This document is the Accepted Manuscript version of a Published Work that appeared in final form in Green Chem., copyright ©
This document is the Accepted Manuscript version of a Published Work that appeared in final form in Green Chem.,
The Royal Society of Chemistry 2017 after peer review and technical editing by the publisher. To access the final edited and
copyright © The see: http://dx.doi.org/10.1039/c6gc00894a. review and technical editing by the publisher. To access the
published work Royal Society of Chemistry 2016 after peer
final edited and published work see: http://dx.doi.org/10.1039/c6gc00894a.

Green Chemistry

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

PAPER

Cite this: Green Chem., 2016, 18,
4682

View Journal | View Issue

Glycerol oxidehydration to pyruvaldehyde over
silver-based catalysts for improved lactic acid
production†
G. M. Lari,a R. García-Muelas,b C. Mondelli,*a N. Lópezb and J. Pérez-Ramírez*a
Sustainable technologies for the valorisation of the burgeoning amounts of glycerol (GLY) obtained as
waste in the production of biodiesel are increasingly sought after. Its conversion into lactic acid (LA) is
appealing due to the versatility of this platform chemical and its high added value. Here, we introduce
Ag-based catalysts for the oxidehydration of GLY to pyruvaldehyde (PAl) and demonstrate the superiority
of this compound in comparison to dihydroxyacetone (DHA) as the intermediate of an alternative
two-step GLY-to-LA process. Evaluation of various metals and carriers identified Ag/Al2O3 as the best performer for PAl production. This was rationalised based on the optimal redox potential of the metal and the
high concentration of Lewis-acid sites and the limited Brønsted acidity of the support. At 623 K and O2/GLY =
0.5, a PAl yield of 80% was attained, which remained stable for 24 h. Characterisation of the used catalyst indicated that the surface of the silver nanoparticles was partially oxidised upon reaction. Density functional theory
(DFT) modelling revealed that the oxidation of acetol obtained from GLY after the initial dehydration step is

Received 30th March 2016,
Accepted 23rd May 2016

kinetically and thermodynamically favoured on a partially oxidised silver surface (AgOx/Ag) compared to metallic (Ag) or fully oxidic (Ag2O) ones. Finally, we show that PAl can be isomerised into LA and methyl lactate

DOI: 10.1039/c6gc00894a

over Sn-containing zeolites with the same rates as DHA but at a 40 K lower temperature (343 vs. 383 K). This

www.rsc.org/greenchem

not only allows for energy savings but also for a remarkably increased catalyst stability.

Introduction
Lactic acid (LA) has been recognised as a primary biobased
platform molecule, in view of its multiple applications in the
healthcare and food industries and its use in the production
of acrylates and biodegradable and biocompatible polymers.1
Especially due to the increasing prominence of polylactic acid,
which is replacing polyethylene terephthalate (PET) in a
growing number of uses,2 the annual demand for LA is
expected to increase to 6 × 105 ton by 2020.3 The current man-

a
Institute for Chemical and Bioengineering, Department of Chemistry and Applied
Biosciences, ETH Zurich, Vladimir-Prelog-Weg 1, CH-8093 Zurich, Switzerland.
E-mail: cecilia.mondelli@chem.ethz.ch, jpr@chem.ethz.ch
b
Institute of Chemical Research of Catalonia, ICIQ, The Barcelona Institute of
Science and Technology, Av. Països Catalans 16, ES-43007 Tarragona, Spain
† Electronic supplementary information (ESI) available: Density of states and calculated core-level shifts (Ag 3d) for the three silver surfaces modelled, characterisation data of the supports, the Sn-containing zeolites and an Ag-based catalyst,
tabulated initial conversion and selectivity data for the glycerol oxidehydration
and acetol dehydrogenation reactions, selectivity to pyruvaldehyde as a function
of the concentration of Brønsted-acid sites in the best Ag-based catalyst, energy
profiles for the regeneration of Ag2O and AgOx/Ag surfaces and representation of
the additionally considered Ag2O/Ag surface and energy profiles for the reaction
over this surface and for its regeneration. Structures from the DFT calculations
are freely available at http://dx.doi.org/10.19061/iochem-bd-1-8. See DOI:
10.1039/c6gc00894a

4682 | Green Chem., 2016, 18, 4682–4692

ufacture route, based on the fermentation of sugars, seems
unable to sustain such a large market, due to its low productivity and the stoichiometric amount of gypsum generated
as waste.3 The application of electrodialysis membranes to split
the salts avoiding the addition of sulphuric acid or the engineering of yeasts able to withstand solutions with lower pH have
been proposed to address the latter issue,4 but the low sugar
concentration of the fermentation broth remains a limiting
factor of the biocatalytic technology. For this reason, alternative
pathways enabling a large-scale, economically and environmentally-friendly production of LA are demanded. One of the most
promising strategies in this direction comprises the conversion
of glycerol (GLY) through oxidation to dihydroxyacetone (DHA)
followed by isomerisation.5 The triol is obtained as a byproduct in the preparation of biodiesel in growing amounts
and its very low price justifies its purification to serve as feedstock for the preparation of added-value chemicals.6 The conversion of glycerol into lactic acid has been conducted either
in one pot over Au/CeO2, Au–Pt/TiO2 and Pt/Sn-MFI catalysts7
or in two separate steps, using an enzyme or an iron-containing zeolite in the first reaction8 and Lewis acids for the
second.9 Zeolites containing tin atoms in framework positions
are superior catalysts for DHA isomerisation,10 which involves
the dehydration of DHA to pyruvaldehyde (PAl), followed by
the addition of a solvent (i.e., water or alcohol) molecule and a

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Green Chemistry

Paper

Lewis-acid sites favours the formation of PAl by oxidationdehydration (oxidehydration), but there is limited information
on the reaction pathway (Scheme 1b) and the exact nature of
the two functionalities.
Herein, we evaluate different metals and carriers used in
oxidation and dehydration reactions, respectively, identifying
silver supported on alumina as an efficient system for the gasphase production of PAl from GLY. Its superiority is explained
based on the optimal redox potential of the active phase and
the Lewis acidity of the oxide. Additionally, we assess the
impact of the reaction conditions and study the stability of the
best-performing catalyst. The dehydrogenation mechanism
and the chemical state of the surface are unravelled by combining DFT simulations and in-depth catalyst characterisation.
Finally, the benefits of using PAl as the intermediate in the
novel GLY-to-LA process are shown by confronting the longterm stability of Sn-containing zeolites in the conversion of
PAl and DHA in water and methanol.

Experimental
Catalyst preparation

Scheme 1 (a) Steps involved in the liquid-phase isomerisation of DHA
to LA in water or methanol and (b) reaction network envisaged for the
gas-phase conversion of GLY to PAl. The dashed arrows identify competitive oxidation and dehydration pathways.

