Solvent effects in the hydrogenation of 2-butanone

Article abstract of DOI:10.1016/j.jcat.2012.01.011

In liquid-phase reaction systems, the role of the solvent is often limited to the simple requirement of dissolving and/or diluting substrates. However, the correct choice, either pure or mixed, can significantly influence both reaction rate and selectivity. For multi-phase heterogeneously catalysed reactions observed variations may be due to changes in mass transfer rates, reaction mechanism, reaction kinetics, adsorption properties and combinations thereof. The liquid-phase hydrogenation of 2-butanone to 2-butanol over a Ru/SiO ₂ catalyst, for example, shows such complex rate behaviour when varying water/isopropyl alcohol (IPA) solvent ratios. In this paper, we outline a strategy which combines measured rate data with physical property measurements and molecular simulation in order to gain a more fundamental understanding of mixed solvent effects for this heterogeneously catalysed reaction. By combining these techniques, the observed complex behaviour of rate against water fraction is shown to be a combination of both mass transfer and chemical effects.

Full text of DOI:10.1016/j.jcat.2012.01.011

Journal of Catalysis 289 (2012) 30–41
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Solvent effects in the hydrogenation of 2-butanone
B.S. Akpa a, C. D’Agostino a, L.F. Gladden a, K. Hindle b, H. Manyar b, J. McGregor a, R. Li a, M. Neurock c,
N. Sinha c, E.H. Stitt d, D. Weber a, J.A. Zeitler a, D.W. Rooney b,
⇑
a Department of Chemical Engineering and Biotechnology, University of Cambridge, New Museums Site, Pembroke Street, Cambridge, CB2 3RA, UK
b CenTACat, School of Chemistry and Chemical Engineering, Queen’s University Belfast, David Keir Building, Stranmillis Road, Belfast BT9 5AG, UK
c Departments of Chemical Engineering, University of Virginia, 102 Engineers’ Way, PO Box 400741, Charlottesville, VA 22904-4741, USA
d Johnson Matthey Catalysts, PO Box 1, Belasis Avenue, Billingham, TS23 1LB, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
In liquid-phase reaction systems, the role of the solvent is often limited to the simple requirement of dis-
solving and/or diluting substrates. However, the correct choice, either pure or mixed, can significantly
influence both reaction rate and selectivity. For multi-phase heterogeneously catalysed reactions
observed variations may be due to changes in mass transfer rates, reaction mechanism, reaction kinetics,
adsorption properties and combinations thereof. The liquid-phase hydrogenation of 2-butanone to 2-
butanol over a Ru/SiO2 catalyst, for example, shows such complex rate behaviour when varying water/
isopropyl alcohol (IPA) solvent ratios. In this paper, we outline a strategy which combines measured rate
data with physical property measurements and molecular simulation in order to gain a more fundamen-
tal understanding of mixed solvent effects for this heterogeneously catalysed reaction. By combining
these techniques, the observed complex behaviour of rate against water fraction is shown to be a com-
bination of both mass transfer and chemical effects.
Received 16 August 2011
Revised 18 January 2012
Accepted 19 January 2012
Available online 27 March 2012
Keywords:
Solvent effects
Mixed solvents
Mass transfer
Heterogeneous catalysis
Hydrogenation
2-Butanone
Ru
Ó 2012 Elsevier Inc. All rights reserved.
DFT
1. Introduction
and products, gas solubility and other mass transfer effects need
to be considered as these can all significantly influence reaction
The ability of a solvent to influence either the rate or selectivity
of a reaction has been known within organic chemistry for
150 years [1]. Over that time there have been a number of attempts
to understand the role of the solvent with methods including
multiple linear regression analysis, factor analysis and principal
component analysis all being used in an attempt to develop some
level of understanding and predictability of these effects. Such
analysis tends to rely on various free energy relationships, and de-
spite the complicated nature of solvents and solutions, they have
shown that they can provide some insight into many chemical
processes [1].
Like the organic reactions above, many heterogeneously cata-
lysed processes are carried out in a solvent for the simple purpose
of dissolving the reactants and keeping the products in solution.
Here too, solvents are known to influence both rate and selectivity
although in such systems the multi-phase nature of the reactions
increases their complexity. Therefore, in addition to factors such
as solvent polarity, dielectric constant and acid/base properties of
the reaction medium, factors such as the solvation of reactants
rates and product selectivities [2]. Other important criteria for con-
sideration include the potential for improved heat transfer and the
influence on deactivation such as reduced carbon laydown on the
catalyst surface [3].
Toukonitty et al. for example investigated solvent effects for the
enantioselective hydrogenation of 1-phenyl-1,2-propanedione
with a cinchonidine modified Pt/Al2O3 catalyst [4]. When using a
range of different solvents, including binary solvent mixtures, it
was observed that while no correlation between the solvent dielec-
tric constant and hydrogenation rate could be found the enantio-
meric excesses decreased non-linearly with an increasing solvent
dielectric constant. This dependence was partly attributed to the
open (3) cinchonidine conformer, although it was primarily taken
into account by applying transition state theory and the Kirkwood
treatment with the result that a model was able to predict the
behaviour of the system as a function of the solvent dielectric con-
stant. Contrasting with this work, Mukherjee and Vannice later re-
ported on solvent effects during the liquid-phase hydrogenation of
citral using Pt/SiO2 and eight nonreactive solvents [5]. Here it was
observed that the rates were affected by the choice of solvent;
however, differences in the product distribution were not signifi-
cant. In this case, it was found that the variation in specific activity
⇑
Corresponding author.
E-mail address: d.rooney@qub.ac.uk (D.W. Rooney).
0021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2012.01.011

