the plug
sions are of the type r = kg(C,). As explained above, when flow assumption. If Tal and T, represent the axial tem-
there is no transport influence and f, is increased from peratures when catalysts with the concentration of active
f,, to f,,, the rate will increase from rl to r2 so that N1 = material f, and f,, are used respectively, and iff,, > f,,
N 2 and s = 1. If there is external heat transfer influence and Twis the temperature of the wall, we have
then iff,, > f,,, TS2> Taland r i >> r;; it should be noted
that due to the exponential nature of the temperature T,, > Tal > Tw
dependence the rate r i increases over r2 more than rl’ for a nonisothermal exothermic reaction, and
increases over rl. Hence N 2 > N1 and s > 1 (see Figure
4). Tap < T,, < Tw
444 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982

interpiare - rmc lntroparticla 0 Transport E'fect,

a coupling of various transport phenomena causes the test
to be apparently obeyed, the KN criterion should be
conducted at a second reaction temperature. If the test,
at the second temperature, is not obeyed, then transport
effects are influencing the reaction.
Of course a negative result for the test will also be ob-
tained if besides nonideal flow behavior interphase or in-
traparticle concentration or temperature gradients are
present.
E. Poisoning and Poor Contacting of the Catalyst.
Poor contacting of the catalyst may be due to the fluid
channelling and bypassing the catalyst. This case and the
case of catalyst poisoning would result in a certain number
6 moles of active material not being used for the reaction.
It will be shown that when this occurs the measured
...........!:......I turnover numbers, using catalyst samples with different
values off,, will not be the same and the KN criterion will
therefore be disobeyed. In two separate experiments let
the measured rates in mol d 8 - l of catalyst be r< and r2/
and the concentration of active material in surface mol g-l
.. I
of catalyst be f , and f,,, and the weights of the catalysts
taken be w1 and w2,respectively.
Hence, the calculated turnover numbers will be

: ...... ..
.................
0
0.0:

In the case when the reaction is carried out under no

poisoning conditions and good catalyst contacting

Figure 5. Interphase and intraparticle transport effects. The chart

indicates the trend of the turnover numbers N1 and N 2 when f,, >
f,,, and shows limits for the values of s when interphase and intra- The "true" rate rl is then related to rl' by
particle effects exist separately or are coupled together.
61
for a nonisothermal endothermic reaction. r,' = rl - N-
W1
As the dependence of the rate on temperature is expo-
nential, conversions in the plug flow reactor, where radial Dividing (28) by f,, and using (26) and (27)
temperature gradients are significant, will change as f , is
N(fml- 6 1 / ~ 1 )
changed but p is kept constant. A similar analysis may N,' =
be obtained for the problem of longitudinal temperature fm1
gradients.
The relation used for tubular reactors (Boudart, 1968; Similarly
Denbigh, 1966) where the plug flow assumption is valid Nfm, -WW2)
may be written as N,' =
fmz

N,' and N,' are then equal only when

where r here is expressed in mol sd g-' of catalyst. Mul-

tiplying the above equation on both sides by f ,
This is not possible except in the case
61 = Cyf,,W1 (29)
and
6, = afm2W2 (30)
To check for this case the weights of the catalysts should
be changed by an unequal amount (say, w1/2 and w2/3)
and the measurement of the rates repeated. If again
If there is any significant deviation from plug flow be- turnover numbers are found to be equal, then there is no
havior due to the influence of transverse and longitudinal catalyst poisoning or poor catalyst contacting.
temperature or concentration gradients, or due to back-
mixing and velocity gradients, the above equation will be Experimental Results and Discussion
invalid. The liquid phase hydrogenation of cyclohexene was
Hence, if in two experiments using different values of studied over supported Pt/Si02 catalysts. The details of
f, but keeping p constant the same conversion is obtained, the apparatus, materials, procedure, and results are given
then the KN criterion is said to hold and deviations from elsewhere (Madon et al., 1978). The reaction was tested
plug flow behavior will be negligible. In the rare event that via the above criterion for any transport effects.
 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982 445

in fm (rvrface mol Pt g-1 catalyst) In i, (ruriace mol pt g-l catalyst)

