You are on page 1of 10

Catalytic and kinetic study of the liquid-phase hydrogenation of acetophenone

over Cu/SiO
2
catalyst
Nicola s M. Bertero, Carlos R. Apestegua, Alberto J. Marchi *
Catalysis Science and Engineering Research Group (GICIC), Instituto de Investigaciones en Catalisis y Petroqumica (INCAPE), UNL-CONICET, Santiago del Estero 2654,
3000 Santa Fe, Argentina
1. Introduction
The selective hydrogenation of aromatic ketones is of great
importance in Fine Chemistry because the resulting phenyl
alcohols nd application as avors, fragrances and intermediates
for the synthesis of more valuable organic molecules [1]. However,
the selective hydrogenation of the C O group of the ketone
molecule is frequently not easily achieved due to the formation of
unwanted products via side reactions such as the aromatic ring
hydrogenation. In addition, the hydrogenolysis of the unsaturated
alcohol to the corresponding aromatic hydrocarbon is other
possible and unwanted reaction. The selective hydrogenation of
acetophenone (AP) to 1-phenylethanol (PhE) has been investigated
on different metals because PhE is widely employed in pharma-
ceutical and fragrance industries [1,2]. For example, in pharma-
cology PhE is an intermediate of analgesic and anti-inammatory
drugs, while in food additives it is widely used in chewing gums
and yogurts because of its characteristic strawberry scent.
A simplied reaction network for the AP conversion is shown in
Fig. 1. PhE and cyclohexylmethylketone (CHMK) are primary
products formed by hydrogenation of the carbonyl group and the
aromatic ring of AP molecule, respectively. Both primary products
can be hydrogenated to 1-cyclohexylethanol (CHE). On the other
hand, the hydrogenolysis of the COH bond in PhE produces
ethylbenzene (EB) that is not a valuable product in this reaction.
The AP hydrogenation has been studied on noble metals such as
Pt, Ru and Pd [39]. Platinum hydrogenates the carbonyl and
phenyl groups of the AP molecule at similar rates producing
comparable amounts of PhE and CHMK. Both products are then
consecutively hydrogenated to CHE [3,4]. Ruthenium exhibits a
high ability for reducing the aromatic ring of AP and is therefore
poorly selective for synthesizing PhE [5,6]. Palladium promotes
initially the selective hydrogenation of C O bond of AP forming
PhE but is also active for consecutively converting PhE to EB by
hydrogenolysis [79]. The AP hydrogenation has been also
investigated on non-noble metals, especially on Ni-based catalysts
[1016]. These studies agree in that nickel promotes the selective
AP hydrogenation to PhE, but the concomitant formation of
variables quantities of CHMK and CHE is also observed. Moreover,
nickel is active for producing EB by PhE hydrogenolysis. Very
few papers have studied the use of Cu-based catalysts for AP
Applied Catalysis A: General 349 (2008) 100109
A R T I C L E I N F O
Article history:
Received 2 June 2008
Received in revised form 10 July 2008
Accepted 15 July 2008
Available online 23 July 2008
Keywords:
Selective hydrogenation
Acetophenone
Copper-based catalyst
Kinetic modeling
A B S T R A C T
The liquid-phase hydrogenation of acetophenone (AP) to 1-phenylethanol (PhE) was studied on Cu(6.8%)/
SiO
2
catalyst. Catalytic tests were carried out in a batch reactor by varying temperature, total pressure
and AP initial concentration between 353373 K, 520 bar, and 0.0380.251 M, respectively, and using
four different solvents: isopropylic alcohol (IPA), cyclohexane, toluene and benzene. The selectivity to
PhE was about 100% irrespective of the solvent used, but the initial AP conversion rate followed the order
IPA > cyclohexane > toluene > benzene. The differences in catalyst activity when changing the solvent
were interpreted by considering the effect of the solventmetal interaction on the relative coverage of
adsorbed reactant species. Experimental data were well interpreted by kinetic modeling only when
assuming that: (i) the adsorption of AP and H
2
is competitive; (ii) AP adsorption is strong; (iii) copper
surface is saturated in AP; (iv) the PhE coverage on the catalyst is negligible. These assumptions were
consistent with the fact that the reaction was negative order with respect to AP and rst order in H
2
. The
highly selective AP hydrogenation to PhE was explained by considering that the strong electrostatic
repulsion between metallic Cu and the phenyl group tilts the AP molecule thereby favoring its adsorption
via the carbonyl group and the formation of the unsaturated alcohol. Also, PhE was not consecutively
converted via hydrogenolysis or other acid catalyzed reactions since the support was inert.
2008 Elsevier B.V. All rights reserved.
* Corresponding author. Fax: +54 342 4531068.
E-mail address: amarchi@q.unl.edu.ar (A.J. Marchi).
Contents lists available at ScienceDirect
Applied Catalysis A: General
j our nal homepage: www. el sevi er . com/ l ocat e/ apcat a
0926-860X/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2008.07.014
hydrogenation. Zaccheria et al. [17] reported that a Cu(8%)/Al
2
O
3
catalyst converts aryl ketones, including AP, into the corresponding
alcohols with selectivities higher than 90%. In contrast, they
observed the formation of signicant amounts of hydrogenolysis
products and ethers when using Cu/SiO
2
, probably because of the
catalyst acidity. Two patents fromSumimoto Chemical Co. claimed
for the use of Cu/SiO
2
catalysts to selectively hydrogenate AP to
PhE [18,19]. High PhE yields from AP would be obtained when
using Cu/SiO
2
catalysts with Cu loadings higher than 65% [18] or
modied by the addition of alkaline or alkaline earth metals [19].
From the above survey on the liquid-phase AP hydrogenation
literature it seems that metallic copper is the best candidate for
selectively obtaining PhE. However, the causes for this superior
PhE selectivity on Cu have not been established yet. Studies on the
AP hydrogenation reaction kinetics and mechanism on copper
catalysts are also lacking. On the other hand, it has to be noted that
the type of solvent used in the AP hydrogenation reaction may
signicantly modify the catalyst activity. For example, Aramenda
et al. [7] studied the AP hydrogenation over Pd-supported catalysts
using different alcohols as solvent and observed that the catalytic
activity decreases when increasing the alcohol dielectric constant.
Bejblova et al. [9] observed large differences in AP conversion
activity on Pd when a high-polar solvent (methanol) was replaced
by a non-polar one (n-hexane). Similarly, it was reported that the
AP hydrogenation activity on Ni Raney was larger in cyclohexane