B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
31
did not correlate with either the solvent dielectric constant or the
dipole moment. Similarly, Gómez-Quero et al. studied solvent ef-
fects during the liquid-phase hydrodehalogenation of haloarenes
in methanol, THF, water/methanol and water/THF mixtures using
a Pd/Al2O3 catalyst [6]. Within this work they noted that higher ini-
tial rates were observed with increasing water content in the sol-
vent mixture and attributed this to an increase in the dielectric
constant of the medium. Such observations were consistent with
an electrophilic mechanism in which the solvent helped to stabilise
the arenium intermediate. An overall dependence of rate on sol-
vent for this reaction could then be established with approximately
80% of the contribution being due to the dielectric constant with
molar volume being a secondary factor. Mixed alcohol/water sol-
vents, the subject of this work, have also been investigated previ-
ously. For example, a minimum in the rate of 2-butene-1,4-diol
hydrogenation (the second step in the hydrogenation of 2-
butyne-1,4-diol over Pd/Al2O3) occurred at 80–90 molH2O% in a
2-propanol/water solvent [7]. Elsewhere, an enhancement in the
rate of reaction in mixed alcohol/water solvents as compared to
the pure alcohol has been observed in 2-butanone hydrogenation
over Pd catalysts [8]; acetophenone hydrogenation over Raney-Ni
[9]; and o-nitrotoluene hydrogenation over Pd/C [10]. In the latter
example, an increase of almost 50% in the reaction rate was re-
ported upon changing solvent from pure methanol to a mixed
methanol/water solvent containing 18% water.
As mentioned above, other factors are also important, for exam-
ple, polar solvents are known to enhance the adsorption of non-
polar reactants and non-polar solvents the adsorption of polar
reactants [11]. More recently, Vanoye et al. observed an interesting
inflection in the initial rates obtained using a mixed ethanol/hep-
tane solvent system for the liquid-phase dehydration of ethanol
to diethylether over heterogeneous sulphonic-acid catalysts [12].
In this case, the authors concluded that the observed inflection in
non-polar solvents could be explained by the formation of a new
liquid phase around the acid site indicating that local structure is
also important. It is quite clear from the above discussion that
the role of the solvent is less well defined than in normal organic
reactions with increases in rate being observed in some systems,
and selectivity in others.
In briefly reviewing selective hydrogenation reactions over
ruthenium catalysts which relate to the work discussed herein, it
is further noted that carbonyl hydrogenations are particularly ac-
tive in the presence of water [13]. Conversely, in the hydrogenation
of benzene over Ru/C, water was found to have an adverse effect on
the reaction rate [14]. Similar effects of water have also been ob-
served by Vaidya and Mahajani in the hydrogenation of n-valeral-
dehyde to n-amyl alcohol over Ru/Al2O3; however, no explanation
was given for its influence [15]. A possible reason for enhanced
activity in water may be due to its ability to dissociate over differ-
ent metal surfaces. The dissociation of water over Ru is well known
in vapour phase studies as well as in aqueous solutions; this leads
to the formation of surface hydroxyl intermediates and protons
which can subsequently influence the reaction [16,17].
While such effects are known, to our knowledge there are no
studies which extensively examine the role of water in a mixed sol-
vent system covering its effect on mass transfer, diffusion and reac-
tion kinetics. Therefore, a detailed study of the influence of water
over diffusion, adsorption, desorption and elementary reaction
steps is necessary to improve our current understanding of the role
played by the solvent in altering reaction activity and selectivity. In
this paper, we attempt to investigate such effects and have probed
the role of water on the hydrogenation of 2-butanone (methyl
ethyl ketone (MEK)) to 2-butanol using a combination of experi-
ments and density functional theory. These results indicate that
the rate of hydrogenation of MEK is strongly correlated to the
solvent composition and that water plays an important role in
altering energetics and kinetics of the elementary steps involved
in the overall hydrogenation.
2. Materials and methods
Catalysts comprised of 1% and 5% Ru/SiO2 and were prepared by
incipient wetness using an aqueous ruthenium (III) chloride trihy-
drate salt (density 2250 100 kg mꢀ3). They were dried at 393 K in
air and subsequently reduced in 5% H2/95% He at 673 K for 3 h. Ki-
netic experiments were carried out in a 380 ml Premex stainless
steel reactor, equipped with a gas inducing impeller with online
hydrogen consumption monitoring. The solvents used were ultra-
pure water (distilled, deionised >18 MX) and/or 2-propanol (Reidel
De Haan >99.5%). Typically, 240 ml of solvent was charged to the
reactor with the required mass of catalyst (0.1 g unless otherwise
stated). After purging with N2, the solvent was heated to the de-
sired temperature before injection of 7.5 ml 2-butanone (Aldrich
99 + %) and 2.5 ml solvent (2-butanone concentration of
0.33 mol lꢀ1). After purging with H2, the reactor was pressurised
and monitored by hydrogen consumption or by sampling using a
GC equipped with a DB-1 capillary column and FID detector. All
gases were BOC research grade. Mass balances performed on reac-
tions were >95% in all cases.
Pulsed field gradient (PFG) NMR measurements of diffusion
coefficients were conducted using a Diff30 diffusion probe with a
10 mm r.f. coil on a Bruker DMX 300 spectrometer, operating at a
1H resonance frequency of 300.13 MHz. Samples were prepared
using 5 mm NMR tubes filled to a height of approximately
20 mm. The PGSTE pulse sequence was employed with the diffu-
sion encoding time
D and gradient pulse duration d fixed at
100 ms and 2 ms, respectively, with a maximum gradient strength,
g, of up to 1 T/m [18]. All measurements were carried out at
303.15 K.
THz Time-Domain Spectroscopy (THz-TDS) studies were carried
out using coherent pulses of broadband terahertz radiation
(0.1–4 THz) generated by photoexcitation of a DC biased semi-
insulating GaAs substrate by 12 fs pulses of a NIR laser (Femtola-
sers, Femtosource cM1, Vienna, Austria, centre wavelength
800 nm). In order to suppress the absorption of water vapour in
the air, the sealed sample chamber was purged by nitrogen gas
to ensure a relative humidity below 2% for all measurements.
Liquid samples were measured in a standard cell (PIKE Technolo-
gies, Watertown, USA) comprising 3 mm z-cut quartz windows
and a 200 lm PTFE spacer. For each sample, 200 time-domain
waveforms were collected and averaged.
Periodic gradient-corrected density functional theoretical (DFT)
[19] calculations as implemented in the Vienna ab initio Simulation
Package (VASP) [20] were used to follow the reaction energies and
activation barriers for different pathways involved in the hydroge-
nation of MEK over a model Ru(0001) surface and to simulate the
effects of the solvent. The Kohn–Sham equations were solved using
a plane-wave basis set with a cut-off energy of 400 eV [21]. The
Perdew–Wang 91 exchange correlation functional was used to de-
scribe non-local gradient corrections [22]. The core electrons and
the nuclei of the atoms were described by Vanderbilt ultrasoft
pseudopotentials [23]. A 3 ꢁ 3 ꢁ 1 k-point grid was used to model
the first Brillouin zone [24]. All the reported calculations were car-
ried out spin-restricted. The wavefunctions were converged to
within 1 ꢁ 10ꢀ4 eV and the geometries were optimised until the
force on each atom was less than 0.05 eV/Å. In order to test the de-
gree of accuracy, we tightened the electronic convergence criterion
from 10ꢀ4 to 10ꢀ6 eV and the structural optimisation criterion
from 0.05 to 0.01 eV/Å and found that the energies for the adsorp-
tion of hydrogen and methyl ethyl ketone on Ru(0001) changed by
less than 0.002 eV and 0.03 eV, respectively. As such, we have used