-I 1 IO -1 I -1 0 -9
I
I I

0 1 .50% Pt/SiOp D=lOO%

02.30% Pt/SiO2 E= 62%
v 0 . 5 3 % Pt/BO2 W 56%

0 1 .50% Pt/SiO2 BlOO% Figure 7. The Koros-Nowak criterion applied in the study of the
0 2.30% Pt/SiO2 W 62% liquid phase hydrogenation of cyclohexene on Pt/SiOz catalysts.
0 0.38% Pt/SiO2 BlOO% Conditions: r corrected for 101.3 kPa Hz,
solvent-cyclohexane =
V 0.53% Pt/Si02 E= 56%
20 cm3,cyclohexene added = 0.5 cm3,

severe internal diffusion, 7 = 1/4, and using eq 4, r cc

Figure 6. The Koros-Nowak criterion applied in the study of the (kDe)'/2.One may express k and De by simple Arrhenius
liquid phase hydrogenation of cyclohexene on Pt/Si02 catalysts. relationships (Satterfield, 1970; Satterfield et al., 1968)
catalyst particle size > 200
Conditions: r corrected for 101.3 kPa H2,
mesh, solvent-cyclohexane = 20 cm3,at 275 K cyclohexene added De = A, exP(-ED,/R,T)
= 2 cm3,at 307 K cyclohexene added = 0.5 cm3.
k = A exp(-E/R,T)
The reaction was carried out with three different size
fractions of Pt/SiO, catalysts: (a) greater than 200 mesh, Hence the observed activation energy for the severely in-
(b) 100 to 200 mesh, and (c) 60 to 100 mesh. Observed ternal diffusion influenced reaction is given by '/z(EDe +
reaction rates on the three separate catalyst size fractions E). Using an activation energy of 12.6 kJ mol-' for the
have been plotted in Figures 6 and 7, which are diagram- diffusion of H2in cyclohexane (Satterfield, 1970; Satter-
matic representations of the KN criterion written as r a field et al., 1968), the activation energy of the reaction in
f," and indicate clearly whether mass transport effects play the 100 to 200 mesh Pt/SiO, particles is '/&4.4 12.6)+
a dominant role. From Figure 6, we ascertain that with = 18.5 kJ mol-'. Though this is close to the observed value
the catalyst size fraction (a) there is negligible transport of 18.9 kJ mol-', it must be stressed, however, that using
influence as s is equal to 1, whereas with the two larger values of activation energy alone to explain diffusional
size fractions (Figure 7) the observed reaction rates are influence could lead to errors.
found to be severely affected by intraparticle concentration Mears and Boudart (1966) studied the liquid phase
gradients, the value of s being 1/2. Calculations to sub- dehydrogenation of isopropyl alcohol on nickel catalysts
stantiate these observations are given elsewhere (Madon, reduced with sodium borohydride and on Raney nickel in
1975). a slurry reactor. The slurry was agitated either by stirring
It is also interesting to note the values of the turnover it with a propeller or, to study the effect of mass transfer
numbers when catalysts with size fraction (a) were used. under different hydrodynamic conditions, by shaking it.
The 1.5% and 0.38% Pt/SiOz samples had the identical As the slurry agitation rate was increased the reaction rate
dispersion of 100% and N was found to be 2.67 s-' and 2.51 increased and then leveled off to a plateau. It is usually
s-' at 275 K and 9.16 s-l and 9.02 s-' at 307 K. The 2.3% inferred (Satterfield, 1970) that there is no influence of
and 0.53% Pt/SiO, samples had dispersions of 62% and external mass transfer on the reaction rates at the plateau
56%, respectively, and N was found to be 2.40 s-' and 2.33 value. Mears and Boudart argued that, at the plateau
s-l at 275 K and 8.67 s-' and 8.51 s-l at 307 K. It should value, though the possibility of external mass transfer
be noted that only catalysts with dispersions that are influence was eliminated, the rates could still be affected
identical or very close must be compared when the KN by pore diffusion. To check the latter point, they prepared
criterion is used. The identity of the turnover numbers nickel catalysts of different surface areas and noted that
at two different temperatures is another way of looking chromium acted as a structural promoter and increased
at the test and indicates, as does Figure 6, that the reaction the specific surface area of the active nickel component.
rates are not influenced by any physical events. Surface areas of the nickel catalysts were determined by
The activation energy of the reaction, using the small the fatty acid adsorption method of Smith and Fuzek
size particles, i.e., when s = 1,was found to be 24.4 kJ mol-l (1946) and by the nitrogen BET technique. At high agi-
whereas using the larger size (100 to 200 mesh) particles tation intensities, the rate constant per unit surface area
when s = 1/2, a value of 18.9 kJ mol-' was obtained. remained essentially constant over a sevenfold variation
The rate expression for the liquid phase hydrogenation in the specific area of the catalyst. Mears and Boudart
of cyclohexene may be written in the form r = kvg(C,). For concluded that this observation provided strong evidence
446 Ind. Eng. Chem. Fundam., Vol. 21, No. 4, 1982

