Kinetics and intraparticle diffusion modelling of a complex multistep reaction: Hydrogenation of acetophenone over a rhodium catalyst

Article abstract of DOI:10.1006/jcat.1998.2019

In a first step the intrinsic kinetics of acetophenone hydrogenation on a Rh/C catalyst was studied in a semi-batch reactor. Tiny catalyst grains of 17 μm in average diameter were used in order to avoid any intraparticle diffusion limitation. Experiments were performed over a wide range of operating conditions and various Langmuir-Hinshelwood kinetic equations were discriminated over the complete conversion range. The model based on nondissociative adsorption of hydrogen and noncompetitive adsorption of the organic species with the gas molecules was found to fit better all the experimental data. The relevance of its optimized parameters was then discussed. In a second step the intraparticle diffusion limitations were also studied in a semi-batch reactor by varying the particle size. Two models based on the complex kinetics previously established were developed, taking into account the catalyst grain size and shape. After an additional adjustment of one of the adsorption constants ratios, they were found to provide a good representation of the data in terms of activity and selectivity.

Full text of DOI:10.1006/jcat.1998.2019

JOURNAL OF CATALYSIS 175, 328–337 (1998)
ARTICLE NO. CA982019
Kinetics and Intraparticle Diffusion Modelling of a Complex Multistep
Reaction: Hydrogenation of Acetophenone over a Rhodium Catalyst
,1
ꢀ
ꢀ
ꢀ
Isabelle Bergault, Pierre Fouilloux, Catherine Joly-Vuillemin, and Henri Delmas†
ꢀ
Laboratoire de Ge´nie des Proce´de´s Catalytiques, CNRS, CPE Lyon—BP2077, 43 Bd du 11 Novembre 1918, 69616 Villeurbanne Cedex, FRANCE;
and †Laboratoire de Ge´nie Chimique, CNRS, 18 Chemin de la Loge, 31078 Toulouse Cedex, FRANCE
Received October 9, 1997; revised December 8, 1997; accepted January 12, 1998
Mostofthekineticstudiesreportedintheliteratureconcern
the selective hydrogenation of acetophenone into phenyl
1-ethanol followed by the hydrogenolysis into ethylben-
zene. The studies performed with this complex reaction
scheme are often based on oversimple hypotheses as power
law kinetics (3, 4, 9–12) and only a few models, based on
the more realistic Langmuir–Hinshelwood hypothesis, are
available (7, 14, 20). They all suppose that the reaction at
the catalyst surface is the rate limiting step. Some authors
propose a competitive adsorption of the species but they
are not able to choose between a dissociative or nondis-
sociative adsorption of hydrogen because, either they per-
formed their experiments at a constant hydrogen pressure
(13), or the agreement between the experimental results
and the computed values obtained with the dissociative and
nondissociative models is not significantly different (7, 14).
In the present work, experiments were performed over a
wide range of operating conditions. Rhodium supported on
carbon was chosen as the catalyst and cyclohexane as a non-
hydroxylated solvent, in order to hydrogenate the aromatic
ring as well as the ketone function. Our aim was, first, to
select a suitable kinetic model fitting our experimental data
and, second, to discuss the methodology to be followed in
the case of a complex reaction scheme. The analysis of the
data was based on the complete kinetics, using the whole
concentration profiles and the hydrogen consumption ver-
sus time all along a given experiment. This method was
proved to be more effective than using only initial kinetics
(15). The model discrimination and parameter optimiza-
tion are reported in this paper and the results discussed.
The intraparticle diffusion effects were then studied and
a model developed in order to predict the reaction rates
and selectivities with the large particles used in trickle-bed
reactors.
In a first step the intrinsic kinetics of acetophenone hydrogena-
tion on a Rh/C catalyst was studied in a semi-batch reactor. Tiny
catalyst grains of 17 ꢀm in average diameter were used in order
to avoid any intraparticle diffusion limitation. Experiments were
performed over a wide range of operating conditions and various
Langmuir–Hinshelwood kinetic equations were discriminated over
the complete conversion range. The model based on nondissocia-
tive adsorption of hydrogen and noncompetitive adsorption of the
organic species with the gas molecules was found to fit better all
the experimental data. The relevance of its optimized parameters
was then discussed. In a second step the intraparticle diffusion lim-
itations were also studied in a semi-batch reactor by varying the
particle size. Two models based on the complex kinetics previously
established were developed, taking into account the catalyst grain
size and shape. After an additional adjustment of one of the ad-
sorption constants ratios, they were found to provide a good rep-
c
resentation of the data in terms of activity and selectivity.
ꢁ 1998
Academic Press
INTRODUCTION
Threephasecatalytichydrogenationsarecommonlyused
in a large number of chemical processes. They are usually
carried out in fixed bed reactors wherein gas and liquid are
flowing through a bed of large catalyst particles, or in reac-
tors containing a slurry of small catalyst particles suspended
in the gas–liquid mixture. The design and scale-up of these
multiphasic devices involve many aspects including com-
plex coupled phenomena and require accurate kinetic laws
as well as reactant transfer calculations.
Acetophenone hydrogenation has already been chosen
by other authors as a model reaction for kinetics studies.
On the one hand, it is of industrial relevance because its
intermediates and products are used in the pharmaceutical
and cosmetic industry. On the other hand, it is a good exam-
ple of a complex multistep reaction. The reported results
(1–8) show clearly the strong dependence of the reaction
scheme towards catalyst nature and operating conditions.
