Single-stage liquid-phase synthesis of methyl isobutyl ketone under mild conditions

Article abstract of DOI:10.1016/j.molcata.2004.04.040

In this study, a catalyst mixture of activated Mg/Al hydrotalcite as solid base and Pd on carbon nanofibers was investigated for the single-stage liquid-phase synthesis of methyl isobutyl ketone (MIBK) from acetone and H ₂ at 313 K and H₂ pressures of 1-20 bar. With this catalyst system acetone and H₂ are selectively converted into MIBK in a single reactor, even at 1.2 bar H₂. Although the hydrogenation of one of the intermediates, mesityl oxide (MO), at 1.2 bar suffers from severe mass transfer limitations in H₂, the catalytic activity in the single-stage synthesis is stable over several hours if the steady-state concentration of MO is kept low. Under the presented conditions the dehydration reaction of diacetone alcohol to MO is rate limiting in the production of MIBK.

Full text of DOI:10.1016/j.molcata.2004.04.040

Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
Single-stage liquid-phase synthesis of methyl
isobutyl ketone under mild conditions
Ferry Winter, A. Jos van Dillen, Krijn P. de Jong∗
Department of Inorganic Chemistry and Catalysis, Debye Institute, Utrecht University,
P.O. Box 80 083, 3508 TB Utrecht, The Netherlands
Received 13 April 2004; accepted 26 April 2004
Available online 2 July 2004
Abstract
In this study, a catalyst mixture of activated Mg/Al hydrotalcite as solid base and Pd on carbon nanofibers was investigated for the
single-stage liquid-phase synthesis of methyl isobutyl ketone (MIBK) from acetone and H2 at 313 K and H2 pressures of 1–20 bar. With this
catalyst system acetone and H2 are selectively converted into MIBK in a single reactor, even at 1.2 bar H2. Although the hydrogenation of one
of the intermediates, mesityl oxide (MO), at 1.2 bar suffers from severe mass transfer limitations in H2, the catalytic activity in the single-stage
synthesis is stable over several hours if the steady-state concentration of MO is kept low. Under the presented conditions the dehydration
reaction of diacetone alcohol to MO is rate limiting in the production of MIBK.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Carbon nanofibers; CNF; Palladium catalyst; Platinum catalyst; Hydrotalcite; Mesityl oxide; Methyl isobutyl ketone; MIBK
1. Introduction
convert acetone and hydrogen directly into MIBK. Several
catalytic systems containing condensation, dehydration and
hydrogenation functionalities have been investigated for the
single-stage process, e.g., Pd on ion-exchange resin [5], Pd
or Pt on ZSM-5 [6–8], Pd/Nb2O5/SiO2 [9,10], Ni/␥-Al2O3
[11] and Ni or Pd on MgO [12,13]. The single-stage pro-
cesses as developed or proposed up to now still require high
reaction temperatures and pressures, and side-reactions oc-
cur, in particular hydrogenation of acetone to 2-propanol
and probably polymerization. As a consequence, the selec-
tivity of the process is lowered and the catalysts deactivate
[14–17].
Several studies have been reported on the use of
hydrotalcite-based catalyst systems for the single-stage
production of MIBK [15,16,18–20], in which the process
is mainly performed at high temperatures and/or hydrogen
pressures. In these studies, the actual condensation catalyst
is a mixed oxide obtained by calcination of the hydrotal-
cite. Recently, extending the pioneering work of Tichit and
co-workers [21–24] our group developed a highly active
catalyst for the liquid-phase condensation of acetone at
low temperatures, i.e., activated Mg–Al hydrotalcite (HT)
[25,26]. Furthermore, this catalyst showed dehydration
functionality of DAA to MO already at room temperature
[2,26].
Methyl isobutyl ketone (MIBK) is an industrially im-
portant chemical, mainly used as a coating solvent. It is
the third largest tonnage product formed from acetone and
is mainly produced via a conventional three-step process,
wherein starting from acetone the three reactions are exe-
cuted separately (Fig. 1). The first step consists of the aldol
condensation of acetone to form diacetone alcohol (DAA).
In a sequential step DAA is dehydrated to mesityl oxide
(MO). In the third step, MIBK is formed by selective hy-
==
drogenation of the C C bond of MO. The first two steps
are base and acid catalysed reactions, respectively, the third
step is generally catalysed by supported noble metals.
The high costs for the conventional three-step process
relate to the yields of the first two steps being controlled
by the thermodynamic equilibria of these reactions. The
equilibrium of the condensation of acetone shifts to DAA
at lower reaction temperatures, whereas for the dehydra-
tion step the equilibrium shifts to MO at higher reaction
temperatures [1–4]. To deal with the thermodynamic con-
straints, single-stage processes have been developed to
∗
Corresponding author. Tel.: +31 30 2536762; fax: +31 30 2511027.
E-mail address: k.p.dejong@chem.uu.nl (K.P. de Jong).