against any influence of internal or external diffusion. k' = defined by eq 20

They argued that at low agitation intensities the reaction m = observed reaction order
rate was reduced as a result of internal (and to a certain n = true reaction order
AO = molar flow rate, mol s-'
extent external) mass transfer due to catalyst agglomera-
tion and clumping, and with increased agitation the re- N = turnover number, s-'
N = molar flux, mol s-l cm-3
action rate increased due to better suspension of the p = defined by eq 5
catalyst in the slurry with breakup of catalyst aggregates. r = reaction rate, mol s-l 8-l (or ~ m - of~ catalyst
)
The test, performed with several catalysts in which the r ( C ) = rate expression, defined by eq 15
surface area of the active component was varied and used R = radius, cm
to conclude that reaction rates at the plateau value were R, = gas constant
free from both internal and external mass transfer, was s = Koros-Nowak criterion number or slope
in effect the diagnostic criterion later proposed by Koros S, = external surface of the catalyst particle, cm2
and Nowak (1967). t = time, s
The Koros-Nowak criterion has also been used suc- T = temperature, K
V , = volume of the catalyst particle, cm3
cessfully on vapor phase reactions: hydrogenation of w = weight of the catalyst, g
ethylene (Schlatter and Boudart, 1972),cyclohexene (Segal x = conversion
et al., 1978), cyclopropane (Boudart et al., 1966), and ox-
ygen (Hanson and Boudart, 1978), decomposition of NO Greek Letters
(Amirnazmi et al., 1973) on supported Pt catalysts, and a = any integer or fraction
ammonia synthesis on supported Fe catalysts (Dumesic a' = volume fraction
et al., 1975). p = heat generation function, defined by eq 10
y = Arrhenius number, defined by eq 9
Conclusion 6 = unused moles of active material, see eq 28 rate
7 = effectiveness factor, defiied as the ratio of the actual rate
An experimental criterion to check for true reaction rates for a catalyst particle to the rate evaluated at external
in heterogeneous catalysis has been developed. If the KN surface conditions
test is obeyed, at two different temperatures, one can be X = thermal conductivity of the catalyst, J cm-' s-l K-'
sure that the observed kinetic data are free of artifacts such h = normalized Thiele modulus, defined by eq 14
as mass or heat transport, catalyst poisoning, and con- v = stoichiometric coefficient
tacting. If the test is not obeyed, it may be difficult to say t = Ci/Ci,
p = active space velocity, defined by eq 1
exactly which physical event is affecting the rate data, I$= Thiele modulus, defined by eq 4
though good characterization of the catalyst used and some x = defined in Table I
understanding of the reaction being studied may help one w = defined in Table I
to understand which artifact influences the reaction rate.
The KN criterion cannot be used for catalysts with a Subscripts
bimodal pore distribution if one uses the dilution method a = property at the axis of a tubular reactor
suggested by Koros and Nowak (1967). b = bulk property
The test may be used with batch, continuous stirred, or i = species 1, 2, 3, ...
s = surface property
plug flow reactors, and during differential or integral rate eq = equilibrium