EXPERIMENTAL
The hydrogenation of acetophenone (Aldrich, 99% min)
was carried out using cyclohexane (RP Normapur Pro-
labo 99.5% min) as a solvent and in the presence of a
1 Corresponding author.
328
0021-9517/98 $25.00
c
Copyright ꢁ 1998 by Academic Press
All rights of reproduction in any form reserved.

HYDROGENATION OF ACETOPHENONE
329
Rh catalyst supported on coconut shell activated carbon
(Degussa, 3% Rh/C, G181 XKB/D). Rh/C was provided
as large particles of an average diameter dp = 3.37 mm.
A scanning microscope diagram showed that the metal is
not uniformly dispersed on the carbon support. The dis-
tribution of the metallic crystallites is of eggshell type, the
core being free of metal. A catalyst particle tortuosity fac-
tor equal to 5, generally considered as an average, was
taken. The other characteristic parameters were its den-
sity ꢁs = 2105 kg ꢂ mꢃ3, apparent density ꢁapp = 980 kg ꢂ mꢃ3
,
=
porosity εp = 0.53, and specific surface area SBET
870 m2 ꢂ gꢃ1. To obtain particle samples of uniform size
the commercial catalyst was crushed and sieved. The finest
particle size (dp = 17 ꢀm) was used for the intrinsic kinetic
study.
Experiments were carried out in a stainless steel agitated
semi-batch reactor with a classical manifold for hydrogen
supply at constant pressure. Its maximum capacity was 150
cm3 and its optimum content in volume was 60 to 75 cm3.
Baffles and magnetic agitation provided a good gas–liquid
mass transfer. The temperature regulation led to temper-
ature variations less than ꢄ1 K. In each experiment, 0.5 g
of catalyst (3% Rh/C, dp = 17 ꢀm) were charged with 45 g
of solvent. Catalyst was pretreated in cyclohexane under
hydrogen pressure (1.8 MPa) during 1 h at 368 K. Then, af-
ter the operating conditions were attained, acetophenone
contained in a reagent supply vessel was introduced into the
gas–liquid–solid mixture. The pressure was maintained con-
stant during an experiment by adjusting the hydrogen feed
to the consumption by means of a pressure regulator. Liq-
uid samples were withdrawn at regular intervals of time and
analysed by gas chromatography, using a polar and capillar
column (type BP20) and the data obtained from the exper-
iments were reproducible within ꢄ5%. We used constant
hydrogen pressures varying from 0.5 to 4.0 MPa, tempera-
tures from 333 to 373 K and acetophenone initial concen-
trations from 0.118 to 1.14 mol/liter. For the intraparticle
diffusion study, experiments were performed at T = 353 K
FIG. 1. Reaction scheme of acetophenone hydrogenation on Rh/C
catalyst.
that the last two represent only 3%. The simplified reaction
scheme of Fig. 1 was then proposed, where we considered
only the compounds with significant concentration.
Figures 2 and 3 show typical hydrogen consumption
curves and the concentration profiles of reaction products
plotted versus time during an experiment. The hydrogen
consumption was measured from the pressure loss in a stor-
age vessel and was systematically checked to be in good
agreement with the mass balance based on the liquid phase
concentrations. From these curves, initial global rates of
hydrogenation were measured, with an accuracy of ꢄ10%
and used for preliminary discussions on model discrimina-
tion. Concentration profiles versus time were required for
software determination of kinetic parameters.
Checking the Chemical Regime
under P = 2.5 MPa, for different particle sizes (17, 50, 100,
H2
In order to determine the intrinsic kinetics of the reac-
tion, it was necessary to ensure that none of the external
500, and 3370 ꢀm). The hydrogenation of pure intermedi-
ates like phenyl 1-ethanol or methylcyclohexylketone, was
also performed in order to elucidate the reaction scheme.
RESULTS
Kinetics of the Reaction Controlled by Chemical Regime
Reaction Scheme and Conversion Measurements
From the GC analysis, six molecules coming from the
hydrogenation of the aromatic ring and/or of the ketone
function were identified: methylcyclohexylketone, phenyl-
1-ethanol, cyclohexyl-1-ethanol, phenyl-1-ethanol, ethyl-
benzene, and ethylcyclohexane. Considering the contribu-
tion of each compound to the total mass balance, it appears
FIG. 2. Hydrogen consumption versus time for PH2 = 2.5 MPa, T =
353 K, mcata = 0.5 g, and CꢅAC = 0.6 mol ꢂ l ꢃ1
.

330
BERGAULT ET AL.
liquid–solid mass transfer (ϕL-S(H ) = kL-S ꢂ aP ꢂ Cꢀ ꢂ VL and
2
H2
ϕL-S(AC) = kL-S ꢂ aP ꢂ CAC ꢂ VL, expressed in kmol ꢂ sꢃ1 with
ϕL-S liquid–solid mass transfer coefficient). The minimum
possible value for kL-S was evaluated, considering that the
Sherwood number (Sh = kL-S ꢂ dp/Di) would equal 2 in this
case. Molecular diffusivities Di were estimated by Wilke
and Chang’s correlation (16). As an example, experimen-
tal results and calculations performed at T = 353 K, PH
=
2
4 MPa, and CꢅAC = 0.6 mol/liter gave
rH = 0.757 mmol ꢂ sꢃ1 < ϕL-S(H ) = 9.7 mmol/sꢃ1
2
2
rAC = 0.11 mmol ꢂ sꢃ1 < ϕL-S(AC) = 5.93 mmol/sꢃ1
.
