Liquid-phase glycerol hydrogenolysis by formic acid over Ni-Cu/Al 2O3 catalysts

Article abstract of DOI:10.1016/j.jcat.2012.03.004

Two series of Ni-Cu/Al₂O₃ catalysts with different Cu/Ni ratio and different total metal contents were prepared and tested in the glycerol hydrogenolysis to 1,2-propanediol under N₂ pressure and using formic acid as hydrogen donor molecule. The results from the deep characterization of the catalysts by N₂-physisorption, transmission electron microscopy (TEM), temperature-programed reduction (TPR), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and temperature-programed desorption of ammonia (TPD-NH₃), together with the results from the activity tests allowed deeper understanding of the role played by Ni, Cu, and acid catalytic sites. The kinetic study performed with the optimized catalyst revealed that the OH groups of glycerol and of the target product, 1,2-propanediol, compete for adsorption on the acid sites of the Al₂O₃ support. 90% glycerol conversion and 82% 1,2-propanediol selectivity after 24 h reaction time were achieved when operating at 493 K.

Full text of DOI:10.1016/j.jcat.2012.03.004

Journal of Catalysis 290 (2012) 79–89
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Liquid-phase glycerol hydrogenolysis by formic acid over Ni–Cu/Al O catalysts
2
3
a,
⇑
a
a
b
b
I. Gandarias , J. Requies , P.L. Arias , U. Armbruster , A. Martin
a
School of Engineering (UPV/EHU), Alameda Urquijo s/n, 48013 Bilbao, Spain
Leibniz-Institut für Katalyse e. V. an der Universität Rostock, Albert-Einstein-Str. 29a, D-18059 Rostock, Germany
b
a r t i c l e i n f o
a b s t r a c t
Article history:
2
Two series of Ni–Cu/Al O3 catalysts with different Cu/Ni ratio and different total metal contents were
Received 29 December 2011
Revised 2 March 2012
Accepted 3 March 2012
Available online 5 April 2012
prepared and tested in the glycerol hydrogenolysis to 1,2-propanediol under N
mic acid as hydrogen donor molecule. The results from the deep characterization of the catalysts by N
physisorption, transmission electron microscopy (TEM), temperature-programed reduction (TPR), X-ray
diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and temperature-programed desorption of
2
pressure and using for-
2
-
3
ammonia (TPD-NH ), together with the results from the activity tests allowed deeper understanding of
the role played by Ni, Cu, and acid catalytic sites. The kinetic study performed with the optimized catalyst
revealed that the OH groups of glycerol and of the target product, 1,2-propanediol, compete for adsorp-
Keywords:
Biomass
Hydrogenolysis
tion on the acid sites of the Al
after 24 h reaction time were achieved when operating at 493 K.
Ó 2012 Elsevier Inc. All rights reserved.
2 3
O support. 90% glycerol conversion and 82% 1,2-propanediol selectivity
1
,2-Propanediol
Transfer hydrogenation
Formic acid
1
. Introduction
studied 2-propanol and ethanol as solvents and hydrogen donor
molecules for the process under inert atmosphere [22] or low
hydrogen pressure [23]. It was recently reported that the reaction
mechanism is different regarding the origin of the active hydrogen
It is widely assumed that the dependence of human well-being
on fossil fuels should be reduced, mainly due to geopolitical, natural
resource scarcity, and environmental factors. Biomass appears as
the only renewable source for liquid fuels and plastics [1]. Main bio-
mass conversion processes that are being currently developed, like
bio-oils from the pyrolysis or high-pressure liquefaction of biomass
is molecular hydrogen or a hydrogen donor molecule. Under H
2
pressure, glycerol is first dehydrated to acetol, which subsequently
undergoes a hydrogenation process to give 1,2-PDO [24] (Scheme
1A). On the other hand, when the hydrogen atoms are produced
from 2-propanol dehydrogenation, glycerol is directly converted
to 1,2-PDO through intermediate alkoxide formation (Scheme
1B). Moreover, it was also observed that the hydrogen donor and
glycerol compete for the same active sites [25]. In a previous work,
not shown here, a semi-continuous process was developed in
which the hydrogen donor molecule (2-propanol, methanol, or for-
mic acid) was continuously fed into the autoclave containing the
[
2,3], sorbitol valorization to polyols [4], or glycerol hydrogenolysis
to propanediols (PDO) [5–17], require oxygen removal reactions. In
these hydrogenation/hydrogenolysis reactions, high hydrogen pres-
sures are required to reach acceptable conversions and selectivities.