1381-1169/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.molcata.2004.04.040

274
F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
O
O
O
O
OH
-H2O
+H2
2
acetone
DAA
MO
MIBK
Fig. 1. Reaction scheme of the self-condensation of acetone to diacetone alcohol (DAA), dehydration of DAA to mesityl oxide (MO) and followed by
selective hydrogenation to methyl isobutyl ketone (MIBK).
Earlier studies reported in literature on the hydrogena-
tion of MO to MIBK showed that at higher temperatures
the reaction proceeds with appreciable initial rates [27,28].
However, with the catalysts explored up to now at low pres-
sure and temperature full conversion of MO to MIBK is
not attained due to deactivation [29]. For hydrogenation
of ␣,␤-unsaturated ketones and aldehydes, palladium-based
catalysts, especially Pd/C catalysts, exhibit the best perfor-
mance in the hydrogenation of the olefinic bond [15,30]. As
a promising alternative for activated carbon we used carbon
nanofibers (CNF) as the support material, because of their
high mechanical strength, high surface area and accessibil-
ity, tuneable surface properties, their purity and the absence
of micropores [31].
In this work we studied the formation of DAA and MO
from acetone using activated HT. Furthermore, a palladium
on CNF catalyst (Pd/CNF) was investigated in the hydro-
genation of MO at H2 pressures ranging between 1 and
20 bar and with different MO concentrations. An exploratory
study was conducted on the performance of this Pd/CNF
catalyst in comparison with other supported palladium cata-
lysts, i.e., Pd on SiO2 (Pd/SiO2) and Pd on graphite (Pd/G),
and with a platinum supported on CNF catalyst (Pt/CNF).
Finally, the single-stage conversion of acetone and H2 into
MIBK under mild conditions was studied using a catalyst
mixture of activated HT and the Pd/CNF catalyst.
2.2. Growth of carbon nanofibers
CNF were grown out of a commercial 57 wt.% Ni/SiO2
catalyst (Engelhard Ni5270P, calcined at 648 K for three
hours)) from syngas [34,35]. About one gram of calcined
Ni/SiO2 was first reduced at 873 K in a fixed bed reac-
tor for 2 h in a 20% H2/N2 flow (400 ml min−1). Next,
the temperature was decreased to 773 K and a mixture of
CO (80 ml min−1), H2 (30 ml min−1) and N2 (290 ml min−1
)
was passed through the catalyst bed for 20 h. The reactor
content was refluxed for 1.5 h in a KOH solution (1 M) and
thoroughly washed in order to remove the silica support.
Next, the fibers were treated for 2 h in boiling concentrated
nitric acid to remove exposed nickel and to introduce ad-
sorption sites, cooled to room temperature, filtered off, thor-
oughly washed with demineralised water and dried at 393 K
for 24 h.
2.3. Preparation of palladium and platinum supported
catalysts
Pd or Pt was deposited on the carbon supports via an ion
adsorption method, earlier applied by Hoogenraad [36]. Typ-
ically, 5 g of oxidized CNF or graphite (Lonza, HSAG100)
was suspended in 100 ml demineralised water. The pH of
the suspension was adjusted at a value between 5 and 6 by
adding ammonium hydroxide solution (25%). Subsequently,
an aqueous solution of Pd(NH3)4(NO3)2 (Alfa Aesar) or
Pt(NH3)4(NO3)2 (Aldrich) was added and the resulting sus-
pension was stirred for 20 h at room temperature under N2
atmosphere. The catalyst precursor was filtered off under
N2 flow, washed and dried at 353 K for 20 h in N2 flow.
The Pd catalyst precursors were reduced in hydrogen flow
at 523 K (5 K min−1) for 2 h, whereas the Pt catalyst pre-
cursor was reduced at 473 K (2 K min−1) in hydrogen flow
for 3 h. A 3 wt.% Pd on SiO2 catalyst obtained from Engel-
hard (Q500-189, unreduced sample) was reduced in hydro-
gen flow at 523 K for 2 h (5 K min−1).
2. Experimental
2.1. Preparation of hydrotalcite materials
HT with a Mg/Al ratio of 2 was prepared via co-precipit-
ation [32]. To an aqueous solution (70 ml) containing
0.35 mol NaOH (Merck) and 0.09 mol Na2CO3 (Acros)
an aqueous solution (45 ml) of 0.1 mol Mg(NO3)2·6H2O
(Acros) and 0.05 mol Al(NO3)2·9H2O (Merck) was added.
The resulting white suspension was heated to 333 K for
24 h under vigorous stirring, after which the precipitate was
filtered off and washed extensively. The sample, further
designated as HT-CO3, was dried for 24 h at 393 K.
2.4. Catalyst characterization
HT-CO3 was activated using the procedure of Roelofs
et al. [33]. HT-CO3 was heated to 723 K (10 K min−1) in
a nitrogen flow for 8 h (HTcalc). HTcalc was rehydrated in
decarbonated water for 1 h under nitrogen atmosphere, fil-
tered off and washed with ethanol to obtain a relatively high
amount of accessible basic sites (HTact).