measurements. A simple analysis of the criterion has been
given, and its use in the laboratory demonstrated. Though Literature Cited
in general the test may be used for any catalyst if an ap- Amirnazmi, A.; Benson, J. E.; Boudart, M. J. Catal. 1973, 30,55.
Anderson, J. B. Chem. Eng. Sci. 1963, 78,147.
propriate inert diluent can be obtained, it is most readily Ark, R. Chem. Eng. Scl. 1957, 6 , 262.
and easily used for supported metal catalysts where the Aris, R. Ind. Eng. Chem. Fundam. 1965, 4, 227.
Benson, J. E.: Boudart, M. J. Catal. 1965, 4 , 704.
support is inert and for heterogenized homogeneous or- Benson, J. E.; Hwang, H. S.; Boudart, M. J. Catal. 1973, 30, 146.
ganometallic catalysts. Bischoff, K. B. AIChE J. 1965, 1 1 , 351.
Bischoff, K. B. Chem. Eng. Sci. 1967, 22, 525.
Boudart, M.; Aidag, A.; Benson, J. E.; Dougharty, N. A,; Harkins, C. G. J.
Nomenclature Catal. 1966, 6 , 92.
Boudart, M. "Kinetics of Chemical Processes"; Prentice-Hall: Englewood
a = external surface area of catalyst per unit volume of the Cliffs, NJ, 1968.
catalyst, cm-l Boudart, M. AIChE J. 1972, 18, 465.
Carberry, J. J. AIChE J. 1961, 7,350.
A' = preexponential factor of the rate constant Carberry, J. J. Ind. Eng. Chem. 1966, 58(10),40.
A = A'fm, A'fa, or A'fw Chambers, R.; Boudart, M. J. Catal. 1966, 6, 141.
A, = preexponential factor in the Arrhenius relationship for Denbigh, K. 0. "Chemical Reactor Theory"; Cambridge University Press:
D. Cambridge, 1966.
Dumesic, J. A.; Topsm, H.; Khammouma, S.; Boudart, M. J. Cafal. 1975,
c ==concentration,mol cm-3 37,503.
DaII = Damkohler number of the second kind Hanson. F. V.; Boudart, M. J. Catal. 1978, 53,56.
DaIv = Damkohler number of the fourth kind Hudglns, R. R. Chem. Eng. Scl. 1968, 23, 93.
Hudgins, R. R . Can. Chem. Eng. 1972, 50,427.
De = effective diffusivity, cm2 s-l Hutchlngs, J.,; Carberry, J. J. AIChE J. 1966, 12, 20.
E = activation energy, kJ mol-' Koros, R. M.; Nowak, E. J. Chem. Eng. Scl. 1967, 22, 470.
f, = concentration of active material, surface moles of active Madon, R. J. Ph.D. Dissertation, Stanford Unlverslly, 1975.
material g-' of catalyst Madon, R. J.; O'Connell, J. P.; Boudart, M. AIChE J. 1978, 2 4 , 904.
Mears. D. E.; Boudart, M. AIChE J. 1966, 72, 313.
f, = concentration of active material, cm2of active material Mears, D. E. Ind. Eng. Chem. Process Des. Dev. 1971a, 10, 541.
g-' of catalyst Mears, D. E. J. Catal. 1971b, 20. 127.
f, = concentration of active material, g of active material g-' Mears, D. E. J. Catal. 1973, 30,283.
of catalyst Petersen, E. E. Chem. Eng. Scl. 19658, 20, 587.
Petersen, E. E. "Chemlcal Reaction Analysis"; Prentlce-Hall: Englewood
g(CJ = g(Cls,C2s,CBs...) = concentration function in the rate Cliffs, NJ. 1965b.
expression SatterfleM, C. N.; Saraf, S. K. Ind. Eng. Chem., Fundam. 1985, 4 , 451.
h = heat transfer coefficient, J cm-2 s-l K-l Satterfleld, C. N.; Ma, Y. H.; Sherwood, T. K. Inst. Chem. Eng. (London).
A H = heat of reaction, J mol-' Symp. Ser. 1966, 28, 22.
Satterfield, C . N. "Mass Transfer In Heterogeneous Catalysis": MIT Press:
k , = mass transfer coefficient, cm s-l Cambridge, MA, 1970.
k = rate constant Schlatter, J. C.; Boudart, M. J. Cafal. 1972, 24, 482.
 Ind. Eng. Chem. Fundam. 1982, 21, 447-451 447

