This	document	is	the	Accepted	Manuscript	version	of	a	Published	Work	that	appeared	in	final	form	in	ChemSoc.Rev,	
copyright	©	The	Royal	Society	of	Chemistry	2017	aGer	peer	review	and	technical	ediIng	by	the	publisher.	To	access	
the	 final	 edited	 and	 published	 work	 see:	 hKps://pubs.rsc.org/en/content/arIclelanding/2017/gc/c7gc02692g#!
divAbstract

Green Chemistry

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

PAPER

Cite this: Green Chem., 2017, 19,
5932

View Journal | View Issue

A mechanism for the selective epimerization of the
glucose mannose pair by Mo-based compounds:
towards catalyst optimization†
M. Rellán-Piñeiro,

a

M. Garcia-Ratés

b

and N. López

*a

The selective C2 epimerization of the glucose/mannose pair on a set of Mo-based catalysts was studied
by means of density functional theory. The process, known as the Bilik reaction, encompasses a 1,2
C-shift of the C3 centers at the sugars. Molybdic acid was initially proposed as a catalyst in this reaction,
and recent experimental studies have shown that the polyoxometalate (POM) Keggin cluster H3PMo12O40
also presents a good performance. In the present work, we propose a reaction mechanism for the epimerization on the Keggin cluster with different heteroatoms and extend it to a larger POM, H6P2Mo18O62,
and the continuous α-MoO3(010) surface. We have found that in the transition state corresponding to the
1,2 C-shift the Mo center acts as an electron buffer that promotes the transformation of the aldehyde
group in C1 into an alkoxy group and the C2 alkoxy into an aldehyde group. As a consequence, the activity
of Mo-containing compounds can be traced back to the reducibility of the Mo center and a simple microReceived 4th September 2017,
Accepted 15th November 2017

kinetic model illustrates that this descriptor generates an activity volcano. This allows the identification of

DOI: 10.1039/c7gc02692g

a new POM that shall be 4.7 times more active than the parent compound. We have thus shown that continuum models linking the properties of molecular cluster-like catalysts and oxide surfaces can be derived

rsc.li/greenchem

and this paves the way towards a unified theory in catalysis.

Introduction
Obtaining compounds with controlled chirality is a challenge
to develop platform chemicals from biomass. Some of the
target compounds could be easily generated from sugars but
in some cases the natural abundance of the parent sugar is
relatively small (rare sugars) whereas their epimers are widely
abundant. The chemical toolbox now allows epimerization
reactions,1 that is, the rearrangement of a single stereogenic
center in a molecule that contains several of them.2 In nature,
epimerases catalyze this reaction at multiple carbon positions
providing a high specificity for products,3,4 but their transfer to
the industrial scale suffers from the very specific conditions of
temperatures and pH as well as difficult recycling.5 Therefore,
there is a need to develop inorganic catalysts that can allow the
successful epimerization avoiding these drawbacks.
Lobry de Bruyn–Alberda van Ekenstein observed that
during the isomerization of aldoses to ketoses promoted by a

a
Institute of Chemical Research of Catalonia, ICIQ, and The Barcelona Institute of
Science and Technology, Av. Països Catalans 16, 43007 Tarragona, Spain.
E-mail: nlopez@iciq.es
b
Max Planck Institute for Chemical Energy Conversion, Stiftstrasse 34-36,
D-45470 Mülheim an der Ruhr, Germany
† Electronic supplementary information (ESI) available. See DOI: 10.1039/
c7gc02692g

5932 | Green Chem., 2017, 19, 5932–5939

base, epimers were obtained as byproducts.6 Lewis acids like
Ni+2 diamines in methanol are active and the equilibrium is
reached only in 5 min at 60 °C.7,8 However, the catalyst is
complex and the reactivity is limited by the low solubility of
sugars in methanol.9 Other Lewis acids like metal(III) chlorides
(CrCl3, AlCl3, InCl3, LaCl3, DyCl3, and YbCl3) epimerize
glucose to fructose, giving mannose as a secondary product.10
Alternatively, the Sn-Beta zeolite has also been proposed as an
active catalyst for epimerization; framework Sn sites are active
in methanol,11 and partially hydrolyzed Sn sites catalyze the
reaction when Na+ is added to the reaction media.12 Borate
salts also activate the Sn-Beta zeolite for epimerization reactions.13,14 Mechanistic investigations by 13C and 1H NMR
showed that epimerization proceeds via 1,2 C-shift and, in
some cases, also by C1–C2 hydride transfer. Generally, these
catalysts perform the reaction via the coordination of the sugar
to only one metal center.
However, the most prominent inorganic catalysts are based
on Mo. Mo-Catalysts have found several applications in
biomass conversion and, most recently, also when employed
as single atoms or small clusters.15–21 In the 1970s, Bilik and
coworkers22 reported the epimerization of aldoses at the C2position catalyzed by molybdates under acidic conditions and
gave name to the reaction.23 In their experiments, the thermodynamic equilibrium between the aldose pair was reached
after 3 h at 90 °C, and the reaction was quite neat. In 1982,

This journal is © The Royal Society of Chemistry 2017

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

Green Chemistry

Scheme 1 Schematic representation of the epimerization of glucose to
mannose. The carbons where the rearrangement occurs are marked in
colors illustrating the chiral nature of the C2 positions.