Powder X-ray diffraction (XRD) patterns were measured
using an Enraf-Nonius CPS 120 powder diffraction appara-
tus with Co K␣ radiation (λ = 1.789 Å). N2 physisorption
measurements were performed using a Micromeritics ASAP
2400 analyser. When micropores are present, the surface area

F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
275
calculated with the t-method is reported instead of the BET
surface area. Volumetric CO2 adsorption measurements in
the range 0–100 mbar were executed at 273 K with a Mi-
cromeritics ASAP 2010C apparatus after drying the sam-
ples at 393 K in vacuo for at least 20 h. The total amount of
accessible sites was determined taking the amount of CO2
chemisorbed to zero pressure by extrapolation of the linear
part of the uptake isotherm. Hydrogen chemisorption data
were recorded using a Micromeritics ASAP 2010C appara-
tus. Prior to the chemisorption measurements at 343 K for the
Pd samples and 307 K for the Pt catalyst, samples were dried
in vacuo at 393 K for 16 h, reduced in H2 flow at 473 K for 2 h
(5 K min−1) and degassed for two hours at the reduction tem-
perature. The H/M ratios are based on the amount of hydro-
gen adsorbed by extrapolating the linear part of the isotherm
of the total amount of adsorbed hydrogen to zero pressure.
TEM images to assess the metal particle size were obtained
with a Philips CM-200 FEG TEM operating at 200 kV. Sam-
ples were, after an ultrasonic treatment in butanol, dispersed
on a holey carbon film. A Philips XL30FEG electron mi-
croscope was used to obtain SEM images. ICP analyses to
determine metal loadings were performed using a Thermo
Jarrel Ash Atom Scan 16 apparatus. After crushing and sus-
pending in ethanol, the catalyst particle size distribution was
measured with an optical particle sizer Accusizer Model 770,
capable of detecting particles in the range of 1–500 ␮m.
catalyst was kept in 1.2 bar H2 for one hour, after which
it was added to a solution containing 120 ml of acetone or
ethanol, 6 g iso-octane as an internal standard (99.5% pure)
and MO (Acros, 99% ␣- and ␤-isomer (ratio 1:13), purified
by distillation) at 313 K. During the reaction the hydrogen
consumption was monitored by an electronic mass flow
controller. For the single-stage synthesis of MIBK at 331 K
and 1.2 bar H2 a mixture of 0.3 g HTact and 0.2 g Pd/CNF
was used. MO hydrogenation experiments at 10 and 20 bar
H2 were performed in a 1 l Parr reactor. Typically, 275 mg
of crushed catalyst was suspended in 670 ml of ethanol or
acetone at 313 K, after which a second solution contain-
ing 80 ml of ethanol or acetone and the desired amount of
MO was added. Next, the reactor was filled with H2 and
pressurized. Aliquots of 1 ml were taken from the reaction
mixture during reaction and analysed using a Chrompack
CP 9001 GC provided with a Chrompack CP 9050
autosampler.
3. Results and discussion
3.1. Characterization of the hydrotalcite
The XRD pattern of HT-CO3 shows the characteristics
of a crystalline layered structure (Fig. 2a). By heating the
sample up to 723 K, the layered structure is destroyed and
a mixed oxide phase is present as can be concluded from
XRD (Fig. 2b). This heat treatment causes a large increase
in surface area as well as in micropore and mesopore volume
and a lowering of the average pore diameter (Table 1) [38].
Upon reconstructing the HT structure by rehydrating HTcalc
in CO2-free water, the micropores disappear and the average
pore diameter increases. The HT structure is restored to
a large extent as can be concluded from XRD (Fig. 2c),
although line broadening is apparent in the XRD pattern of
HTact when compared to that of HT-CO3.
2.5. Catalytic reaction procedures
Condensation and dehydration to MO experiments were
performed under N2 atmosphere with 0.30 g of activated HT
catalyst. MO hydrogenation experiments were performed
in a semi-batch slurry reactor operating at a constant pres-
sure of 1.2 bar H2 at 313 K [37]. The set-up consisted of a
thermostatted stirred double-walled glass reactor, equipped
with baffles and a gas-tight mechanical stirrer. For a typical
hydrogenation experiment, 50 mg of crushed hydrogenation
(c)
(b)
(a)
5
15
25
35
45
55
65
75
85
95
105
2
degrees
θ
Fig. 2. XRD patterns of (a) HT-CO3, (b) HTcalc and (c) HTact
.

276
F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
Table 1
Table 2
Results from N2-physisorption of HT samples
Results from N2 physisorption
Sample
Surface
area
(m2 g−1
Total pore
volume
Micropore
volume
Average pore
diameter
(nm)
Sample
Surface
area
(m2 g−1
Total pore
volume
Micropore
volume
(ml g−1
)
)
(ml g−1
)
(ml g−1
)
)
(ml g−1
)
HT-CO3
HTcalc
HTact
58
109a
132
0.37
0.54
0.61
0.00
0.01
0.00
25
11
19
Oxidized CNF
Graphite
Pd/SiO2
137
111
278a
0.38
0.25
0.93
0.00
0.00
0.02
a
a
Surface area calculated with t-method.
