See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/230455120

The Thermal Conductivity of Porous Catalyst Pellets

Article  in  The Canadian Journal of Chemical Engineering · February 1979

DOI: 10.1002/cjce.5450570104

CITATIONS READS

9 466

2 authors, including:

Mahfooz Soomro
Western Sydney University
9 PUBLICATIONS   72 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

I am writing a paper on recycled aggregates, I would like to cite your article as reference, kindly send me the article. Regards, Dr. Soomro View project

All content following this page was uploaded by Mahfooz Soomro on 30 June 2018.

The user has requested enhancement of the downloaded file.

 The Thermal Conductivity of Porous
Catalyst Pellets
112. SOOMRO and R . HUGHES
Department of Chemical Engineering, University of Salford, Salford, M54 W7 England
Thermal conductivities of ?-alumina and alumina supported
nickel catalysts prepared by impregnation and coprecipitation
have been measured by a transient method. Results obtained
were compared with predictive models and it was found that a
simple modification' of a one parameter model gave good agree-
ment with the results for the support and impregnated catalysts
but did not give satisfactory predictions for the coprecipitated
catalyst.
T he thermal conductivity of porous solids is of im-
portance in many fields and has led to a number
of investigations, both experimental and predictive.
For catalysts used in fixed bed reactors where the
catalyst pellet may be of moderate dimensions a
knowledge of the effective thermal conductivity is
essential for estimating the temperature rise within
the pellet and also in the reactor; i t is especially im-
portant when the dynamics of a catalytic reactor a r e
considered.
A number of models have been proposed for the
prediction of the thermal conductivity of porous
materials, including those of Woodside and Messmer"',
Luikov and co-workers'2', Kr~ipiczka'~',Buttc4' and
Harriott'". Although some of these give good esti-
mates of the true thermal conductivity of porous solids
there is still appreciable uncertainty in the accuracy
of these models, so experimental measurements are
still necessary. This is particularly so when the pore
structure is complex such as that occurring with
catalyst pellets. The problem is made more difficult
when supported metal catalysts are used since a
knowledge of the precise distribution of t h e metal and
support is seldom available.
Basically, two methods a r e available f o r determin-
ing the thermal conductivities of porous catalysts ; the
transient method and the steady state method. The
first systematic measurements of the thermal conduc-
tivities of porous catalysts was made by Sehr"' using
both methods. Steady state measurements were made
by Masamune and Smith"' of the thermal conductivity
of microporous silver particles pressed into pellet form.
This study was particularly valuable in pointing to the
importance of contact area between the particles; if
the contact area is low even an intrinsically high
thermal conductivity material such as silver will give
a low overall effective thermal conductivit.y when
pressed into pellet form. Other measurements of the
thermal conductivity of porous solids have been re-
ported by Mischke and Smithc8', KlingCs', Somerton(Io',
Schroeder"" and Cunningham et a1(12'.
Previous work in this d e p a r t ~ e n t " ~ has
. ~ ~ )concen-
trated on the measurement of the thermal conduc-
tivities of silica and silica supported metal catalysts.
The present work extends this to alumina based
catalysts using a transient measurement method. The
results are compared with various models and a
modified form of the Woodside and Messmer model is
developed which is found to be better agreement for
this class of catalysts compared with other models.
24 The Canadian Journal of Chemical Engineering, Vol. 57, February, 1979
On a mesurC, par une mkthode transitoire, les conductivitb
thermiques de I'alumine-7 ainsi que de catalyseurs au nickel
support& sur alumine et prCparb soit par imprkgnation soit par
coprkcipitation. Les rbultats ont ktk comparb avec les prkdic-
tions de divers modhles et on a trouvC qn'une modification
simple d'un modhle i un seul paramhtre permettait d'obtenir
une bonne reprksentation des rksultats obtenus pour le support
et pour le catalyseur obtenu par imprkgnation, mais ne reprB
sentait pas de facon satisfaisante le comportement du catalyseur
coprkcipitk.
Models used f o r comparison purposes
The simplest models used for predictive purposes
involve just one parameter, the porosity of the pellet.
Assuming a simple two phase structure of a continu-
ous and discontinuous phase, these may be arranged
in series or parallel with respect to heat flow to give
a minimum and maximum value respectively for t h e
effective thermal conductivity of the porous compact.
If t h e two phases a r e distributed randomly, then
Woodside and Messmer") postulated t h a t the thermal
conductivity could be represented by the expression :
K, = K j (K*,/Kj)'-' . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
Where and K . refer to fluid and solid conducti-
vity and E is the pellet voidage. Equation (1) applies
strictly f o r a mono-disperse pore size distribution i.e.
only gas phase conduction in the macropores plus solid
phase conduction is considered. When attempts a r e
made to apply this equation to a bidisperse pore size
distribution involving macro and micro porous struc-
tures typical of catalyst pellets produced by powder
compression, very high predicted values are ob-
tained(15'.
The model of Butt'*' extends the random pore model
of Wakao & Smith'"' to predict thermal conductivities
of catalyst pellets. This model requires a knowledge of
mean micro- and macropore diameters in addition t o
the void fractions f o r pellet and microporous particles.
The essential breakthrough in predictive methods was
made by Harriott"' who showed t h a t simple thermal
conductivity/void fraction relations must be applied
twice. The effective conductivity of t h e microporous
particles must f i r s t be estimated using the micropore
voidage and this must then be used with the macro-
pore void fraction to estimate the thermal conductivity
of the pellet. Harriott"' identified three regions of
model application depending on the area of contact
between t h e microporous particles, which he termed
the packed bed model, the consolidated particles model
and the porous solid model.
This model, however, also requires t h a t mean pore
diameters be known o r a r e capable of estimation, in
order to calculate t h e gas conduction contribution
within the pores. The gas phase contribution is often
small, however, and it occurred to us t h a t a simpler
one parameter model such as t h a t of Woodside and
Messmer could be applied in t u r n to t h e microporous