Segal, E.; Madon, R. J.; Boudart, M. J . Catai. 1976, 52, 45. Weisz, P. B. Z . Phys. Chem. (Frankfurt am Main) 1957, 1 7 , 1.
Sinfelt, J. H. C9fel. Rev. 1989, 3 , 175. Weisz, P. B.; Prater, C. D. Ab.. Cafal. Relat. Sub/. 1954, 6, 143.
Smith, H. A.; Fuzek, J. F. J . Am. & e m . Soc.1946, 68, 229. Welsz, P. B.: Hicks, J. S. Chem. Eng. Sci. 1962, 77, 265.
Stewart, W. E.; Vllladsen, J. V. AIChEJ. 1969, 75, 28. Whyte, T. E. Catal. Rev. 1973, 8 , 117.
c .-.
TaJbl, D. (3.: Simons, J. B.; Carbeny, J. J. Ind. Eng. Chem. Fundam. 1966,
3. I 1 I .
Receiued for reuiew September 28, 1981
Temkin, M. I . Klnet. Kafal. 1962, 3 , 509.
Wanke, S. E.: Dougharty, N. A. J . Cafai. 1972, 24. 367. Accepted May 7, 1982

Mechanism of Stickiness in Hygroscopic, Amorphous Powders

Galen E. Downton,' Jos6 L. Flores-Luna,2 and C. Judson King'

Department of Chemical Engineering, University of Callfornia, Berkeiey, California 94720

Sticking, caking, and agglomeration tendencies of hygroscopic, amorphous powders are interpreted in terms of
a mechanism of viscous flow driven by surface energy during particle contact. The mechanism predicts that
stickiness should occur for combinations of temperature and moisture content corresponding to viscosities of the
amorphous material below some relatively constant value, which for short-time contact should be within the range
loe to 10' Pas. The viscosity is affected by both moisture content and temperature, thereby explaining the inverse
relationship between sticky-point temperature and moisture content for such powders. Both bulk-solution and
powdered specimens of a 7:l w/w mixture of sucrose and fructose were prepared, with moisture contents in the
range of 2 to 7% w/w. Viscosity measurements of the bulk-solution specimens were made with both a falirg-baii
method and a rheogoniometer. Viscosities corresponding to experimentally measured sticky-point temperatures
fell consistently within the range 0.3 X lo7 to 4.0 X lo7 Paos, thereby lending support to the proposed mechanism
and quantitative model.

Introduction operating problems, and powder-handlingdifficulties. For

Dehydration of liquid materials is widely used to pro- heat-sensitive products this can also lead to overheating,
duce stable, easily handled and stored products in an resulting in unpleasant sensory characteristics and/or
economical fashion. Several techniques, such as spray degradation. These problems are preventing some
drying, freeze drying, and drum drying, are available for materials-notably fruit juices which contain large
producing dried powders from a wide range of raw mate- amounts of hygroscopic, amorphous sugars-from being
rials including liquid foods, plastics, detergents, fertilizers, successfully dehydrated.
and pharmaceuticals. One phenomenon frequently en- Coping with stickiness in spray dryers and in powdered
countered during production and storage of these dried products and achievement of successful agglomeration have
powders is stickiness. largely been matters of trial-and-error experimentation to
A familiar example of stickiness is caking during storage. find conditions which avoid or control the sticky charac-
Hygroscopic dried foods such as instant coffee or drink teristics of a given material. Cooling the dryer walls and
mixes tend to form hard cakes when subjected to high flushing them with cool air have been employed to avoid
temperatures and/or humidities. The powder is no longer stickiness in spray dryers (Lazar et al., 1956; Gupta,
free-flowing and, because the cakes are less porous, they 1978a,b). Various additives, often of high molecular
reconstitute less readily. Another example of stickiness, weight, have also been used to combat stickiness in spray
which can be turned into an attribute, is agglomeration, drying (Brennan et al., 1971; Gupta, 1978a,b) and in
a process commonly used with dehydrated powders to powders (Peleg and Mannheim, 1969,1973,1977).
obtain larger, porous particles with better rehydration and There is as yet no quantitative model which describes
handling characteristics and/or more pleasing and con- the mechanism of stickiness in dried powders and explains
sistent color, shape, and appearance. This operation is the success of failure of past approaches to the problem.
redly one of controlled stickiness; particles are deliberately Previous mechanistic speculations are nearly all qualitative
subjected to steam, moist air, or a fine mist of water, as and incomplete. White and Cakebread (1966) recognized
well as repeated contact, so that surfaces become sticky dried liquid foods as amorphous glasses, i.e., metastable,
and clusters form via particle collisions and adherence. In supercooled liquids below their glass-transition tempera-
spray drying, stickiness can be a major problem, which ture, Tg.At temperatures below Tg, the viscosity is ex-
occurs when particles which are insufficiently dry collide tremely high, of the order of 10l2Pa.s (Jones, 1956). A t
with one another or with the walls of the drying apparatus higher temperatures and/or humidities, the glass assumes
and become stuck. This can lead to lower product yields, more of a liquid nature, with a lower viscosity. Stickiness
defects, such as lumpiness and caking, were attributed to
this incipient liquid state.
Procter and Gamble Company, Coffee Division, The Winton Peleg and Mannheim (1977) attributed caking of onion
Hill Technical Center, 6210 Center Hill Road, Cincinnati, OH powder to moisture absorption and proposed a humidi-
45224. ty-based caking mechanism. In their model, water absorbs
'Mezquital de Oro, Apdo. Postal 138, Hermosillo, Sonora, on particle surfaces, forming a saturated solution and
Mexico. thereby making the particles sticky and capable of forming
0196-4313/82/1021-0447$01.25/0 0 1982 American Chemical Society