Surface area calculated with t-method.
needed for applying the metal precursor. Graphite shows
platelets with only the external surface accessible for large
molecules (not shown). The graphite platelets are randomly
stacked to form a house of cards structure with a meso- and
macro pore structure.
Volumetric CO2-adsorption measurements were per-
formed to get information about the total amount of acces-
sible active basic sites of HTact. With the sample prepared
as described, typically 6.6 ml STP CO2 g−HT1 was found, cor-
−1
responding to 0.29 mmol g of basic sites. According to
HT
The results of N2-physisorption of the support materials
are given in Table 2. CNF as well as the graphite support
show a much lower surface area and pore volume than the
Pd/SiO2 material. No micropores were present in the CNF
and graphite support materials, whereas micropores were
observed with Pd/SiO2.
these results, we calculated that these sites represent about
6 mol% of the total amount of aluminium present in the
structure of HT. This is in line with the conclusions stated
by Roelofs et al. [25], that only the active sites at the edge
area of the stacked platelets are accessible.
3.2. Characterization of the support material
3.3. Characterization of palladium and platinum
supported catalysts
In Fig. 3 an SEM image of CNF is displayed. The fibers
are tightly interwoven and the skeins of the fibers form a
porous structure. The diameter of the fibers is in the range
of 20–80 nm. XRD and TEM showed that the graphite-like
structure of CNF had not been affected after removal of the
growth catalyst and the activation treatment with nitric acid
In Table 3 the characterisation results of the various hy-
drogenation catalysts are presented. After reduction at 523 K
of the Pd/CNF catalyst, TEM examination shows palladium
particles on the fibers of about 1–4 nm (Fig. 4). Examination
Fig. 3. SEM image of Pd/CNF.

F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
277
Table 3
Characterization results for the various hydrogenation catalysts
3.4. Condensation and dehydration to MO over HTact
The catalytic properties of HTact were investigated in the
formation of MO from acetone at 313 K (Fig. 5). As can
be seen, the condensation of acetone to DAA proceeds very
Sample
Metal
loading particle size
Metal
H2-chemisorption
Particle size
in ethanol
(␮m)
H/M d (nm)
(wt.%)
Pd/CNF 0.90
Pd/G 0.46
Pd/SiO2 2.76
Pt/CNF 1.24
(TEM, nm)
rapid over HTact and equilibrium (0.45 molDAA l
−1) is at-
1–4
1–3
1–3
∼1
0.43 2.6
0.63 1.8
0.75 1.5
0.88 1.3
24
35
23
n.d.
tained within the first hour. The rate of dehydration of DAA
to MO is much lower. We found an init−ia1l rate in the dehy-
dration reaction of 9.3 × 10−3 mol MO gHT h−1, which is six
times higher than that measured by Roelofs et al. at 296 K
[2,26]. The activity in the dehydration reaction lowers some-
what in time, which is most likely due to the gradually in-
creasing concentration of MO and water.
by TEM of Pd/SiO2 as well as of Pd/G reduced at 523 K
shows small Pd particles of about 1–3 nm. Moreover, EDX
analysis during TEM on Pd/G even shows Pd at surfaces
where with TEM no particles are observed, indicating that
also particles smaller than 1 nm must be present. Pt/CNF
showed particles down to 1 nm after reduction at 473 K, but
also with this sample EDX reveals the presence of smaller
Pt particles. The results obtained by H2-chemisorption for
the various catalysts correspond with the results obtained by
TEM analysis.
To get an indication of the macroscopic size of the catalyst
particles measurements were performed after severely crush-
ing the various catalysts and suspending them in ethanol
(Table 3). Pd/CNF and Pd/SiO2 show a volume averaged
particle size of 24 and 23 ␮m, respectively. However, the
particle size distribution is rather broad with most of the par-
ticles in the range of 2–10 ␮m, whereas the largest part of
the volume fraction is in the range 6–30 ␮m. Pd/G shows a
somewhat higher average particle size, i.e., 35 ␮m, caused
by the presence of some particles larger than 70 ␮m. How-
ever, with this catalyst the large portion of the volume frac-
tion can be found in the range 2–22 ␮m.
3.5. Hydrogenation of MO in ethanol
In Fig. 6 the results of the hydrogenation at 313 K of
0.08 mol l−1 MO at 1.2 bar H2 and 0.35 mol l−1 MO at
1.2 and 10 bar H2 over Pd/CNF are depicted. Under the
given conditions this catalyst exhibits high (initial) activ-
ities and high selectivities (Table 4). The reaction with
MO0 = 0.08 mol l−1 at 1.2 bar resulted in complete con-
version within 40 min. Although the reaction with MO0
= 0.35 mol l−1 at 1.2 bar started with a somewhat higher
rate (Table 4), probably due to the higher initial MO con-
centration, a decrease in rate was observed right from the
start resulting in a total deactivation after about 3 h (Fig. 6).