1,2-hydride shift (Scheme 1a).11 Kinetic studies demonstrated
that the overall rate is determined by the dehydration step,11
and that the direct conversion of PAl is up to one order of magnitude faster than that of DHA under the same conditions.12
Based on these observations, an alternative process with PAl as
the intermediate would be advantageous to diminish the mass
of catalyst or the temperature applied to the isomerisation step.
Most importantly, milder conditions are expected to minimise
one of the biggest drawbacks of the stannosilicates, i.e., their
fast deactivation under continuous-flow operation.12 In this
respect, it has been shown that the hydrothermal stability of
zeolites increases at lower temperatures13 and the highly oxygenated reactant and intermediates are less prone to condensation
reactions leading to fouling.14
The selective conversion of GLY to PAl has been scarcely
investigated. A patent reported the activity of transition
metals, comprising vanadium, molybdenum, copper, silver
and tin, supported onto an aluminosilicate.15 Additionally, PAl
has been observed as a byproduct in the liquid-phase conversion of GLY under oxidising conditions16 and in its gas-phase
oxidative upgrading to DHA or acrylic acid (AAc), respectively.8b,17 It has been noted that the coexistence of redox and

This journal is © The Royal Society of Chemistry 2016

Supported metal catalysts. Supported catalysts containing
5 wt% of V, Ni, Fe, Sn, Cu, Ru, Ag, Pd or Pt were prepared by
dry impregnation of γ-alumina (Alfa Aesar, 42 576, 99.997%,
labelled as Al2O3-a) using NH4VNO3 (Fluka, >99%), NiCl2·6H2O
(Alfa Aesar, 99.95%), Fe(NO3)3·9H2O (Sigma-Aldrich, >98%),
SnSO4 (ABCR, 95%), Cu(NO3)2·3H2O (Sigma-Aldrich,
98–103%), RuCl3·xH2O (ABCR, 99.9%), AgNO3 (ABCR, 99.9%),
Pd(NO3)2·xH2O (ABCR, 99.9%), H2PtCl6 (Sigma-Aldrich, 8 wt%
water solution) and HAuCl4·3H2O (ABCR, 99.99%) as the metal
precursors. An aliquot of the salt (corresponding to 0.05 g of
the metal) was dissolved in deionised water (0.50 cm3) and the
solution was added to 0.95 g of the dried support. The wet
powder was mixed for 5 min and freeze-dried at 3 × 10−4 bar
for 16 h. Additional catalysts containing 5 wt% Ag were
obtained following the same procedure using zeolites with
MFI framework and a nominal Si/Al ratio of 40 (Zeolyst,
CBV8014, denoted as Z40) or 940 (Tosoh Corp., HSZ-8090H0A,
denoted as Z940), other γ-aluminas (Sigma-Aldrich, acidic,
basic and neutral, Brokmann grade I, labelled as Al2O3-b,
Al2O3-c and Al2O3-d, respectively), silica (Evonik, Sipernat 120),
amorphous silica-alumina (Sigma-Aldrich, grade 135) and
silicon carbide (Sigma-Aldrich, particle size <100 nm) as
carriers. Finally, a large batch (ca. 50 g) of 5 wt% Ag/Al2O3-a
was prepared by spray deposition18 using a Büchi Mini
Spray Dryer B-290. Silver nitrate (3.95 g) was dissolved in
deionised water (500 cm3) under magnetic stirring at room temperature, followed by the addition of the support (47.5 g). The
resulting suspension was pumped at 10 cm3 min−1 into the twofluid nozzle (1.4 mm diameter), using an air flow of 0.5 m3 h−1
to form the spray. The inlet temperature was set at 493 K,
the aspiration rate at 35 m3 h−1 and the outlet temperature
at 353 K. The dried particles were separated using a cyclone.

Green Chem., 2016, 18, 4682–4692 | 4683

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Paper

Tin zeolites. Z940 was used as received for post-synthetic
incorporation of tin.19 The zeolite (3.3 g) was added to an
aqueous solution of NaOH (0.3 M, 100 cm3) and SnSO4 (0.04 M)
and the mixture was hold at 338 K for 30 min in an
Easymax™ 102 reactor system (Mettler Toledo). Thereafter, the
treatment was quenched using 2 dm3 of deionised water and
the suspended material was filtered and washed until neutral
pH of the filtrate was achieved. The solid obtained was ion
exchanged in aqueous NH4NO3 (0.1 M, 6 h, 298 K, 100 cm3 per
gram of dried zeolite, 3 times), dried (338 K, 16 h) and calcined. This material was labelled as Sn-MFI. Another catalyst
coded as [Sn]BEA was prepared by hydrothermal synthesis following a reported procedure.20 Briefly, tetraethyl orthosilicate
(Sigma-Aldrich, 99.9%, 30 g) and tetraethyl ammonium
hydroxide (Sigma-Aldrich, 35 wt%, 33 g) were mixed for
90 min. A solution of SnCl4·5H2O in water (0.45 g in 2.75 cm3)
was added and the stirring was continued until complete evaporation of the ethanol formed upon hydrolysis. Then, HF (SigmaAldrich, 48 wt% in water, 3.2 g) was added to the clear solution
under stirring, resulting in the formation of a thick gel. Thereafter,
commercial zeolite beta (Tosoh Corporation, HSZ-980HOA, bulk
Si/Al = 220, 0.36 g) suspended in water (0.36 g in 2.75 cm3) was
added to act as a nucleation seed and the gel was transferred into
a 100 cm3 Teflon-lined autoclave and heated at 413 K for 11 days.
The [Sn]BEA solid obtained was subjected to the same filtration,
calcination and ion exchange procedures as described above.
Catalyst characterisation
The metal content in the samples was determined by X-ray
fluorescence (XRF) spectroscopy using an EDAX Orbis MicroXRF analyser equipped with a Rh source operated at a voltage
of 35 kV and a current of 500 μA. Powder X-ray diffraction
(XRD) was performed using a PANalytical X’Pert PRO-MPD
diffractometer with Ni-filtered Cu Kα radiation (λ =
0.1541 nm), acquiring data in the 5–70° 2θ range with a step
size of 0.05° and a counting time of 8 s per step. N2 sorption
was conducted at 77 K using a Micromeritics Tristar analyser.
Prior to the measurements, the samples were evacuated at
573 K and 10−4 bar for 3 h. Fourier transform infrared (FTIR)
spectroscopy of adsorbed pyridine (Sigma-Aldrich, >99%) was
carried out using a Bruker IFS66 spectrometer equipped with a
liquid-N2 cooled mercury cadmium telluride (MCT) detector.
Sample wafers (ca. 1 cm2, 0.02 g) were evacuated (10−6 bar)
for 4 h at 693 K, prior to adsorption at room temperature.
Gaseous and weakly adsorbed molecules were subsequently
removed by evacuation at 373 K (for selected samples) and at
473 K for 30 min. Spectra were recorded in the range of
650–4000 cm−1 with a resolution of 4 cm−1 and co-addition of
32 scans. The concentration of Brønsted- and Lewis-acid sites
was calculated from the area of the pyridine bands at 1545 and
1454 cm−1, respectively, using their corresponding extinction
coefficients εBrønsted = 1.67 and εLewis = 2.94 cm μmol−1. Temperature-programmed reduction with hydrogen (H2-TPR) was
performed using a Micromeritics Autochem II 2920 unit connected to a MKS Cirrus 2 quadrupole mass spectrometer. The
samples (0.1 g) were pre-treated in He flow (20 cm3 min−1) at