Molecular hydrogen is easily ignited in contact with air and shows
high diffusivity; therefore, it presents considerable hazards when
working at high pressures. In addition, most of the nowadays avail-
able hydrogen gas is produced from fossil fuels by energy intensive
processes [18]. In situ generation of the hydrogen can be a promising
alternative to avoid the inherent drawbacks of working with
molecular hydrogen. Nevertheless, the so far published results on
liquid-phase glycerol hydrogenolysis generating the hydrogen by
simultaneous glycerol reforming [19–21] are still far to the best
2 3
glycerol aqueous solution and Ni–Cu/Al O catalyst. It was ob-
served that there was an optimum feeding rate for each donor that
maximized 1,2-PDO production, and that the bests results in terms
of glycerol conversion and 1,2-PDO selectivity were obtained using
formic acid.
An additional advantage of using formic acid as hydrogen donor
is that it can be obtained from renewable resources through the
Biofine Process. This process transforms non-food biomass feed-
stock by acid-catalyzed hydrolysis to give levulinic acid and formic
acid, together with some furfural and a char residue [26]. Formic
2
results reported under H pressure [12,15] (see Table 1).
Another interesting option that allows working under inert
atmosphere is to use hydrogen donor molecules. However, cata-
lytic transfer hydrogenation reactions have not been extensively
reported in the glycerol hydrogenolysis process. Musolino et al.
acid is also considered a promising H
through CO hydrogenation [27,28]. Formic acid can be catalyti-
cally dehydrogenated to give hydrogen and CO . A parallel dehy-
dration reaction gives water and CO, which can be further
converted to CO and hydrogen through water–gas shift reaction
2
storage as it can be obtained
2
2
⇑
Corresponding author. Address: Escuela Técnica Superior de Ingeniería,
Alameda Urquijo s/n, P.C., 48013 Bilbao, Spain. Fax: +34 946 014 179.
E-mail address: inaki_gandarias@ehu.es (I. Gandarias).
2
0
021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2012.03.004

8
0
I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
Table 1
Selected examples of glycerol hydrogenolysis under H
2
atmosphere and by in situ generation of hydrogen through simultaneous glycerol reforming.
Catalyst
Atmos.
P (MPa)
Temp (K)
Time (h)
Conv. (%)
Selectivity to 1,2-PDO (%)
Refs.
2
5
0
.7 Pt/NaY
Pt/Al
.6 Ir/C + NaOH (1 M)
Inert
Inert
Inert
4.2
4.1
3.0
2.0
3.0
503
493
453
473
453
15
6
8
24
20
85.4
50.1
76.0
65.3
80
64.0
47.2
9.2
89.6
98.2
[17]
[18]
[19]
[12]
[15]
2 3 2 3
O + 5 Ru/Al O
Copper-chromite
2 8.6
Cu0.4/Mg5.6Al O
H
H
2
2
Scheme 1. (A) Reaction scheme of glycerol hydrogenolysis and degradation under H
2
pressure [24]. (B) Direct glycerol hydrogenolysis to 1,2-PDO using 2-PO as a hydrogen
donor molecule [25].
[
29]. The selectivity to CO
2
and hydrogen depends on the operating
samples were digested in a microwave oven (Anton Paar Multi-
wave, Perkin Elmer) with a mixture of aqua regia and hydrofluoric
acid.
condition and the catalyst used.
This paper studies the role of supported Ni, Cu, and acid cata-
lytic sites in the liquid-phase glycerol hydrogenolysis to 1,2-PDO
when formic acid is used as the hydrogen donor molecule.
Surface area, pore volume, and pore size distributions were
2
determined by N physisorption at 77 K on a Micromeritics ASAP
2
010 instrument. Prior to the analysis, all samples were degassed
at 473 K at 0.1 mbar for 4 h. The surface area was calculated using
the Brunauer, Emmett, and Teller (BET) method, while pore size
distributions were calculated using the Barrett, Joyner, and Halen-
da (BJH) method applied to the desorption leg of the isotherms.
The temperature-programed reduction of the samples was
studied on a Micromeritics Autochem AC2920 apparatus. The
reduction of previously calcined catalyst samples was carried out
under a gaseous mixture of 5 vol.% of hydrogen in argon, at a heat-
ing rate of 10 K/min from room temperature to 1173 K and a gas
flow rate of 50 mL/min. TPR was finished by an isothermal period
of 1 h at 1173 K. The hydrogen consumption was monitored by a
calibrated thermal conductivity detector.
2
. Experimental
2.1. Catalysts preparation
Ni–Cu/Al
Aluminium isopropoxide (Aldrich) was dissolved in deionized
water (9 mL of H O per gram of aluminium isopropoxide) by vigor-
ous stirring of the solution at 313 K. The pH was measured and
kept between 3.8 and 4.2 adding the required amounts of HNO
1.0 M). Simultaneously, nickel (II) nitrate hexahydrate (Aldrich)
2 3
O catalysts were prepared by the sol–gel method.
2
3
(
and/or copper (II) nitrate hemi pentahydrate (Alfa Aesar) were dis-
solved in ethanol. The precursor solution was slowly added to the
aluminium isopropoxide solution. The mixture was stirred for
The XPS measurements were performed with an ESCALAB
2
20iXL (ThermoScientific) using monochromatic AlKa radiation.