FIG. 3. Concentration profiles versus time for PH2 = 2.5 MPa, T =
353 K, mcata = 0.5 g, and CAꢅ C = 0.6 mol ꢂ lꢃ1, (᭿) AC, (᭜) MCC, (᭡) CE,
(᭹) PE.
So the experimental rates are far lower than those totally
limited by liquid–solid mass transfer and cannot be affected
by diffusion to the grain surface.
or internal mass transfer resistance was limiting in our re-
action conditions. For this purpose, gas–liquid and liquid–
solid mass transfer, as well as internal diffusion limitations
were examined separately.
When gas–liquid mass transfer is totally limiting, the hy-
drogen consumption becomes independent from the cata-
lyst loading and the hydrogen flux through the gas–liquid
interface is expressed as (kG-L, aG-L gas–liquid mass transfer
coefficient and gas–liquid interphase area; VL liquid vol-
ume):
Theinternaldiffusionlimitationswereexperimentallyin-
vestigated. From experiments carried out for different cata-
lyst particle sizes at T = 353 K, the hydrogen consumption
was plotted versus time (Fig. 5). The obtained diagrams in-
dicate clearly that the internal diffusion is more and more
limiting with the increase in particle size. But the specific
hydrogenation rate is not increased farther for particles un-
der 50 ꢀm in diameter, proving that the internal diffusion
is a fortiori not limiting for the kinetic studies performed
with particles of 17 ꢀm.
ꢀ
ꢁ
ϕG-L = kG-LaG-L CHꢀ ꢃ CH ,1 VL (kmol ꢂ sꢃ1).
2
2
Background of the Kinetic Equations
Our kinetic models were based on the Langmuir–
Hinshelwood formalism. Different steps are involved in a
catalytic transformation such as adsorption of the reactants,
reaction on the active sites at the catalyst surface and des-
orption of the products. We supposed successively that each
ofthesestepswastherate-limitingstep. Thecompetitivead-
sorptionoftheorganicspeciesandhydrogen, involvingonly
one type of catalytic site, as well as a noncompetitive one,
involving two different types of sites, were also considered.
We also took into account the possibility of being dissocia-
tive or nondissociative for the adsorption of hydrogen.
Finally, 16 models were written, and for each of these
models it was supposed that the five reactions of the scheme
presented in Fig. 1 were following the same Langmuir–
Hinshelwood mechanism. A first selection was based on
the examination of Figs. 6 and 7 that show a nearly zero-
order dependence with respect to the initial acetophenone
concentration, and a varying order (from one to zero) with
respect to hydrogen pressure. Only the four models which
suppose that the reaction at the catalytic surface is the rate-
limiting step are able to predict such dependences. At this
stage, a more sophisticated analysis of the results based on
the complete kinetic curve is required in order to discrimi-
nate between a competitive or noncompetitive adsorption
of the species and a dissociative or nondissociative adsorp-
tion of hydrogen.
The influence of the mass of catalyst on the initial hy-
drogenation rate was then investigated for the most ꢅdras-
tic operating conditions (PH = 4 MPa, T = 373 K, CAC
=
2
0.6 mol/liter). The results, plotted on Fig. 4, show the lin-
ear dependence of the initial rate with the catalyst weight,
proving that the gas–liquid mass transfer is not limiting.
In order to determine if the liquid–solid mass trans-
fer resistance was limiting, the observed reactants con-
sumption rates (rH and rAC) were compared to the min-
2
imum transfer rates obtainable in the case of a limiting
FIG. 4. Initial rate of hydrogenation as a function of catalyst loading;
PH = 4 MPa, T = 373 K, CꢅAC = 0.6 mol ꢂ lꢃ1
.
2

HYDROGENATION OF ACETOPHENONE
331
FIG. 5. Hydrogen pressure drop in the storage vessel versus time: Influence of the particle size. PH = 2.5 MPa, T = 353 K, mcata = 1 g, and CꢅAC = 0.6
2
mol ꢂ lꢃ1 (particle diameter dp being respectively equal to 17, 50, 100, 500, and 3370 ꢀm for curves 1, 2, 3, 4, and 5).
Kinetic Equations
The rates are (ki rate constants and Ki adsorption con-
stants):
Mass balances corresponding to the four remaining mod-
els representing steps 1, 2, 3, 4, 5 of the reaction scheme
(Fig. 1) with respective rates r1, r2, r3, r4, r5 are written as
r1 = k1KACAC ꢀ X,
r2 = k2KACAC ꢀ X,
r3 = k3KMCCMCC ꢀ X, r4 = k4KPEPE ꢀ X,
r5 = k5KPEPE ꢀ X.