4684 | Green Chem., 2016, 18, 4682–4692

Green Chemistry

423 K for 1 h, cooled to 308 K and finally heated up to 1273 K
(10 K min−1) while monitoring the consumption of H2 and the
formation of H2O. Quantification was achieved by integration
of the area of the reduction peaks, using the results obtained
in the H2-TPR of known amounts of CuO for calibration.
Thermogravimetric analysis (TGA) of the used catalysts was
conducted using a Mettler Toledo TGA/DSC1 instrument. The
samples (0.015 g) were heated in the range of 323–1023 K
(10 K min−1) under an air flow (40 cm3 min−1). The weight loss
between 523–1023 K was ascribed to the removal of carbonaceous deposits. X-ray photoelectron spectroscopy (XPS) was
carried out under UHV (10−8 mbar) using a VG Thermo
Escalab 220i-XL spectrometer equipped with a high-intensity
twin anode (Al Kα radiation). In the case of the used Ag/Al2O3-a
sample, the measurement was repeated after sputtering with
4 kV Ar+ ions rastered over a 2 × 2 mm area. The spectra were
charge-corrected with respect to the Al 2p signal at 74.0 eV. TEM
images were acquired using an FEI Talos instrument operated at
200 kV. Powdered samples were deposited on Mo-grids as such.
Catalyst testing
Gas-phase conversion of GLY to PAl. The oxidehydration of
GLY to PAl was studied at ambient pressure in a continuousflow fixed-bed reactor (Scheme 2a) comprising: (i) mass flow
controllers for feeding N2 (PanGas, 99.999%), H2 (PanGas,
99.999%) and O2 (PanGas, 99.995%), (ii) a syringe pump for
feeding an aqueous GLY (Sigma-Aldrich, 99%) solution, (iii) a
tubular quartz micro-reactor (i.d. = 6 mm) heated in an oven,
(iv) a liquid-gas separator immersed in an ice bath and (v) a
bag for the collection of gaseous samples. The supported
metal catalysts (0.1 g, particle size = 0.2–0.4 mm) were loaded
into the reactor, which was then filled with 2-mm glass
spheres in order to facilitate the evaporation of the liquid feed.
The system was heated at the reaction temperature (623 K) in
an N2 flow (100 cm3 min−1) and allowed to equilibrate for
30 min. Thereafter, the gas composition was changed to 20 vol%
H2/N2 (100 cm3 min−1) for 60 min to reduce the metal.
Finally, O2/N2 (typically 1.5 vol%, 100 cm3 min−1) was admitted
along with a 10 wt% aqueous GLY solution (0.1 cm3 min−1).
Liquid samples were periodically collected at the reactor outlet
and analysed by high-performance liquid chromatography
(HPLC) using a Merk LaChrome system equipped with a
HPX-87H column kept at 308 K and refraction index and
UV-Vis detectors. A 0.005 M aqueous H2SO4 solution flowing at
0.600 cm3 min−1 was used as the eluent. Calibration curves
were measured in the 0.1–10 wt% range using GLY (SigmaAldrich, 99%), Ac (Alfa Aesar, 95%), PAl (Acros Organics, 40 wt%
aqueous solution), PAc (ABCR, 98%), Acr (Fluka, >95%), AAc
(Acros Organics, 99.5%), DHA (Merck, >98%), GAl (SigmaAldrich, >90%) and GAc (Acros Organics, >99.7%). CO and
CO2 in the gas samples were analysed by gas chromatography
using a HP5890 unit equipped with a gas-sampling valve, a
Restek-Shin Carbon ST80/100 column and a thermal conductivity detector. Helium was used as the carrier gas. The concentrations of the carbon oxides were determined from their
absolute peak areas using calibration curves measured in the

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Green Chemistry

Paper

Scheme 2 Reactor setups used for (a) the continuous gas-phase conversion of GLY into PAl and (b) the continuous liquid-phase conversion of DHA
and PAl into LA.

0–10 wt% range. The conversion of GLY (XGLY) and selectivity
to the product i (Si) were calculated as:
X GLY ¼ 1 À ðnGLY;out =nGLY;in Þ
Si ¼ ni;out =ðnGLY;in À nGLY;out Þ
where n refers to the moles of GLY or compound i and in/out
to the reactor inlet/outlet. The carbon balance was obtained as
the ratio between the moles of carbon in the liquid and
gaseous products and the moles of carbon fed. The experimental error, determined by three repetitions of the same catalytic run, was within 5%. No GLY conversion was observed in
blank tests conducted with only quartz particles and the glass
beads loaded into the reactor.
Liquid-phase conversion of DHA and PAl to LA or ML. Batch
catalytic experiments for the conversion of DHA or PAl into LA
were carried out under autogenous pressure in 15 cm3 thickwalled glass vials (Ace, pressure tubes, front seal) dipped in an
oil bath heated at 333–423 K. The vials were loaded with ca.
10 cm3 of a 0.33 M aqueous DHA solution (Merk, 98%) or PAl
(Acros Organics, 40 wt% in water). Then, Sn-MFI (ca. 0.01 g)
was added to achieve a substrate/tin molar ratio of 1000. The
mixture was allowed to react under vigorous stirring for
15 min. Thereafter, the reactions were quenched using an ice
bath and the catalysts were removed by means of a Chromafil
Xtra 0.25 μm syringe filter. The continuous conversion of DHA
and PAl into LA or ML was studied in a homemade continuous-flow reactor setup (Scheme 2b) comprising (i) an HPLC
pump (Gilson-306), (ii) a stainless-steel tubular reactor with a
pre-column (Swagelok SS-T4-S-035, i.d. = 4.6 mm), both heated
in a tubular oven, (iii) a backpressure regulator (Swagelok,
LH2981001) and (iv) an automatic sampler. The reactor was
loaded with the Sn-containing zeolite catalysts (0.10 g) diluted
with quartz (1 g, particle size = 0.25–0.36 mm) and a liquid
feed (0.2 cm3 min−1) was admitted which contained the substrate (0.33 M) dissolved in water or methanol. Thereafter, the
system was pressurised to 20 bar and heated to 383 K or 343 K

This journal is © The Royal Society of Chemistry 2016

in the case of the conversion of DHA or PAl, respectively.
Samples were periodically collected at the outlet of the reactor
and analysed by HPLC using the same instrument and
method as described for the analysis of the condensate
obtained from the conversion of GLY. An additional calibration
curve was measured in the 0.1–10 wt% range for L-lactic acid
(ABCR, 98%). For the reactions performed in methanol, the
amount of ML produced was determined using a gas chromatograph (HP 6890) equipped with an HP-5 capillary column
and a flame ionisation detector. Quantification was accomplished by integration of its peak using 2,2,4-trimethylpentane
(Fluka, 99.5%) as the internal standard. Conversion and
selectivity data were calculated as mentioned above.
Computational details
Simulations were conducted using the Vienna Ab-initio Simulation Package (VASP)21 and the RPBE functional.22 Core electrons were described by projector augmented wave (PAW)
pseudopotentials,23 while valence electrons by a kinetic energy
cutoff of 450 eV. Although its true structure has been debated
for long,24 experimental observations herein gathered indicate
that silver forms a partially oxidised ultrathin film in O2-rich
environments that is commensurate with the p(4 × 4) Ag(111)
unit cell. Therefore, three different models were used to
describe the silver catalyst to account for poor, medium and
rich oxygen conditions, respectively (Fig. 1). Metallic silver,
which dominates at low O2/GLY ratios, was represented by

Fig. 1 Representation of the three (111) surfaces used as the models for
(a) metallic Ag, (b) the partially oxidised metal AgOx/Ag and (c) the bulk
oxide Ag2O.