32
B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
Table 1
the lower criterion to carry out the calculations reported herein.
Transition states were isolated using the climbing nudged elastic
band approach, which optimises the forces on all of the atoms
for a particular image and over a series of images chosen along
the minimum energy path until the force perpendicular to the
reaction coordinate was within 0.05 eV/Å [25]. Frequency calcula-
tions were carried out for each transition state to ensure that there
was a single imaginary frequency corresponding to vibration mode
along the reaction coordinate. In the transition state calculations in
the presence of a solvent, the solvent molecules are allowed to
move along the reaction coordinate. This enables us to model the
possible stabilisation of transition state by the solvent molecules.
The Ru(0001) surface was modelled by using a 3 ꢁ 3 unit cell
comprised of four metal layers along with 16 Å vacuum spacing nor-
mal to the metal surface to separate the metal slabs. The aqueous
medium was simulated by filling the vacuum region within the unit
cell with 24 water molecules in order to simulate a density of water
of 1 g/cm3. The initial starting structures were adapted from previ-
ous long time (30 ps) ab initio molecular dynamic simulations car-
ried out for water on Cu(111) and Pt(111) [45]. For the reaction
carried out in 2-propanol, 6 isopropyl alcohol (IPA) molecules were
used in order to create a density of 0.785 g/cm3. The initial structure
of the solvent, represented by explicit solvent molecules, was ob-
tained by carrying out a series of simulated annealing runs using
ab initio molecular dynamics (AIMD) as implemented in VASP. The
Comparison of reaction rate at 303.15 K, 3 bar H2 and 1400 rpm using 5 wt%
Ru/SiO2
a
( measured at 298.15 K).
Rate (mol minꢀ1 gꢀ1) ꢁ 104
31.90
4.55
0.95
1.25
Dielectric constant
Water
IPA
Methanol
Heptane
76.7
19.9a
31.6
1.9
a significant variation in initial rate with water clearly producing
a value over 33 times larger than that observed in methanol. Such
results indicate that under the conditions used here no correlation
between the dielectric constant and initial rate exists. While it can-
not be guaranteed that these experiments were carried out in the
kinetic regime, the fact that water has both the highest rate and
lowest hydrogen gas solubility suggests that removal of mass
transfer resistance, as will be demonstrated later, would only exag-
gerate this difference further.
When water/IPA mixtures of varying mole fractions were tested
as solvents in the hydrogenation of MEK using 5 wt% Ru/SiO2, the
results showed that the observed rate increased with water mole
fraction (xH2O), until a maximum was attained after which it de-
creased until xH2O ꢃ 0.9 when it again started to increase (Fig. 1).
This trend was confirmed at two different reaction conditions
and it is worthy of note that a similar trend was observed by Hu
et al. for the hydrogenation of 2-butyne-1,4-diol using the same
solvent mixture [30]. Again the mixture dielectric constant cannot
be used to correlate the rate behaviour as it is known to decrease
linearly with increasing IPA concentration [31].
AIMD simulations (at 400 K,
Dt = 1 fs) were performed for both sys-
tems within an NVT ensemble to obtain an optimised, low energy
structure for the interface. The simulations were all run for 1200
time steps or 1.2 ps. A Nose–Hoover thermostat was used to regu-
late the temperature during the run [26,27]. The resulting solvent/
metal interface structures established from AIMD simulations were
then fully optimised using the conjugate gradient methods until the
forces on all of the atoms were all less than 0.05 eV/Å. The simula-
tions represent a very simplified model of the solvent/metal inter-
face carried out at 0 K. They do not capture the dynamic changes
in the solvent structure that is present under actual catalytic condi-
tions or the effects of temperature as this would require hundreds of
thousands of ab initio molecular dynamics simulations carried out to
long times to provide the appropriate sampling which is well be-
yond what is currently possible. The simulations thus provide what
may be a first-order effect that captures the role of explicit solvent
but does not include the effects of dynamics or temperature.
The diffusion of MEK, IPA and water in IPA/water/MEK mixtures
was studied using classical molecular dynamics (MD) as imple-
mented in the Discover module of Materials Studio [28]. The MD
simulations were performed on a system consisting of 64 water
(for pure water) molecules in a cubic simulation cell with each side
equal to 12.74 Å, which yielded liquid densities of 0.925 g/cm3 for
water at 298 K. NVT simulations were carried out with periodic
boundary conditions at 293.15 K in this work. Again the Nose–
Hoover algorithm was used to maintain the temperature at the
specified set point [26,27]. All the simulations were run for
10,000 time steps with each time step of 1 fs. Diffusion coefficients
were obtained by calculating the mean squared displacements
(MSD) from the simulations using the Einstein relation [29]:
3.1. Gas–liquid mass transfer
As mentioned previously, the observed effect may be due to
external and/or internal mass transfer limitations and hence these
must be taken into consideration. Estimation of the gas–liquid
mass transfer efficiency (gG–L) can be expressed as:
gG—L ¼ 1 ꢀ CaG—L
ð1Þ
where CaG–L is the Carberry number for gas–liquid mass transfer
which is defined as the ratio between the observed reaction rate
and the maximum transfer rate both based on the liquid volume.
This value is determined from the measured volumetric gas–liquid
2.2x10-3
2.0x10-3
1.8x10-3
1.6x10-3
1.4x10-3
1.2x10-3
1.0x10-3
8.0x10-4
6.0x10-4
4.0x10-4
2.0x10-4
3.0x10-2
2.5x10-2
2.0x10-2
1.5x10-2
1.0x10-2
5.0x10-3
D
E
1
6t
2
D ¼
½rðtÞ ꢀ rð0Þꢂ
ð1Þ
T = 333K, P = 3 barg
T = 303K, P = 1 barg
where r(t) denotes the coordinates of the centre-of-mass of the mol-
ecules at time (t).
0
20
40
60
80
100
3. Results and discussion
molH O
%
2
As shown in Table 1, tests using a range of four different sol-
vents with 5 wt% Ru/SiO2 at 303.15 K and 1400 rpm demonstrate
Fig. 1. Reaction rates for MEK hydrogenation in different water/IPA mixtures at Ç
333.15 K, 3 barg j 303.15 K, 1 barg (on secondary axis).

B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
33
Table 2
mass transfer coefficients (kLa) obtained via the pressure step
Calculated effectiveness factors for 5% Ru/SiO2.
method and the hydrogen solubility [32,33]. While it can be seen
that there is a marked increase in the kLa at xH2O > 0.50 it decreases
sharply as xH2O ? 1, (Fig. 2). A similar effect of IPA addition to water
on bubble size was observed by Hu et al. [30]; namely a minimum
bubble diameter was observed at xH2O ꢄ 0.98.
xH2O
gG–L
gL–S (H2)
gL–S (MEK)
gpore (H2)
gpore (MEK)
0
1.00
0.99
0.98
0.98
0.97
0.96
0.89
1.00
1.00
1.00
0.99
0.99
0.99
0.98
1.00
1.00
1.00
1.00
1.00
1.00
1.00
0.86
0.59
0.33
0.21
0.15
0.11
0.08
0.98
0.94
0.90
0.91
0.91
0.92
0.90
0.12
0.40
0.64
0.80
0.92
0.99
Hu et al. attributed this observation to a surfactant type effect
with preferential adsorption of alcohol or diol at the gas–liquid
interface at low concentrations. It was considered that the hydro-
phobic organic parts were directed into the gas phase with the
hydrophilic OH groups being directed into the aqueous phase
resulting in polarisation of the interface which hindered drain-
age/rupture of the film separating two bubbles [30]. This indicates
that the changes in bubble size, which directly relate to the mass
transfer rate, would be stronger functions of the surface excess en-
ergy and solution viscosity rather than the traditional Weber num-
bers used to estimate mean bubble diameters. Using the data
published by Park et al. these can be compared with the kLa values,
and as shown in Fig. 2 a reasonable correlation is observed [31].
However, when these values were used to estimate the gas–liquid
mass transfer efficiency using the methods reported by Dietrich
et al., all values for the mixed water/IPA system were greater than
95% except for the pure water system which yielded a value of 89%
(Table 2) thereby indicating that external gas to liquid mass trans-
fer did not account for the observed rate trends [32].
With 1% Ru/SiO2, all transport processes had calculated effectiveness factors >0.95.
Prior to estimating the mass transfer rate, one must have
knowledge of both the bulk diffusion coefficient as well as hydro-
gen solubility in the mixed solvents. Here the hydrogen concentra-
tion in water/IPA mixtures was evaluated from the data and
equations given by Nischhenkova [36] while the Wilke–Chang cor-
relation [37] was used for hydrogen diffusivity estimation in water
and the Schiebel equation [38] for diffusivity in IPA. For mixed sol-
vents the correlation developed by Perkins and Geankoplis [39]
was used, where the viscosity of the mixture was calculated from
the equation developed by Kendall and Monroe [40]. Despite the
approximations used for bulk diffusivity, the estimated and calcu-
lated data were both used to determine the liquid–solid Carberry
number which was then utilised to estimate the liquid–solid mass
transfer efficiency according to Eq. (5)
3.2. Liquid–solid mass transfer
n
gLꢀS ¼ ð1 ꢀ CaLꢀS
Þ
ð5Þ
Fishwick et al. showed that calculation of liquid to solid mass
transfer rate is complicated by large local differences in the particle
slip velocities within the reactor [34]. However, the liquid to solid
Carberry number may be used to estimate such resistances [31].
Here calculation of the liquid–solid mass transfer coefficient, kLs,
is estimated via the dimensionless Sherwood number which is it-
self calculated for laboratory stirred tank reactors using the follow-
ing correlation [35]
In each case, the external mass transfer efficiencies were shown
to be P 0.98 for all solvent compositions when both H2 and MEK
(Table 1) are treated as the limiting species and therefore liquid–
solid mass transfer was not considered to be limiting in this
system.
At present we are not aware of any experimental results for
hydrogen diffusivity in mixed solvents which could be used to help
validate the aforementioned empirical equations; however, such a
comparison can be made for 2-butanone using the data measured
by PFG-NMR across the entire solvent composition range, as shown
in Fig. 3. A minimum in the rate of diffusion of 6.59 ꢁ 10ꢀ10 m2 sꢀ1 is
seen at ꢄ85 molH2O%. This minimum implies that molecular mobil-
ity is hindered and hence liquid structuring is enhanced at this con-
centration. The previously reported diffusion coefficient of 2-
propanol in the binary 2-propanol/water system follows a similar
Sh ¼ 2 þ 0:4Re1=4Sc1=3
ð2Þ
with the dimensionless Reynolds and Schmidt numbers calculated
using Eqs. (3) and (4), respectively.
d2i N
l
q
Re ¼
Sc ¼
ð3Þ
ð4Þ
l
qD
20
18
16
14
12
10
8
PFG-NMR
MD
Calculation
0.18
0.16
0.14
0.12
0.10
0.08
0.06
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
-0.5
kLa
Viscosity
Surface excess
6
4
2
0
0.0
0.2
0.4
0.6
0.8
1.0
x
0
20
40
molH O
60
80
100
H2O
%
2
Fig. 3. Diffusion coefficient of 2-butanone (0.33 mol lꢀ1), in 2-butanone/2-propa-
nol/water mixtures at 303.15 K using PFG-NMR, MD calculations and literature
correlations.
Fig. 2. Volumetric mass transfer coefficient for water/IPA mixtures. All have
0.33 mol lꢀ1 of MEK except xH2O = 1.0.