dAC
dt
mcata
V1
= (ꢃr1 ꢃ r2) ꢀ
,
t is time;
In the case where the mechanism is noncompetitive and
nondissociative X = ꢂ, noncompetitive and dissociative
X = ꢃ, competitive and nondissociative X = ꢄ , competitive
and dissociative X = ꢅ with:
dPE
dt
mcata
V1
= (r1 ꢃ r4 ꢃ r5) ꢀ
= (r2 + r5 ꢃ r3) ꢀ
,
,
PH = const;
2
dMCC
dt
mcata
V1
PH
He
2
CH,1 = CꢀH,1
=
;
ꢂ =
1
KH PH
2 2
He Henry’s const;
ꢀ
ꢁ
ꢁ
ꢀ
(1 + KACAC + KPEPE + KMCCMCC + KCECE)
1 + KH PH
2
2
dCE
dt
mcata
= (r3 + r4)
;
ꢃ =
V1
= mcata(r1 + 3r2 + r3 + 3r4 + 2r5).
p
KHPH
1
2
dnH
2
ꢀ ꢀ
p
(1 + KACAC + KPEPE + KMCCMCC + KCECE)
1+ KHPH
dt
2
FIG. 6. Initial hydrogenation rate as a function of initial acetophe-
none concentration: Phe´ = 2.5 MPa, (᭿) T = 333 K, (᭜) T = 353 K, and
(᭡) T = 373 K.
FIG. 7. Initial hydrogenation rate as a function of hydrogen pressure:
(᭿) T = 333 K, (᭜) T = 353 K, and (᭡) T = 373 K.

332
BERGAULT ET AL.
KH PH
2
2
is quite commonly encountered in the literature (17–19).
ꢄ =
ꢀ
ꢁ
2
It was also found that the adsorption constants could
not be accurately determined; multiplying each of them
by a constant factor gave the same set of optimized ki-
netic constants and an identical minimum value for the
best-fit function. A similar result had already been found
by (13), (15), and (33), indicating that only the adsorp-
tion constant ratios could be optimized with good preci-
sion. The models were then rewritten, introducing the ad-
sorption constant ratios. Qi = Ki/KPE ꢂ KPE was arbitrarily
fixed at a value of 100 liter ꢂ molꢃ1 and it was checked,
a posteriori, that our optimization results were not influ-
enced by this value. For a temperature of 353 K, 8i—the
minimal best-fit function value—amounted to 341 and 351,
respectively, for a competitive and noncompetitive disso-
ciative adsorption, but 216 (competitive) and 121 (noncom-
petitive) for nondissociative adsorption.
1 + KACAC + KPEPE + KMCCMCC + KCECE + KH PH
2
2
ꢅ =
p
KHPH
2
q
.
ꢀ
ꢁ
2
1 + KACAC + KPEPE + KMCCMCC + KCECE + KHPH
2
Optimization of the Kinetic Parameters
The model discrimination and parameter optimization
were carried out using a commercial dynamic simulation
software (MKF, Cheminform, Saint Petersbourg, Russia).
The nonlinear differential equation systems were solved
following an iterative method based on the linearisation
of the equations. Parameter optimization was performed
using the algorithm of the Simplex, based on the minimisa-
tion of the best-fit function value, defined as
ꢀ
ꢁ
P
Either with the noncompetitive model or the competitive
one, a dissociative adsorption of hydrogen did not provide
a good fit of the experimental data, the effects of hydrogen
2
Yci ꢃ Yei xp
100
Npts
i
8i =
ꢀ
,
ꢀ
ꢁ
2
Ymi axerror
pressure being not conveniently represented. At low pres-
p
where Yci is the calculated response value, Yei xp is the ex-
perimental value, and Ymi ax is the maximum experimental
value. The term “error” is the relative error on the experi-
mental value Ymi ax (in our case, the error was fixed at 0.05,
the value being representative of the chromatographic er-
ror). Npts is the number of points considered for the op-
timization. A statistical analysis was systematically per-
formedandtheconfidenceintervalsofeachparameterwere
calculated.
sure, KHP becomes much smaller than one and the
H2
p
dissociative models predict rates proportional to KHPH
2
and consequently a 0.5 order dependence with hydrogen
pressure which is not in good agreement with the exper-
imental observations (first order at low pressure). Only
the nondissociative models provided a good representation
of the hydrogen pressure effects. Considering all the data,
the noncompetitive model provided the best fit (8i = 121
instead of 216 for the competitive one). Similar compu-
tations performed at 333 and at 373 K led to the same
model. Activation energies and hydrogen adsorption heat
were evaluated, neglecting the effect of temperature on the
adsorption constants ratios.
Results of the Optimization
First of all, the influence of the adsorption constants
values on the optimization results was studied for the
competitive as well as for the noncompetitive models. In
In conclusion, the intrinsic kinetics of acetophenone hy-
drogenation is well represented by a noncompetitive and
nondissociative model, in our range of operating condi-
tions. Table 1 gives the optimum values of the kinetic pa-
rameters and their standard deviation and Fig. 8 shows the
good agreement between typical experimental concentra-
tion profiles and the selected model.