Green Chem., 2016, 18, 4682–4692 | 4685

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Paper

Green Chemistry

4 layers of a clean 4 × 4 Ag(111) surface. For intermediate
O2/GLY ratios (∼0.5), the p(4 × 4) model reported elsewhere24b
was used. This can be generated from metallic Ag(111) by
removing 4 metal atoms and adding 6 oxygen atoms to the
topmost layer. The lattice parameter of the bulk metal was
4.22 Å, in good agreement with the experimental value of
4.09 Å.25 Thus, both metallic and oxygen-covered silver surfaces have a length of 11.94 Å. Finally, for high O2/GLY ratios,
the p(2 × 2) Ag2O(111) surface was used. Here, the on-surface
distance between the cells is 13.92 Å, very similar to the one of
Ag(111). For completeness, a model comprising a silver oxide
layer epitaxially-grown on metallic silver, Ag2O/Ag(111), was also
investigated. Details to this model are provided in the ESI.† For
all of the surfaces, the two topmost layers were relaxed while
keeping the two bottom ones fixed to the bulk distances. The
density of states for the clean surfaces is shown in Fig. S1.†
The Brillouin zone was sampled by a 3 × 3 × 1 Γ-centred
k-points mesh generated through the Monkhorst–Pack
method.26 To avoid spurious interactions between the periodic
images, a vacuum region between the slabs of 15 Å and dipole
correction along z were included.27 To find the transition
states, the climbing image nudged elastic band (CI-NEB)
method was employed.28 The optimisation thresholds of 10−5
and 10−3 eV were applied for electronic and ionic relaxations,
respectively. The assessment of the minima and transition
states was performed by diagonalising the numerical Hessian
matrix obtained by ±0.02 Å displacements. In the calculation
of frequencies, the surface atoms were kept frozen, except for
the oxygen atoms on AgOx/Ag or Ag2O that were part of the
reaction. All of the DFT energies reported along the manuscript include zero-point energies.

Results and discussion
Catalytic oxidehydration of GLY to PAl
Evaluation of alumina-supported metal catalysts. A series of
alumina-supported metal catalysts were prepared to assess the
role of the metal in the conversion of GLY into PAl. The metals
comprised V, Ni, Fe, Sn, Cu, Ru, Ag, Pd, Pt and Au and were
selected among noble and non-noble transition metals based

Table 1

on their reported catalytic activity in the dehydrogenation of
alcohols. They are characterised by different redox potentials,
spanning from −1.13 to 1.68 V, relatively to the semi-reactions
reported in Table 1.25 The metal content in the final materials
was similar and was comprised between 4.4 and 5.1 wt% in all
cases except for Ni and Sn. For these, the loading was limited
to 3.1 and 1.9 wt%, respectively, due to the poor water solubility of the precursors. The choice of the alumina carrier was
based on literature reports indicating that the reaction proceeds through the dehydration of GLY to Ac, followed by its
dehydrogenation, and that the first reaction is catalysed by
Lewis-acid sites.3 Indeed, testing of the bare support yielded a
GLY conversion of 53% with high Ac selectivity (85%,
Table S1†). The most prominent by-product was Acr. The
impregnation of the support with the metal precursors led to
only minor modification in the surface area and did not significantly alter its Lewis acidity. In contrast, its mild Brønsted
acidity (Table 1) generally decreased, likely due to the replacement of protons at ion-exchange positions with the metal
cations, except for the Cu-, Ru- and Pd-based catalysts. For
these, a slight increase was noticed, which could be associated
with the higher electronegativity of the metals and thus with
the stronger acidic nature of H atoms of hydroxyl groups.
Fig. 2a shows the catalytic data collected upon testing of
the catalysts in the gas-phase oxidehydration of GLY at 623 K.
With the exception of the Pt- and Au-based samples, all of the
materials attained a high GLY conversion level. This suggests
that a certain degree of oxidation of the metal is advantageous,
which cannot be achieved with the most noble metals. The
product distribution, evidencing a low Ac selectivity at high
PAl yields and vice versa, is in line with the two-step reaction
path proposed in Scheme 1b. Only minor amounts of DHA
(<1%), Gal (<1%) and GAc (<2%) were detected (Table S1†),
indicating that the direct dehydrogenation of the hydroxyl
groups of GLY is hindered under these conditions. Indeed, we
have recently reported the oxidation of GLY to DHA to happen
at much higher O2/GLY ratios over steamed iron-containing
silicalite featuring negligible acidity.8b The competitive dehydration reaction leading to Acr, which can be further converted into AAc under oxidising conditions, was relevant only
in the case of Sn-, Cu- and Ru-containing materials. This was

Characterisation data for the alumina-supported metal catalysts

Catalyst

Metal contenta (wt%)

SBET b (m2 g−1)

Vpore c (cm3 g−1)

CLewis d (µmol g−1)

CBrønsted d (µmol g−1)

Eh e (V)

Semi-reaction

V/Al2O3-a
Ni/Al2O3-a
Fe/Al2O3-a
Sn/Al2O3-a
Cu/Al2O3-a
Ru/Al2O3-a
Ag/Al2O3-a
Pd/Al2O3-a
Pt/Al2O3-a
Au/Al2O3-a

4.9
3.1
4.8
1.9
4.5
5.1
4.8
5.0
4.1
4.4

74
76
78
78
81
72
74
79
72
80

0.39
0.45
0.42
0.44
0.38
0.40
0.45
0.45
0.35
0.46

65
61
72
80
68
76
72
70
58
69

8
7
3
4
18
16
8
21
26
4

−1.13
−0.72
−0.44
−0.14
0.34
0.70
0.80
0.90
1.20
1.83

V2+ + 2e− → V
Ni2+ + 2e− → Ni
Fe2+ + 2e− → Fe
Sn2+ + 2e− → Sn
Cu2+ + 2e− → Cu
Ru4+ + 4e− → Ru
Ag+ + e− → Ag
Ru4+ + 4e− → Sn
Pd2+ + 2e− → Pd
Au+ + e− → Au

a

XRF spectroscopy. b BET method. c Volume adsorbed at p/p0 = 0.99. d FTIR spectroscopy of adsorbed pyridine, after desorption at 473 K. e Redox
potential of the semi-reaction reported in the column on the right.28