(non-polar) than in methanol, ethanol or n-propanol [13].
In this paper, we present a detailed study of the liquid-phase
hydrogenation of acetophenone over a Cu(6.8%)/SiO
2
catalyst.
Experimental data were obtained by varying the operative
conditions (temperature, H
2
pressure, AP concentration) and
using four different solvents (isopropyl alcohol, cyclohexane,
benzene and toluene). Results were interpreted by kinetic
modeling employing heterogeneous LangmuirHinshelwood
HougenWatson (LHHW) and non-stationary models. A reaction
mechanismbased on the interaction between reactants and Cu
0
is
proposed to explain the selective AP hydrogenation to PhE. The
effect of solvent on catalyst activity is interpreted by considering
solventreactant and solventmetal interactions.
2. Experimental
2.1. Catalyst preparation
A Cu/SiO
2
catalyst with 6.8 wt.% of copper was prepared by
incipient wetness impregnation method. Copper was deposited on
commercial silica (Grace G62, 99.7%, Na: 0.1%, SO
4
2
: 0.1%, others:
0.1%) by adding dropwise a 0.57 M aqueous solution of Cu(NO
3
)
2
(Merck, 98%). The solid was dried in an oven at 373 K for 12 h and
then calcined in air owat 673 K for 3 h. Prior to catalytic tests, the
catalyst was activated ex situ in H
2
ow at 543 K for 2 h and then
quickly transferred to the reactor in an inert atmosphere of N
2
in
order to avoid metal reoxidation.
2.2. Characterization of the catalyst
The crystalline structure of the sample was determined by X-
ray diffraction (XRD) using a Shimadzu XD-1 diffractometer and
Ni-ltered Cu Ka radiation in the range of 2u = 10808 at a scan
speed of 28/min. Crystallites size were calculated from the CuO
(1 1 1) diffraction lines using the DebyeScherrer equation.
BET surface area (S
g
), pore volume (V
P
) and mean pore diameter
(d
P
) were measured by N
2
physisorptionat 77 Kin a Quantochrome
Corporation NOVA-1000 sorptometer. Copper content was deter-
mined by atomic absorption spectroscopy (AAS).
The reducibility of the calcined sample was determined by
temperature-programmed reduction (TPR) using a Micromeritics
AutoChem II 2920 V2.00 equipment. The TPR prole was obtained
using a H
2
(5%)/Ar gaseous mixture at 60 cm
3
/min STP and a sample
size of about 100 mg. The sample was heated from 298 to 773 K at
10 K/min. H
2
uptake was measured using a TCD detector. Because
water was formed during sample reduction, the gas exiting from
the reactor was passed through a cold trap before entering the
thermal conductivity detector.
Sample acidity was determined by temperature-programmed
desorption (TPD) of NH
3
preadsorbed at 373 K. Samples (100 mg)
were treated in He (60 cm
3
/min) at 723 K for 2 h and then exposed
to a 1% NH
3
/He stream for 40 min at 373 K. Weakly adsorbed NH
3
was removed by ushing with He at 373 K during 2 h. Temperature
was then increased at a rate of 10 K/min and the NH
3
concentration
in the efuent was measured by mass spectrometry in a Baltzers
Omnistar unit.
The metal copper dispersion was determined by titration with
N
2
O at 363 K using a stoichiometry of (Cu
0
)
s
/N
2
O = 2, where (Cu
0
)
s
implies a Cu
0
atom on surface [20,21]. Pre-reduced samples were
exposed to pulses of N
2
O(10%)/Ar. The number of chemisorbed
oxygen atoms was calculated from the consumption of N
2
O
measured by mass spectrometry (MS) in a Baltzers Omnistar unit.
2.3. Catalytic tests
The liquid-phase hydrogenation of AP (Aldrich, 99%) was
carried out in a 600 ml autoclave (Parr 4843) equipped with a
mechanic stirrer. The temperature, total pressure and AP initial
concentration were varied between 353 and 373 K, 5-20 bar and
0.08380.2514 M, respectively. The solvents used were isopropylic
alcohol (Aldrich, 98%), cyclohexane (Merck, 99%), toluene (Aldrich,
99%) and benzene (Merck, 99%). The autoclave was loaded with
150 ml of solvent and 1 g of catalyst in a N
2
inert atmosphere. The
reaction system was stirred and heated to reaction temperature at
2 K/min. Then, 1.54.5 ml of AP were injected to the reactor and
the pressure was rapidly increased to 3.7, 8.7 or 18.7 bar with
hydrogen. The batch reactor was assumed to be perfectly mixed. It
was veried that the stirring speed (600 rpm) and the catalyst
particle size (<100 mm) used insured the kinetic control of the
reaction; i.e., diffusional limitations were negligible.
The concentrations of unreacted AP and reaction products were
followed during the reaction by ex situ gas chromatography using
an Agilent 6850 chromatograph equipped with ame ionization
detector (heated at 523 K), temperature programmer, and a 30 m
Innowax capillary column with a 0.25 mm coating. Liquid samples
Fig. 1. Reaction network for acetophenone hydrogenation over metal catalysts.
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 101
were withdrawn fromthe reactor by using a loop under pressure in
order to avoid ushing. Data were collected every 515 min at the
beginning of the reaction and then every 3060 min. In all the
cases, the only reaction product detected was PhE. AP conversion
(X
AP
, mol of AP reacted/mol of AP fed) was calculated as X
AP
=
(C
0
AP
C
AP
)=C
0
AP
; where C
0
AP
is the initial concentration of acet-
ophenone and C
AP
is the acetophenone concentration at reaction
time t.
3. Results
3.1. Catalyst characterization
The specic surface area (S
g
= 220 m
2
/g) and pore volume
(V
P
= 0.42 cm
3
/g) determined for the Cu/SiO
2
sample were similar
to those corresponding to SiO
2
(S
g
= 230 m
2
/g, V
P
= 0.49 cm
3
/g),
thereby indicating that the addition of copper did not modify
signicantly the textural properties of the support. On the other
hand, the TPD prole of NH
3
on Cu/SiO
2
(not shown here) did not
exhibit any NH
3
desorption peak, thereby showing that the sample
does not contain acid surface sites able to retain ammonia at 373 K.
The XRD pattern of the silica-supported copper sample after
decomposition in air at 673 K is shown in Fig. 2a. The sample
exhibited a single crystalline phase of tenorite-like CuO (ASTM
Standard 5-0661) with a large average crystallite size (19 nm).
The TPR prole of the calcined Cu/SiO
2
sample exhibited only a
broad peak, with a maximum at 536 K, arising from the reduction
of CuO (Fig. 2b). The wide peak observed here for CuO reduction
would indicate a heterogeneous size distribution of the CuO
particles, as it has been suggested in the literatures [22,23]. The
metallic Cu dispersion on SiO
2
, determined by the dissociative
adsorption of N
2
O at 363 K, was 1.2% that corresponds to a mean
Cu
0
particle size of 175 nm. This large value of the metallic Cu
crystallites is consistent with the large CuO crystallite size
determined by XRD and probably reects a weak interaction