each case, the adsorption constants of the organic species
P
were found to be high (>50 liter ꢂ molꢃ1); the sum i KiCi
had to be much greater than one in order to predict the
zero order with respect to the acetophenone initial con-
centration. This fact corresponds to a situation where all
of the catalytic sites are occupied by the species, which
TABLE 1
The Values of the Kinetic Parameters
ln kꢅ1 (k1ꢅ in mol ꢂ kgꢃ1 ꢂ sꢃ1
ln kꢅ2 (k2ꢅ in mol ꢂ kgꢃ1 ꢂ sꢃ1
ln kꢅ3 (k3ꢅ in mol ꢂ kgꢃ1 ꢂ sꢃ1
ln kꢅ4 (k4ꢅ in mol ꢂ kgꢃ1 ꢂ sꢃ1
ln kꢅ5 (k5ꢅ in mol ꢂ kgꢃ1 ꢂ sꢃ1
)
)
)
)
)
26.98 ꢄ 0.01
25.41 ꢄ 0.17
20.16 ꢄ 0.29
27.77 ꢄ 0.38
30.79 ꢄ 1.08
9.97.10ꢃ9 ꢄ 0.02.10ꢃ9
1.68
E1 (kJ ꢂ molꢃ1
)
)
)
)
)
79.65 ꢄ 0.17
80.79 ꢄ 2.58
69.47 ꢄ 0.89
84.62 ꢄ 1.02
100.26 ꢄ 3.27
ꢃ49.51 ꢄ 0.1
0.63
E2 (kJ ꢂ molꢃ1
E3 (kJ ꢂ molꢃ1
E4 (kJ ꢂ molꢃ1
E5 (kJ ꢂ molꢃ1
KꢅH (MPaꢃ1
)
1HH (kJ ꢂ molꢃ1
)
2
Q
K
AC (ꢃ)
Q
Q
MCC (ꢃ)
CE (ꢃ)
PE (m3 ꢂ kmolꢃ1
)
100
5, 7.10ꢃ2

HYDROGENATION OF ACETOPHENONE
333
with the intrinsic kinetics of the reaction over a wide range
of operating conditions and is useful as such for reactor
design.
The second point of this discussion is in relation with
the reliability of the estimated parameters. The statisti-
cal analysis provides good confidence interval values, in-
dicating that each parameter is accurately obtained. The
influence of the KPE value, fixed at 100 liter ꢂ molꢃ1 for the
study, was evaluated and found to be negligible on the other
optimized parameters, as well as on the best-fit function
values, as soon as KPE is greater than 50. The adsorption
constant ratios are following the expected tendency, in-
creasing with the reactivity of the molecules towards the
surface of the metal, due to their degree of unsaturation
FIG. 8. Experimental and simulated profiles: hydrogenation of ace-
tophenone at T = 333 K, under 2.55 MPa hydrogen pressure and for
[CꢅAC] = 0.631 mol ꢂ lꢃ1
.
(KAC > KPE > KMCC > KCE). Concerning the value of KH
2
Discussion
expressed in usual units (KH = 4 liter/mol at T = 353 K),
2
the literature data are so scattered that the value found is in
the convenient range (from 10ꢃ2 liter/mol to 400 liter/mol
from (18, 26, 30, 31)). The heat of adsorption optimized
for hydrogen (1HH = ꢃ49.51 kJ/mol) is slightly larger than
those given in the bibliography, but the difference is accept-
able (17, 19, 31). The estimated activation energies are also
in good agreement with the published results (17–19, 30).
Their values are very close to one another for the different
steps of the reaction scheme, which agrees fully with the
fact that no influence of temperature on the selectivities
was observed.
The last point concerns the interaction effects of the
species. The selected model and the corresponding set of
parameterswellrepresenttheintrinsickineticsofacetophe-
none hydrogenation; they should also fit the kinetics of
hydrogenation of intermediates such as phenyl 1-ethanol
and methylcyclohexylketone. To verify this fact, experi-
ments were then performed with pure phenyl 1-ethanol or
pure methylcyclohexylketone in the same range of oper-
ating conditions. The experimental results were simulated
using the noncompetitive and nondissociative model and
the parameter values given in Table 1. They are suitable
to represent phenyl 1-ethanol hydrogenation (maximum
disagreement 10%) but not well adapted to the kinetics
of methylcyclohexylketone (68% discrepancy with exper-
imental data). It was also checked that another set of
parameters could not provide a good fitting for both hydro-
genation of acetophenone and hydrogenation of methyl-
cyclohexylketone. Vice versa, the optimization of kinetic
parameters on hydrogenation of intermediates cannot be
directly used for hydrogenation of acetophenone because
they do not provide the best fit. They can only be used as
initial values. Similar limitations had already been observed
(15) for hydrogenation of nitriles. The interaction between
the adsorbed species are probably not negligible and more
sophisticated models, including these effects and account-
ing for the complexity of the catalytic surface, should be
developed.
The first point to be discussed is the validity of the se-
lected model. When referring to the few kinetics studies
on the hydrogenation of acetophenone, even if the com-
petitive model is often preferred in the case of the selec-
tive hydrogenation of acetophenone into phenyl 1-ethanol
(7, 14). The noncompetitive model is also proposed (20), in
the case of a complex reaction scheme involving the hydro-
genation of the aromatic ring as well as the ketone func-
tion. It is obvious that the nature of the unsaturated func-
tion has a drastic influence on the adsorbates behaviour.
When the reaction involves the hydrogenation of a double
bond, or of a carbonyle in a linear molecule, the proposed
model often considers the competitive adsorption of the
species with hydrogen (21–25). But noncompetitive models
seem to better represent the hydrogenation of the aromatic
ring (17, 19, 26–28). This observation is in good agreement
with the fact that the ring is always strongly adsorbed at
the catalyst surface (26). Recently a noncompetitive model
(33, 34) was proposed for the hydrogenation of several aro-
matic molecules. Nethertheless, the authors indicate that
the hydrocarbon adsorption is probably halfway between
competitive and noncompetitive because of the small size
of the hydrogen molecule. Concerning hydrogen, dissocia-
tively adsorbed species often coexist with nondissociated
ones (bridged and on top species). Among the above ref-
erences on the hydrogenation of the aromatic ring, some
authors consider a nondissociative adsorption of hydrogen.