01
01 02 03 04 05
la
Figure 1 - Experiniental thermal conductivities of different
catalysts.
0 = Silica/l3c/o Alumina;
= 0.5% Ni on Alumina;
= 62c/o~N i on Alumina;
0 = 7-Alumina;
0 = 5 % N i on Alumina;
@ = 75% N i on Alumina;
= 75% N i on Alumina (coprecipitated).
particles and the catalyst pellet to give simple estima-
tions which might still be in reasonable accord with
experimental measurements. Woodside a n j Messmer's
model was therefore modified a s follows.
The thermal conductivity of the microporous par-
ticles K,' was first computed
i.e. K,'= KJ(K~/K,)'-'~.
.........................
where ei is the void fraction of the particles in the
pellet.
The effective thermal conductivity f o r the pellet
was then computed using K,' a s the new 'solid' con-
ductivity for the macroporous pellet.
K, = K,(K3'/Kj)1+f~.
............................
with E. a s the macropore voidage.
Combining Equations ( 2 ) and ( 3 ) we g e t :
K, = KJ(K,/KJ)
(l-ei) . . . . . . . . . . . . . . . . . . .. ( 4 )
so that the effective thermal conductivity may be
estimated from a knowledge of the thermal conducti-
vity of solid and fluid and the appropriate void frac-
tions for the micro-and macropore regions.
Initially, a n alternative method of estimating the
thermal conductivity was employed. Instead of apply-
ing the Woodside & Messmer"' equation t o the micro-
porous particles, the conductivity of these was first
estimated using the working graph proposed by
Krupiczka in his paperc3'. The value of X,' for the
microporous particles obtained in this manner was
then substituted in the Woodside and Messmer relation
for the pellet (Equation ( 3 ) above) to get the overall
pellet conductivity. This method is designated a s
The Canadian Iournnl of Chemical Engineering, Vol. 5 7 , February, 1979
TABLE1
EFFECTOF NICKELCONTENT ON EFFECTIVE THERMAL
CONDUCTIVITY(IMPREGNATED CATALYSTS)
o/o Nickel I KdW/m.K)
0 I 0.184
0.5 I ,215
5 I 0.304
62 I 0.358
75 0.448
Modification I in the Results section whilst the double
application of the Woodside & Messmer relations a s
derived above is designated Modification 2.
Experimental
The thermal conductivities of porous pellets were
measured by the transient method in which the tem-
06 perature-time record of thermocouples placed a t the
centre and surface of a spherical pellet was measured
subsequent to a step change in surface temperature.
If A T o is the initial difference in the temperature
readings between the centre and surface of the spher-
ical pellet and a is the thermal diffusivity, the solu-
tion of the conduction equation for constant a is"6'
+exp (- q)) , (5)
where A T is the temperature difference after time 8
and R is the pellet radius. To avoid errors in TO
i t is preferable to apply the equation to subsequent
. sets of two readings a t times greater than zero; thus
neglecting terms beyond the first, we get
.(6)
.(2) If the density and specific heat of the pellet are
known, the effective thermal conductivity may be cal-
culated from the thermal diffusivity, a.
Spherical pellets were prepared from alumina, silica-
alumina and alumina impregnated with various
amounts of nickel. A co-precipitated nickel/alumina
catalyst was also used. The pellet assembly was made
.(3) by first pressing two half-cylindrical pellets, inserting
a platinum/l3 % rhodium-platinum thermocouple wire
0.0254 mm dia. with the junction located exactly a t
the centre of the assembly and finally pressing both
halves together. The resulting cylindrical pellet was
abraded to a sphere of 12.7 mm diameter. Another
similar thermocouple was positioned a t the pellet
surface.
The completed pellet was immersed rapidly in a bath
of molten wax maintained at 120 * 1 ° C and the tem-
perature a t the surface and centre recorded as a func-
tion of time.
The total pore volume of the pellet was measured
using the penetration of a liquid solution of 13%
Cetane in carbon tetrachloride. Micropore volume was
determined by titrating a 25 g sample of the powder
with water until the fluidity was destroyed. Thus the