NMR-based mechanistic studies24 with labeled 13C1 and 13C2
aldoses showed a stereospecific carbon-skeleton rearrangement during the reaction formally leading to the C2 epimerization by C1–C2 carbon shift (Scheme 1).
The mechanistic studies with molybdates24–26 indicate that
the reaction is intramolecular and takes place through the
open form of the aldose. The accepted molecular mechanism
presented in Scheme 2 requires two Mo-centers to occur. The
hydrated aldose coordinates through the four hydroxyl groups
C1–C4 (the oxygen atoms on C1 and C4 are attached to Mo
atoms, while the oxygen atoms on C2 and C3 form a bridge
between Mo).23,27 Then, in the transition state, the C3–C2
breaks while C3–C1 is simultaneously formed. This process is
accompanied by the dehydration of C1 and hydration of C2.
The chain rearrangement leads to the inversion of configuration in C2 obtaining the epialdose-dimolybdate complex. The
presence of a terminal aldehyde group and hydroxyl groups in
C2 and C3 is essential, and the alcohol group is not needed in
position C4, although it enhances the rate. Using Density
Functional Theory, DFT, only the epimerization step has been
studied with Mo oxide dimers and tetramers and due to rigidity, the former was found to be better for this elementary
step.28
Molecular oxides, like the molybdenum-based polyoxometalate (POM) H3PMo12O40, are also active and highly selective catalysts, more than 90%, in aqueous solution and acid media
( pH = 2.5–3.5).19 By 13C NMR the C1–C2 carbon shift reported
in the Bilik reaction was found and XPS identified electron
transfer from glucose to the Mo octahedra units. The kinetic
studies reported apparent activation barriers between 95 and
100 kJ mol−1, thus slightly lower than those reported for
molybdic acid, 126 kJ mol−1.29 More importantly, the catalyst

Scheme 2

Paper

was stable under reaction conditions and could be recycled.
Later, this catalyst was employed to transform glucose to
mannose in aqueous solution as the first step to obtain mannitol,30 and POM overperformed bulk molybdenum oxide, which
suffered complete leaching. Other closely related applications
like tandem reactions for the production of alkyl lactates from
ketohexoses31 in ethanol have been identified on bulk MoO3.
Finally, niobium molybdates (LiNbMoO6 and HNbMoO6) have
been proposed as heterogeneous catalysts for the reaction.32
In this work, we study by first principles methods the epimerization reaction of glucose to mannose catalyzed by
different Mo oxide species including two POMs:33 the Keggin34
H3PMo12O40 (with a set of different heteroatoms), the
Dawson H6P2Mo18O62 clusters and the bulk material
α-MoO3(010),17 both as pristine and defective surfaces. We
propose a mechanism for the Bilik reaction which explains the
experimental observations and requires a single center.
Finally, we have found that the reducibility of the Mo reaction
centers is the unique descriptor for the activity of Mo oxides
irrespective of their molecular or continuous nature.

Computational methods
Density Functional Theory (DFT) calculations were carried out
with the 5.3.3 version of Vienna ab initio simulation package
(VASP) for the molecular and continuous Mo-based catalyst.35,36 The Perdew–Burke–Ernzerhof (PBE) functional37 was
chosen to describe the exchange correlation energy. To correctly describe the on-site Coulomb interactions of localized
4d electrons of molybdenum atoms the Hubbard U correction,38 Ueff, with a value of 3.5 eV was introduced. The Ueff
value was optimized to reproduce the correct reaction energy
and electron localization for the dissociative H2 adsorption.
Gaussian smearing with smearing parameter σ = 0.1 eV was
used. The van der Waals contributions were accounted for
with Grimme’s semiempirical vdW-D2 correction.39 Projectoraugmented wave (PAW) pseudopotentials40 were used to
replace the inner electrons. The kinetic cut-off energy to
expand the valence electrons in plane-waves was set to 450 eV
and spin polarization was taken into account.
For all the molecules present in this study, including polyoxometalates, HnXMo12O40, a 19.5 × 20 × 20.5 Å3 box was used.
For the H6P2Mo18O62 cluster the simulation box volume was

Accepted mechanism for the epimerization of aldose pairs, Bilik reaction, catalyzed by molybdic acid. Adapted from ref. 23.

This journal is © The Royal Society of Chemistry 2017

Green Chem., 2017, 19, 5932–5939 | 5933

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

Paper

increased to 25.5 × 25 × 24.5 Å3. All cells were large enough to
avoid the interaction between structures of neighboring unit
cells, which were sampled at the Gamma point. All atoms were
allowed to relax. Benchmark calculations for these settings
were done with the Gaussian code obtaining very similar
values, see Table SI1.† A slab model was used to test the reactivity on the α-MoO3(010) surface. A (3 × 3) supercell containing two bilayers with 3 × 3 × 1 k-point sampling was considered. The upper bilayer and adsorbates were relaxed. A
vacuum gap of 15 Å was added between slabs. The use of asymmetric slabs causes spurious terms, which were avoided with
the use of a dipole correction along the z axis.41 The reaction
path was tested both on the clean surface and on the terminal
oxygen, Ot, defective surface with one vacancy per supercell.
The climbing image nudged elastic band (CI-NEB)42,43 method
was used to find the transition states, which were then refined
with the improved dimer method (IDM).44,45 A vibrational analysis was applied to confirm all the minima and saddle points,
and all energies were corrected by ZPVE. Solvation was
included for [H3−nPMo12O40]−n, n between 0 and 3, through
the VASP-MGCM continuum model46,47 for an aqueous
environment. To balance the charge of the Keggin cluster a
cation with the equivalent valence Rb+1, Sr+2 or Y+3 was added
to the system. To minimize spurious electrostatic interactions
the distance between the anion and the cation was set to its
maxima within the box. For the three systems the multiplicity
of the POM was singlet. All the structures are published in the
ioChem-BD database48 and can be retrieved in ref. 49.