The first to zero-order dependence with hydrogen pressure
observed in our case would be in good agreement with the
adsorption of at least one molecular hydrogen species, per-
haps in minor quantity but presenting a high reactivity. The
hydrogenation of the aromatic ring by addition of hydro-
genmolecularentitieshasalreadybeendemonstratedinthe
case of the benzene molecule in a gas phase (29). Finally,
the noncompetitive and nondissociative model selected for
hydrogenation of acetophenone over a Rh/C catalyst is in
reasonable agreement with the bibliographic results. What-
ever the questions on its unicity, it gives a good formal fit

334
BERGAULT ET AL.
In conclusion, this study shows the complexity of model
TABLE 2
discrimination and parameter optimization for a multistep
reaction scheme. In the case where a more realistic model
would be wanted the adsorbed species lateral interactions
could not be neglected and should be taken into account.
Nevertheless, the noncompetitive and nondissociative hy-
potheses selected here to represent the intrinsic kinetics of
acetophenone hydrogenation provide a good representa-
tion of all the concentration profiles in a batch reactor over
a wide range of operating conditions within the interval of
experimental precision. In particular this model remains
perfectly adapted for reactor calculations.
Set of Equations Solved Numerically
Kinetics Modelling in a Diffusion Regime
Model Assumptions
The catalyst was purchased as 3.37-mm particles to be
used in fixed bed reactors like trickle beds. That is why we
needed to study the effects of the particle size on the activ-
ity. For this purpose, experiments were performed in a semi-
batch reactor for different particle sizes. The hydrogen con-
sumption plotted versus time indicates clearly a decrease in
activity in function of the particle diameters due to inter-
nal diffusion (see Fig. 5). The concentration profiles have
also been represented versus time for samples of uniform
particle size. They show that selectivities were affected to
some extent by intraparticle diffusion limitation: when the
particle size was varied from 17 to 3370 ꢀm, the maximum
phenyl 1-ethanol yield is altered from 65 to 55%. In order
to predict the behaviour of multiphase reactors, wherein
internal mass transfer limitations are large, an intraparticle
diffusion model was developed using the intrinsic kinetics
established above.
Because of the catalyst characteristics described in the
experimental part, the large particles (dp = 3.37 mm) were
considered as spherical and impregnated with rhodium
metal with an “eggshell” distribution. The fraction of the
catalyst particles containing metal is called the active frac-
tion (noted Factive) and depends on the thickness of the im-
pregnation layer (noted Ep0). In this layer, the rate per mass
of active catalyst (containing metal) r0i, is equal to the rate
given by the kinetic study ri (given per total mass of cata-
lyst) divided by Factive. Small particles were obtained by
crushing and sieving the largest of the commercial catalyst.
They were also considered as spherical and it was supposed
that only a part of them (equal to Factive) contained metal
contributing to the reaction, the other part being free of
rhodium and totally inactive.
Ep), was always small, compared to the largest diameter
dp, and could be modelled by a plane layer. The sets of
equations in Table 2 were solved numerically using an or-
thogonal collocation method (32) for the mass balance on
the solid and a fourth-order Runge–Kutta method for the
mass balance on the liquid.
Two models were developed: one being written for the
totally active small particle (dp ꢆ 100 ꢀm) and considering
a spherical geometry, and the other for the large particles
(dp = 3.37 mm) and considering a plane geometry for the
active layer. It was controlled a posteriori by electron mi-
Rate Expressions (mol/kg Active Catalyst/s)
r0 = ri/Factive with ri function of the concentrations inside
croscopy that the active layer of the catalyst grains (noted theisolid whatever the size of the particle.

HYDROGENATION OF ACETOPHENONE
335
FIG. 10. Comparison of simulated and experimental profiles: Ep0
=
1000 ꢀm; dp = 100 ꢀm; CAꢅ C = 0.604 mol ꢂ lꢃ1; T = 353 K; PH2 = 2.5 MPa;
mcata = 1.0 g; (᭿) AC; (᭜) MCC; (᭹) PE; (᭡) CE; ——— simulation.
FIG. 9. Ep0 value effects on the simulations: dp = 3,37 mm; CAꢅ C
=
0,604 mol ꢂ lꢃ1; T = 353 K; PH = 2,5 MPa; mcata = 1.0 g; (᭿) AC; (᭜) MCC;
(᭡) PE; (᭹) CE; ———. S2imulated curve for Ep0 = 1000 ꢀm and ----
Ep0 = 300 ꢀm.
Effectiveness factors ꢆ1 and ꢆ2 are always lower than one
varying from 0.9 to 0.78 for dp = 50 ꢀm and from 0.7 to 0.45
for dp = 100 ꢀm, as long as acetophenone concentration is
not equal to zero. These results indicate clearly the increas-
ing intraparticle diffusion limitation with the particle size.