micropore volume is determined directly whilst the
macropore volume is obtained by difference between
the total pore volume and micropore volume.
25
 0.4
-
0.:
OU
E
2
v
Q, 0.2
Y
'"\
0.I
0.I 0.2 0.3
Ea
Figure 2 - Comparison of experimental and predicted thermal
conductivities for 0.5 ch Ni on impregnated r-Alumina;
0 = Experimental;
0 = Porous solid model;
- - _= Consolidated particles model;
8 = Modification 1;
0 = Modification 2.
Results and Discussion
Catalyst pellets of pure alumina and containing dif-
ferent amounts of nickel on alumina were prepared
with different densities by the methods outlined
above. The thermal conductivity was thus obtained a s
a function of pellet macroporosity. All the measure-
ments were made with the catalyst pores filled with
a i r a t atmospheric pressure.
Preliminary experiments were made to determine
the thermal conductivity of 13% ' Silica-Alumina
pellets which had previously been measured in this
' a similar technique. Results were in
d e ~ a r t m e n t " ~by
good' agreement with the results obtained by calibra-
tion previously so the present results were deemed to
be satisfactory.
Thermal conductivity data are presented as func-
tions of the macroporosity and nickel content and
a r e compared with the two modifications of the Wood-
side and Messmer model"' and the porous solid and
consolidated particles models of Harriott'". The latter
represents higher contact areas giving conductivities
mere appropriate to the present study.
The experimental results for pure y-alumina and
f o r varying amounts of nickel present on the same
y-alumina support are summarised in Figure 1.
Thermal conductivities are presented as functions of
the macroporosity of the pellet. Also shown in this
figure, for comparison purposes, is the thermal con-
ductivity of silica - 13% alumina. In general, the
thermal conductivity decreases with macroporosity,
reflecting the decreasing solid contribution to overall
26 The Canadian Journal of Chemical Engineering, Vol. 5 7 , February, 1979
I I
0.1 a2 0.3
la
Figure 3 - Comparison of experimental and predicted thermal
conductivities for 75% Ni impregnated on r-Alumina
0 = Experimental;
0 = Porous solid model;
_ _ - - Consolidated particles model;
8 = Modification 1;
0 = Modification 2.
conductivity as the pellet becomes less dense. The low-
est thermal conductivity is shown by the silica-alumina
pellet which is to be expected in view of the lower ther-
mal conductivity of pure solid silica (8.11 W/m*K)
compared t o that of pure solid alumina 19.86 (W/m=K).
For the alumina and nickel impregnated aluminas the
effective thermal conductivity increased broadly with
nickel content. This increase is not, however, in direct
proportion to the nickel content. This is demonstrated
in Table 1 which compares the effective thermal con-
ductivity of alumina and impregnated aluminas a t a
given void fraction of .macropores = 0.23). Addi-
tion of nickel to alumina gives an initial sharp rise
in effective thermal conductivity of the pellet; a
similar increase is shown for nickel additions above
62%. In t h e intermediate region between 5 and 62%
no results were obtained so an unequivocal conclusion
cannot be reached, but i t would appear that in this
intermediate region the effective thermal conductivity
of the pellet is insensitive to the amount of nickel
present. Further work is obviously desirable in this
range of nickel content.
The coprecipitated catalyst had a larger macropore
w i d fraction as compared with the impregnated
catalyst of similar loading, so that it was not possible
to compare these two methods of catalyst preparation
directly. This difference in macropore void fraction is
partially a consequence of the difference in prepara-
tion, since in impregnated catalysts a t high nickel
loadings there will be a tendency to reduce both macro-
pore and micropore voidage due to t h e process of im-
pregnation. F o r the coprecipitated catalyst t h e ex-
 TABLE
2
VALUES OF PARAMETERS
USED IN THE MODELS