Results
The energy profile for the epimerization reaction starting by
open glucose was calculated for the reaction on H3PMo12O40,
H6P2Mo18O62 and the pristine and defective α-MoO3(010) surfaces. The structures of these compounds are shown in Fig. 1.

Fig. 1 Mo-Based systems where the Bilik reaction was investigated:
(a) Keggin POM H3PMo12O40; (b) Dawson POM H6P2Mo18O62, two different
types of Mo centers are indicated by 1 and 2; (c) α-MoO3; (d) α-MoO3−x.
Mo atoms in green, O in red, P in orange and H in white.

5934 | Green Chem., 2017, 19, 5932–5939

Green Chemistry

All Mo centers of deprotonated [PMo12O40]3− and pristine
α-MoO3(010) surfaces are equivalent whereas H6P2Mo18O62 has
two different Mo environments (labelled as 1 and 2 in Fig. 1b).
The molybdenum atoms in these three catalysts present a
formal oxidation state of +6. In the defective α-MoO3−x(010)
surface two Mo+6 centers are reduced to Mo+5.50–52
The coordination environments of the Mo centers in
H3PMo12O40 do not allow the two center glucose coordination
observed in glucose-dimolybdate complexes.23,27 Glucose coordinates to a single Mo, thus being similar to metal chlorides,10
which results in the mechanism and reaction profiles shown
in Fig. 2. Detailed reaction energies and energy barriers are
shown in Table SI2;† and geometric parameters are provided
in Table SI3.† Reaction free energy profiles including
vibrational terms are shown in Fig. SI2.†
The overall reaction is almost thermoneutral, with the formation of mannose slightly favored by 0.09 eV. The barrier for
non-catalyzed gas-phase epimerization of glucose to mannose
is 1.37 eV. For H3PMo12O40, the open form of the glucose is
dissociatively adsorbed on a Mo center (R1) through the TSdiss.
In the adsorbed state, Gdiss, the hydroxyl group of C2 is
bonded to a Mo center whereas the C1-aldehyde group interacts with the same Mo center forming a bidentate complex.
This step is endothermic by 0.57 eV, with an activation barrier
of 1.04 eV. Although glucose adsorption through other
hydroxyl groups is possible, Fig. SI3,† it will not lead to the
desired reactivity. The adsorbed glucose undergoes 1,2 C-shift
(R2) through TSepim where the C2–C3 bond is broken and C1–C3
is simultaneously formed. Concomitantly the C1 aldehyde transforms into an alkoxy group and C2 alkoxy into an aldehyde
group. The corresponding final state is the dissociated
mannose, Mdiss, bonded to the Mo center through the oxygen of

Fig. 2 (a) Proposed mechanism10 for the epimerization reaction on
H3PMo12O40. (b) Optimized structures for reaction on H3PMo12O62.
(c) Reaction profile for the Bilik reaction for glucose on H3PMo12O40,
H6P2Mo18O62, α-MoO3(010) and α-MoO3−x(010). Dehydration step on
α-MoO3−x(010) is represented by the blue line.

This journal is © The Royal Society of Chemistry 2017

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

Green Chemistry

the alkoxy group of C1, with the aldehyde group of C2 also
interacting with the same Mo center. The forward energy
barrier is 0.99 eV. Finally, the proton is transferred back to Mdiss
driving the associative desorption (R3) of the mannose through
TSasso. This process is exothermic by 0.64 eV with a barrier of
0.40 eV.
The reaction mechanism starts with the open form of the
glucose, but in water aldoses are mainly in their ring configurations. The presence of Mo-based compounds in reaction
media can catalyze glucose opening, see Fig. SI4.†
The Dawson cluster has two different Mo positions,
Fig. 1b. Position 1 exhibits a better performance in both,
binding and energy barriers for the reaction and, therefore,
only the reactivity at this site will be discussed. The dissociative adsorption of glucose (R1) is endothermic by 0.75 eV, and
the barrier is 1.07 eV. The 1,2 C-shift step (R2) is nearly thermoneutral, with Mdiss formation slightly favored by 0.05 eV.
The energy barrier is 1.10 eV and the associative desorption of
mannose (R3) is exothermic, −0.79 eV, with a small barrier
of 0.31 eV.
We extended the reaction path calculated for POMs to the
bulk α-MoO3 on the pristine and defective (010)
surfaces.15–18,21 The mechanism on the regular surface follows
that found for POMs. However, the first step is only slightly
endothermic by 0.20 eV, with a relatively low barrier of 1.17 eV.
The reaction energy for the epimerization step is −0.04 eV,
practically thermoneutral, and the forward activation energy is
0.94 eV. The desorption of mannose is exothermic, with a reaction energy of −0.24 eV and a barrier of 0.85 eV. Oxides might
lose oxygen and thus the reaction on the reduced surface
needs to be investigated. Unlike the Mo+6 centers, glucose
adsorption on the Mo+5 center of α-MoO3−x(010) through the
–OH group of C2 is non-dissociative, and it is highly exothermic, −1.26 eV. The adsorbed glucose is dissociated by transferring the hydrogen atom of the interacting –OH group to a
neighboring Ot. This –OH dissociation is slightly endothermic
by 0.14 eV. The energy barrier to this step is 0.53 eV. Most
likely, this step corresponds to a proton coupled electron transfer as the images along the path belong to two different electron localizations. Again, the 1,2 C-shift step is close to
thermoneutral, −0.04 eV, and its barrier is 0.87 eV. After this
step, the hydrogen is transferred back from the surface to the
mannose oxygen bonded to the Mo center. This process is
exothermic by 0.23 eV with an energy barrier of 0.55 eV.
Desorption of the mannose molecule formed in this step is
highly endothermic, with a reaction energy of 1.40 eV.
However, defective MoO3−x has been reported as an active catalyst for the hydrodeoxygenation (HDO) of oxygenated compounds as alcohols and polyalcohols through the breaking of
C–O bonds.15,16,18 Glucose has five –OH groups that could
undergo HDO. We have calculated the dehydration step for the
–O5H group of glucose obtaining an energy barrier of 1.01 eV,
Fig. SI5.† Similar values are expected for the other –OH
groups. This result compromises the selectivity for epimerization on the defective MoO3−x(010) surface towards dehydrated
compounds and discards it as a catalyst for the reaction.