ꢆ3 , ꢆ4, and ꢆ5 (related to the reactions of the intermediates
PE and MCC) are greater than one at the beginning and
decrease later to values lower or equal to one. Intermedi-
ates are formed inside the pores of the catalyst. Their con-
centrations are, at the beginning of the experiment, higher
inside the particle than at the surface, leading to effective-
ness factors greater than one. The more limiting the inter-
nal diffusion, the higher is this overconcentration. When
the acetophenone disappears, the intermediate formation
slows down and their profiles inside the particle become
concave. Then, the effectiveness factors decrease to values
lower than one.
Results and Discussion
First of all, the thickness of the impregnation layer Ep0
was evaluated. Electron microscope diagrams provide only
an approximate value of Ep0 because, on the one hand, the
particles used are not perfectly spherical as considered in
the model and, on the other hand, very small and highly
active rhodium crystallites can possibly be present at the
center of the catalyst but are not detected by the electron
micrographs. A good estimation of Ep0 value being essential
for the rate and effectiveness factors computations, experi-
ments using different particle sizes were simulated for sev-
eral Ep0 values. For particle diameters smaller than 100 ꢀm,
the simulations were found to be insensitive to the Ep0 value
andtheconcentrationprofileswereaswellrepresentedwith
Ep0 = 300 ꢀm as with Ep0 = 1000 ꢀm. Conversely, simula-
tions were found to be highly sensitive to the Ep0 value in
the case of large particles as is shown in Fig. 9. This result
seems logical since the intraparticle diffusion is highly lim-
iting for this particle size. With Ep0 = 1000 ꢀm providing the
best fit to all our experimental datas, this value was used for
the whole study.
For large particles, the computed effectiveness factors ꢆ1
and ꢆ2 (Fig. 12) were very low (from 0.07 to 0.045), which
is consistent with the observed hydrogenation rates. ꢆ3, ꢆ4,
and ꢆ5 are greater than one at the beginning and decrease
rapidly to values lower than one (0.05 to 0.15), indicating
the strong limitation by intraparticle diffusion.
Experiments performed with different particle sizes were
simulated and the results compared to the experimental
data. In order to confirm that intraparticle diffusion was
not limiting for the kinetic study, every experiment per-
formed with particles of 17 ꢀm diameter was simulated
and the effectiveness factors were calculated. They were
found to be equal to 1 ꢄ 0.05 in every case, proving the ki-
netic regime and in good agreement with the experimental
results on the influence of the particle size between 17 and
50 ꢀm (see Fig. 5). For small particles (dp = 50 ꢀm and dp =
100 ꢀm), simulated profiles were found to be in good agree-
ment with the experimental ones. As an example, Figs. 10
and 11 show the results obtained for dp = 100 ꢀm. Only
the maximum obtained for the phenyl 1-ethanol profile is
slightly underestimated by the model, but the difference
remains acceptable.
FIG. 11. Effectiveness factors (᭿) ꢆ1ꢃꢆ2; (♦) ꢆ3, and (᭺) ꢆ4ꢃꢆ5 for
Ep0 = 1000 ꢀm; dp = 100 ꢀm; CꢅAC = 0.604 mol ꢂ lꢃ1; T = 353 K; PH
2.5 MPa; mcata = 1.0 g.
=
2

336
BERGAULT ET AL.
FIG. 12. Effectiveness factors (᭿) ꢆ1ꢃꢆ2; (4) ꢆ3, and (♦) ꢆ4ꢃꢆ5 for Ep0 = 1000 ꢀm; dp = 3370 ꢀm; CꢅAC = 0.604 mol ꢂ lꢃ1; T = 353 K; PH2 = 2.5 MPa;
mcata = 1.0 g.
Simulated profiles were also found to be in good agree-
ment with the experiments despite the fact that the phenyl
1-ethanol concentration is once again slightly underesti-
mated (25%). One way to improve the fit of the PE profile
for all the particle sizes is to adjust the value of the adsorp-
tion constants ratio QAC by increasing it from QAC = 1.67
(given by the intrinsic kinetics study) to QAC = 4 (Fig. 13).
This adjustement can probably be considered as a conse-
quence of the species interaction effect. As was previously
observed during the kinetics study, some parameter values
such as the adsorption constants could possibly be affected
by the concentration of the species in the solution. The in-
teraction between adsorbed species are probably not negli-
gible and this phenomenon is not accounted for in the used
Langmuir–Hinshelwood equations used. Nevertheless, by
adjusting one of the adsorption constant ratios, the pro-
posed intraparticle diffusion models provide a very good
representation of the experimental data in a wide range of
particle sizes in terms of activity and selectivity. They will be
helpful to predict the performances of multiphase reactors
using large particles such as fixed beds.
GENERAL CONCLUSION
The kinetics of acetophenone hydrogenation on a Rh/C
catalyst was experimentally studied in the conditions of
chemical and diffusion regimes. The discrimination be-
tween different possible reaction mechanisms and the pa-
rameter optimization were made by means of a computer
program, especially in the case of diffusion limiting transfer.
This is useful for the modelling of fixed bed reactors where
large catalyst particles are used. A noncompetitive and
nondissociative model based on a Langmuir–Hinshelwood
type surface reaction was compared to the experiments. In
this way it was possible to obtain a mathematical repre-
sentation of the experimental results within the limits of
experimental precision. Nevertheless, some hydrogenation
experiments on pure intermediates of the reaction, in par-
ticular the strongly adsorbed phenyl-1 ethanol, showed that
this model, neglecting the adsorbed species interaction ef-
fects, was not sophisticated enough to be universal.