Deactivation of the catalyst, however, was avoided or at
least suppressed to a large extent when the hydrogen pres-
sure was increased up to 10 bar. From these results it may
be concluded that at relatively high MO concentrations
(polymerisation) products are formed which irreversibly
deactivate the metal sites. A high excess concentration of
Fig. 4. TEM image of Pd/CNF after reduction at 523 K.

278
F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
0.5
0.4
0.3
MO
DAA
0.2
0.1
0.0
0
1
2
3
4
5
6
7
time (h)
Fig. 5. DAA and MO production at 313 K with HTact
.
mass transfer limitation in hydrogen appeared from the ex-
periments with catalyst samples, which were shortly and
prolongedly crushed to obtain samples with different parti-
cle sizes. With MO0 = 0.08 mol l−1 the initial reaction rate
increased almost twofold with decreasing catalyst particle
size. We also calculated the Weisz–Prater number (Φ) for an
nth order reaction [41] that indicates the absence of internal
mass transfer limitations if:
100
80
60
40
20
0
0.08 M - 1.2 bar
0.35 M - 1.2 bar
0.35 M - 10 bar
ꢀ
ꢁ
ρprwob,spdp2
n + 1
Φ =
< 0.15
36DeCs
2
The parameters used for the calculation of Φ (in H2) for
the reaction over Pd/CNF in ethanol as solvent at 1.2 bar
H2 are given in Table 5. We found a value for Φ of 1.3
which is amply above the value of 0.15. The initial rate
when started with MO0 = 0.35 mol l−1 at 10 bar is 1.11 ×
10−3 molMO gcat−1 s−1 and from this value, with Chydrogen
= 38.3 mol m−3 [42] and the other relevant parameters of
Table 5, a value for the Weisz–Prater number of 0.66 could
be derived. Obviously, although this value still exceeds the
criterion and mass transfer limitation still occurs, this did
not bring about deactivation. At 10 bar hydrogen is avail-
able throughout the catalyst particle thereby protecting the
Pd active phase from deactivation. We cannot, however,
exclude deactivation of part of the metal sites at rela-
tively high MO concentrations, i.e., in the first stage of the
reaction.
0
30
60
90
time (min)
120
150
180
Fig. 6. MO (0.08 and 0.35 M) hydrogenation at 1.2 and 10 bar H2 over
Pd/CNF at 313 K in ethanol.
MO at the active sites could occur if hydrogen is rapidly
depleted in the catalyst particles due to a high rate of
hydrogenation in combination with its low solubility.
Nijhuis et al. [39] and Ros et al. [40] exploring the use
of small rhodium particles on CNF for the hydrogenation of
cyclohexene, recently found that at low hydrogen pressures
(∼1 bar) hydrogenation reactions with comparably high re-
action rates give rise to internal mass transfer limitations
in hydrogen. A first indication that the MO hydrogenation
at 1.2 bar hydrogen over our Pd/CNF catalyst suffers from
Table 4
Activity and selectivity for MO hydrogenation in ethanol
Sample
MO concentration
H2 pressure
(bar)
Initial activity × 104
TOFa (s−1
)
Selectivity to
MIBK (%)
(mol l−1
)
(molMO gcat−1 s−1
)
Pd/CNF
0.08
0.35
1.2
10
2.54
11.1
6.9
30.1
>99
>99
Pd/G
Pd/SiO2
Pt/CNF
0.08
0.08
0.08
1.2
1.2
1.2
0.69
2.84
0.99
2.5
1.5
1.8
>99
>99
94
a
TOF is given for the first 120 s in mol MO hydrogenated, per mol metal surface atoms in the catalyst per second.

F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
279
Table 5
surface groups on the carbon nanofibers can have a tremen-
dous effect on the activity and selectivity in hydrogenation
reactions. It could well be that the presence of these func-
tional groups introduced by treatment in nitric acid, results
in a more active catalyst. More research is needed to fully
elucidate these effects, preferentially without the presence
of mass transfer limitations in H2 enabling comparison be-
tween the intrinsic activities of the catalysts more quantita-
tively.
Values of the parameters used to validate the Weisz–Prater criterion for
the presence of internal diffusion limitations
Parameter
Value
0.25
obs
r
Observed reaction rate per−u1nit mass
w,p
of catalyst particle (mol kgcat s−1
Catalyst particle diameter (m)
)
dp
ρp
24 × 10−6
Density of catalyst particle (kg m−3
)
1020
Chydrogen
Da
ε
Concentration of reactant (mol m−3
Diffusion coefficient (m2 s−1
)
4.6
)
2.3 × 10−9
Void fraction of catalyst particle
Tortuosity factor
0.46
3.6. Hydrogenation of MO in acetone
τ
2
De
Effective diffusion coefficient of
7.2 × 10−10
reactant (m2 s−1
)
In order to investigate to what extent the hydrogenation of
MO under mild conditions over these catalysts is applicable
in the single-stage process of MIBK starting from acetone
and hydrogen, hydrogenation experiments were performed
in acetone as solvent and reactant. Hydrogenation of MO0
= 0.08 mol l−1 in acetone over Pd/CNF did not bring about
acetone hydrogenation up to 20 bar over Pd/CNF. Moreover,
the selectivity of the hydrogenation of MO to MIBK re-
mained >99%.