34
B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
trend [41]. Comparing the measured data to calculated values (also
shown in Fig. 3) a significant difference between the calculated
and the experimental data is evident. However, even when using
these calculated data, the results still indicate that liquid–solid mass
transfer for MEK was not limiting. If similar errors are imposed on
the hydrogen diffusivity, the mass transfer efficiency is reduced
but is still high enough for it to be considered as not limiting. There-
fore, while in this case it is unlikely that liquid–solid diffusion is
important, this analysis does highlight deficiencies in the empirical
correlations employed here for diffusivity estimation.
There are of course a number of alternative correlations which
could be used to estimate the diffusivity in pure or mixed solvents,
although these will not be reported here as even an order of mag-
nitude change here in diffusivity estimates does not alter the con-
clusion that external mass transfer rates are not limiting in this
case. The diffusivity of MEK in different ternary mixtures of IPA,
water and MEK was, however, estimated using classical molecular
dynamics simulations. The compositions of the ternary mixtures
were varied by changing the relative number of molecules in the
system. Here we considered ternary mixtures with 0% (IPA 92.7%
and MEK 7.3%), 41.2% (IPA 51.5% and MEK 7.3%), 74.2% (IPA
18.5% and MEK 7.3%) and 96.9% (MEK 3.1%) water (mole fractions).
The number of water, IPA and MEK molecules per unit cell in these
systems were chosen as (0:38:3), (17:21:3), (30:8:3) and (62:0:2),
respectively. The diffusivity of MEK in these mixtures was then
calculated based on the mean squared displacement of MEK mole-
cules over 10 ps. As shown in Fig. 3, the diffusivity of MEK in the
ternary solution changes with the composition. It can be observed
that the molecular dynamics results for diffusivity are in agree-
ment, at least qualitatively, with the PFG-NMR results and that
using this method also suggests a minimum diffusivity of MEK
when xH2O is close to 0.74.
As it is known that the catalyst is not poisoned by product, the
Weisz-Prater criteria were employed to calculate the effectiveness
factor, and subsequently used to give an estimate of the true reac-
tion rates [42]. In general if the calculated value of /WP ꢅ 1, then
the reaction is absent of internal mass transfer limitations. This va-
lue changes depending on the order of the reaction and a very good
review of this method and its application is given by Mukherjee
and Vannice [43]. Using the calculated efficiencies it is possible
to modify the observed reaction rates and to therefore estimate a
‘true’ rate which is free from mass transfer effects. The result of this
manipulation is shown in Fig. 5 where it can be seen that the cor-
rected reaction rate now increases with xH2O. This increase pro-
ceeds linearly up to xH2O ꢃ 0.8, after which it increases much
more rapidly until xH2O = 1.0. It is possible that the increasing rate
over the lower water concentrations could now be attributed to
previous discussions, e.g. changes in the bulk dielectric constant
or related to changes in the adsorption equilibria on the catalyst
surface. Obviously the change at xH2O ꢃ 0.8 suggests that water is
playing a key role in enhancing the reaction rate above this con-
centration. It should be noted that the result of this manipulation
is, in this case, much more sensitive to estimated values of the dif-
fusion constant and particle size. For example a 20% increase in the
diffusion coefficient will contribute to approximately a 10% in-
crease in the rate, whereas a 20% increase in the particle diameter
will result in a 17% decrease in the rate (Note, these sensitivities
are evaluated when the efficiency was originally estimated at
24%). This sensitivity is of course one of the main reasons for oper-
ating in the mass transfer free region; and therefore, in order to
confirm that this effect was not an artefact of introducing the effi-
ciency correction, the reactions were repeated using a catalyst with
a lower loading and smaller particle size, i.e. 1% Ru/SiO2 and
20 lm. In this case all the calculated mass transfer limitations,
using the above correlations, were negligible and this catalyst
showed the trend directly. It can also be observed that the 1% cat-
alyst was an order of magnitude slower than the 5% catalyst, which
may be attributed here to differences in the dispersion or errors in
the efficiency calculations. Transmission electron microscopy
(TEM) studies were performed to obtain Ruthenium particle size
distribution data for both 1% and 5% Ru/SiO2 catalysts. This TEM
data collected for both catalysts showed that the 1% Ru/SiO2 cata-
lyst has a smaller ruthenium particle sizes (ꢄ1.5 nm) with a higher
metal dispersion than the 5% Ru/SiO2 catalyst which showed aver-
age particle sizes of ꢄ3 nm, as shown in Fig. 6. The above clearly
3.3. Internal mass transfer
In Fig. 4, tests with varying catalyst particle sizes showed that
the observed reaction rate increased with decreasing particle size,
inferring pore diffusion limitations. Here the particle size was
determined using known sieve fractions and only the initial rate
is reported as it is known that the particle size reduces during
the experiment due to the shear forces from the impeller.
0.020
0.018
0.016
0.014
0.012
0.010
0.008
0.006
0.004
0.002
0.000
0.0025
0.025
0.0020
0.020
0.0015
0.0010
0.0005
0.0000
0.015
0.010
0.005
0.000
0.0
0.2
0.4
0.6
0.8
1.0
X
H
O
2
0-45
45-75
75-125
150-250
Fig. 5. T = 303.15 K, P = 1 barg, corrected rate for 5% Ru/SiO2 using calculated
internal mass transfer effectiveness for hydrogen as well as the uncorrected for 1%
Particle Size (µm)
Ru/SiO2
.
Fig. 4. Observed initial rate with varying particle size.