3
0 min at 313 K and then introduced into an ultrasonic apparatus
for another 30 min. The mixture was then rested for 48 h at
28 K and subsequently for another 12 h at 375 K. The product ob-
tained was crushed (granule size between 300 and 500 m) and
calcined from room temperature to 723 K at a heating rate of
K/min. This final temperature was maintained for 4 h. Previous
The binding energies were referred to the C1s peak at 284.8 eV.
The spectra were fitted with Gaussian–Lorentzian curves to deter-
mine the peak maxima and areas. The concentrations of the ele-
ments in the near-surface region were obtained after division by
the element-specific Scofield factors and the transmission function
of the spectrometer.
3
l
2
to the reaction test, the catalyst powder (0.9 g) was pre-reduced
in a tubular oven from room temperature to 923 K at a heating rate
of 10 K/min under a hydrogen flow of 75 mL/min. The final temper-
ature was maintained for 1 h. After cooling down to room temper-
ature, the reduced sample was transferred, avoiding the contact
with air, to a glovebox and stored there under Ar atmosphere be-
fore use.
X-ray powder diffraction (XRD) measurements were carried out
on a STADI P automated transmission diffractometer (STOE, Darms-
1
tadt) with CuK
a
radiation and Ge monochromator. The pattern
was scanned in the 2h range of 5–60° (step width 0.5°, 100 s per
step) and recorded with a STOE position sensitive detector. The
samples were prepared for flat plates. Phase analysis was carried
out with the Win Xpow software package, including the powder
diffraction file (PDF).
2.2. Catalyst characterization
For TEM characterization,
equipped with a LaB6 filament and a supertwin lens operating at
00 kV was used. Bright-field images were acquired using a CCD
a Phillips CM200 microscope
The catalysts were chemically analyzed using the ICP-OES spec-
2
trometer Optima 3000XL (Perkin Elmer). Before measurement,

I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
81
camera (TVIPS GmbH). Powders of the catalysts samples were dis-
persed in ethanol with ultrasound, and drops were deposited on
copper grids coated with amorphous carbon films.
The acidity of the calcined and reduced samples was deter-
mined by ammonia temperature-programed desorption (TPD)
measurements. The sample was pre-treated in a He stream at
by means of a pneumatic 6-port valve with a loop size of 100
and injected via a splitter into two separate analytical lines (Line
1: Poraplot Q, 25 m  0.53 mm  0.20 m, flame ionization detec-
tor (FID) for analysis of CO, CH and CO via a methanizer; line 2:
HP PLOT molesieve, 25 m  0.53 mm, thermal conductivity detec-
tor (TCD) for analysis of N , H , CH , and CO) using Ar as carrier
ll
l
2
4
2
2
4
6
5
73 K for 0.5 h and then in situ reduced using a gas mixture of
vol.% of H in Ar from 323 to 723 K at a heating rate of 10 K/
gas. The liquid phase was analyzed in another gas chromatograph
(Shimadzu 17A GC-FID with autosampler; Chrompack FFAP col-
2
min and a gas flow rate of 100 mL/min. Next, the sample was
cooled down to 373 K and saturated with ammonia using a gas
umn, 25 m  0.32 mm  0.3
lm, carrier He) using 1,4-butanediol
as internal standard. Glycerol conversion and selectivity values
were calculated on carbon basis. The carbon balance was cross-
checked by total organic carbon (TOC) analysis of the liquid sample
after each activity test using the model TOC-VCPN (Shimadzu).
1 mL of the liquid sample from the reactor was diluted in 100 mL
before the TOC analysis.
mixture of 5 vol.% NH
lowing catalyst equilibration in a helium flow, NH
3
in He (flow rate 50 mL/min) for 0.5 h. Fol-
was desorbed
3
using a linear heating rate of 10 K/min up to 723 K. The area under
the curve was integrated to determine the total acidity of the sam-
ple from its NH desorption profile.
3
Total carbon content in used catalysts was determined using a
Truespace CHNS analyzer (LECO instruments, Ltd.).
3
. Results and discussion
2.3. Activity test
3.1. Optimization of the Cu/Ni ratio
Fig. 1 shows a scheme of the activity test set-up utilized. Inside
2 3
In order to optimize the Cu/Ni ratio, six different Ni–Cu/Al O
the glovebox, under Ar atmosphere, the reduced catalyst was intro-
duced into a basket and then into the autoclave (Parr, 300 mL stain-
less steel). Next, the autoclave was covered with parafilm, taken out
catalysts were prepared by sol–gel method with a constant total
nominal metal content of 35 wt.% and different proportions of Ni
and Cu. These catalysts were deeply characterized and tested in
the glycerol hydrogenolysis reaction using formic acid as hydrogen
donor molecule.
from the glovebox to the activity plant and tightened under N
After purging and leakage check, the reactor temperature was set to
73 or 493 K, and the N pressure was increased to 15 bar. A 130 mL
2
flow.