Material I e< I
I
€a

I
y-Alumina 0.452 0.2335 0.5896 4.575 x 10-2 5.367 x 10-4
(MOD 1)
0.476
0.502
0.527 1.0744 x
(MOD 2)
0.5% NiIA1203 0.449 0.2374 0.5888 4.564 x I 5.52 X
0.473
0.499
0.5235
0.1964
0.153
0.111
1 (MOD 1 )
i.ofSX
(MOD 2)
10-3

5% Ni/Ala03 0.424 0.27 0.5814 4.471 x 5.68 x 10-4

0.447 0.231 (MOD 1)
1.12 x 10-3
I 0.4714
0.495 I 0.189
0.149 (MOD 2)

1 1
_ _ _ ~
62% Ni/A1203 0.284 0.3208 0.418 3.23 x 1.150 x
0.2845 (MOD 1)
0.245 2.16 x 10-3
(MOD 2)
~

75% Ni/A1203 0.215 0.401 0.338 2.95 x 10+ 1.42 x 10-3

(Imcregnated) 0.2277 0.349 (MOD 1 )
0.247 0.311 3.984 x 10-3
0.274 0.2358 (MOD 2)
75% Ni/A1203 I 0.5952 0.3585 2.95 x 1.34 x 10-3
(Co-precipitated) 0.5663 (MOD 1)
0.162 0.549 3.53 x 10-3
0.167 0.533 (MOD 2)