However, concerning deactivation very different re-
sults were obtained when the hydrogenation with MO0
= 0.08 mol l−1 was performed at 1.2 bar H2 in acetone in-
stead of in ethanol. Hydrogenation at 1.2 bar H2 led to a
rapid deactivation of all catalysts (Fig. 7). A side reaction
of MO in which acetone might be involved, because with
MO0 = 0.08 mol l−1 in ethanol this fast deactivation does
not occur. In a next experiment it was tested whether ace-
tone itself deactivates the catalyst. To this end the catalyst
was suspended in acetone at 1.2 bar H2 one hour before
adding MO. It turned out that the catalyst was still active,
which demonstrates that probably polymerisation of MO
and/or a reaction of MO with acetone cause deactivation.
It is clear that this deactivation is not just a feature of the
carbon supported catalysts, since hydrogenating MO under
these conditions over Pd/SiO2 results in deactivation too
(Fig. 7).
Φb
Weisz–Prater
1.3
a
Diffusion constant is determined by using the method from
Wilke-Chang [48,49].
b
Calculated with the assumption that the reaction is first order.
In Table 4 the results obtained with MO0 = 0.08 mol l−1 at
313 K and 1.2 bar H2 for all four catalysts are given. Pt/CNF
exhibits a comparable activity in the hydrogenation of MO
as the palladium catalysts, however, with a selectivity sig-
nificantly lower than measured with the palladium catalysts
of 94 and >99%, respectively. Although the reaction is re-
stricted by mass transfer of H2, the average catalyst particle
size of all catalysts is rather similar making an exploratory
study on their performances meaningful. Pd/CNF exhibits
an initial activity 3-fold that of Pd/G and an activity rather
similar to that obtained with Pd/SiO2. Comparison of the
turn over frequencies (TOF) obtained with these catalysts
shows the distinct activity of Pd/CNF to be even more pro-
nounced. The TOF obtained with Pd/CNF is a factor 3 and 5
higher than these obtained with Pd/G and Pd/SiO2, respec-
tively. The reason for this difference can be the effect of the
presence of oxygen-containing surface groups on the carbon
support [43–45] or of the metal particle size [46]. However,
the metal particle sizes do not differ significantly. As was
already stated by Toebes et al. [45,47], oxygen containing
1.2 bar Pd/CNF
1.2 bar Pd/SiO2
10 bar Pd/CNF
100
80
60
40
20
0
0
10
20
30
40
50
60
time (min)
Fig. 7. MO (0.08 mol l−1) hydrogenation at 313 K in acetone at 1.2 and 10 bar H2 over Pd/CNF and at 1.2 bar H2 over Pd/SiO2.

280
F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
Table 6
The activity of Pd/CNF is maintained when a hydrogen
Activity and TOF for the single-stage production of MIBK over a catalyst
mixture of HTact and Pd/CNF at 331 K and 1.2 bar H2
pressure as high as 10 bar is applied and full conversion is ob-
tained, which demonstrates that a rapid decrease of the MO
concentration or a good availability of hydrogen throughout
the catalyst particles is needed to avoid deactivation. The
initial rate observed at 10 bar H2 over Pd/CNF was found to
be 1.5 × 10−4 molMO gcat−1 s−1 and a Weisz–Prater num-
ber of 0.04 was calculated, indicating the absence of mass
transfer limitations in H2 when performing the reaction in
acetone at 10 bar H2.
Activitya
TOFa (s−1
)
(mol product g−1
h
)
−1
HT
DAA production
MO production
MIBK formation
1.34#
0.026
0.023
1.26#
0.024
0.022
a
Activity and TOF calculated over the first 30 min or when marked
with (#) over the first 5 min over HTact. Acti−vi1ty and TOF in the MIBK
formation over Pd/CNF are 0.036 molMIBK gpd/cnf h−1 and 0.27 s−1, re-
spectively.
3.7. Combination of Pd/CNF and HTact in the single-stage
production of MIBK
found with GC-MS, which could, together with water, affect
the catalytic activity of HTact. It was investigated separately
whether the Pd/CNF displays any dehydration functional-
ity. It was found that under these conditions the dehydration
The suitability of a mixture of HTact and Pd/CNF in the
single-stage process of MIBK starting from acetone and H2
was tested at 1.2 bar H2 and 331 K (Fig. 8). In Table 6 the ac-
tivity and TOF values over the first 30 min in the condensa-
tion, dehydration and hydrogenation reaction are given. The
rate of the dehydration reaction is calculated from the sum
of the amount of MIBK and the amount of MO formed dur-
ing reaction. At this reaction temperature, the condensation
reaction of acetone to form DAA proceeds extremely fast
as equilibrium is attained within 5 min. The rate of conden-
reaction proceeds exclusively over HTact
.