4
2
of 4 wt.% glycerol aqueous solution were placed in a steel feed cyl-
inder and the pressure inside the feed cylinder was increased to
3
.1.1. Elemental analysis and N
Table 2 summarizes the theoretical and real metal contents to-
2
-physisorption
4
5 bar. Next, the stirring speed was set constant at 550 rpm. The
reaction start time (t ) was established when the line connecting
the feed cylinder and the reactor was opened. N was directed
gether with the main textural properties of the calcined supported
catalysts. Each catalyst is named with the rounded nominal con-
tent of each metal. As it can be observed, the real Ni and Cu con-
tents measured by ICP-OES were lower than the theoretical ones
in all the samples. This means that with the used sol–gel method,
it was not possible to fix the desired metal amounts in the cata-
lysts, which points at a systematic deviation in preparation proce-
dure. Measured amounts ranged from 27.7 to 30.1 wt.%. However,
it is interesting to observe that the theoretical and real Cu/Ni ratio
values were quite similar.
0
2
through the feed cylinder to the reactor till the pressure of the sys-
tem reached 45 bar, guaranteeing that all the feed went into the
reactor. When the operating pressure was reached, the line con-
necting the feed cylinder and the reactor was closed. From the
beginning to the end of the activity test, an aqueous solution of for-
mic acid (7.0 or 12.5 wt.%) was pumped into the autoclave at a con-
À1
À1
stant rate of 0.02 mL/min. These feed rates of 2:1 mmol g
h
of
catal
formic acid for the activity tests performed at 473 K and
2
Concerning the N physisorption results, all the samples pre-
À1
À1
3:6 mmol g
h
for the activity tests performed at 493 K were
catal
sented Type IV isotherms, with the typical hysteresis loops of mes-
oporous materials. Moreover, there is an interesting trend, as both
BET area and total pore volume increased with increasing Ni pro-
portion in the catalyst, slightly decreasing for the catalyst contain-
ing only Ni.
optimized in a previous work. During the reaction, increments in
the system pressure were observed due to the formation of CO
and H from the decomposition of the pumped formic acid. Never-
theless, final pressure never exceeded 57 bar.
2
2
After 16-h reaction time, the reactor was cooled down and the
gas phase was analyzed by directly connecting the reactor to the
gas chromatograph (Hewlett–Packard 5890). Samples were taken
3.1.2. Temperature-programed reduction
The active Ni and Cu sites in glycerol hydrogenolysis are re-
duced sites [25]. In order to set the temperature for the pre-reduc-
tion of the samples, the TPR analysis of 7Ni28Cu/Al and
8Ni7Cu/Al was carried out. These two catalysts were selected
2 3
O
2
2 3
O
as they were the ones, among the catalyst having both Ni and Cu,
with the highest and the lowest Cu/Ni ratio. The TPR spectra of the
investigated samples are shown in Fig. 2.
Calcined 7Ni28Cu/Al
for copper ( and b) at 503 and 549 K. The presence of more than
one reduction signal in TPR profiles of Ni–Cu/Al catalysts was
already observed and explained by different species [30–32]. Some
authors hypothesized the presence of CuAl besides pure CuO
30,31], which undergoes reduction at higher temperatures than
CuO [33]. Nevertheless, the formation of CuAl was not expected
2 3
O catalyst presented two reduction peaks
a
2 3
O
2 4
O
[
2 4
O
in the studied catalysts, as it is only formed at calcination temper-
atures above 973 K [30], and these catalysts were calcined at 723 K.
Moreover, in the XPS analysis of 7Ni28Cu/Al
2
O
3
and 28Ni8Cu/
in the
Fig. 1. Scheme of the activity test set-up.
Al (shown in Section 3.1.3), the presence of CuAl O
2
O
3
2 4

8
2
I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
Table 2
Characteristics of the catalysts with constant total nominal metal content.
Sample/
Al
Ni content wt.%
Theoretical (Real)
Cu content wt.%
Theoretical (Real)
Ni + Cu wt.%
Theoretical (Real)
Cu/Ni ratio
Theoretical (Real)
S
BET
2
Total pore volume
(cm /g)
Average pore
size (nm)
a
a
a
a
3
2
O
3
(m /g)
0Ni35Cu
7Ni28Cu
13Ni22Cu
20Ni15Cu
28Ni7Cu
35Ni0Cu
0.0
0.0)
35.0
(30.0)
35.0
(30.0)
–
–
103.5
0.074
0.167
0.207
0.247
0.282
0.213
4.53
3.83
3.84
3.66
3.54
3.33
(
(
6.6
5.1)
28.4
(23.8)
35.0
(28.9)
4.30
(4.67)
157.9
201.2
229.6
281.6
240.2
13.3
10.4)
21.7
(17.3)
35.0
(27.7)
1.63
(1.66)
(
(
(
(
20.3
17.3)
14.7
(12.8)
35.0
(30.1)
0.72
(0.74)
27.5
22.7)
7.5
(6.2)
35.0
(28.9)
0.27
(0.27)
35.0
27.9)
0.0
(0.0)
35.0
(27.9)
0.00
(0.00)
a
Chemical composition determined by ICP.