between the copper crystallites and the support.
3.2. Catalytic results
3.2.1. Solvent selection
The solvent effect on catalyst activity and selectivity for AP
hydrogenation was studied using a polar protic solvent (isopropylic
alcohol), a naphtenic non-polar solvent (cyclohexane) and two
aromatic non-polar solvents (toluene and benzene). Fig. 3 compares
the evolution of AP conversion as a function of time when using the
above-mentioned solvents. It is inferred that the catalyst activity
pattern follows the order: IPA > cyclohexane > toluene > benzene.
In contrast, the solvent had no effect on catalyst selectivity. In fact,
the selectivity toPhE was higher than99%during the entire catalytic
runs, irrespective of the solvent used.
The results in Fig. 3 clearly showthat the Cu/SiO
2
activity for AP
hydrogenation depends on the solvent used. The modication of
the catalyst activity when changing the solvent is frequently
observed in liquid-phase catalyzed reactions, but it is hardly
explained in terms of simple reaction parameters. Previous work
[2430] recognizes the complex role of the solvent on solid-
catalyzed reactions because the solvent inuence may be related
with reactantsolvent, productsolvent and/or catalystsolvent
interactions. While it is not the intent of this work to ascertain the
fundamentals of the solvent effect on the reaction mechanism, we
observe in Fig. 3 that the Cu/SiO
2
activity is hampered when using
aromatic non-polar solvents, such as benzene or toluene. The AP
conversion rate on Cu/SiO
2
, in fact, is clearly higher in cyclohexane
or IPA. Because it is expected that cyclohexane had a lower
interaction either with reactant and product molecules or with
catalyst surface in comparison to isopropylic alcohol, we selected
cyclohexane as the solvent to perform our catalytic tests.
3.2.2. Acetophenone hydrogenation
The effect of the hydrogen pressure on AP hydrogenation over
Cu/SiO
2
was studied in cyclohexane at 363 K and C
0
AP
= 0:168M:
The catalytic tests were conducted at total pressures of 5, 10 and
20 bar, which implies hydrogen partial pressures of 3.7, 8.7 and
Fig. 2. Characterization of calcined Cu/SiO
2
sample: (a) X-ray diffraction pattern; (b) TPR prole.
Fig. 3. Solvent effect on AP conversion over Cu/SiO
2
[T = 363 K, p
H
2
= 8:7bar;
W
CAT
= 1 g, C
0
AP
= 0:168M; V
SOLV
= 150 ml].
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 102
18.7 bar, respectively, by considering that at 363 K the solvent
vapor pressure is 1.3 bar (calculated with Antoine equation), and
the vapor pressures of AP and products are negligible. The obtained
X
AP
vs. time plots are presented in Fig. 4a. It is observed that the
catalytic activity increases with the hydrogen partial pressure. The
local slopes of the curves in Fig. 4a give the AP conversion rate at a
specic value of AP conversion and reaction time. The initial AP
conversion rates (r
0
AP
; mol APh
1
g
1
) were calculated by poly-
nomial differentiation of the curves at zero time. The reaction order
with respect to hydrogen was calculated by considering for r
0
AP
a
power-law rate equation:
r
0
AP
= k(P
H
2
)
a
(C
0
AP
)
b
(1)
Reaction order a, determined by both linear and non-linear
regression from Eq. (1) with C
0
AP
constant, was close to one.
The effect of the AP initial concentration on catalytic activity
was studied at 363 K and P
H
2
= 8:7bar: The initial concentration of
AP was varied between 0.084 and 0.251 M. The results are given in
Fig. 4b and show that the catalyst activity decreases by increasing
the initial AP concentration. The reaction order with respect to AP
was determined by calculating r
0
AP
from data represented in Fig. 4b
and applying both linear and non-linear regression with Eq. (1).
The value calculated for b was negative and about 0.33.
The inuence of temperature on catalytic activity was
investigated between 353 and 373 K, at P
H
2
= 8:7bar and C
0
AP
=
0:168M: Fig. 4c shows the increase of catalytic activity with
temperature. The apparent activation energy (E
A
) was determined
by numerical regression using an Arrhenius-type function. A value
of E
A
= 54.9 KJ mol
1
was obtained.
Data in Fig. 4 show effects that changes in H
2
partial pressure,
initial concentration of AP and temperature have on catalyst
activity. Regarding the catalyst selectivity, in all the catalytic runs
of Fig. 4 the selectivity to PhE was higher than 99%, which shows
that the Cu/SiO
2
catalyst is highly selective for hydrogenating AP to
PhE in this range of experimental conditions.
In summary, our catalytic results show that the reaction order
with respect to AP is negative for AP hydrogenation on Cu/SiO
2
,
thereby indicating that the interaction of AP with surface-active
sites is very strong. Moreover, taken into account that SiO
2
is a non-
reactive support and that the interaction between the large Cu
0
crystallites and SiO
2
is weak, it would be expected that AP
hydrogenation over Cu/SiO
2
takes place essentially via a mono-
functional mechanism on metallic copper. On these bases, we
developed several kinetic models in order to interpret and explain
the patterns of selectivity and activity experimentally determined
on Cu/SiO
2
for AP hydrogenation.
3.3. Kinetic modeling
3.3.1. LHHW models
Based on the previously discussed results, we considered here
the following hypothesis for the formulation of LHHW hetero-
geneous models:
(1) The adsorption of AP and H
2
may be either competitive (models
with only one active site) or non-competitive (models with two
different metal catalytic sites).
(2) The H
2
adsorption is dissociative [3133].
(3) PhE adsorption/desorption steps are reversible quasi-equili-
brium steps.
(4) The hydrogenation surface reaction is irreversible (total
conversion of AP to PhE was obtained in the experimental
runs).
(5) Hydrogen concentration in the liquid phase is constant,
because of the constant hydrogen partial pressure during the
entire experiment, the high solvent volume, and efcient
mixing.
Considering the former hypothesis with active sites S
1
and S
2
,
and non-competitive H
2
chemisorption (hydrogen adsorbed on
metallic site S
1
), the elementary steps shown in Eqs. (2)(5)
represent the general reaction mechanism:
H
2
2S
1
=2HS
1
; r
1
= k
H
2
p
H
2
(C
S
1
)
2
k
1
H
2
(C
HS
1
)
2
(2)
AP S
2
=APS
2
; r
2
= k
AP
C
AP
C
S
2
k
1
AP
C
APS
2
(3)
2HS
1
APS
2
=PhES
2
2S
1
; r
3
= k
S
(C
HS
1
)
2
C
APS
2
(4)
PhES
2
=PhE S
2
;
1
K
PhE
=
C
PhE
C
S
2
C
PhES
2
(5)
If the adsorptions of H
2
and AP are competitive, then S
1
= S
2
= S.
The general system of differential equations to be solved, based
on Eqs. (2)(5), is:
dC
+
AP
dt
=
1
C
0
AP