4686 | Green Chem., 2016, 18, 4682–4692

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Green Chemistry

Paper

explained by their comparatively high concentration of
Brønsted acid sites, which are well-known active centres for
the selective formation of this molecule from GLY. Pt- and Pdcontaining solids, which evidenced moderate acidity upon pyridine adsorption did not show this behaviour likely because of
their fully reduced status under the reaction conditions and
the consequent absence of hydroxyl groups on their surface.
Overall, the selectivity to the dehydration product (Ac) was
higher for those materials containing a more noble metal.
Assuming that GLY dehydration mostly takes place on the
support and the Ac dehydrogenation is catalysed by the metal,
this implies a lower activity of these metals in the dehydrogenation step. On the other hand, over-oxidation of PAl to PAc
was favoured on non-noble metals. Furthermore, up to 8% of
the GLY fed was converted into CO and CO2. Overall, the Agbased catalyst was superior to the other systems due to a good
balance between activity and selectivity to PAl. This is visualised
in Fig. 2b, which highlights a maximum for the PAl yield (80%)
over this material. Notably, this yield matches the state-of-theart data claimed in a patent over another Ag-based solid.15
Accordingly, this catalyst was chosen for further studies.
Support selection for silver catalysts. A series of silver catalysts was prepared using supports with distinct acid–base properties (Table S2†), to explore whether the catalyst activity
could be further enhanced by optimising the first step. The
carriers comprised 3 additional aluminas, leading to solids
featuring different Lewis acidity, 2 zeolites with MFI framework topology and a Si/Al ratio of ca. 40 or 940 as well as amorphous silica and silica-alumina, generating catalysts with very
diverse Brønsted and Lewis acidity (Table 2). The specific
surface area of the carriers was in line with that generally
reported for these materials and was not substantially altered
by the deposition of the silver phase, suggesting a good metal
dispersion. An increased Lewis-acidic character was shown to
be beneficial to raise the conversion of the substrate while
keeping a high selectivity (Fig. 2c). Indeed, for all the aluminasupported samples a linear conversion-vs.-Lewis acidity trend
was observed. The strength of the Lewis-acid sites appeared to
play no significant role, since it varied significantly among
materials displaying a similar selectivity (Table 2). Brønstedacid sites proved extremely detrimental for the selectivity of

Fig. 2 (a) GLY conversion and selectivities to Ac, PAl and PAc as a function of the metal reduction potential for alumina-supported catalysts
and (b) the corresponding PAl formation rate. (c) Reaction performance
vs. the Lewis-acid sites concentration for alumina-, Z940- and SiO2supported silver catalysts. Reaction conditions: T = 623 K, P = 1 bar,
CGLY = 10 wt%, LHSV = 6 h−1 and O2/GLY = 0.5.

Table 2

Characterisation data for the Ag-based catalysts

Support

Aga
(wt%)

SBET b
(m2 g−1)

Vpore c
(cm3 g−1)

CLewis d
(µmol g−1)

CBrønsted d
(µmol g−1)

Lewis sites
strengthe

Brønsted sites
strengthe

Ag/Al2O3-a
Ag/Al2O3-b
Ag/Al2O3-c
Ag/Al2O3-d
Ag/SiO2
Ag/ASA
Ag/Z940
Ag/Z40

4.8
4.8
4.9
4.8
4.9
5.1
4.7
5.0

74
70
73
76
121
80
325
403

0.45
0.38
0.51
0.56
0.35
0.61
0.19
0.21

72
53
25
18
1
51
21
89

8
2
0
0
0
54
0
151

0.91
0.86
0.60
0.75
1.00
0.41
0.80
0.57

0.45
0.52
—
—
—
0.88
—
0.79

a
e

XRF spectroscopy. b BET method. c Volume adsorbed at p/p0 = 0.99. d FTIR spectroscopy of adsorbed pyridine, after desorption at 473 K.
Defined as the ratio between the sites probed by pyridine after desorption at 473 and at 373 K.

This journal is © The Royal Society of Chemistry 2016

Green Chem., 2016, 18, 4682–4692 | 4687

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Paper

the reaction (Fig. S2†). While promoting a high GLY conversion, the strongly Brønsted-acidic Z40 and ASA support limited
the formation of the product of interest since they favoured
the dehydration of GLY to Acr (Table S1†). Based on these
results, a bigger batch of 5 wt% Ag/Al2O3-a was prepared to
enable an in-depth characterisation and testing at variable
reaction conditions. Spray deposition was used as a synthesis
method instead of impregnation due to the higher reproducibility of the properties of the solids obtained by this technique.19 The catalyst showed the same catalytic performance
as that obtained by impregnation. The XRD pattern of the
solid in as-prepared form evidenced, after subtraction of the
diffractogram of the bare support, weak reflections specific to
Ag(I) oxide (Fig. 3a). After activation of this material by
reduction in diluted hydrogen, the diffraction lines of Ag2O
disappeared and two reflections appeared at 38 and 64 2θ,
which are attributed to metallic silver nanoparticles of about

Green Chemistry

13 nm diameter. H2-TPR analysis confirmed the stoichiometry
of the silver oxide in the freshly-synthesised sample (Fig. 3b).
After activation, negligible hydrogen consumption was
measured, indicating the complete reduction to the metallic
state. Transmission electron micrographs (Fig. 4a and b)
revealed that the Ag2O particles were very small (ca. 5 nm) and
that the metallic phase featured larger aggregates (ca. 12 nm).
The latter evidence is in agreement with the XRD data and
indicates that clustering happened upon reduction (Fig. 3a).
Finally, the nature of the metal surface was investigated by XPS
(Fig. 3c). A signal centred at 368.2 eV was observed for the asprepared sample, which is attributed to the 3d transition of
Ag(I) species.29 After reduction of the catalyst, the position of
the 3d signal shifted to 367.2 eV, as expected for metallic
silver. This difference is also in good agreement with the data
calculated by DFT (Table S3†), which predicted that the signal
of oxidic silver would appear at a 0.7 eV higher binding energy
than that of metallic silver.
Impact of the reaction conditions for Ag/Al2O3-a. The dependence of the reaction performance on the temperature and on
the O2-GLY ratio was studied in the 573–673 K and 0–1.5
ranges, respectively, using the 5 wt% Ag/Al2O3-a catalyst prepared by spray deposition. As expected, the conversion
increased at higher temperatures (Fig. 5a). The selectivity to
PAl was preserved until 623 K. Above this temperature, the
over-oxidation to PAc increased. This effect is in line with the
fact that the oxidation of the aldehyde also happens homogeneously following a non-catalytic radical pathway, the rate of
which strongly depends on temperature.30 In the absence of
oxygen, Ac was the most prominent product (Fig. 5b) and the
Acr yield was comparable to that obtained in a blank run with
the bare support. This indicates that the rate of dehydrogenation is negligible if O2 is not supplied. As detailed below, it
additionally hints the inactivity of the metal in its reduced state
or the necessity of reducing the surface coverage of the hydrogen
abstracted from the substrate using an oxidant. The PAl yield
steeply increased upon addition of O2 reaching ca. 80% at
O2/GLY = 0.5, i.e., the stoichiometric ratio. Above this value, it
diminished due to the more pronounced over-oxidation.
The stability of the catalyst was tested at the optimal temperature and O2/GLY ratio in a 24 h run (Fig. 5c). The PAl yield
decreased by only ca. 10% in this test highlighting the robustness of the material. This minor loss of performance was
ascribed to the deposition of carbonaceous species (4.8 wt% C
on the used catalyst) and a slight sintering of the silver particles from ca. 12 to 15 nm (Fig. 4c). It should be noted that a
catalyst featuring a distinct Brønsted acidity would undergo
rapid deactivation by fouling due to the promoted formation
of dehydration and condensation products.
DFT simulations on Ac dehydrogenation to PAl