This journal is © The Royal Society of Chemistry 2017

Paper

Theory also explains why in experiments vicinal hydroxyl
groups have an important effect on the epimerization reaction.23 In our path, the relevance of the C2 alcohol group is
evident, since it takes part directly in the mechanism, as
glucose adsorbs on the Mo center through this alcohol group
and its transformation into a ketone allows the chain
rearrangement. In order to study the influence of –O3H and
–O4H groups, we calculated the intermediates and transition
states for the 1,2 C-shift step for glucoses without these groups
on H3PMo12O40. The energy profile of this step for three
systems is shown in Fig. SI6.† The presence of –O3H and –O4H
has a weak influence on the adsorption energy of molecules;
they are between 0.47–0.57 eV for the three systems. However,
the energy barriers increase from 0.99 eV, with glucose, to 1.08 eV
without –O4H and to 1.23 eV without –O3H, in agreement
with the experiments with molybdates,24 Table SI4.† Although
these –OH groups do not play a direct role in the interaction
with the POM, the presence of –O3H has a great influence on
the stabilization of the transition state. In the adsorbed
glucose the C3–O3 bond distance is 1.413 Å. This distance is
reduced to 1.350 Å at the transition state, indicating the formation of a partial double bond C3GO3H that stabilizes the
transition state. This effect could also be present in the reaction catalyzed by molybdates. In adsorbed glucose a hydrogen
bond is formed between –O3H and –O4H with a distance of
2.432 Å. In the transition state structure, C3 has partial sp2 geometry reducing this distance to 2.259 Å, thus stabilizing the
structure. In adsorbed glucose without –O4H, the –O3H group
forms a strong hydrogen bond with the –O5H group with a distance of 1.800 Å. In contrast, for glucose, this distance
increases to 2.000 Å at the transition state due to the geometry
change in C3, increasing the activation energy. Thus, computed results explain previous experimental observations for
molybdates.
The reaction is carried out in aqueous solution, and we
have investigated the influence of the solvent only for the
H3PMo12O40 system. In gas-phase, the dissociative adsorption
of sugars involves high energy barriers, but the presence of
water molecules acting as hydrogen shuttles reduces this
barrier to 0.54 and 0.55 eV for glucose and mannose, respectively, Fig. SI7.† In addition, in aqueous solution the Keggin
cluster has a high solubility and is fully dissociated.53 We have
investigated how deprotonation and solvation affects the 1,2
C-shift. Binding energy for dissociated glucose and mannose
and the energy barrier for this elementary step on solvated
[HnMo12O40]3−n are shown in Fig. 3 and Table SI5.† The dissociative binding energy of glucose and mannose on solvated
H3PMo12O40 is not affected by the presence of water. However,
the energy barrier for 1,2 C-shift is reduced by 0.24 eV. This is
due to the polarity of water that stabilizes the transition state
structure. Binding and activation energies for 1,2 C-shift
increase almost linearly with the overall charge of the deprotonated POMs, Fig. SI8,† the binding energy ranging from 0.57
to 1.05 eV and the activation barrier from 0.79 to 1.00 eV. The
presence of a polar solvent smoothed this effect by the stabilization of the charge. The completely deprotonated POM,

Green Chem., 2017, 19, 5932–5939 | 5935

View Article Online

Paper

Green Chemistry
Table 1 Binding energy, BE, in eV, of glucose and mannose on
HnXMo12O40 (relative to gas-phase glucose and POM), activation energy,
Ea, eV, for 1,2 C-shift, and Bader analysis of the charge transfer from the
heteroanion to the Mo12O36 skeleton