composition was computed from XPS peak intensities, and the re-
sults are presented in Table 3. These values are compared to the
bulk content of Ni and Cu as obtained from ICP results with Al
and O serving as stoichiometrical balance. In order to estimate
Peak α
Peak β
28Ni7Cu
Peak β
Peak α
2 3
the Al O content, it was assumed that all the Ni was in the form
of NiO and that all the Cu was in the form of CuO. As it can be ob-
served in Table 3, for both catalysts, the Ni and Cu surface mass
fractions are significantly lower than in the bulk. This seems to
indicate that Ni and Cu are not well dispersed on the surface of
the support, probably forming relatively big particles, which is
coherent with the high metal content of the catalysts. Those Cu
and Ni atoms far from the surface of these big particles are hard
to detect by XPS measurements. It is interesting to point out that
the Cu/Ni surface ratio for both catalysts is around two times smal-
ler than the Cu/Ni bulk ratio. This suggests that Ni is better dis-
persed than Cu, forming smaller particles, as it was confirmed by
the XRD analysis of the same calcined catalysts (see Section 3.1.4).
The surface C content detected in the catalysts (7–12%) can be
7Ni28Cu
3
00 400 500 600 700 800 900 1000 1100 1200
Temperature (K)
2 3 2 3
Fig. 2. TPR profiles of calcined 7Ni28Cu/Al O and 28Ni7Cu/Al O catalysts.
related to the adsorption of ambient hydrocarbons and CO
the high surface area Al support during calcination in air at
23 K. In fact, the 284.8 eV binding energy observed for the C 1s
electrons (see Table 4) is characteristic of the adventious C that
2
by
near-surface region was not detected. Also, in the XRD analysis of
2 4
calcined catalysts, diffraction reflections stemming from CuAl O
were not detected (see Section 3.1.4). Some other authors attrib-
uted the lower TPR signal to the reduction of highly dispersed
2 3
O
7
2
comes from CO and ambient hydrocarbons.
CuO species (peak
of bulk CuO (peak b) [32,34,35]. This second explanation fits better
with the obtained TPR profiles. In the case of 7Ni28Cu/Al cata-
a), and the higher TPR signal to the reduction
2 3
O
lyst, the peak corresponding to bulk CuO (peak b) presents higher
intensity as compared to the peak related to dispersed CuO (peak
Table 3
Bulk atomic composition obtained from ICP results and surface atomic composition
obtained from XPS results.
a
). The significant formation of bulk CuO is coherent with the high
Cu content of this catalyst. On the other side, in the case of
8Ni7Cu/Al catalyst, the peak at higher temperature corre-
sponding to bulk CuO (peak b) shows lower intensity than the peak
related to dispersed CuO (peak ). As the Cu content in this catalyst
Sample/Al
2
O
3
O (%)
Al (%)
C (%)
Ni (%)
Cu (%)
Cu/Ni
2
2 3
O
7Ni28Cu
a
Bulk
37.4
60.0
33.7
27.5
–
7.0
5.10
1.9
23.8
3.6
4.7
1.9
a
Surface
is comparatively smaller, the proportion of dispersed CuO is
enhanced.
Concerning the reduction of nickel, for both catalysts a broad
peak reduction profile in the temperature range 650–1050 K was
recorded. As expected, the intensity was higher for the catalyst
with higher Ni content. This broad peak suggests the presence of
2
8Ni7Cu
Bulk
a
37.6
52.7
33.5
23.8
–
12.3
22.7
9.5
6.2
1.6
0.27
0.16
Surface
a
Atomic bulk composition obtained from ICP results for Ni and Cu, including
stoichiometrical balance by O and Al.
a mixture of NiO
port [36].
x
species having different interaction with the sup-
Based on TPR results, the reduction temperature for catalyst
activation was set to 923 K.
Table 4
Binding energies (eV) and relative abundance (%) of core electrons of C 1s, Cu 2p and
Ni 2p.
Sample/Al
2
O
3
C 1s (%)
Cu 2p3/2 (%)
Ni 2p3/2 (%)
3
.1.3. X-ray photoelectron spectra
The same two calcined catalysts, 7Ni28Cu/Al O and 28Ni7Cu/
2 3
7
2
Ni28Cu
8Ni7Cu
284.8 (83)
284.7 (90)
933.5 (100)
934.0 (100)
855.2 (100)
854.8 (100)
2 3
Al O , analyzed by TPR, were analyzed by XPS. Atomic surface

I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
83
The Cu and Ni chemical species found on the catalysts surfaces,
and their proportions were also evaluated. As observed in Table 4,
in the near-surface region of the samples only NiO and CuO were
found.