It appears that after three hours the reaction is in pseudo
steady-state, as the MO concentration did not increase during
the course of the reaction and the MIBK production rate
remained constant. These observations strongly indicate that
the production rate of MIBK is mainly determined by the
rate of dehydration. Furthermore, the TOF in the MIBK
formation over Pd/CNF is about 10-fold the TOF in the
dehydration over HTact (Table 6).
−1
−1
sation over HTact was calculated to be 1.3 mol
g
h
DAA
HT
(Table 6) over the first minutes and the DAA concentration
stayed at equilibrium throughout the experiment.
From the arguments mentioned above we can conclude
that most likely the dehydration reaction over HTact is the
rate-determining step in the single-stage synthesis of MIBK
under the described mild conditions. Above results show
the potential of this catalyst system for the single-stage
liquid-phase synthesis of MIBK from acetone and hydro-
gen under mild conditions. Note that the system is far from
optimised. Further work will concentrate on the condensa-
tion and dehydration reaction over activated hydrotalcites
in order to improve the MIBK productivity of this catalyst
system.
During the experiment the MO concentration remained
low at a value of 0.003 mol l−1 by further hydrogenation
to MIBK. Within the first three hours the productivity of
the catalyst system decreased somewhat, but next a stable
MIBK production rate was found for several hours. Deacti-
vation of the hydrogenation catalyst was not apparent, which
is due to the low steady-state concentration of MO during
the experiment even at this low H2 pressure. The dehydra-
tion rate over HTact after three hours was found to be 6.2
× 10−3 mol MO gH−T1 h−1, with a selectivity to the desired
products (DAA, MO and MIBK) of around 95%. The main
side products are due to secondary condensations as was
4. Conclusions
Results obtained with the HTac–Pd/CNF catalyst mixture
demonstrate the suitability of this catalyst system for the
single-stage liquid-phase production of MIBK from acetone
and hydrogen under mild conditions. Due to the balance be-
tween the formation rate of MO and the rate of its hydro-
genation, from the start the steady state concentration of MO
is low and remains low even at a hydrogen pressure as low
as 1.2 bar, due to which deactivating species starting from
MO are not formed. The catalytic activity is stable over sev-
eral hours. In this process dehydration of DAA to MO over
HTact is rate limiting.
0.30
0.25
0.20
0.15
0.10
0.05
0.00
DAA
MIBK
MO
0
1
2
3
4
5
6
time (h)
Acknowledgements
Fig. 8. Formation of DAA, MO and MIBK in the single-stage liquid-phase
synthesis of MIBK from acetone and H2 (1.2 bar) at 331 K over 0.3 g
HTact and 0.2 g Pd/CNF in 1.8 mol acetone.
The authors are grateful to C. van der Spek, V. Koot, A.
Mens and M. Versluijs-Helder for their technical assistance.

F. Winter et al. / Journal of Molecular Catalysis A: Chemical 219 (2004) 273–281
281
[22] D. Tichit, M.H. Lhouty, A. Guida, B. Chiche, F. Figueras, A. Auroux,
D. Bartalini, E. Garrone, J. Catal. 151 (1995) 50.
[23] D. Tichit, M.N. Benanni, F. Figueras, R. Tessier, J. Kervennal, Appl.
Clay Sci. 13 (1998) 401.
[24] A. Guida, L.M. Hassane, D. Tichit, F. Figueras, P. Geneste, Appl.
Catal. A: Gen. 164 (1997) 251.
P. Groenendijk (DOW) is kindly thanked for helping per-
form the catalysis measurements at higher pressures. The
Netherlands Technology Foundation (CW/STW 790.35.733)
is acknowledged for financial support.
[25] J.C.A.A. Roelofs, D.J. Lensveld, A.J. van Dillen, K.P. de Jong, J.
Catal. 203 (2001) 184.
[26] J.C.A.A. Roelofs, A.J. van Dillen, K.P. de Jong, Catal. Today 60
(2000) 297.
References
[27] M.L.A. von Holleben, M. Zucolotto, C.A. Zini, E.R. Oliveira, Tetra-
hedron 50 (1994) 973.
[1] G.G. Podrebarac, F.T.T. Ng, G.L. Rempel, Chem. Eng. Sci. 52 (1997)
2991.
[28] J. Wisniak, M. Herskowitz, D. Roffe, S. Smilovitz, Ind. Eng. Chem.
Prod. Res. Dev. 15 (1976) 163.
[2] J.C.A.A. Roelofs, PhD thesis, Utrecht University, The Netherlands,
2001.
[29] A. Alba, M.A. Aramendia, V. Borau, C. Jimenez, J.M. Marinas,
React. Kinet. Catal. Lett. 25 (1984) 45.
[3] Y.K. Kim, J.D. Hatfield, J. Chem. Eng. Data 30 (1985) 149.
[4] E.C. Craven, J. Appl. Chem. 13 (1963) 71.
[5] C.U. Pittman Jr., Y.F. Liang, J. Org. Chem. 45 (1980) 5048.
[6] D. Bombos, Gr. Bozga, M. Bombos, A. Stefan, I. Stanciu, Rom.