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

X

[PMo12O40]−3, has an energy barrier for the epimerization
elementary step of 1.00 eV, very similar to the barrier calculated for H3PMo12O40 in gas phase. Geometrical details are
shown in Table SI6.† In order to have a better understanding
of the role of the solvent in this step we performed a set of calculations with one and two explicit water molecules, Fig. SI9
and Table SI7.† One water molecule was adsorbed on the –OtH
group bonded to the Mo reactive center through a hydrogen
bond, the water molecule being the hydrogen donor. The presence of this water molecule reduces the activation energy by
0.12 eV. This water molecule weakens the Ot–Mo bond allowing a better stabilization of the transition state. Alternatively, if
the water molecule is placed between –O3H and –O4H, being
the hydrogen acceptor of –O3H and the donor for –O4H, the
activation energy is reduced by 0.10 eV as water assists the formation of the partial double bond C3GO3H. We also calculated
the system including simultaneously these two water molecules, and then the activation energy is reduced by 0.19 eV,
and thus the value is very similar to that calculated with the
solvation VASP-MGCM model.
Finally, we tested the influence of the heteroatom in the
HnXMo12O40 (X = P, As, Si, Ge) in the 1,2 C-shift elementary
step. It is reported that the heteroatom does not participate
directly in the reactivity, but these heteroatoms present
different formal charges and thus induce different charge distribution over the POM structure which finely tunes the redox
properties of the POM framework.54,55 Intermediates and transition states on HnXMo12O40 (X = S, P, As, Si, Ge) for this
elementary step have been calculated. Energy profiles for this
step are shown in Fig. SI10.† As is shown in Table 1 the heteroatom has a relatively mild influence. Binding energies range
from 0.51 to 0.67 eV for glucose, and between 0.49 and 0.62 eV
for mannose, see Tables SI8 and SI9.† Energy barriers for the
glucose to mannose epimerization are between 0.95 and
1.07 eV. Binding and activation energies increase with the
number of valence electrons in the heteroatom, and, for a
given valence with the heteroatom size.
The lowest energy barriers for the epimerization step are on
solid α-MoO3(010), both pristine and defective surfaces. This

5936 | Green Chem., 2017, 19, 5932–5939

BE (M)

S
P
As
Si
G
Fig. 3 Charge effects in the binding, BE, and activation energies, Ea,
both in eV, for the 1,2 C-shift step at the Keggin cluster including solvation effects. Solvated POMs and solvated open form of glucose as
reference.

BE (G)

Ea TSepim
(G → M)

0.51
0.57
0.61
0.65
0.67

0.49
0.55
0.57
0.59
0.62

0.95
0.99
1.00
1.02
1.07

Formal
charge
XO4n−

Bader
charge
XO4n−

Transferred
charge

−2
−3
−3
−4
−4

−1.61
−2.23
−2.19
−2.72
−2.69

−0.39
−0.77
−0.89
−1.28
−1.31

energy barrier on α-MoO3−x(010) is 0.09 eV lower than that on
α-MoO3(010) and 0.16 eV lower than that on H3PMo12O40.
However the adsorption energies of glucose and mannose on
α-MoO3−x(010) are highly exothermic, −1.26 and −1.40 eV
respectively, and therefore, the desorption of glucose/mannose
is the most energy demanding step. α-MoO3(010),
H3PMo12O40, and H6P2Mo18O62 have similar energy barriers
for the dissociative adsorption of the glucose/mannose pair,
Fig. 2. However, the binding energies of dissociated molecules
and the energy barrier for the epimerization step increase in
the order: α-MoO3(010) < H3PMo12O40 < H6P2Mo18O62.
As stability seems to be an issue for the Mo-based catalyst,
we have investigated the thermodynamic stability of crystalline
structures in terms of their cohesive energy, i.e. the energy
required to separate the crystalline material into isolated molecular units. We have calculated the cohesive energy for
[PMo12O40]−3, the POM species present in aqueous solution,
and for bulk α-MoO3 in order to compare their stabilities. The
monomer molybdic acid, H2MoO4, was used as a reference, as
it is the smallest molybdenum compound present in aqueous
solution. The dissolution of MoO3 and [PMo12O40]−3 can be
written as:
nMoO3 þ nH2 O ! nH2 MoO4

ð1Þ

½PMo12 O40 ŠÀ3 þ 12H2 O ! 12H2 MoO4 þ PO4 3À

ð2Þ

The calculated cohesive energies are 1.06 eV per MoO3 for
bulk α-MoO3, and 1.78 eV per MoO3 for [PMo12O40]−3. This
higher thermodynamic stability of the POM is due to the
electrostatic interactions between the encapsulated anion and
the Mo12O36 framework that makes it stable against leaching
in aqueous solution. The same strategy is used for the stability
of layered niobium molybdates where the presence of Li+ or H+
atoms in the interlayer region has a similar effect as the encapsulated ion in POMs, avoiding leaching in aqueous solution by
improving electrostatics.32

Discussion: descriptor analysis
Direct deprotonation of the POM affects the MoO-framework
charge and similarly POMs with different heteroatoms present
slightly different MoO-framework charges due to the charge