3.1.4. X-ray diffraction
Fig. 3 shows the XRD patterns of the calcined catalysts before
reduction. For high contents of Cu, sharp reflections that can be as-
signed to CuO (JPCDS; 41-0254) were detected. The intensity of
these reflections declined with decreasing Cu content in the cata-
lysts; the reflections were not visible when Cu content was lower
2 4
than 7 wt.%. Diffraction lines due to the CuAl O , usually detected
at 2h = 31.3° and 37.9° (JPCDS; 33-0448), were not observed, which
agrees with the XPS results where only CuO was detected in the
near-surface region. In the case of high Ni content, reflections at
2h = 37.0° and 43.3° attributed to (1 0 1) and (0 1 2) planes of
NiO (JPCDS; 44-1159) could be observed. These reflections are
smaller and broader as compared to the reflections related to
CuO, indicating that NiO crystallite size was smaller than CuO crys-
tallite size. When the Ni content was 13 wt.% or less, reflections re-
lated to NiO were not detected.
NiO and CuO crystallite sizes were obtained by applying the
Scherrer equation. For CuO, sizes between 38 and 53 nm were ob-
tained, whereas for NiO, sizes between 5 and 12 nm were calculated.
Therefore, the dispersion of Ni was significantly better than the dis-
persion of Cu, which agrees with the above shown XPS results.
Fig. 4. XRD patterns of reduced 13Ni22Cu/Al
constant nominal metal content and different Cu/Ni metal ratio.
2
O
3
and 20Ni15Cu/Al
2
O
3
catalysts with
particles over the alumina support was observed. As an example,
in Fig. 5, the TEM images of 35Ni0Cu/Al , 28Ni7Cu/Al , and
Ni28Cu/Al are displayed.
2
O
3
2 3
O
7
2 3
O
3.1.6. Activity test results
The six calcined and reduced Ni–Cu/Al
2 3
O catalysts with con-
XRD analysis of reduced samples 13Ni22Cu/Al
2
O
3
and
stant total nominal metal content and different Cu/Ni ratio were
tested in glycerol hydrogenolysis with formic acid as hydrogen do-
nor molecule. Fig. 6 illustrates the main test results, in terms of
glycerol conversion, and 1,2-PDO selectivity and yield, achieved
with each catalyst at 473 and 493 K. For both temperatures, there
was a clear correlation between glycerol conversion and the Cu/Ni
ratio of the catalysts: with increasing Ni proportion in the catalyst,
the glycerol conversion increased. Therefore, it is clear that the Ni
plays a key role in the transfer hydrogenation of glycerol utilizing
the active hydrogen coming from formic acid. Nevertheless, there
was a significant decrease in 1,2-PDO selectivity in the tests per-
formed with the catalyst containing only Ni, namely 35Ni0Cu/
Al O . Hence, the presence of Cu seems to be necessary for obtain-
2 3
ing high 1,2-PDO selectivities.
At this point, a deeper analysis of the product distribution as a
function of the Cu/Ni ratio of the catalysts was required. For this
purpose, Table 5 presents the product distributions and carbon bal-
ances of the activity tests. For clarity, only the results obtained at
2
0Ni15Cu/Al was also carried out. The samples were reduced
2 3
O
following the same procedure as described in Section 2.1. Inside
the glovebox, part of the reduced catalyst powder was introduced
into a glass capillary (diameter 0.7 mm; wall thickness 0.01 mm),
which was afterwards sealed by melting in order to avoid contact
with air. Next, the capillary was introduced into the XRD equip-
ment for analysis. As observed in Fig. 4 for both reduced samples,
metallic Cu, Ni, and a Cu–Ni alloy were detected. Obviously, the
formation of the Cu–Ni alloy took place during the reduction under
2
H flow at 923 K. Small proportions of the CuO and NiO phases that
were initially present in the calcined samples were not reduced, as
demonstrated by the reflections at about 37.5° and 61–62°. Finally,
the broad reflections at 2h = 66.7° correspond to a well-dispersed
c
2 3
-Al O phase.
3.1.5. TEM images
The six calcined catalysts were also characterized by TEM. For
all the samples, a heterogeneous size distribution of metal oxide
4
93 K are displayed, but the same discussion could be done with
the ones obtained at 473 K. It is interesting to observe that there
is an increase in the selectivity to products with less than 3 C
atoms, which means products coming from C–C bond hydrogenol-
ysis, with increasing Ni content on the catalyst. The overall selec-
tivity for such cleavage products rises to 50.5% in the test
performed with the catalyst containing only Ni.