dC
AP
dt
=
1
C
0
AP
r (6)
dC
+
PhE
dt
=
1
C
0
AP

dC
PhE
dt
=
1
C
0
AP
r (7)
where r is the rate of the limiting step in the reaction mechanism
and C
+
i
= C
i
=C
0
AP
the relative concentration of the i component.
By assuming different rate-limiting steps (r.l.s.), (adsorption
of H
2
, adsorption of AP or surface chemical reaction) and two
Fig. 4. Effect of (a) H
2
partial pressure [T = 363 K, C
0
AP
= 0:168M]; (b) AP initial concentration [T = 363 K, p
H
2
= 8:7bar]; (c) temperature [ p
H
2
= 8:7bar; C
0
AP
= 0:168M[ on the
catalyst activity. W
CAT
= 1 g, V
SOLV
= 150 ml (cyclohexane).
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 103
possibilities for the adsorption of H
2
and AP (competitive and non-
competitive), we analyzed seven different LHHW models. Details
of these models are given in Table 1. The obtained expressions for
the reaction rate r in Table 1 were simplied by grouping kinetic
and equilibrium constants as parameters P
j
, where j identies the
group of constants.
3.3.2. Non-stationary model
We also use non-stationary kinetic models for interpreting
our experimental data. In contrast with LHHW kinetics, the non-
stationary models do not assume the existence of rate-
limiting steps, so that they consider all the surface reaction and
adsorption/desorption rates involved in the reaction mechanism
[34].
The catalytic surface coverage for the i-th component (AP or
PhE) is given by:
du
i
dt
= k
i
C
i
u
F
k
/
i
u
i

X
j
n
i j
r
j
(8)
where u
i
is the coverage of species i, k
i
and k
/
i
are the rate constants
of adsorption and desorption, respectively, u
F
, the fraction of
uncovered catalyst surface, r
j
the rate of j-th surface reaction, and
n
ij
the stoichometric coefcient of component i in j-th surface
reaction. Adsorbate conversion surface reactions are considered
elementary and thus expressed as rst-order reactions (in our case
i = AP and j = 1):
r
j
= k
S
u
i
(9)
The overall component concentration change is:
d

C
i
dt
= C
S
X
j
n
i j
r
j
(10)
where

C
i
is the overall concentration of i-component in the
mixture, disregarding whether it is adsorbed or not on the catalyst,
and C
S
is the total concentration of surface-active sites. The
concentration of the free i-component in the uid phase (C
i
) is then
given by:
C
i
=

C
i
C
S
u
i
(11)
The balance of surface-active sites on the catalyst is represented
by the following equation:
u
F

X
i
u
i
= 1 (12)
All the equations of this model were written here in terms of the
relative concentration of the i-component (C
+
i
= C
i
=C
0
AP
). The active
sites concentration, C
s
, was determined from the metallic Cu
dispersion value and the catalyst concentration.
3.3.3. Numerical resolution and statisticals
The systems of differential equations (Eqs. (6) and (7) and (8)
(10)) were solved numerically using the RungeKuttaMerson
algorithm. The model parameter estimation was performed by
non-linear regression, using a LevenbergMarquardt algorithm,
which minimizes the objective function (Eq. (13)):
S =
X
i
(C
+
i;t
C
+CALC
i;t
)
2
(13)
where C
+
i;t
and C
+CALC
i;t
are the experimental and calculated concen-
trations, respectively, i is the chemical compound, and t is the
reaction time.
The coefcient of determination (r
2
) gives the tting quality
(i.e., the percentage of explanation of the total data variation
around the average observed value) and was calculated using the
following equation:
r
2
=
P
n
i=1
(C
+CALC
i


C
+
)
2
P
n
i=1
(C
+
i


C
+
)
2
(14)
The model adequacy and the discrimination between models were
determined using the model selection criterion (MSC), according to
the following equation.
MSC = ln
P
n
i=1
(C
+
i


C
+
)
2
P
n
i=1
(C
+
i
C
+CALC
i
)
2
" #

2p
n

(15)
where n is the number of experimental data, p is the amount of
parameters tted,