perimental thermal conductivity increases rapidly with
decrease of the macropore void fraction. This effect
was previously o b s e r ~ e d " ~with
' aoprecipitated nickel/
silica catalysts and i t would seem to be generally true
that a coprecipitated catalyst will have a higher
effective thermal conductivity than an impregnated
catalyst a t a given void fraction. The difference may
be attributable to different locations of the nickel
deposits in the two cases. I n impregnated catalysts the
nickel will reside for the most part inside the micro-
pores. F o r a coprecipitated catalyst, however, the
nickel may be expected to form composite particles
with the alumina support, so that when the particles
are pressed together to form a pellet there is increased
probability of nickel deposits contacting one another
and increasing the thermal conductivity. This effect
is likely to be of more importance a t low macropore
void fractions a s has been observed.
Comparison with predictive models
The experimental results were compared with four
models. These were the two modifications of the
Woodside and Messmer"), designated as Modification
1 and Modification 2 , a s described previously and also
the Consolidated particle and Porous solid models of
Harriott"'. The choice of t h e last two was governed
by the fact that these had previously been shown to
give better agreement than other models for silica
and silica based c a t a l y s t ~ " ~ " ~The
' . results of these
comparisons of experimental data with these models
are illustrated in Figures 2 and 3 which a r e represent-
ative of the results obtained for all t h e pellets in-
vestigated. In general, the lowest thermal conductivity
is given by the consolidated particles model whilst the
highest values a r e predicted by Modification 2 of the
Woodside and Messmer model. The other two models
predich intermediate values between these two ex-
tremes but lie closer t o the lower values a s predicted
by the consolidated particles model.
Values of €1, E., K , and K,' used in the models are
listed in Table 2 . The values of K,' quoted are cal-
culated from the appropriate value of eiPart.the par-
ticle micropore void fraction which is constant, un-
like E{ which varies with the macropore void fraction.
In principle K,' could be estimated from the intercept
if the plot of experimental values of K , versus macro-
pore void fraction but due to the uncertainty of the
shape of the curve a t E . less than 0.1 this procedure is
less satisfactory than a straightforward estimate from
Edwrt and the appropriate K , and Kf values. The K ,
values were taken a s the weight average for nickel
oxide and A1~03.
Figure 2 compares the experimental results with
the various model predictions for the 0.5% nickel im-
pregnated catalyst. Over the limited range of macm-
pore void fraction attainable using instrumented pel-
lets all models approximate to a straight line depend-
ency of effective thermal conductivity on macropore
void fraction. I t can be seen from Figure 2 however,
that Modification 2 of the Woodside and Messmer
model gave the closest agreement to t h e experimen-
tally measured thermal conductivities with a maximum
deviation of 15%. In contrast, all the other models
predicted effective thermal conductivities from 1/2 to
1/3 of the experimentally determined values. In addi-
tion, these three models did not reflect the sharp fall
in effective thermal conductivity with increasing
macropore void fraction whereas Modification 2
closely approximated this over the macropore void
range considered.
Figure 3 is a similar comparison made for the 75%
nickel impregnated catalyst. Again, Modification 2
gave t h e closest agreement with the experimental
results, even though the predicted results were con-