REFERENCES
1. Mushenko, D. V., Lebedeva, E. G., Khimich, V. P., Chagina, V. S., and
Barinov, N. S., Zh. Prikl. Khim. 39, 2766 (1966).
2. Mushenko, D. V., Lebedeva, E. G., Chagina, V. S., Khimich, V. P., and
Barinov, N. S., Zh. Prikl. Khim. 39, 2769 (1966).
3. Barinov, N. S., Mushenko, D. V., and Lebedeva, E. G., Zh. Prikl. Khim.
42, 2398 (1969).
4. Barinov, N. S., Lebedeva, E. G., and Mushenko, D. V., Zh. Prikl. Khim.
42, 2613 (1969).
5. Barinov, N. S., and Mushenko, D. V., Zh. Prikl. Khim. 46, 940 (1973).
6. Barinov, N. S., and Mushenko, D. V., Izvest. Otdel. Khim. Nauk. 6, 507
(1973).
7. Geneste, P., and Lozano, Y., C.R. Acad. Sc. Paris C 280, 1137
(1975).
8. Masson, J., Cividino, P., Bonnier, J. M., and Fouilloux, P., Heteroge-
neous Catalysis and Fine Chemicals II 59, 245 (1991).
9. Mushenko, D. V., Barinov, N. S., and Lebedeva, E. G., Zh. Organ.
Khim. 7, 314 (1971).
FIG. 13. Experimental and simulated concentration profiles in semi-
batch reactor for dp = 3370 ꢀm; CꢅAC = 0.604 mol ꢂ lꢃ1; T = 353 K; PH2 = 2.5
MPa; mcata = 1.0 g; (᭿) AC; (᭜) MCC; (᭡) CE; (᭹) PE; (᭿) simulations
with Qac = 4; ——— simulations with Qac = 1.67.

HYDROGENATION OF ACETOPHENONE
337
10. Csomontanyi, G., Netta, M., and Balmez, M., Revue roumaine de
chimie 18, 1367 (1973).
23. Chaudhari, R. V., Parande, M. J., Ramachandran, P. A., Brahme, P. H.,
Vadgaonkar, H. G., and Jaganathan, R., AIChE. J. 31, 1891 (1985).
11. Serebryakov, B. R., Smirnova, N. A., and Garbuzov, V. G., Azerb. 24. Chaudhari, R. V., Jaganathan, R., Kolhe, S., Emig, G., and Hofmann,
Khim. Zh. 6, 135 (1977).
H., Chem. Eng. Sci. 41, 3073 (1986).
25. Smedler, G., Can. J. Chem. Eng. 67, 51 (1989).
26. Kawakami, K., Ura, K., and Kusunoki, K., J. Chem. Eng. Japan 9, 392
(1976).
12. Del Amo, F. B., Ingenieria Quimica (Madrid) 23, 115 (1991).
13. Cerveny, L., Dobrovolna, Z., Belohlav, Z., and Kluson, P., Collect.
Czech. Chem. Commun. 61, 764 (1996).
14. Ziyatdinov, A. S., Stepanenko, V. V., Chernykh, I. S., Leonova, 27. Broderick, D. H., and Gates, B. C., AIChE J. 27, 663 (1981).
E. B., Pisarenko, V. N., and Kafarov, V. V., Zh. Priklad. Khim. 61,
28. Kut, O. M., Yu¨celen, F., and Gut, G., J. Chem. Tech. Biotechnol. 39,
565 (1988).
107 (1987).
15. Joly-Vuillemin, C., Gavroy, D., Cordier, G., De Bellefon, C., and 29. Mirodatos, C., Dalmon, J. A., and Martin, G. A., J. Catal. 105, 405
Delmas, H., Chem. Eng. Sci. 49, 4839 (1994).
16. Wilke, C. R., and Chang, P., AIChE. J. 1, 264 (1955).
17. Zwicky, J. J., and Gut, G., Chem. Eng. Sci. 33, 1363 (1978).
18. Gut, G., Kosinka, J., Prabucki, A., and Schuerch, A., Chem. Eng. Sci.
34, 1051 (1979).
(1987).
30. Turek, F., Chakrabarti, R. K., Lange, R., Geike, R., and Flock, W.,
Chem. Eng. Sci. 38, 275 (1983).
31. Gupta, A. K., Bhattacharyya, and Saraf, S. K., Indian J. Technol. 23,
184 (1985).
19. Turek, F., Geike, R., and Lange, R., Chem. Eng. Processes. 20, 213
(1986).
20. Smirnova, N. A., Adamyan, V. L., Edigavora, E. I., and Likhushkin,
I. A., Azerb. Khim. Zh. 4, 44 (1987).
21. Kishida, S., and Teranishi, S., J. Catal. 12, 90 (1968).
22. Lemcoff, N. O., J. Catal. 46, 356 (1977).
32. Villadsen, J. V., and Stewart, W. E., Chem. Eng. Sci. 22, 1483 (1976).
33. Toppinen, S., Salmi, T., Rantaquia¨, T. K., and Aittanaa, J., Ind. Eng.
Chem. Res. 36, 2101 (1997).
34. Salmi, T., Toppinen, S., Rantaquia¨, T. K., and Aittanaa, J., in “Proc.
First European Congress on Chemical Engineering, Florence, Italy,
May 4–7, 1997,” p. 2947.