C
+
is the average relative concentration and
C
+CALC
i;t
and C
+
i;t
are the predicted and the experimental values,
respectively. When various different models are compared, the
most signicant is that which leads to the highest MSC value.
3.3.4. Kinetic modeling results
3.3.4.1. LHHW models. The mathematical expressions obtained for
AP conversion rate following different LHHWmodels are presented
in Table 1. Models 1, 2, 3 and 5 of Table 1 were not considered
because they cannot explain the negative order with respect to AP
and/or the rst order in H
2
determined from our experimental
Table 1
LHHW kinetic models used in this work
Model Rate Simplied rate
Model 1
r.l.s.: adsorption of H
2
, non-competitive H
2
AP adsorption r = (C
2
S
1
k
H
2
) p
H
2
r = P
1
Model 2
r.l.s.: adsorption of AP, non-competitive H
2
AP adsorption r =
(C
S
2
k
AP
)C
AP
1K
PhE
C
PhE
r =
P
2
C
AP
1P
3
C
PhE
Model 3
r.l.s.: surface chemical reaction, non-competitive H
2
AP adsorption r =
(C
2
S
1
C
S
2
k
S
K
H
2
K
AP
) p
H
2
C
AP
(1K
AP
C
AP
K
PhE
C
PhE
)(1

K
H
2
p
H
2
p
)
2
r =
P
4
C
AP
1P
5
C
AP
P
3
C
PhE
Model 4
r.l.s.: adsorption of H
2
, competitive H
2
AP adsorption r =
(C
S
k
H
2
) p
H
2
(1K
AP
C
AP
K
PhE
C
PhE
)
2
r =
P
6
(1P
5
C
AP
P
3
C
PhE
)
2
Model 5
r.l.s.: adsorption of AP, competitive H
2
AP adsorption r =
(C
S
k
AP
)C
AP
1

K
H
2
p
H
2
p
K
PhE
C
PhE
r =
P
7
C
AP
1P
8
P
3
C
PhE
Model 6
r.l.s.: surface chemical reaction, competitive H
2
AP adsorption r =
(C
S
k
S
K
H
2
K
AP
) p
H2
C
AP
(1

K
H
2
p
H
2
p
K
AP
C
AP
K
PhE
C
PhE
)
3
r =
P
9
C
AP
(1P
8
P
5
C
AP
P
3
C
PhE
)
3
Model 7
r.l.s.: surface chemical reaction, competitive H
2
AP adsorption, saturated copper surface r =
(C
S
k
S
K
H
2
K
AP
) p
H
2
C
AP
(

K
H
2
p
H
2
p
K
AP
C
AP
K
PhE
C
PhE
)
3
r =
P
10
C
AP
(1P
11
C
AP
P
12
C
PhE
)
3
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 104
data. Thus, calculations were performed using only Models 4, 6 and
7 which full these requirements; results are presented in Table 2.
A reasonable tting was observed when applying Model 4
(Table 2, Model 4.a) but the estimate for P
5
was negative. Then, we
recalculated by assuming P
3
= 0, whichimplies to consider that PhE
is not adsorbed on the metal copper surface. However the estimate
for P
5
was again negative (Table 2, Model 4.b) and Model 4 was
therefore rejected.
Estimates of P
3
and P
8
not signicantly different fromzero were
obtained when applying Model 6 (Table 2, Model 6.a). This model
gave a satisfactory t by removing P
3
(Table 2, Model 6.b), but the
estimate for P
8
was again not signicantly different from zero.
Then, and taken into account that the estimate for P
8
was
signicantly lower than the estimate for P
5
, we recalculated by
assuming (K
H
2
p
H
2
)
0:5
K
AP
C
AP
: A good tting was obtained,
with positive estimates for P
9
and P
5
, and high values of r
2
and MSC,
as shown in Table 2 (Model 6.c). Furthermore, calculations using
Model 6.c resulted in a good tting of the experimental data
obtained in different operative conditions. For example, the results
obtained for three reaction temperatures gave values of MSC
higher than 6 in all cases whereas the sums of squared deviations
were in the order of 10
3
. Fig. 5a shows a good agreement between
the experimental data and the Model 6.c predictions when the
relative concentrations of AP and PhE are represented as functions
of time. Besides, the distribution of residuals followed an
acceptable random trend when represented as a function of time
(Fig. 5b). This is consistent with the hypothesis of random error
included in non-linear regression used for parameter estimations
and gives additional support to the model adequacy.
Model 7 gave also a good tting of experimental data (Table 2,
Model 7.a). However, estimate for P
12
= K
PhE
=