This journal is © The Royal Society of Chemistry 2017

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

Green Chemistry

Paper

transfer between the anions and the MoO6 POM units, see
Table 1 and Table SI10.† Binding energies and the energy
barrier for the 1,2 C-shift increase linearly with the POM MoOframework charge, Fig. SI11.† The reduction potentials follow
a similar trend as they decrease linearly with the degree of
deprotonation.55–57
Analyzing our reaction network we find that during 1,2
C-shift the C1 aldehyde group is transformed into an alkoxy
group and a C2 alkoxy group in an aldehyde group allowing
the carbon chain rearrangement. Thus, the large electronic
rearrangement in this step has been investigated by calculating
the charge density difference of the TSepim with respect to the
[C6H11O6]− anion and the protonated POM. This difference is
shown in Fig. 4a illustrating the charge depletion from the
lone pair of the isolated glucose and accumulation at the Mo
center. This result indicates that Mo reducibility58,59 is the controlling factor in the reaction. However, the direct calculation
of single atom reducibility on the surface of MoO3 is elusive
and thus alternatives are needed.
Iglesia et al. have proposed the H-atom addition energy
(HAE) as the descriptor for C–H activation for methanol oxidation on Keggin clusters60,61 (other geometric and electronic
parameters are discussed in descriptors58,59 in Table SI11†).
We used this parameter for glucose epimerization. HAE is calculated as the energy to add a hydrogen atom to an oxygen on
the catalyst:
HAE ¼ EcatþH• À Ecat

ð3Þ

This descriptor combines both the reducibility of the
metal center, by the addition of an electron, and the basicity
of the oxygen atom, as a proton is retained on a neighboring
oxygen. To minimize the influence of the different basicities
of Mo-neighboring O atoms several O atoms were sampled,
always ensuring that electron localization occurs at the
desired center. The catalysts in the analysis were HnXMo12O40
(X: S, P, As, Si, Ge), H6P2Mo18O62, α-MoO3(010) and
α-MoO3−x(010). The results shown in Fig. 4b and c indicate
that both the reactant ( product) adsorption and the activation energy for the 1,2 C-shift elementary step depend on
the HAE.
In addition, for POMs the reducibility can be extracted from
the previous calculations as follows:
Red ¼ HAE À ðEcatþHþ À Ecat Þ

ð4Þ

where the second term eliminates the contribution from the
basicity. As is shown in Fig. SI12† adsorption and activation
energies correlate with the reducibility for the six studied
POMs, showing how a higher reducibility results in a lower
energy barrier. Thus, the redox properties of the catalyst are
ultimately responsible for catalytic behavior, and reducibility
is the single descriptor that correctly defines adsorption and
activation energies for this process. The electronic descriptor
has two advantages: (i) it maps the geometry of the Mo–O distance (Fig. SI13†) and more importantly (ii) it is very easy to

This journal is © The Royal Society of Chemistry 2017

Fig. 4 (a) Structure of TSepim for H3PMo12O40 (left) and TSepim charge
density difference (right). The charge density difference was calculated
with respect to the anion [C6H11O6]− and cation (H4PMo12O40)+ at the
TSepim geometry; same color code as Fig. 1. Blue color means depletion of
charge density and yellow color means accumulation. (b) Binding energy,
BE, of dissociated glucose and mannose as a function of HAE. Eads(G) =
−2.72 HAE − 10.08 eV, R2 = 0.96, MAE = 0.03 eV. Eads(M) = −2.65 HAE −
9.74 eV, R2 = 0.93, MAE = 0.03 eV. Adsorption on defective MoO3−x(010)
surface is not included in the fittings. (c) Activation energy, Ea, of TSepim as
a function of HAE. Ea(G → M) = −0.79HAE − 2.06 eV, R2 = 0.88, MAE =
0.02 eV. Ea(M → G) = −0.81HAE − 2.13 eV, R2 = 0.81, MAE = 0.03 eV.

measure experimentally as cyclic voltammetry could easily
determine the reducibility.62
A microkinetic model was built in order to obtain the
analytical equation for the reaction rate. The energies for the
adsorption and 1,2 C-shifts are obtained from the linear fittings
in Fig. 4b and c. From the reaction profile (Fig. 2) the epimeriza-

Green Chem., 2017, 19, 5932–5939 | 5937

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

Paper

Green Chemistry

Fig. 5 Volcano plot of the normalized reaction rate as a function of the
HAE. Conditions: T = 100 °C, [G] = 0.90 M, [M] = 0.10 M. Red points mark
the calculated rate with eqn (5) for the catalyst used in the analysis.
Catalysts in the shadow area present drawbacks for epimerization reaction.

tion step is the rate-determining step and using the quasi-equilibrium approximation the reaction rate can be expressed as:
½M Š
1
r ¼ k2 KG ½GŠð1 À
Þ
Kglobal ½GŠ 1 þ KG ½GŠ þ K1 ½MŠ
M

ð5Þ

where Ki are the adsorption constants for glucose and
mannose, i = G, M; k2 is the coefficient for 1,2 C-shift (R2)
starting from glucose, Kglobal is the constant for the epimerization reaction, and [M] and [G] are the corresponding concentrations of glucose and mannose. Dissociative adsorption
and 1,2 C-shift energy barrier can be written in terms of the
HAE. At 100 °C, for [G]i = 1 M and a conversion of 10% the
reaction rate shows a volcano shape with the maximum
placed in HAE = −3.72 eV, with TOF = 10.1 s−1. In Fig. 5 the
normalized reaction rate is plotted as a function of the HAE.
The calculated rate for the defective surface is displaced from
the rate fitting, as its adsorption energy does not adjust to
the linear fitting. Pristine and defective MoO3 surfaces are
placed in a shadow zone, indicating that they present drawbacks
for epimerization reaction; low stability in aqueous solution30–32
or selectivity. Therefore, H2SMo12O40 is the best catalyst for epimerization, and overperforming the activity of H3PMo12O40.
Calculated apparent activation energy using the reaction rate
obtained from eqn (5), for H3PMo12O40 is 1.04 eV, in good agreement with experimental value: 0.99 ± 0.04 eV.19 The apparent
activation energy calculated for H2SMo12O40 is 1.00 eV,
Fig. SI14.† The plot shows a narrow zone of reducibility where
the Mo-based catalysts are active for epimerization. POMs are
placed in the upper end of this zone and, therefore, a more
active catalyst could be developed.