Fig. 3 Characterisation of the as-prepared, activated and used Ag/
Al2O3-a catalyst by (a) XRD, (b) H2-TPR and (c) XPS.

4688 | Green Chem., 2016, 18, 4682–4692

Since the catalyst testing evidenced a strong influence of the
addition of oxygen on the oxidehydration of GLY, the oxidative
conversion of Ac to PAl was investigated in greater detail by
DFT. The reaction occurs as described hereon. Firstly, Ac
adsorbs on the surface (1), which abstracts an H atom cleaving

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Green Chemistry

Paper

Fig. 4 TEM micrographs for the (a) as-prepared, (b) activated and (c) used Ag/Al2O3-a catalyst. The particle size distributions of the Ag-based phase
are reported in the insets.

the more labile O–H bond (2).31 The latter is performed by an
Ag atom on a metallic surface or by an O atom in the case of
AgOx/Ag and Ag2O. Then, the vicinal C–H bond is broken (3),
forming adsorbed PAl. The reaction proceeds with the desorption of PAl (4). When oxygen is present on the surface, catalyst
regeneration occurs through the formation of water by combination of two hydroxyl groups (5). The subsequent desorption
of water creates an oxygen vacancy (6), that is healed by the
adsorption of an oxygen molecule (7). Albeit comprising multiple steps, the latter is represented as an elementary step for
the sake of clarity.
CH3 COCH2 OH ðgÞ þ * ! CH3 COCH2 OH*
CH3 COCH2 OH* þ * ! CH3 COCH2 O* þ H*

ð3Þ

CH3 COCHO* ! CH3 COCHO ðgÞ þ *

ð4Þ

OH* þ H* ! H2 O* þ *

ð5Þ

H2 O* ! H2 O ðgÞ þ *

ð6Þ

1
O2 ðgÞ þ * ! O*
2

This journal is © The Royal Society of Chemistry 2016

ð2Þ

CH3 COCH2 O* þ * ! CH3 COCHO* þ H*

Fig. 5 GLY conversion and product selectivity over Ag/Al2O3-a as a
function of (a) the temperature, (b) the O2/GLY ratio and (c) the time on
stream. Reaction conditions: T = 623 K, P = 1 bar, CGLY = 10 wt%, LHSV
= 6 h−1 and O2/GLY = 0.5.

ð1Þ

ð7Þ

The energy profiles corresponding to the steps (1)–(4) are
shown in Fig. 6, while those relative to steps (5)–(7) are
depicted in Fig. S3.† On pure Ag, the adsorption of Ac is
endothermic and the barrier for step (2) is higher than on the
other two model surfaces, indicating that it shall proceed with
very low rates. The second dehydrogenation (3) occurs simultaneously with PAl desorption (4). On Ag2O, the adsorption of
Ac is more exothermic than on the other systems (1), indicating that the bulk oxide traps Ac more effectively. The first dehydrogenation (2) is less energy demanding than for pure Ag and
the second dehydrogenation (3) proceeds rapidly and is followed by the exothermic desorption of the desired product
PAl. However, water formation (5) is relatively difficult, thus
hindering the regeneration of the catalyst. On AgOx/Ag, the low
reaction and activation energies of step (2) suggest that
CH3COCH2O* is in equilibrium with adsorbed Ac. Reaction (3)
has a low activation energy and is very exothermic. Since step

Green Chem., 2016, 18, 4682–4692 | 4689

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Paper

Fig. 6

Green Chemistry

Energy profiles for the dehydrogenation of Ac to PAl and the corresponding structures calculated over the AgOx/Ag surface.

(7) is less energy demanding, the surface will regenerate faster
than in the case of bulk Ag2O. For the oxidic surface and the
bulk oxide, PAl desorption is always exothermic, thus preventing the on-surface over-oxidation to PAc. It should be noted
that step (2) is an acid–base reaction that could also be catalysed
by the support. Spillover of the oxygenate fragment to the metal
phase would enable the occurrence of the subsequent steps
(3)–(7). However, control experiments with a SiC-supported Ag
catalyst excluded the participation of the alumina carrier in the
abstraction of the first H atom (Table S4†). The results obtained
over the Ag2O/Ag surface (Fig. S4 and S5†) are close to those
described for the bulk Ag2O system and are discussed in the ESI.†
Based on the analysis of the energy diagrams, the high PAl
yield found at O2/GLY = 0.5 is explained by its quick elimination from the surface and the easy catalyst regeneration,
while the low PAl yields measured for low and high O2/GLY
ratios are due to a very slow C–H stripping and a hindered
regeneration of the metal sites, which remain hydroxylated,
respectively. These findings hence rationalise the volcanoshape observed in Fig. 5b.
Characterisation data for the Ag/Al2O3-a catalyst indicate
the surface oxidation of the metal nanoparticles after use
under the optimised conditions. This corresponds to the most
active surface (AgOx/Ag) analysed for the reaction. Indeed, XRD
(Fig. 3a) evidenced a weakening and sharpening of the diffraction peaks attributed to metallic silver, pointing to the formation of bigger particles and of a secondary Ag-based phase,
which remained undetected possibly due to the its limited
amount and/or high dispersion. H2-TPR analysis (Fig. 3b)
showed a reduction peak centred at ca. 523 K, indicating that
ca. 13% of the metal was oxidised to Ag(I) oxide. The location
of this phase was investigated through XPS by measuring
spectra after removing a progressively thicker layer of material
from the surface by sputtering with Ar+ (Fig. 3c). The 3d signal
of silver appeared at 368.0 eV in the spectrum of the used
untreated sample and stepwise shifted to 367.2 eV in those