B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
35
shows that when operating under mass transfer limitations the ob-
served rate is a non-linear function of solvent composition. Even
when removing such mass transfer limitations, through estimation
of the efficiencies or through choice of catalyst, there still remains
an inflexion in the rate at high water concentrations. The change in
rate at ꢃ0.85 mole fraction of water suggests a structural change in
the system as has been previously reported using THz-TDS data
and 2-proponol/water mixtures [41]. Herein, the measured THz-
TDS absorption coefficient, shown in Fig. 7a, is compared with
the rate of reaction. A clear correlation between reaction rate and
spectroscopic data is observed. The absorption coefficient as mea-
sured is directly related to hydrogen-bond dynamics and is consid-
ered as a probe of the structure of the solvent. Qualitatively it is
apparent that above 85–90 molH2O% the alcohol-water mixture
rapidly tends to the absorption coefficient and hence structure of
bulk water. This can be ascribed to the fact that below this concen-
tration, water molecules are largely accounted for within 2-propa-
nol/water networks, with increasing bulk-like water above this
concentration.
It is valid to consider whether studies on alcohol/water mix-
tures have a direct relevance for catalytic systems when studied
in the absence of the reactant molecules. As we and others have
shown, the addition of a small concentration of alcohol into bulk
water dramatically alters the properties of the liquid. The addition
of reactant could therefore feasibly have a similar influence on
alcohol/water mixtures. In order to address this, we have studied
the influence of the addition of 0.33 mol lꢀ1 2-butanone to 2-pro-
panol water mixtures across the entire composition range. It
should be noted that 2-butanone is a hydrogen-bond acceptor
only, unlike 2-propanol and water which are both hydrogen-bond
acceptors and hydrogen-bond donors. Fig. 7b shows the relative
absorption coefficient of the ternary mixture: a greater relative
absorption coefficient indicates more retarded rotational dynamics
and an increase in structuring. Here the maximum relative
absorption coefficient is shifted to a slightly lower value ꢄ85 -
molH2O% than that observed in the absence of MEK [41] and
therefore it can be concluded that the existence of 2-butanone does
not influence the structures of such mixtures significantly; how-
ever, an increase in the magnitude of the relative absorption with
respect to the binary mixture is observed. This indicates that the
presence of 2-butanone further hinders the rotational dynamics
of the system. That 2-butanone interacts with water but not with
2-propanol is confirmed by the fact that the relative absorption
coefficient is zero at 0% water but non-zero at 100% water (with
0.33 mol lꢀ1 2-butanone).
It is clear that the concentration at which the maximum in the
relative absorption coefficient is observed is not only coincident
with rapid acceleration in the reaction rate (Fig. 5) but also with
well-established excess thermodynamic properties of IPA/H2O
mixtures. For instance, the negative excess enthalpy, i.e. excess
with respect to an ideal non-interacting mixture, reaches a maxi-
mum at xH2O ꢄ 0.90. This effect is due to the presence of hydro-
gen-bonding interactions between the water and alcohol
molecules, which in turn influence the structural dynamics of the
solvent.
The above results clearly show that water has a significant ef-
fect on the reaction rate. For the data determined using 5% Ru/
SiO2 at 303.15 K and 1 bar of hydrogen this rate increased by ꢃ7
(pre-correction) to ꢃ75 (post-correction) times that of the IPA sol-
vent. This is much larger than the case of the 1% catalyst which
only showed an order of magnitude increase in reaction rate and
as discussed indicates that the efficiency corrections used here
are likely to have over predicted the ‘real’ reaction rate. Neverthe-
less, there is a clear increase when using water and this increase
appears to correlate with bulk solvent structure. However, if one
normalises the observed reaction rate to the hydrogen concentra-
tion (assuming first order, as identified for the water case), this dif-
ference increases further with the 1% Ru/SiO2 (i.e. non-modified
Fig. 6. HR-TEM images and particle size distribution of (a) 1% Ru/SiO2 and (b) 5% Ru/SiO2 catalysts.

36
B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
(a)
(b)
50
250
40
30
20
10
0
200
150
100
50
0
20
40
60
80 100
0
20
40
60
%
80 100
molH O
%
molH O
2
2
Fig. 7. (a) Absorption coefficient of 2-propanol/water mixtures at 1 THz [41]. (b) Relative absorption coefficient of 2-butanone/2-propanol/water mixtures (303 K) at 1 THz as
evaluated from THz-TDS measurements.
rate) now giving a normalised rate which is 63 times higher in
water when compared to IPA. Such a significant increase in reac-
tion rate necessitates further investigation on how such structural
differences may relate to the observed rate changes between
systems.
of hydrogen into the Ru–C bond as is shown in Fig. 9c. The transi-
tion state is very similar to that shown in Fig. 9d for the C–H bond
formation in the hydroxy route. The activation barrier for this step
was calculated to be 58 kJ molꢀ1 (Table 3b). The subsequent hydro-
genation of the butoxy intermediate to form 2-butanol proceeds
via the addition of hydrogen to the Ru–O bound. Again the transi-
tion state for this reaction which is shown in Fig. 9d is very similar
to that found for the first hydrogenation step in the hydroxy route
(Fig. 9a) as both involve the formation of an O–H bond. The activa-
tion barrier for this step, however, was calculated to be
124 kJ molꢀ1 which is significantly higher than the other three ele-
mentary steps discussed; as such, this may limit the reaction rate.
While the activation barrier for the first hydrogenation step in the
alkoxy route is lower than that found in the hydroxy route, the sec-
ond step in the alkoxy route is significantly higher than that in the
hydroxy route. We carried out kinetic Monte Carlo simulations for
the hydrogenation of MEK over Ru in the vapour phase and found
that the product (2-butanol) is mainly formed via the alkoxy route
[44].
In the presence of the solvent, the hydroxy route was found to
be more favourable than that the alkoxy route. The structures for
all the reactants, products and transition states were calculated
to be very similar to those found for the vapour phase reaction.
MEK adsorbs at the metal–water interface by displacing a water
molecule from the metal surface. Two of the water molecules with-
in the first solvation shell of the adsorbed MEK form hydrogen
bonds with the MEK (Fig. 10a). Similar to the results in the vapour
phase, the hydroxybutyl intermediate, formed by the reaction of
MEK and hydrogen, binds through the carbon atom of its initial
carbonyl group. The hydroxyl group that results is directed to-
wards the solution phase where it is stabilised by forming hydro-
gen bonds with water molecules in solution, as is shown in
Fig. 10b. The Ru–O bond that forms between the hydroxybutyl
and the Ru surface is 2.19 Å as compared to 2.06 Å for MEK which
indicates a stronger interaction of the intermediate with the aque-
ous solvent than with the surface [45].
3.4. Role of solvent
In order to begin to understand the influence of water on the in-
creased rates of reaction, we carried out density functional theoret-
ical calculations to model the adsorption and hydrogenation of
MEK over Ru(0001) in the vapour phase as well as in 2-propanol
and aqueous solvents. A detailed description of these and related
simulations is given elsewhere [44–46].
MEK hydrogenation follows Langmuir–Hinshelwood kinetics
where both MEK and hydrogen chemisorb on the metal surface
and react via a series of hydrogen addition steps through a Hori-
uti–Polanyi like mechanism [47]. MEK adsorbs atop of a Ru atom
of the Ru(0001) surface through its oxygen atom in an g1 config-
uration (Fig. 8a) with an adsorption energy of 33 kJ molꢀ1. The
hydrogenation of MEK can then proceed via two different
pathways.
In the first path (hydroxy route), an adsorbed hydrogen atom
adds to the oxygen of MEK to form a hydroxybutyl intermediate
(Fig. 9b) which is bound di-
r to the surface through its carbon
and oxygen atoms (Fig. 8b). This reaction proceeds via the addition
of hydrogen to the Ru–O bond in the three centre transition state
as depicted (Fig. 9a). The Ru–O bond increases from 2.13 Å in
MEK to 2.26 Å in the transition state as O–H bond begins to form
(1.40 Å). The activation barrier for this step was calculated to be
64 kJ molꢀ1 (Table 3a). The subsequent hydrogenation of the
hydroxybutyl intermediate to form 2-butanol (Fig. 8d) was found
have a barrier of 72 kJ molꢀ1. The transition state for this step in-
volves the addition of hydrogen to the Ru–C bond as is shown in
Fig. 9d.
In the second path (alkoxy route), hydrogen first adds to the
carbon atom to form the 2-butoxy intermediate (Fig. 9c) which is
bound to a three-fold hollow site through its oxygen atom
(Fig. 8c). The transition state for this step involves the insertion
The accessibility of the hydroxyl group to the bulk solvent
allows it to participate in hydrogen bonding with a water molecule
in the bulk solution and leads to extra stabilisation over the