Chem. Papers 54 (2000) 171.
[30] V. Ponec, Appl. Catal. A: Gen. 149 (1997) 27.
[31] K.P. de Jong, J.W. Geus, Catal. Rev. -Sci. Eng. 42 (2000) 481.
[32] S. Miyata, T. Kumura, Chem. Lett. (1973) 843.
[33] J.C.A.A. Roelofs, A.J. van Dillen, K.P. de Jong, Catal. Lett. 74
(2001) 91.
[7] L. Melo, G. Gianetto, F. Alvarez, P. Magnoux, M. Guisnet, Catal.
Lett. 44 (1997) 201.
[34] M.L. Toebes, J.M.P. van Heeswijk, J.H. Bitter, A.J. van Dillen, K.P.
de Jong, Carbon 42 (2004) 307.
[8] P.Y. Chen, S.J. Chu, W.C. Lin, K.C. Wu, C.L. Yang, Stud. Surf. Sci
.Catal. 83 (1994) 481.
[35] M.L. Toebes, J.H. Bitter, A.J. van Dillen, K.P. de Jong, Catal. Today
76 (2002) 33.
[9] Y.Z. Chen, B.J. Liaw, H.R. Tan, K.L. Shen, Appl. Catal. A: Gen.
205 (2001) 61.
[36] M.S. Hoogenraad, Ph.D. thesis, Utrecht University, The Netherlands,
1995.
[37] T.G. Ros, Ph.D. thesis, Utrecht University, The Netherlands, 2002.
[38] W.T. Reichle, S.Y. Kang, D.S. Everhardt, J. Catal. 101 (1986) 352.
[39] T.A. Nijhuis, G. van Koten, F. Kapteijn, J.A. Moulijn, Catal. Today
79/80 (2003) 315.
[10] Y. Higashio, T. Nakayama, Catal. Today 28 (1996) 127.
[11] L.M. Gandia, M. Montes, J. Mol. Catal. 43 (1994) 347.
[12] L.M. Gandia, M. Montes, Appl. Catal. A: Gen. 101 (1993) L1.
[13] K.H. Lin, A.N. Ko, Appl. Catal. A: Gen. 147 (1996) L259.
[14] G.S. Salvapati, K.V. Ramanamurty, M. Janardanarao, J. Mol. Catal.
54 (1989) 9.
[40] T.G. Ros, D.E. Keller, A.J. van Dillen, J.W. Geus, D.C. Konings-
berger, J. Catal. 211 (2002) 85.
[15] A.A. Nikolopoulos, G.B. Howe, B.W.-L. Jang, R. Subramanian, J.J.
Spivey, D.J. Olsen, T.J. Devon, R.D. Culp, in: M.E. Ford (Ed.),
Chemical Industries, Catalysis of Organic Reactions, vol. 82, New
York, 2001, p. 533.
[16] A.A. Nikolopoulos, B.W.-L. Jang, R. Subramanian, J.J. Spivey, D.J.
Olsen, T.J. Devon, R.D. Culp, ACS Symposium Series 767 (Green
Chemical Processes), Charleston, 2000, p. 194.
[17] S. Lippert, W. Baumann, K. Thomke, J. Mol. Catal. 69 (1991)
199.
[18] Y.Z. Chen, B.J. Liaw, H.R. Tan, K.L. Shen, Appl. Catal. A: Gen.
169 (1998) 207.
[19] N. Das, D. Tichit, R. Durand, P. Graffin, B. Coq, Catal. Lett. 71
(2001) 181.
[41] R.A. van Santen, P.W.N.M. van Leeuwen, J.A. Moulijn, B.A. Averill,
Catalysis: An integrated Approach, Second, Revised and Enlarged
Edition, Elsevier, Amsterdam 2nd edition, 1999 pages.
[42] C.L. Young, Solubility Data Series, vol. 5/6, Hydrogen and Deu-
terium, Pergamon Press, Oxford, 1981.
[43] H.P. Boehm, Carbon 32 (1994) 759.
[44] S.S. Barton, M.J.B. Evans, E. Halliop, J.A.F. MacDonald, Carbon
35 (1997) 1361.
[45] M.L. Toebes, Ph.D. thesis, Utrecht University, The Netherlands, 2004.
[46] B. Coq, F. Figueras, Coord. Chem. Rev. 178/180 (1998) 1753.
[47] M.L. Toebes, F.F. Prinsloo, J.H. Bitter, A.J. van Dillen, K.P. de Jong,
J. Catal. 214 (2003) 78.
[48] R.H. Perry, D.W. Green, Perry’s Chemical Engineers’ Handbook,
McGraw-Hill, 7th edition, New York, 1997.
[20] R. Unnikrishan, S. Narayanan, J. Mol. Catal. A: Chem. 144 (1999)
173.
[21] K.K. Rao, M. Gravelle, J. Sanchez-Valente, F. Figueras, J. Catal.
173 (1998) 115.
[49] C.R. Wilke, P. Chang, AIChE J. (1955) 264.