Conclusions
We have identified the mechanism behind the epimerization,
the Bilik reaction, in the glucose/mannose pair on Mo-con-

5938 | Green Chem., 2017, 19, 5932–5939

taining catalysts. The mechanism encompasses the adsorption of the aldose, the 1,2 C-shift and the recombination for
desorption of the product. The same mechanism is valid for
molecular and surface Mo oxides and no significant differences are found for the heteroatom or the curvature of the
surface. The descriptor for the reaction corresponds to the
reducibility of the Mo atoms on the surface/molecular catalyst that plays a crucial role in the 1,2 C-shift. When the reaction is written in terms of the HAE-descriptor (H addition
energy) it is possible to optimize the composition thus indicating the most active system. The molecular catalyst also
presents an excellent behavior in terms of the stability due to
the electrostatic interaction between the heteroatom anion
and the MoOx shell. The bulk oxide shows a low cohesive
energy. However, electrostatic interactions between the
charged heteroatom unit and the Mo12O36 layer are larger
which indicates the preeminent stability of the molecular
catalyst. Finally, a more active POM, H2SMo12O40, is
proposed.

Conflicts of interest
There are no conflicts to declare.

Acknowledgements
The authors acknowledge the Spanish Ministerio de Economía
y Competitividad (MINECO) for financial support (CTQ 201568770-R, Severo Ochoa Excellence Accreditation 2014–2018
SEV-2013-0319, and Severo Ochoa predoctoral grant SVP-2014068237). In addition, we thank BSC-RES for providing computational resources.

References
1 Y.-C. Park, E. J. Oh, J.-H. Jo, Y.-S. Jin and J.-H. Seo, Curr.
Opin. Biotechnol., 2016, 37, 105–113.
2 I. Delidovich and R. Palkovits, ChemSusChem, 2016, 9, 547–
561.
3 M. E. Tanner, Acc. Chem. Res., 2002, 35, 237–246.
4 J. Samuel and M. E. Tanner, Nat. Prod. Rep., 2002, 19, 261–
277.
5 S.-C. Lu and S.-C. Lin, Enzyme Microb. Technol., 2012, 50,
65–70.
6 S. J. Angyal, Carbohydr. Res., 1997, 300, 279–281.
7 T. Tanase, F. Shimizu, S. Yano and S. Yoshikawa, J. Chem.
Soc., Chem. Commun., 1986, 1001–1003.
8 T. Tanase, F. Shimizu, M. Kuse, S. Yano, M. Hidai and
S. Yoshikawa, Inorg. Chem., 1988, 27, 4085–4094.
9 A. M. Peres and E. A. Macedo, Ind. Eng. Chem. Res., 1997,
36, 2816–2820.
10 H. Nguyen, V. Nikolakis and D. G. Vlachos, ACS Catal.,
2016, 6, 1497–1504.

This journal is © The Royal Society of Chemistry 2017

View Article Online

Published on 16 November 2017. Downloaded by Universitat Rovira I Vigili on 4/12/2019 2:42:40 PM.

Green Chemistry

11 R. Bermejo-Deval, R. Gounder and M. E. Davis, ACS Catal.,
2012, 2, 2705–2713.
12 R. Bermejo-Deval, M. Orazov, R. Gounder, S.-J. Hwang and
M. E. Davis, ACS Catal., 2014, 4, 2288–2297.
13 W. R. Gunther, Y. Wang, Y. Ji, V. K. Michaelis, S. T. Hunt,
R. G. Griffin and Y. Roman-Leshkov, Nat. Commun., 2012,
3, 2122.
14 W. R. Gunther, Q. Duong and Y. Román-Leshkov, J. Mol.
Catal. A: Chem., 2013, 379, 294–302.
15 T. Prasomsri, T. Nimmanwudipong and Y. Roman-Leshkov,
Energy Environ. Sci., 2013, 6, 1732–1738.
16 T. Prasomsri, M. Shetty, K. Murugappan and Y. RomanLeshkov, Energy Environ. Sci., 2014, 7, 2660–2669.
17 M. Rellán-Piñeiro and N. López, ChemSusChem, 2015, 8,
2231–2239.
18 V. Zacharopoulou, E. S. Vasiliadou and A. A. Lemonidou,
Green Chem., 2015, 17, 903–912.
19 F. Ju, D. VanderVelde and E. Nikolla, ACS Catal., 2014, 4,
1358–1364.
20 T. L. Lohr, A. R. Mouat, N. M. Schweitzer, P. C. Stair,
M. Delferro and T. J. Marks, Energy Environ. Sci., 2017, 10,
1558–1562.
21 T. Choksi and J. Greeley, ACS Catal., 2016, 6, 7260–7277.
22 V. Bilik, Chem. Zvesti, 1972, 26, 183–186.
23 L. Petruš, M. Petrušová and Z. Hricovíniová, in Glycoscience:
Epimerisation, Isomerisation and Rearrangement Reactions of
Carbohydrates, ed. A. E. Stütz, Springer Berlin Heidelberg,
Berlin, Heidelberg, 2001, pp. 15–41.
24 M. L. Hayes, N. J. Pennings, A. S. Serianni and R. Barker,
J. Am. Chem. Soc., 1982, 104, 6764–6769.
25 J. E. Godfrey and J. M. Waters, Cryst. Struct. Commun., 1975,
4, 5–8.
26 G. E. Taylor and J. M. Waters, Tetrahedron Lett., 1981, 22,
1277–1278.
27 M. Matulova and V. Bilik, Chem. Pap., 1990, 44, 257–265.
28 B. K. Chethana, D. Lee and S. H. Mushrif, J. Mol. Catal. A:
Chem., 2015, 410, 66–73.
29 A. Cybulski, B. F. M. Kuster and G. B. Marin, J. Mol. Catal.,
1991, 68, 87–103.
30 G. M. Lari, O. G. Gröninger, Q. Li, C. Mondelli, N. López
and J. Pérez-Ramírez, ChemSusChem, 2016, 9, 3407–3418.
31 M. Orazov and M. E. Davis, Proc. Natl. Acad. Sci. U. S. A.,
2015, 112, 11777–11782.
32 A. Takagaki, S. Furusato, R. Kikuchi and S. T. Oyama,
ChemSusChem, 2015, 8, 3769–3772.
33 M. T. Pope and A. Müller, Angew. Chem., Int. Ed. Engl.,
1991, 30, 34–48.
34 J. F. Keggin, Proc. R. Soc. London, Ser. A, 1934, 144, 75–100.
35 G. Kresse and J. Furthmuller, Comput. Mater. Sci., 1996, 6,
15–50.
36 G. Kresse and J. Furthmueller, Phys. Rev. B: Condens. Matter
Mater. Phys., 1996, 54, 11169–11186.
37 J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett.,
1996, 77, 3865–3868.