B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
37
Fig. 8. DFT-optimised adsorption configurations for (a) MEK (b) hydroxybutyl intermediate (c) butoxy intermediate and (d) 2-butanol over Ru(0001) in the vapour phase.
Fig. 9. DFT-calculated transition state structures for the hydrogenation of (a) MEK to form the hydroxybutyl intermediate, (b) the hydroxybutyl intermediate to butanol, (c)
MEK to butoxy intermediate and (d) the butoxy intermediate to 2-butanol over Ru(0001) in the vapour phase.
Table 3a
Table 3b
DFT predicted reaction energy and activation barriers for the hydrogenation of MEK to
hydroxybutyl and hydroxybutyl to 2-butanol in vapour, water and IPA.
DFT predicted reaction energy and activation barriers for the hydrogenation of MEK to
butoxy and butoxy to 2-butanol in vapour, water, and IPA.
R1R2Cꢆ ¼ Oꢆ þ Hꢆ ? R1R2Cꢆ—OH R1R2Cꢆ—OH þ Hꢆ ? R1R2CH—OꢆH
R1R2Cꢆ ¼ Oꢆ þ Hꢆ ? R1R2CH—Oꢆ R1R2CH—Oꢆ þ H ? R1R2CH—OꢆH
Solvent
Solvent
D
Erxn (kJ/mol)
Eact (kJ/mol)
D
Erxn (kJ/mol)
Eact (kJ/mol)
D
Erxn (kJ/mol)
Eact (kJ/mol)
D
Erxn (kJ/mol)
Eact (kJ/mol)
Vapour
Water
IPA
21
64
20
32
ꢀ18
ꢀ20
ꢀ20
72
68
70
Vapour ꢀ47
58
58
62
50
4
6
123
115
116
ꢀ23
Water
IPA
ꢀ47
ꢀ42
ꢀ16
R1 ¼ CH3CH2; R2CH3.
R1 ¼ CH3CH2; R2CH3.

38
B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
Fig. 10. DFT-optimised adsorption configurations for (a) MEK, (b) hydroxybutyl intermediate, (c) butoxy intermediate and (d) 2-butanol over Ru(0001) in the aqueous phase.
For clarity, the hydrogen atom involved in hydrogenation has been shown in yellow. (For interpretation of the references to color in this figure legend, the reader is referred to
the web version of this article.)
Fig. 11. DFT-calculated transition state structures for the hydrogenation of (a) MEK to hydroxybutyl intermediate, (b) the hydroxybutyl intermediate to 2-butanol, (c) MEK to
butoxy intermediate and (d) the butoxy intermediate to 2-butanol over Ru(0001) in the aqueous phase. For clarity, the hydrogen atom involved in hydrogenation has been
shown in yellow. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
reactant state (MEK). In the reaction to form the hydroxybutyl
intermediate, the transition state takes on the classic three centre
form with O–H, Ru–O and Ru–H bond lengths of 1.37, 2.18 and
1.7 Å, respectively. Both the hydrogen and the oxygen in the tran-
sition state (Fig. 11a) interact with water molecules within their lo-
cal solvation sphere (including water molecules at the surface as
well as in the bulk solvent) which lowers the activation barrier
to 20 kJ molꢀ1 from 64 kJ molꢀ1 in the vapour phase reaction. The
transition state for the subsequent hydrogenation of the hydrox-
ybutyl intermediate to form 2-butanol in water, which is shown
in Fig. 11b, was also very similar to that in the vapour phase. The
activation barrier in the solution phase was calculated to be

B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
39
activation barrier for this step is lower in water as compared to
IPA by 12 kJ molꢀ1. This is the result of the loss of an addition
hydrogen bond with the solution phase. A more detailed descrip-
tion of the transition state structures is reported elsewhere [45].
In summary, DFT predicts that the activation barriers for hydroge-
nation of MEK follow the order of: vapour > IPA > water. Given that
heptane is the closest representation of the vapour phase, this
trend is in excellent agreement with the reported experimental
data for the solvents (Table 1) particularly when normalised to
hydrogen solubility. As reported here, hydrogenation rates at
303 K were found to be an order of magnitude greater in water
than in IPA. While this can be attributed to the lower activation
barrier for hydrogenation of MEK in water (20 kJ molꢀ1) than IPA
(33 kJ molꢀ1), it does not, however, account for the rapid increase
in reaction rate at xH2O ꢃ 0.8.
Fig. 12. A snapshot from ab initio molecular dynamics simulation of proton
diffusion in pure water. Two water molecules participate in formation of an H5Oþ
In addition to dissolving molecular hydrogen, it is known that
the solvent is also capable of transporting charged reaction species.
The ease of transport of these species can ultimately be decisive in
determining the rate of the reaction. In order to further explain the
peculiar behaviour in the hydrogenation rate of MEK at higher con-
centrations of water (Fig. 1), we carried out ab initio molecular
dynamics to monitor proton diffusivities in IPA-water solvents
with different compositions (xH2O = 1 and 0.87). Proton transfer
was chosen here as a very simple probe of the diffusion of polar
or ionic species in aqueous media and in addition since some
hydrogenation reactions in aqueous media have been suggested
to involve protons that form locally. The simulations were run
out to 5 ps (using time step of 1 fs) in order to allow the proton
to diffuse through the solution. It was observed that the proton
readily shuttles through a chain of water molecules by forming a
2
intermediate as proton shuttles through the hydrogen-bonding network.
65 kJ molꢀ1 which is also lower than that reported above for the
vapour phase (72 kJ molꢀ1).
The hydrogenation of MEK to form the butoxy intermediate in-
volves the formation of a C–H bond in the transition state (Fig. 11c)
which is much less polar than the formation of the O–H bond for
the formation of the hydroxybutyl. As such, there is far less hydro-
gen-bonding stabilisation of the transition state to form the butoxy
as compared to that for the hydroxybutyl. The activation barrier for
this step in presence of the water was calculated to be 58 kJ molꢀ1
,
which is identical to that found for the vapour phase reaction. The
transition state for the subsequent hydrogenation of the butoxy
intermediate in water was calculated to be 120 kJ molꢀ1 which is
very similar to the value of 124 kJ molꢀ1 calculated for the vapour
phase reaction (Fig. 12d). Unlike the adsorbed MEK, the alkoxide
reactant is already quite polar and stabilised by water in the reac-
tant state. The number of hydrogen bonds with water on the
surface and in solution remains the same between the reactant
and the transition state. As such, there is only a very small stabili-
sation (4 kJ molꢀ1) of the transition over the reactant state due to
the presence of water.
A comparison of both mechanisms in the vapour and the
solution phase reveals that the presence of water favours the for-
mation of the O–H bond formation over that of the C–H bond, espe-
cially in the hydroxy path. While both the alkoxy and hydroxy
routes have similar activation barriers for the reactions carried
out in the vapour phase, the hydroxy path is clearly favoured over
the alkoxy route when the reactions are carried out in solution as a
result of the stronger stabilisation of the lower barrier to form the
O–H bond over the C–H bond and the increased stability of the
hydroxybutyl intermediate over the butoxy intermediate.
Previously we have shown that the apparent barriers for the
hydrogenation of ketones and aldehydes to alcohols can be de-
scribed by the rate which depends on the intrinsic hydrogenation
of either the hydroxyalkyl or the alkoxide intermediates [44]. The
results presented here for both mechanisms reveal that water
clearly lowers the effective barrier for the hydrogenation of MEK.
The solution phase also induces a change in the governing reaction
path over that found in the vapour phase.
The same reaction steps were studied using IPA as a solvent. The
calculated activation barriers for the first hydrogenation step in IPA
for the hydroxy and alkoxy mechanisms were calculated to be
32 kJ molꢀ1 and 68 kJ molꢀ1, respectively. As observed in the case
of water, the activation barrier is lower for the hydroxy route as
compared to the vapour phase reaction but there is very little dif-
ference in barrier for the alkoxy route. The relative stabilisation of
transition state for the formation of hydroxybutyl intermediate in
presence of IPA is not as strong as in the case of water. Thus, the
Zundel ion (H5Oþ) like intermediate and a successive sequence of
2
proton transfer steps (Fig. 12). In pure water (xH2O = 1), the diffu-
sivity of the proton was found to be 2 ꢁ 10ꢀ8 m2 sꢀ1 which is in
close agreement with other molecular dynamics and experimental
studies [48]. However, in the water–IPA mixture (xH2O = 0.87)
the diffusivity was found to be over 27 times lower
(7.52 ꢁ 10ꢀ10 m2 sꢀ1). These simulations show that the presence
of IPA disrupts the highly interconnected hydrogen-bonding net-
work between water molecules which subsequently inhibits facile
proton transport and thus lowers the diffusivity. These results indi-
cate that proton transfer increases considerably as solvent compo-
sition changes from xH2O = 0.87 to xH2O = 1 and that this may
significantly enhance hydrogenation activity. This is in agreement
with our experimental results for the hydrogenation of MEK in
mixed solvents (Fig. 1) as well as the THz-TDS data (Fig. 7).
It is clear that hydrogen-bonding structure and dynamics can
play various roles in determining the catalytic behaviour of a sys-
tem. The diffusion of protons through solution, a key step in cata-
lytic hydrogenation described in the present work, and in
electrocatalytic reduction reactions such as those occurring in pro-
ton exchange membrane fuel cells, is coupled to these dynamics
[49,50]. Recent work has shown that the efficiency of this proton
diffusion appears to be coupled to hydrogen-bonding dynamics
and structure of the liquid [51,52]. This diffusion or ‘proton-shut-
tling’ is often described by the Grotthus mechanism [47], which
is discussed in detail elsewhere [47,53,54]. Briefly, it involves the
effective transfer of a proton through the breaking and formation
of hydrogen bonds: specifically, isomerisation between H3O+ and
H5Oþ which couples to hydrogen-bonding dynamics in the second
2
solvation shell of the H3O+. Recent THz-TDS studies have, however,
suggested that 15 water molecules may actually be involved in the
proton diffusion process [55]. Bond-order analysis of molecular
dynamics simulations has shown that the timescale of this proton
transfer is of the order of a few ps [56]. This timescale also corre-
sponds to the lifetime of the alcohol and water clusters in alco-
hol-water mixtures observed by neutron diffraction studies [57]