4690 | Green Chem., 2016, 18, 4682–4692

after irradiation with the ion beam. This finding implies the
formation of an oxidised silver shell around a metallic core
upon reaction. Since the oxygen-to-silver ratio at the surface
was lower than that of Ag2O, as indicated by the slightly
shifted XPS peak, silver might not be fully oxidised to Ag(I) in
this outer layer. The partial oxidation of the Ag phase upon
reaction was confirmed by H2-TPR analysis of another sample,
i.e., Ag/Al2O3-d (15%, Fig. S6†). Additionally, characterisation
of the Ag/Al2O3-a catalyst used at O2/GLY = 1.5 by the same
technique revealed a 2.5-times greater extent of oxidation of
silver (32%, Fig. 3b), highlighting the sensitivity of the active
phase to the oxygen feed content. The change in oxidation
state of the surface silver atoms is analogous to that observed
for Ru-based catalysts applied in the transfer hydrogenation of
furfural to 2-methylfurane, where the active phase has been
reported to include both oxidic and metallic species. The
similarity of the two systems is attributed to the similar Eh
values of the two metals (Fig. 2b).32 It should also be noticed
that the partial oxygen pressures and temperatures employed
in the reaction herein studied are similar to those commonly
applied in ethylene epoxidation.24,33 This transformation com-

Fig. 7 Rate of LA formation from DHA and PAl over Sn-MFI in water.
Reaction conditions: CDHA, PAl = 0.33 M and DHA, PAl/Sn = 1000.

This journal is © The Royal Society of Chemistry 2016

View Article Online

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Green Chemistry

Paper

Fig. 8 (a) Percentage of the initial rate of LA formation over Sn-MFI in water and (b) of ML formation over Sn-MFI and [Sn]BEA in methanol. Reaction
conditions: TDHA = 383 K, TPAl = 343 K, P = 20 bar, CDHA, PAl = 0.33 M, LHSV = 120 h−1.

prises the most common application of silver in heterogeneous
catalysis and the metal has been found to undergo surface oxidation upon this reaction as well.
Continuous liquid-phase conversion of PAl to LA
The rate of formation of LA and methyl lactate (ML) from DHA
over Sn-containing Lewis-acid zeolites has been reported to be
limited by the dehydration of the ketone to PAl (Scheme 1a).12
Therefore, the use of the latter as a substrate would enable
the same productivity at reduced operation temperatures,
thus decreasing the energy demand of the process. Additionally, it is expected that temperature-dependent deactivation
processes, including fouling and hydrothermal modifications of the zeolite, would be hindered, lowering the catalyst
consumption and the operating costs. To confirm the advantages of the use of PAl as an alternative substrate to DHA for
the production of LA and methyl lactate (ML), the state-of-theart [Sn]BEA zeolite as well as a Sn-MFI material prepared by
alkaline-assisted stannation were evaluated in the two reactions. The latter catalyst was selected in view of the rapidity
and scalability of its preparation which contrast the limited
industrial amenability of the synthesis of the former
material.19 The crystallinity, textural properties and the coordination of the tin sites, probed by diffuse reflectance UV-Vis
and acetonitrile-FTIR spectroscopies (Table S5 and Fig. S7†),
of the stannosilicates were in good agreement with previous
data published for these materials.12 At first, Sn-MFI was
tested in the conversion of DHA and PAl in batch experiments
at different temperatures in order to choose conditions suitable to compare initial turnover frequencies for the two substrates. The results (Fig. 7) indicate that the rate of LA
formation from DHA at 383 K, a typical temperature for this
reaction, could be matched using PAl at only 343 K. Hence,
continuous experiments were performed at these two temperatures. When Sn-MFI was operated in water (Fig. 8a) in a 24 h
test, the loss of LA yield was limited to only 23% in the case of
PAl, as opposed to 75% when the reaction was performed
using DHA. An even more pronounced difference in the deactivation pattern was noticed when methanol was used as the
solvent (Fig. 8b). We have previously observed that, for the

This journal is © The Royal Society of Chemistry 2016

alcohol-based conversion of DHA, the 10-membered ring MFI
zeolites suffer from much faster coking with respect to BEA
zeolites featuring the larger 12-membered rings, which
renders the catalyst almost inactive after 12 h on stream.
Remarkably, at the lower temperature employed for the conversion of PAl, more than 40% of the initial activity of Sn-MFI was
retained after use for 24 h. An appreciable gain in stability was
observed also in the case of [Sn]BEA.

Conclusions
In this study, we uncovered silver supported on alumina as a
highly efficient catalyst for the oxidehydration of waste glycerol
to pyruvaldehyde and the advantages of using this compound
instead of dihydroxyacetone for the sustainable manufacture
of lactic acid and methyl lactate. Silver was chosen amongst
other transition metals thanks to its optimal reduction potential, which enabled high activity and selectivity. In contrast,
metals with a more noble character were less active and those
with a higher tendency to oxidise favoured the further oxidation of the product. A maximal pyruvaldehyde yield of ca.
65% was attained, which remained stable over 24 h on stream.
Characterisation of the used catalyst evidenced the formation
of an oxide layer at the surface of the metal nanoparticles,
which was demonstrated by DFT to be the active surface for
the dehydrogenation of the acetol molecule obtained by dehydration of glycerol. In fact, the reaction was found to be
thermodynamically limited on metallic silver and kinetically
hindered on a pure oxide phase, due to the higher activation
energy of the proton abstraction and the more difficult regeneration of the oxidic sites. Finally, we confronted the use of
pyruvaldehyde and dihydroxyacetone for the preparation of
lactic acid and methyl lactate over tin zeolites. Remarkably,
with the former substrate it was possible to achieve comparable rates at a 40 K lower reaction temperature, greatly reducing the deactivation of the stannosilicates. Accordingly, a
glycerol to lactic acid process featuring pyruvaldehyde as intermediate is expected to bring substantial environmental and
economic benefits compared to the best process currently envi-

Green Chem., 2016, 18, 4682–4692 | 4691

View Article Online

Paper

Green Chemistry

saged for the valorisation of this feedstock, which is based on
the exploitation of dihydroxyacetone.
12

Published on 23 May 2016. Downloaded by Universitat Rovira I Vigili on 4/17/2019 5:03:31 PM.

Acknowledgements
This work was sponsored by the Swiss National Science Foundation (Project Number 200020-159760). The Scientific Center
for Optical and Electron Microscopy at ETH Zurich, ScopeM, is
acknowledged for the use of their facilities. Dr S. Mitchell is
thanked for the TEM measurements and Dr R. Hauert for the
XPS analyses. Prof. N. López and R. García-Muelas are grateful
to MINECO (CTQ2015-68770-R) for financial support and
BSC-RES for providing generous computer resources.