The Canadian Iournal of Chemical Engineering, Vol. 5 7 , February, 1979 27

sistently higher throughout the macropore range
taken. However, for this catalyst pellet both the por-
ous solid model and Modification 1 of Woodside and
Messmer’s model gave better agreement than for the
0.5% nickel impregnated catalyst and furthermore
showed a similar value for the decrease of effective
thermal conductivity with macropore void fraction to
t h a t obtained experimentally.
It should be pointed out however, t h a t when at-
tempts were made to compare the results for the pre-
dictive models with those obtained experimentally for
the coprecipitated catalyst, none of the models em-
ployed was satisfactory. Predicted values ranged be-
tween 20 and 40% of the experimental value a t the
lowest macropore void fraction to 35 tro 50% a t the
highest value of G. Additionally, not one of the models
predicted the upward concave shape obtained experi-
mentally (Figure 11, all four showing a linear de-
crease of effective thermal conductivity with macro-
pore void fraction. Of the models used, Modification 2
was closest to the experimental values.
Comparison with results for silica based catalysts
It is fairly well e s t a b l i ~ h e d ( ’ t~h)a t the model of
Woodside and Messmer in its unmodified form tends
to give much higher predictive values compared tb
experimental when used for porous pellets of the type
used as catalysts. It is interesting t h a t in the com-
parisons made in this work, the second modification
ccnsistently gave higher predicted values compared t o
the other 3 models. The question then arises a s to
whether Modification 2 is generally better than other
models or is this agreement obtained unique for
alumina based catalysts?
Therefore, Modification 2 was used to predict
values of the effective thermal conductivity for silica
and silica supported nickel catalysts which had been
measured previously by a n identical experimental
technique by Sharma et al.(’3). Comparisons were con-
fined to impregnated catalysts so t h a t in addition to
pure silica, nickel impregnated catalysts containing
0.5%, 5% and 62% Ni on silica were used. I t was
found that in all cases Modification 2 predicted higher
values than the experimental ones; only a t very large
macropore void fractions (ea > 0.4) were the results
a t all comparable, i.e. within 30% of the experimental
values. Perhaps, a more serious discrepancy of this
model when app!ied to silica based catalysts is t h a t
the predicted decrease of the K , with was very much
greater than t h a t observed experimentally. This, to-
gether with the observations listed above, means t h a t
a low macropore void fractions (en- 0.15) the %
difference between predicted and observed values
could be about 60%.
The difference in the applicability of the modified
Woodside and Messmer method to the two types of
catalyst may reside in the different structures of the
particles in the two cases. The silica used in the pre-
vious study was a material of high surface area
whereas the alumina surface area was much less.
Thus the miciyoporosity of the silica particles was
0.455 dm’/kg whereas that of the alumina was 0.404
dm3/kg. The particles of silica will have a much more
broken structure which will interrupt the heat flow
more than that of the alumina particles. Since a n
essential feature of Modification 2 a s adopted in this
work is the application of the model to the particles
in the f i r s t place, any difference in particle micro-
pore void fraction has a fundamental effect on the
calculated values of Ke. I t would seem t h a t application
of the Woodside and Messmer model to the micropores
28 The Canadian journal of Chemical Engineering, Vol. 5 7 , February, 1979
View publication stats
of the particles becomes less reliable a s the micro-
porosity of the particles increases.
conclusions
Experimental thermal conductivities of alumina
supported catalyst pellets have been made with various
amounts of nickel using a transient-method. Most of
the experiments were done for impregnated catalysts
but one coprecipitated catalyst was also investigated.
The experimental thermal conductivities were com-
pared with the predictions of models and it was found
that a n adaptation of the simple one parameter model
of Woodside and Messmer gave predictions in good
agreement with experimental For t h e impregnated
catalysts. The coprecipitated catalyst did not f i t any
model adequately; this is almost certainly due to the
different structures of this type of catalyst and until
better structural models of catalysts can be developed
this uncertainty in prediction will probably remain.
The agreement predictive and experimental results
may not hold for supports of higher microporosity as
is seen from an examination of previous results f o r
silica based catalysts. The present model should give
reliable estimates, however, for alumina based cata-
lysts.
The variation of thermal conductivity with nickel
content appears to be most sensitive to nickel loading
a t low and high nickel concentrations. I t would clearly
be of interest t o investigate the intermediate range
and establish whether the thermal conductivity is in
fact insensitive to nickel loading in this region.
Nomenclature
K, = effective thermal conductivity of Fellet
K, = thermal conductivity of fluid
K, = thermal conductivity of solid
K‘ = thermal conductivity of micro -0rous particles
R = radius of rellet
T = ternreratlire
a = thermal diffusivity
€a = macro-ore voidage
el = voidage of Darticles in pellet
~i p z r !. = yoidage of particles
0 = time
References
(1) Woodside, W. and Messmer, J.H., J. Appl. Phys. 32,
1688 (1961).
(2) Luikov, AiV., Shashkov, A.G., Vasiliev, L. and (.
Fraiman, Y.E., Int. J. Heat Mass Transfer 1, 117
(1968).
(3) Krupiczka, R., Int. Chem. Eng. 7, 122 (1967)
(4) Butt, J.B., AIChE. 11, 106 (1965).
(5) Harriott, P., Chem. Eng. 10, 65 (1975).
’ (6) Sehr, R.A., Chem. Eng. Sci. 9, 145 (1958).
( 7 ) Masamune, S. and Smith, J.M., Ind. Eng. Chem.
Fundam. 2, 136 (1963).
(8) Mischke, R.A. and Smith, J.M., Ind. Eng. Chem.
Fundam. 1, 258 (1962).
(9) Kling, V.G., Allg. Waerme techn. 3, 167 (1952).
. . Somerton. W.H. and Aime, M.M., J. Pet. Technol. 6 1
(10)
(1958).
(11) Schroeder. J.. Rev. Sci. Instrum. 34. El5 (1963).
i l 2 ) Cunningham,’ R.A., Carberry, J.J. ‘and Smith, J.M.,
AIChE J. 11, 636 (1965).
(13) Sharma, C.S., Harriott, P. and Hughes, R., Chem.
Eng. J. 10, 73 (1975).
(14) Sharma, C.S. and Hughes R., Can. J. Chem. Eng. 54,
358 (1976).
(15) Sharma, C.S., M.Sc., Diss., Univ. Salford (1973).
(16) Carslaw, H.S. and Jaeger, J.C., ‘The Conduction of
Heat in Solids’, Oxford Univ. Press, London (1947).
Manuscript received June 19, 1978; Accepted for publi-
catior! October 18, 1978.
* * *