40
B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
and the timescale of the processes probed by THz-TDS. The struc-
ture of alcohol–water mixtures additionally dictates the diffusion
of solvated electrons, a crucial step in both electrocatalytic and
non-catalytic electrochemical reactions [58]. For example, aqueous
solutions of methanol, ethanol, 1-propanol, 2-propanol and 1-
butanol, 2-butanol and tert-butanol all show a maximum in the
free ion yield upon irradiation at alcohol concentrations of a few
mol% [54–63]. For aqueous 2-propanol, this maximum occurs at
ꢄ97 molH2O% [54].
spectroscopy and reaction kinetic data to better understand com-
plex behaviour in catalytic systems.
Acknowledgments
The authors would like to thank the EPSRC under Grant Number
GR/S43702/01 and Johnson Matthey for funding this work. The
molecular modelling work was done using computational time at
the National Centre for Computational Sciences (NCCS) at Oak
Ridge National Laboratory and the Environmental Molecular Sci-
ences Laboratory (EMSL) at Pacific Northwest National Laboratory.
Both of these are national scientific user facilities sponsored by the
Department of Energy’s Office of Science. The authors are also
grateful for the helpful discussions with Professor Robert J. Davis
(University of Virginia).
4. Conclusions
The results described herein have shown that there is a signifi-
cant variation in the observed rate of the MEK hydrogenation reac-
tion when using 5% Ru/SiO2 in different solvents. In particular, a
complex behaviour is observed for varying water/IPA mole frac-
tions. Using experimental and estimated mass transfer rates we
have shown that while gas–liquid mass transfer also varies signif-
icantly with water mole fraction, it, like the liquid–solid mass
transfer, is not significant at the scale of the reactor used here.
Internal mass transfer was, however, shown to be significant.
Correcting the observed rate for the calculated internal effective-
ness using literature correlations revealed that the rate increased
as the water mole fraction increased. This was confirmed by using
a lower loaded catalyst (1%), although a comparison with the 5%
catalyst indicates that the mass transfer corrections used here
did appear to over predict the ‘true’ reaction rate. Results from
density functional theory calculations showed that water can sig-
nificantly lower the activation energy for the reaction as compared
to the reactions in isopropyl alcohol or the vapour phase, and, in
addition, can alter the preferred hydrogenation mechanism. The
above rationalises why water is a significantly better solvent than
IPA or indeed other alcohols and alkanes. This, however, did not ex-
plain why the intrinsic rate is not ostensibly linear as a function of
solvent composition, but rather goes through a sharp increase at
high water concentrations (>90 mol%). For gas liquid systems, Hu
et al. observed a significant effect at much lower IPA concentra-
tions than this – a few mole% and this was explained by the sur-
face-active behaviour of the alcohol at the gas liquid interface
[30]. This explanation does not appear germane in this context.
In order to explore the probable explanation, the structure of the
IPA-water mixture was explored and a correlation observed but
one which did not correspond completely to the composition of
minimum diffusivity or azeotrope. Therefore, ab initio molecular
dynamics techniques were employed which showed that water
was also found to facilitate diffusion of protons as well as the dif-
fusion of MEK demonstrating significant enhancements for con-
centration greater than 90% water.
References
[1] M.H. Abraham, P. Grellier, J.M. Abboud, R.M. Doherty, R.W. Taft, Can. J. Chem.
66 (1988) 2673.
[2] L. Gilbert, C. Mercier, in: M. Guisnet, J. Barbier, J. Barrault, G. Bouchoulo, D.
Duprez, G. Pérot, C. Montassier (Eds.), Heterogeneous Catalysis and Fine
Chemicals III, Elsevier, Amsterdam, 1993, p. 51.
[3] S. Asplund, C. Fornell, A. Holmgren, S. Irandoust, Catal. Today 24 (1995) 181.
[4] E. Toukoniitty, P. Mäki-Arvela, J. Kuusisto, V. Nieminen, J. Päivärinta, M.
Hotokka, T. Salmi, D.Y. Murzin, J. Mol. Catal. A: Chem. 192 (2003) 135.
[5] S. Mukherjee, A. Vannice, J. Catal. 243 (2006) 108.
[6] S. Gómez-Quero, E. Díaz, F. Cárdenas-Lizana, M. Keane, Chem. Eng. Sci. 65
(2010) 3786.
[7] B. Hu, R.P. Fishwick, A.W. Pacek, J.M. Winterbottom, J. Wood, E.H. Stitt, A.W.
Nienow, Chem. Eng. Sci. 62 (2007) 5392.
[8] J. Wood, J.A. Bennett, N.J. Creamer, L.E. Macaskie, L. Bodeness, in: 8th World
Congress of Chem. Eng., Montreal, Canada, 2009.
[9] J. Masson, P. Cividino, J. Court, Appl. Catal. A: Gen. 161 (1–2) (1997) 191.
[10] R.A. Rajadhyaksha, S.L. Karwa, Chem. Eng. Sci. 41 (7) (1986) 1765.
[11] U.K. Singh, M.A. Vannice, Appl. Catal. A: Gen. 213 (2001) 1.
[12] L. Vanoye, M. Zanota, A. Desgranges, A. Favre-Reguillon, C. De Bellefon, Appl.
Catal. A: Gen. 394 (2011) 276.
[13] P. Kluson, L. Cerveny, Appl. Catal. A: Gen. 128 (1995) 13.
[14] M. Hronec, Z. Cvengrošová, M. Králik, G. Palma, B. Corain, J. Mol. Catal. A:
Chem. 105 (1996) 25.
[15] P.D. Vaidya, V.V. Mahajani, Chem. Eng. Sci. (60) (2005) 1881.
[16] J. Weissenrieder, A. Mikkelsen, J.N. Andersen, P.J. Feibelman, G. Held, Phys. Rev.
Lett. 93 (19) (2004) 196102.
[17] S.K. Desai, M. Neurock, Phys. Rev. B 68 (7) (2003) 075420.
[18] J.E. Tanner, J. Chem. Phys. 52 (1970) 2523.
[19] P. Hohenberg, W. Kohn, Phys. Rev. 136 (1964) B864 LP.
[20] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.
[21] W. Kohn, L.J. Sham, Phys. Rev. 140 (1965) A1133 LP.
[22] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C.
Fiolhais, Phys. Rev. B 46 (1992) 6671.
[23] D. Vanderbilt, Phys. Rev. B 32 (1985) 8412.
[24] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) LP 5188.
[25] G. Mills, H. Jonsson, G.K. Schenter, Surf. Sci. 324 (1995) 305.
[26] W.G. Hoover, Phys. Rev. A 31 (1985) 1695.
[27] S. Nose, J. Chem. Phys. 81 (1984) 511.
[28] Materials Studio, v4.4.0.0 ed. Accelrys Software Inc., San Diego, CA, 2008.
[29] M.P. Allen, D.J. Tildesley, Computer Simulation of Liquids, Oxford University
Press, Oxford, 1989.
[30] B. Hu, A.W. Nienow, A.W. Pacek, E.H. Stitt, Ind. Eng. Chem. Res. 46 (13) (2007)
4451.
[31] J.G. Park, S.H. Lee, J.S. Ryu, Y.K. Hong, T.G. Kim, A.A. Busnaina, J. Electrochem.
Soc. 153 (9) (2006) G811.
[32] E. Dietrich, C. Mathieu, H. Delmas, J. Jenck, Chem. Eng. Sci 47 (1992) 3597.
[33] F. Kapteijn, E. Crezee, B.W. Hoffer, R.J. Berger, M. Makkee, J.A. Moulijn, Appl.
Catal. A: Gen. 251 (2003) 1.
[34] R.P. Fishwick, J.M. Winterbottom, E.H. Stitt, Catal. Today 79 (2003) 195.
[35] Y. Sano, N. Yamaguchi, T. Adachi, J. Chem. Eng. Jpn. 7 (1974) 255.
[36] L.G. Nischhenkova, M.V. Ulitin, V.N. Gorelov, Izv. Vys. Ucheb. Zaved. Khim. I
Khim. Tekhnol. 33 (1990) 22.
[37] C.R. Wilke, P. Chang, Am. Inst. Chem. Eng. J. 1 (1955) 264.
[38] E. Scheibel, Ind. Eng. Chem. 46 (1954) 2007.
The above discussions highlight the potential beneficial role of
water in both lowering the activation barrier as well as increasing
the proton diffusion coefficient and combined these effects appear
to correlate well with the experimental data reported. However, it
should also be noted that significant deactivation is observed in the
presence of water [64]. At high concentrations of water, this deac-
tivation occurs at a sufficiently fast time scale so that it should be
included in any reaction model. However, it did not affect the ini-
tial rate data presented here.
[39] L.R. Perkins, C.J. Geankoplis, Chem. Eng. Sci. 24 (1969) 1035.
[40] J. Kendall, K. Monroe, J. Am. Chem. Soc. 39 (1917) 1787.
[41] J. McGregor, R. Li, J.A. Zeitler, C. D’Agostino, J. Collins, M.D. Mantle, H.G.
Manyar, J.D. Holbrey, T. Youngs, C. Hardacre, E.H. Stitt, L.F. Gladden, Phys.
Chem. Chem. Phys., submitted for publication.
The results reported herein therefore suggest that catalytic per-
formance can be tailored through controlling the structural
dynamics of the solvent, e.g. through changing composition.
Overall, this work confirms and highlights the importance of the
solvent system in modifying the mass transfer, diffusion and
reaction mechanism during the hydrogenation of MEK. Further-
more, it demonstrates that it is possible to link DFT calculations,
[42] P.B. Weisz, C.D. Prater, Adv. Catal. 6 (1954) 144.
[43] S. Mukherjee, M.A. Vannice, J. Catal. 243 (2006) 108.
[44] N.K. Sinha, M. Neurock, J. Catal., submitted for publication.
[45] N.K. Sinha, M. Neurock, Chem. Cat. Chem. submitted for publication.