In the light of these results, it can be suggested that Ni is active
in the hydrogenolysis of glycerol using the active hydrogen atoms
coming from formic acid. In the absence of Cu, the hydrogenolysis
of C–C and C–O bonds occurred in a similar extent. When Cu was
present in the catalyst together with Ni, the selectivity to cleavage
products significantly decreased (see Table 5). How does Cu reduce
the activity of Ni in the C–C bond cleavage?
It is well known that in bimetallic catalysts, the presence of a
second metal may induce significant changes in both activity and
selectivity for catalytic reactions. For instance, the selectivity ef-
fects induced by alloying Ni with Cu were observed in cyclopen-
tane hydrogenolysis [37] and in ethane hydrogenolysis [38]. The
variation in the catalytic performance is often explained by the
ensemble theory, which considers that the addition of an inactive
2 3
Fig. 3. XRD patterns of calcined Ni–Cu/Al O catalysts with constant nominal metal
content and different Cu/Ni metal ratio.

8
4
I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
2 3 2 3 2 3 2 3
Fig. 5. TEM images of calcined Ni–Cu/Al O samples. (A) 35Ni0Cu/Al O , (B) 28Ni7Cu/Al O , (C) 7Ni28Cu/Al O .
metal results in a dilution of the surface-active metal atoms, and
then, in a decrease of the active ensemble size [39,40]. Cu is also
active in the C–O bond hydrogenolysis but not in the C–C bond
hydrogenolysis; therefore, it can be considered as an inert metal
in the C–C bond cleavage reaction [41]. Above shown XRD results
of reduced samples confirm the formation of a NiCu alloy in the
formed at the end of the test exponentially decreased with reduc-
ing the Ni surface proportion (and therefore increasing the Cu pro-
portion). This indicates that glycerol C–C bond cleavage requires a
whole ensemble of contiguous Ni atoms. The formation of a Ni–Cu
alloy reduces Ni ensemble size, and as a result, the C–C hydrogen-
olysis activity of the catalyst drastically diminishes.
bimetallic NiCu/Al
intercalated. This causes a reduction in the size of the Ni ensembles
available on the active surface.
2
O
3
catalysts. In the alloy, Ni and Cu atoms are
Hence, the presence of both metals is required for obtaining
high yields to 1,2-PDO. Ni provides comparatively high hydrogen-
olysis activity of C–C and C–O bonds and the formation of Ni–Cu
alloy limits C–C hydrogenolysis and promotes C–O hydrogenolysis.
Indeed, there is an optimum in Cu/Ni ratio (0.72) that maximizes
1,2-PDO production, as for both reaction temperatures the maxi-
In order to check whether glycerol C–C bond cleavage reaction
is sensitive to the size of the Ni ensemble, reduced samples of
the six catalysts were analyzed by XPS and the surface atomic com-
position of Ni, Cu, Al, O, and C was obtained from XPS peaks inten-
sities. The total mmol of C < 3 products (ethylene glycol, ethanol,
methanol, and methane) measured at the end of each activity test
were correlated to the relative proportion of Ni and Cu surface
atoms. As it can be seen in Fig. 7, the total mmol of C < 3 products
mum in 1,2-PDO yield was obtained with the 20Ni15Cu/Al
2 3
O cat-
alyst (see Fig. 6C).
The performance of formic acid during the reaction must be also
discussed. In a previous work, not shown here, we observed that
2
formic acid readily reacted to CO and hydrogen, with 100% con-
Fig. 6. Glycerol conversion, 1,2-PDO selectivity and yield obtained at 473 and 493 K for each Ni–Cu/Al
2
O
3
catalyst: 16-h reaction time, 45 bar N
2
pressure, 135 mL glycerol
À1
À1
aqueous solution (4 wt.%), 0.9 g of catalyst, 2.1 or 3:6 mmol g
h
formic acid feed rate.
catal

I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
85
Table 5
Product selectivities and carbon balances for the tests performed at 493 K. 16-h reaction time, 45 bar N
2
pressure, 135 mL glycerol aqueous solution (4 wt.%), 0.9 g of catalyst,
À1
À1
3
:6 mmol g
h
formic acid feed rate.
catal
Catalyst/Al
2
O
3
Conv. (%)
Selectivity (%)
Liquid finalb C (%)GC
Liquid finalc C (%)TOC
Solid finald C (%)CHN
Total finale C (%)
1
,2-PDO
Acetol
1-PO
C < 3a
0Ni35Cu
7Ni28Cu
13Ni22Cu
20Ni15Cu
28Ni7Cu
35Ni0Cu
29.2
40.9
47.5
49.3
47.7
53.0
75.5
82.2
78.1
75.4
74.3
42.0
11.2
5.0
3.0
4.1
2.4
1.8
5.0
4.5
7.4
9.2
9.2
3.3
8.3
8.3
11.4
11.3
14.0
50.5
94.9
97.0
94.6
99.1
99.3
76.0
92.6
97.4
94.9
102.5
102.1
75.9
1.05
1.14
1.12
1.15
1.06
0.57
96.5
103.2
97.6
101.7
101.9
93.4
a
b
c
Ethylene glycol, ethanol, and methane.