K
H
2
p
H
2
q
was
negative. Besides, estimates for P
11
= K
AP
=

K
H
2
p
H
2
q
and P
10
=
k
s
C
s
K
AP
=

K
H
2
p
H
2
q
not signicantly different from zero were
obtained. When it is assumed that PhE is not adsorbed on the
metal copper surface, P
12
= 0, Model 7 turns into Model 7.b, which
is identical to Model 6.c (see simplied rate, Table 1). The last
means that estimates for P
10
and P
11
(Model 7.b) equal to those
calculated for P
9
and P
5
(Model 6.c), respectively, should be
obtained. Thus, no difference between Model 6.c and 7.b from
mathematical and statistical point of view can be concluded.
In order to nd more information that permits discriminate
between Models 6.c and 7.b, the variation of k
s
C
s
and P
11
(or P
5
)
with temperature was plotted (Fig. 6). As it can be seen from
Fig. 6a, while k
s
C
s
is increasing with temperature, P
11
(or P
5
) is
practically constant between 353 and 373 K (Fig. 6b). If we
consider that P
5
= K
AP
(Model 6.c), this would mean that the AP
adsorption heat on metal copper is almost zero. This is contra-
dicting the experimental results which showthat AP adsorption on
metal copper is very strong. Instead, as P
11
is a ratio of K
AP
and K
H
2
(Model 7.b), it is likely that variation with temperature of K
H
2
compensates the variation of K
AP
, giving as result an almost
constant value for the parameter estimate in the temperature
range considered. Then, Model 7.b seems to be more consistent
than Model 6.c from a physical point of view.
In summary, the best LHHW model to interpret the AP
hydrogenation results on Cu/SiO
2
catalyst considers that the
adsorption of AP and H
2
is competitive and the rate-limiting step is
the surface chemical reaction. Furthermore, the model assumes
negligible PhE coverage, strong adsorption of AP on copper and
total coverage of metal copper surface.
3.3.4.2. Non-stationary model. Taken into account the results
obtained with LHHW models, the tting of the experimental data
using the non-stationary model was performed assuming that PhE
does not adsorb on the metal copper surface. A very good
approximation of the experimental data was observed when non-
stationary model was applied (Table 3, Model 8.a), but the estimate
for k
1
AP
was negative. Thus, we recalculated by assuming that
k
AP
k
1
AP
, i.e., the AP adsorption rate constant would be much
higher than that of desorption which is in agreement with the
negative order (strong adsorption) with respect to AP determined
from our experimental data. A good tting was obtained (Table 3,
Model 8.b) and estimates of k
AP
and k were signicantly different
from zero. Moreover, calculations using Model 8.b gave a good
tting of all the experimental data obtained in different operative
conditions. In fact, the results obtained for three temperatures
showthat in all cases the values of the statistical S were lower than
1.56 10
3
and of MSC higher than 7.5.
Fig. 7a shows that a very good agreement is veried between
experimental data and Model 8.b predictions. The differences
(C
+
OBS
C
+
CALC
) were always <2 10
2
, which corresponds to a
relative residual of 5% or less, and were lower than those obtained
using the LHHWModel 6.c. Furthermore, these residuals showed a
clear random behavior (see Fig. 7b). Finally, by comparing Tables 2
and 3 it can be inferred that the non-stationary model led to MSC
Table 2
Kinetic modeling results using LHHW models
Model Estimated parameters S r
2
MSC
4.a P
3
= 2.29 10
1
6.06 2.53 10
2
0.993 4.69
P
5
= 2.15 10
1
3.89
P
6
= 3.73 10
3
3.71 10
2
4.b P
3
= 0 2.53 10
2
0.993 4.76
P
5
= 3.61 10
1
6.96 10
2
P
6
= 2.47 10
3
2.60 10
4
6.a P
3
= 7.51 10
3
1.90 10
1
3.73 10
3
0.999 6.53
P
5
= 8.37 10
1
2.42 10
1
P
8
= 3.20 10
3
9.04 10
2
P
9
= 2.71 10
2
1.25 10
2
6.b P
3
= 0 3.74 10
3
0.999 6.60
P
5
= 8.35 10
1
1.96 10
1
P
8
= 1.72 10
3
8.95 10
2
P
9
= 2.68 10
2
1.11 10
2
6.c P
3
= 0 3.73 10
3
0.999 6.68
P
5
= 8.21 10
1
8.68 10
2
P
8
= 0
P
9
= 2.62 10
2
2.89 10
3
7.a P
10
= 6.69 10
3
2.01 10
2
3.73 10
3
0.999 6.61
P
11
= 1.55 10
1
1.18
P
12
= 3.66 10
1
6.34 10
1
7.b P
12
= 0 3.73 10
3
0.999 6.68
P
10
= 2.62 10
2
2.89 10
3
P
11
= 8.21 10
1
8.68 10
2
Reaction conditions: T = 363 K, p
H2
= 8.7 bar, W
CAT
= 1 g, C
0
AP
= 0:168M,
V
SOLV
= 150 ml (cyclohexane).
Table 3
Kinetic modeling results using non-stationary models
Model Estimated parameters S r
2
MSC
8.a k
AP
= 459.87 3061.50 1.56 10
3
0.999 7.47
k
/
AP
= 14:87 324:92
k
S
= 62.82 2.78
8.b k
AP
= 602.91 105.84 1.56 10
3
0.999 7.55
k
/
AP
= 0
k
S
= 62.82 2.17
Reaction conditions: T = 363 K, p
H
2
= 8:7bar; W
CAT
= 1 g, C
0
AP
= 0:168M;
V
SOLV
= 150 ml (cyclohexane).
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 105
values higher than those resulting when applying the LHHW
Model 6.c or Model 7.b.
On the other hand, from the values of k
S
(surface reaction
constant) determined at different temperatures (Fig. 6c) we
calculated, by numerical regression, the apparent activation
energy and obtained E
A
= 55.3 kJ mol
1
that is in good agreement
with the value determined directly from experimental data
(54.9 kJ mol
1
). As well, considering the k
AP
estimates at different
temperatures, an apparent activation energy of 53.8 kJ mol
1
for
the AP adsorption on copper was calculated. This value is of the
same order to the one obtained for the surface chemical reac-
tion, suggesting an important interaction between AP and metal
surface.
In summary, calculations using the non-stationary model
resulted in an acceptable tting of the experimental results by
assuming k
AP
k
1
AP
. This assumption is in agreement with the fact
that the best tting of experimental data using LHHW models was