This journal is © The Royal Society of Chemistry 2017

Paper

38 S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys
and A. P. Sutton, Phys. Rev. B: Condens. Matter Mater. Phys.,
1998, 57, 1505–1509.
39 S. Grimme, J. Comput. Chem., 2006, 27, 1787–1799.
40 G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter
Mater. Phys., 1999, 59, 1758–1775.
41 G. Makov and M. C. Payne, Phys. Rev. B: Condens. Matter
Mater. Phys., 1995, 51, 4014–4022.
42 G. Mills, H. Jonsson and G. K. Schenter, Surf. Sci., 1995,
324, 305–337.
43 G. Henkelman, B. P. Uberuaga and H. Jónsson, J. Chem.
Phys., 2000, 113, 9901–9904.
44 G. Henkelman and H. Jónsson, J. Chem. Phys., 1999, 111,
7010–7022.
45 A. Heyden, A. T. Bell and F. J. Keil, J. Chem. Phys., 2005,
123, 224101.
46 M. Garcia-Ratés and N. López, J. Chem. Theory Comput.,
2016, 12, 1331–1341.
47 M. Garcia-Ratés, R. García-Muelas and N. López, J. Phys.
Chem. C, 2017, 121, 13803–13809.
48 M. Álvarez-Moreno, C. de Graaf, N. López, F. Maseras,
J. M. Poblet and C. Bo, J. Chem. Inf. Model., 2015, 55, 95–
103.
49 https://doi.org/10.19061/iochem-bd-1-44.
50 S. S. Sunu, E. Prabhu, V. Jayaraman, K. I. Gnanasekar,
T. K. Seshagiri and T. Gnanasekaran, Sens. Actuators, B,
2004, 101, 161–174.
51 A. Borgschulte, O. Sambalova, R. Delmelle, S. Jenatsch,
R. Hany and F. Nüesch, Sci. Rep., 2017, 7,
40761.
52 K. Inzani, M. Nematollahi, F. Vullum-Bruer, T. Grande,
T. W. Reenaas and S. M. Selbach, Phys. Chem. Chem. Phys.,
2017, 19, 9232–9245.
53 I. V. Kozhevnikov, Chem. Rev., 1998, 98, 171–198.
54 J. Macht, M. J. Janik, M. Neurock and E. Iglesia, J. Am.
Chem. Soc., 2008, 130, 10369–10379.
55 I.-M. Mbomekallé, X. López, J. M. Poblet, F. Sécheresse,
B. Keita and L. Nadjo, Inorg. Chem., 2010, 49, 7001–
7006.
56 K. Maeda, H. Katano, T. Osakai, S. Himeno and A. Saito,
J. Electroanal. Chem., 1995, 389, 167–173.
57 M. Sadakane and E. Steckhan, Chem. Rev., 1998, 98, 219–
238.
58 M. Capdevila-Cortada and N. López, ACS Catal., 2015, 5,
6473–6480.
59 M. Capdevila-Cortada, G. Vilé, D. Teschner, J. PérezRamírez and N. López, Appl. Catal., B, 2016, 197,
299–312.
60 P. Deshlahra and E. Iglesia, J. Phys. Chem. C, 2014, 118,
26115–26129.
61 P. Deshlahra, R. T. Carr, S.-H. Chai and E. Iglesia, ACS
Catal., 2015, 5, 666–682.
62 Notice that Mo5+ centers are outliers in the fit for
adsorption via electrostatic contributions.

Green Chem., 2017, 19, 5932–5939 | 5939