B.S. Akpa et al. / Journal of Catalysis 289 (2012) 30–41
41
[46] N.K. Sinha, Ph.D. Thesis, Theoretical Insights into the Influence of Solvent on
Heterogeneous Catalytic Hydrogenation and Hydrogenolysis Reactions,
University of Virginia, 2011.
[56] H. Lapid, N. Agmon, M.K. Petersen, G.A. Voth, J. Chem. Phys. 122 (2005) 1.
[57] L. Dougan, S.P. Bates, R. Hargreaves, J.P. Fox, J. Crain, J.L. Finney, V. Reat, A.K.
Soper, J. Chem. Phys. 121 (13) (2004) 6456.
[47] J. Horiuti, M. Polanyi, Trans. Farad. Soc. 30 (1934) 1164.
[48] B. Mauro, I. Tamio, T. Kiyoyuki, Chem. Phys. Chem. 6 (2005) 1775.
[49] M.J. Janik, C.D. Taylor, M. Neurock, Top. Catal 46 (3–4) (2007) 306.
[50] M.J. Janik, C.D. Taylor, M. Neurock, J. Electrochem. Soc. 156 (1) (2009) B126.
[51] N. Agmon, Chem. Phys. Lett. 244 (5) (1995) 456.
[52] C.D. Taylor, M. Neurock, Curr. Opin. Solid State Mater. Sci. 9 (1-2) (2005) 49.
[53] D. Marx, M.E. Tuckerman, J. Hutter, M. Parrinello, Nature 397 (6720) (1999)
601.
[58] J. Cygler, G.R. Freeman, Can. J. Chem. – Revue Can. Chimie 62 (7) (1984) 1265.
[59] A.D. Leu, K.N. Jha, G.R. Freeman, Can. J. Chem. – Revue Can. Chimie 60 (18)
(1982) 2342.
[60] Y. Maham, G.R. Freeman, J. Phys. Chem. 89 (20) (1985) 4347.
[61] Y. Maham, G.R. Freeman, J. Phys. Chem. 91 (6) (1987) 1561.
[62] P.C. Senanayake, G.R. Freeman, J. Chem. Phys. 87 (12) (1987) 7007.
[63] Y. Maham, G.R. Freeman, Can. J. Chem. – Revue Can. Chimie 66 (7) (1988) 1706.
[64] H.G. Manyar, D. Weber, H. Daly, J.M. Thompson, D.W. Rooney, L.F. Gladden,
E.H. Stitt, J.J. Delgado, S. Bernal, C. Hardacre, J. Catal. 265 (1) (2009) 80.
[54] M. Tuckerman, K. Laasonen, M. Sprik, M. Parrinello, J. Phys. Chem. 99 (16)
(2002) 5749.
[55] K.J. Tielrooij, R.L.A. Timmer, H.J. Bakker, M. Bonn, Phys. Rev. Lett. 102 (19)
(2009) 198303.