obtained with Model 7.b that considers a strong adsorption of AP
on copper. Both models require, therefore, the existence of a strong
interaction between AP and the metallic copper surface to properly
predict the catalytic data.
4. Discussion
Our catalytic results showed that AP is selectively hydro-
genated on Cu/SiO
2
yielding almost 100% PhE under the operative
Fig. 5. Acetophenone hydrogenation on Cu/SiO
2
: (a) experimental (symbols) and modeling results (full lines) using LHHW Model 6.c or 7.b; (b) Evolution of residuals for AP
and PhE [p
H2
= 8.7 bar, W
CAT
= 1 g, C
0
AP
= 0:168M; V
SOLV
= 150 ml (cyclohexane)].
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 106
conditions used in this work. This selective conversion of AP to PhE
demonstrates that the promotion of parallel or consecutive
reaction pathways such as the aromatic ring hydrogenation or
the hydrogenolysis of COH bond in the PhE molecule is negligible
on Cu/SiO
2
in the conditions used in this work. Characterization of
Cu/SiO
2
by TPD of NH
3
proved that it does not contain surface acid
sites and this can easily explain the absence of by-products formed
via hydrogenolysis or cracking pathways. In contrast, a more
detailed discussion is required to justify the selective hydrogena-
tion of the C O bond of the AP molecule on copper.
Our experimental data on Cu/SiO
2
were properly predicted by
kinetic modeling only when it was considered that AP was strongly
adsorbed on copper. Consistently, results obtained by varying the
initial AP concentration showed that the reaction was negative
order with respect to AP. The APCu interaction may be interpreted
by taking into account the attractive and repulsive forces existing
between the emergent d-orbitals of the metal and the adsorbate. In
fact, the factors governing intrinsic metal selectivities have been
qualitatively related to the electronic lling of d-orbitals using
extended Hu ckel calculations [35]. In particular, the repulsive
forces existing between metal d-orbitals and the phenyl group of
AP would increase with the lling of d-orbitals. The d-orbitals in
metallic copper are completely lled and it is expected then that
the repulsion with the phenyl group will be higher for Cu than for
other non-noble metals having partially lled d-orbitals such as Ni.
The strong electrostatic repulsion between metallic Cu and the
phenyl group would tilt the AP molecule thereby favoring its
adsorption via the carbonyl group and the formation of the
unsaturated alcohol (Fig. 8). It must be noted that the repulsion
between copper and the phenyl group is also consistent with the
fact that the best LHHW model to interpret the AP hydrogenation
results (Model 7.b) assumes that the PhE adsorption on copper is
negligible. The rapid desorption of PhE from Cu drastically
decreases the possibility of the consecutive hydrogenation of
PhE to CHE that easily occurs on Ru [5,6] or the PhE hydrogenolysis
to EB that is observed on Pd [79].
AP molecules may interact through its C O bond with the
copper surface via on-top, di-s
CO
or p
CO
adsorption modes, but it is
expected that di-s
CO
or p
CO
adsorptions led to the formation of
more stable adsorbed species [35]. Fig. 8 represents the surface
copper atoms covered by AP adsorbed via di-s
CO
and p
CO
adsorptions and by atomic hydrogen formed by the dissociative
H
2
adsorption. From the estimates obtained with LHHW-type
Model 7.b, it can be approximately inferred that K
AP
is about one
order of magnitude higher than K
H
2
. Then, Model 7.b assumes that
the adsorption of AP and hydrogen is competitive and that AP
adsorbs on copper much stronger than hydrogen. These kinetic
modeling assumptions imply that the AP hydrogenation rate may
decrease with increasing initial AP concentration because the
density of Cu-active sites available for H
2
adsorption will be
drastically diminished by the AP covering. We effectively
determined that the AP hydrogenation rate is negative order with
respect to AP. Additionally; the adsorption modes of AP over metal
copper surface may play a role in the hydrogenation rate. It is likely
that the ratio of molecules adsorbed in the di-s
CO
mode respect to
p
CO
mode varies with AP concentration and reaction conditions. If
the most stable and less reactive mode of adsorption prevails, then
the hydrogenation rate will diminish.
The interaction between copper and reactants described above
gives also insight on the effect that changing the solvent has on
catalyst activity. The AP hydrogenation rate was lower in aromatic
solvents in comparison to cyclohexane or IPA. It is known that the
aromatic solvents may interact with the metallic surface of non-
noble metals such as Cu or Ni [2427]. In our case, this solvent
metal interaction would not hamper the strong adsorption of AP on
copper, but can decrease the concentration of Cu surface sites
available for the dissociative chemisorption of hydrogen. For a
given H
2
partial pressure, the diminution of the surface metal
fraction covered by adsorbed hydrogen should decrease the AP
hydrogenation rate because the reaction is positive order in
hydrogen. In contrast, the interaction of cyclohexane with copper
is very weak and would not interfere with the H
2
chemisorption.
Thus, differences in the solventmetal interaction strength can
explain the higher catalyst activity for AP hydrogenation observed
here when the reaction is carried out in cyclohexane in comparison
Fig. 6. Variation of parameters with temperature for AP hydrogenation over Cu/SiO
2
catalyst. (a) k
s
C
s
fromLHHW-type Model 7.b; (b) P
11
= K
AP
=