13

14
15
16

17

References

18

1 (a) T. Werpy and G. R. Petersen, Top Value Added Chemicals
from Biomass: Vol. I - Results of Screening for Potential Candidates from Sugars and Synthesis Gas, US DOE, 2004;
(b) J. J. Bozell and G. R. Petersen, Green Chem., 2010, 12,
539–554; (c) R. E. Drumright, P. R. Gruber and
D. E. Henton, Adv. Mater., 2000, 12, 1841–1846;
(d) S. Cohen, T. Yoshioka, M. Lucarelli, L. Hwang and
R. Langer, Pharm. Res., 1991, 8, 713–720; (e) J. C. SerranoRuiz, A. Pineda, A. M. Balu, R. Luque, J. M. Campelo,
A. A. Romero and J. M. Ramos-Fernández, Catal. Today,
2012, 195, 162–168.
2 J. Lunt, Polym. Degrad. Stab., 1998, 59, 145–152.
3 M. Dusselier, P. Van Wouwe, A. Dewaele, E. Makshina and
B. F. Sels, Energy Environ. Sci., 2013, 6, 1415–1442.
4 (a) R. Datta and M. Henry, J. Chem. Technol. Biotechnol.,
2006, 81, 1119–1129; (b) Y.-J. Wee, J.-N. Kim and H.-W. Ryu,
Food Technol. Biotechnol., 2006, 44, 163–172; (c) M. Singhvi,
G. Gurjar, V. Gupta, D. Gokhale, et al., RSC Adv., 2015, 5,
2024.
5 D. T. Johnson and K. A. Taconi, Environ. Prog., 2007, 26,
338–348.
6 R. Ciriminna, C. Della Pina, M. Rossi and M. Pagliaro,
Eur. J. Lipid Sci. Technol., 2014, 116, 1432–1439.
7 (a) P. Lakshmanan, P. P. Upare, N.-T. Le, Y. K. Hwang,
D. W. Hwang, U.-H. Lee, H. R. Kim and J.-S. Chang, Appl.
Catal., A, 2013, 468, 260–268; (b) Y. Shen, S. Zhang, H. Li,
Y. Ren and H. Liu, Chem. – Eur. J., 2010, 16, 7368–7371;
(c) H. J. Cho, C.-C. Chang and W. Fan, Green Chem., 2014,
16, 3428–3433.
8 (a) D. Hekmat, R. Bauer and J. Fricke, Bioprocess Biosyst.
Eng., 2003, 26, 109–116; (b) G. M. Lari, C. Mondelli and
J. Pérez-Ramírez, ACS Catal., 2015, 5, 1453–1461.
9 P. Y. Dapsens, C. Mondelli and J. Pérez-Ramírez, Chem. Soc.
Rev., 2015, 44, 7025–7043.
10 C. M. Osmundsen, M. S. Holm, S. Dahl and E. Taarning,
Proc. R. Soc. London, Ser. A, 2012, 468, 2000–2016.
11 F. de Clippel, M. Dusselier, R. Van Rompaey, P. Vanelderen,
J. Dijkmans, E. Makshina, L. Giebeler, S. Oswald,

19

4692 | Green Chem., 2016, 18, 4682–4692

20
21

22
23

24

25
26
27
28

29

30
31
32

33

G. V. Baron, J. F. M. Denayer, P. P. Pescarmona, P. A. Jacobs
and B. F. Sels, J. Am. Chem. Soc., 2012, 134, 10089–10101.
G. M. Lari, P. Y. Dapsens, D. Scholz, S. Mitchell,
C. Mondelli and J. Pérez-Ramírez, Green Chem., 2016, 18,
1249–1260.
R. M. Ravenelle, F. Schuessler, A. D’Amico, N. Danilina,
J. A. van Bokhoven, J. A. Lercher, C. W. Jones and
C. Sievers, J. Phys. Chem. C, 2010, 114, 19582–19595.
P. T. Williams and S. Besler, Renewable Energy, 1996, 7,
233–250.
H. Baltes and E. I. Leupold, US Pat 4302609, 1981.
R. K. Pazhavelikkakath Purushothaman, J. van Haveren,
I. Melián-Cabrera, E. R. H. van Eck and H. J. Heeres, ChemSusChem, 2014, 7, 1140–1147.
F. Auneau, S. Noël, G. Aubert, M. Besson, L. Djakovitch and
C. Pinel, Catal. Commun., 2011, 16, 144–149.
A. H. Øygarden, J. Pérez-Ramírez, D. Waller, K. Schöffel
and D. Brackenbury, WO 2004/110622, 2004.
P. Y. Dapsens, C. Mondelli, J. Jagielski, R. Hauert and
J. Pérez-Ramírez, Catal. Sci. Technol., 2014, 4, 2302–2311.
S. Valencia and A. Corma, US Pat 5968473, 1999.
(a) G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996,
6, 15–50; (b) G. Kresse and J. Furthmüller, Phys. Rev. B:
Condens. Matter, 1996, 54, 11169–11186.
B. Hammer, L. B. Hansen and J. K. Nørskov, Phys. Rev. B:
Condens. Matter, 1999, 59, 7413–7421.
(a) P. E. Blöchl, Phys. Rev. B: Condens. Matter, 1994, 50,
17953; (b) G. Kresse and J. Joubert, Phys. Rev. B: Condens.
Matter, 1999, 59, 1758.
(a) M. Schmid, A. Reicho, A. Stierle, I. Costina, J. Klikovits,
P. Kostelnik, O. Dubay, G. Kresse, J. Gustafson, E. Lundgren,
J. N. Andersen, H. Dosch and P. Varga, Phys. Rev. Lett., 2006,
96, 146102; (b) J. Schnadt, A. Michaelides, J. Knudsen,
R. T. Vang, K. Reuter, E. Lægsgaard, M. Scheffler and
F. Besenbacher, Phys. Rev. Lett., 2006, 96, 146101.
W. M. Haynes, CRC Handbook of Chemistry and Physics,
CRC Press, Boca Raton, 96th edn, 2015.
H. J. Monkhorst and J. D. Pack, Phys. Rev. B: Solid State,
1976, 13, 5188–5192.
G. Makov and M. C. Payne, Phys. Rev. B: Condens. Matter,
1995, 51, 4014–4022.
(a) G. Henkelman and H. Jónsson, J. Chem. Phys., 2000,
113, 9978–9985; (b) G. Henkelman, B. P. Uberuaga and
H. Jónsson, J. Chem. Phys., 2000, 113, 9901–9904.
(a) A. M. Ferraria, A. P. Carapeto and A. M. Botelho do
Rego, Vacuum, 2012, 86, 1988–1991; (b) S. W. Gaarenstroom
and N. Winograd, J. Chem. Phys., 1977, 67, 3500–3506.
M. Besson and P. Gallezot, Catal. Today, 2000, 57, 127–141.
Y. Kwon, S. C. S. Lai, P. Rodriguez and M. T. M. Koper, J.
Am. Chem. Soc., 2011, 133, 6914–6917.
M. J. Gilkey, P. Panagiotopoulou, A. V. Mironenko,
G. R. Jenness, D. G. Vlachos and B. Xu, ACS Catal., 2015, 5,
3988–3994.
J. P. Dever, K. F. George, W. C. Hoffman and H. Soo, Ethylene Oxide, in Kirk-Othmer Encyclopedia of Chemical Technology, John Wiley & Sons, Inc., Hoboken, 2004, pp. 632–673.

This journal is © The Royal Society of Chemistry 2016