K
H
2
p
H
2
q
fromLHHW-
type Model 7.b; (c) k
s
and k
AP
from non-stationary Model 8.b.
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 107
to benzene or toluene. The highest catalyst activity for AP
hydrogenation was determined when IPA was used as solvent.
Previous work has showed that IPA is adsorbed dissociatively on
metallic copper forming isopropoxide and atomic hydrogen as
surface species [2830]. IPA may therefore act as an extra source of
surface atomic hydrogen thereby increasing the AP hydrogenation
rate. It has also to be noted that IPAprobably interact with AP in the
liquid phase via the formation of hydrogen-bridge bonds [8]. This
interaction may cause the activation of the carbonyl group via the
polarization of the C O bond and consequently increase the
Fig. 7. Acetophenone hydrogenation on Cu/SiO
2
: (a) experimental (symbols) and modeling results (full lines) using non-stationary Model 8.b; (b) evolution of residuals for AP
and PhE [ p
H
2
= 8:7bar; W
CAT
= 1 g, C
0
AP
= 0:168M, V
SOLV
= 150 ml (cyclohexane)].
Fig. 8. Representation of hydrogen and acetophenone adsorbed species on the metallic copper surface of Cu/SiO
2
catalyst.
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 108
catalyst activity for AP hydrogenation. Thus, in the case of IPA both
the solventreactant and the solventcatalyst interactions may
contribute to increase the AP hydrogenation rate. Finally, we
remark that Cu/SiO
2
was completely selective to PhE, irrespective
of the solvent used in the catalytic test. This show that the strong
adsorption of AP on copper via its C Obond, which is the essential
requisite for selectively obtaining PhE, is not modied by the
nature of the solvent employed in the reaction.
5. Conclusions
1-Phenylethanol yields of about 100% are obtained by liquid-
phase acetophenone hydrogenation on a Cu(6.8%)/SiO
2
catalyst.
Several factors contribute to achieve the complete conversion of
acetophenone to 1-phenylethanol: (i) the strong electrostatic
repulsion between metallic Cu and the phenyl group tilts the
acetophenone molecule thereby favoring its adsorption via the
carbonyl groupandthe formationof the unsaturatedalcohol; (ii) the
adsorption of 1-phenylethanol on copper is negligible; (iii) 1-
phenylethanol is not converted to unwanted products via hydro-
genolysis or cracking pathways because silica is an inert support.
The acetophenone conversion reaction is negative order with
respect to acetophenone and rst order in hydrogen. Consistently,
the experimental data are well interpreted by using heterogeneous
LHHW and non-stationary models that assume that acetophenone
competes with hydrogen for surface copper-active sites but
adsorbs on the metal much stronger than hydrogen.
The Cu/SiO
2
activity for acetophenone hydrogenation is higher
in cyclohexane than in benzene or toluene. This is because the
solventcopper interaction is stronger using aromatic solvents and
thereby decreases the concentration of Cu surface sites available
for the dissociative chemisorption of hydrogen. But the highest Cu/
SiO
2
activity is obtained in isopropyl alcohol, probably because
isopropyl alcohol adsorbs dissociatively on metallic copper and
forms extra atomic hydrogen that increases the acetophenone
hydrogenation rate.
Acknowledgements
We thank the Universidad Nacional del Litoral (UNL), Consejo
Nacional de Investigaciones Cientcas y Te cnicas (CONICET), and
Agencia Nacional de Promocio n Cientca y Tecnolo gica (ANPCyT),
Argentina, for the nancial support of this work.
References
[1] K. Bauer, D. Garbe, 3rd ed., Ullmanns Encyclopedia, vol. A11, VCH, New York,
1988, p. 141.
[2] P. Rylander, Hydrogenation Methods, Academic Press, Inc., London, 1985, p. 66.
[3] G.F. Santori, A.G. Moglioni, V. Vetere, G.Y. Moltrasio Iglesias, M.L. Casella, O.A.
Ferretti, Appl. Catal. A: Gen. 269 (2004) 215.
[4] C. Chen, H. Chen, W. Cheng, Appl. Catal. A: Gen. 248 (2003) 117.
[5] L. Cerveny, Z. Belohlav, M.N.H. Hamed, Res. Chem. Intermed. 22 (1996) 15.
[6] M. Casagrande, L. Storaro, A. Talon, M. Lenarda, R. Frattini, E. Rodrguez-Castello n,
P. Maire-les-Torres, J. Mol. Catal. A: Chem. 188 (2002) 133.
[7] M.A. Aramenda, V. Borau, J.F. Go mez, A. Herrera, C. Jime nez, J.M. Marinas, J. Catal.
140 (1993) 335.
[8] A. Drelinkiewicza, A. Waksmundzka, W. Makowski, J.W. Sobczak, A. Kro l, A. Zie ba,
Catal. Lett. 94 (2004) 143.
[9] M. Bejblova , P. Zamostny, L. Cerveny, J. Cejka, Collect. Czech. Chem. Commun. 68
(2003) 1969.
[10] J.M. Bonnier, J.P. Damon, J. Masson, Appl. Catal. 42 (1988) 285.
[11] J. Masson, S. Vidal, P. Cividino, P. Fouilloux, J. Court, Appl. Catal. A: Gen. 99 (1993)
147.
[12] J. Masson, P. Cividino, J. Court, Appl. Catal. A: Gen. 161 (1997) 191.
[13] J. Masson, P. Cividino, J.M. Bonnier, P. Fouillox, Stud. Surf. Sci. Catal. (1991) 245.
[14] M.V. Rajashekharam, I. Bergault, P. Fouilloux, D. Schweich, H. Delmas, R.V.
Chaudhari, Catal. Today 48 (1999) 83.
[15] R.V. Malyala, C.V. Rode, M. Arai, S.G. Hegde, R.V. Chaudhari, Appl. Catal. A: Gen.
193 (2000) 71.
[16] J.M. Bonnier, J. Court, P.T. Wierzchowski, Appl. Catal. 53 (1989) 217.
[17] F. Zaccheria, N. Ravasio, R. Psaro, A. Fusi, Tetrahedron Lett. 46 (2005) 3695.
[18] Oku et al., Sumitomo Chemical Company, US 6410806, 2002.
[19] S. Ito, T. Hibi, Sumitomo Chemical Company, US 5663458, 1997.
[20] A. Dandekar, M.A. Vannice, J. Catal. 178 (1998) 621.
[21] M.H. Kim, J.R. Ebner, R.M. Friedman, M.A. Vannice, J. Catal. 208 (2002) 381.
[22] A.J. Marchi, C.R. Apestegua, Appl. Clay Sci. 13 (1998) 35.
[23] A.J. Marchi, D.A. Gordo, A.F. Trasarti, C.R. Apestegua, Appl. Catal. A: Gen. 249
(2003) 53.
[24] A.K. Myers, G.R. Schoofs, J.B. Benziger, J. Phys. Chem. 91 (1987) 2230.
[25] C.M. Friend, E.L. Muetterties, J. Am. Chem. Soc. 103 (1981) 773.
[26] Y. Yu, J.J. Chessick, A.C. Zettlemoyer, J. Phys. Chem. 63 (1959) 1626.
[27] A.K. Myers, J.B. Benziger, Langmuir 3 (1987) 414.
[28] L.J. Shorthouse, A.J. Roberts, R. Raval, Surf. Sci. 480 (2001) 37.
[29] M. Bowker, R.J. Madix, Surf. Sci. 116 (1982) 549.
[30] S.C. Street, A.J. Gellman, Surf. Sci. 372 (1997) 223.
[31] G.H. Guvelioglu, P. Ma, X. He, Phys. Rev. B 73 (2006) 155436.
[32] G. Cilpa, G. Chambaud, Surf. Sci. 601 (2007) 320.
[33] A.J. Marchi, J.F. Paris, N.M. Bertero, C.R. Apestegua, Ind. Eng. Chem. Res. 46 (2007)
7657.
[34] Z. Belohlav, P. Zamostny, Can. J. Chem. Eng. 78 (2000) 513.
[35] F. Delbecq, P. Sautet, J. Catal. 152 (1995) 217.
N.M. Bertero et al. / Applied Catalysis A: General 349 (2008) 100109 109

You might also like