(
(
(
(
Final mass of C in the liquid phase obtained by GC-FID analysis/initial mass of C in the reactor) Â 100.
Final mass of C in the liquid phase obtained by TOC analysis/initial mass of C in the reactor) Â 100.
Mass of C in the catalyst at the end of the test obtained by CHN analysis/initial mass of C in the reactor) Â 100.
d
e
Final mass of C in the liquid and gas phase obtained by GC-FID analysis, plus mass of C in the catalyst obtained by CHN elemental analysis/initial mass of C in the
reactor) Â 100.
ble amounts of polyglycerols were formed during the reaction.
These compounds are not detectable with capillary GC, but TOC
is sensitive for any carbonaceous compound. If such compounds
were formed, then they should be found in the solid or on the
catalyst.
Carbon content of used catalysts obtained by CHN analysis of
solid residue gives an idea of the amount of coke (or polyglycerols)
formed during the reaction. The carbon content measured in the
catalysts represented only 1% of the initial glycerol carbon (see Ta-
ble 5), and in terms of catalyst weight, the measured carbon repre-
sented less than 2.7 wt.% in all cases. Moreover, part of the carbon
amount detected in spent samples was already present in the fresh
catalysts from the preparation procedure (see XPS results in Table
3). Therefore, the obtained results indicate that negligible amounts
of coke were formed during the reaction.
3.2. Optimization of the total metal content
Fig. 7. Total mmol of C < 3 products formed after 16-h reaction time as a function of
the surface Ni/(surface Ni + surface Cu) ratio, which was obtained by XPS analysis of
the six reduced catalysts. 45 bar N
2
pressure, 135 mL glycerol aqueous solution
2 3
Another series of five different Ni–Cu/Al O catalysts was pre-
À1
À1
(
4 wt.%), 0.9 g of catalyst, 3:6 mmol g
h
formic acid feed rate, 493 K.
catal
pared by sol–gel method with the optimized Cu/Ni ratio of 0.72,
derived from the results reported in the previous section, and dif-
ferent total metal content. These catalysts were also characterized
and tested in glycerol hydrogenolysis using formic acid as hydro-
gen donor molecule.
version in all the tests. Formic acid was not detectable with the GC-
FID equipment used for liquid-phase analysis. Therefore, in order
to ensure that in all the performed tests formic acid was com-
pletely converted, five liquid samples stemming from different
activity tests were selected and analyzed by GC–MS (Shimadzu
GC–MS QP2010 SE; CP-Sil 5 CB column suited for formaldehyde,
2
3.2.1. Elemental analysis and N physisorption
Table 6 summarizes the theoretical and real metal content to-
gether with the main textural properties of the calcined catalysts.
As seen for the above-described first catalyst series, the real Ni
and Cu contents were lower than the theoretical ones in all the
samples, whereas the real Cu/Ni ratio values were closer to the the-
oretical ones in those catalysts with high metal content.
6
0 m  0.32 mm  0.45 mm) as well as one sample of the solution
pumped. Formic acid was only detected in the solution pumped.
Concerning the gas phase, negligible CO amounts were detected,
indicating that if CO was formed via formic acid dehydration, it
was readily converted to CO
2
by water–gas shift reaction. Molecu-
2
Concerning the N physisorption results, all the samples pre-
lar hydrogen was detected in the gas phase. Those hydrogen atoms,
coming from formic acid decomposition, that were not transferred
to glycerol combined to form H . The hydrogen balance (mol of H
2
coming from formic acid = mol of H in the gas phase + mol of H
transferred to liquid products) closed within a 10% error in all
the experiments.
Table 5 also shows the carbon balances for the different activity
tests. For the liquid phase, it was assumed that all the carbon de-
tected in the TOC measurements came from glycerol, as all the for-
sented Type IV isotherms, with hysteresis loops typical for meso-
porous materials. Within this series, there is no clear trend in
BET area and total pore volume as a function of total metal content.
Nevertheless, as expected, the catalysts with higher metal con-
tents, and therefore lower alumina contents, presented lower BET
areas. This is explained by the surface coverage and plugging of
the pores.
3.2.2. X-ray diffraction
mic acid was converted to CO
that all the CO detected came from formic acid and that all the
CH detected came from glycerol decomposition (no Fischer–Trop-
sch reaction). As observed in Table 5, there is a quite good agree-
ment between final liquid carbon amounts obtained by GC-FID
and TOC analysis (used as a cross-check). This proves that negligi-
2
. For the gas phase, it was considered
Fig. 8 illustrates the XRD patterns of the calcined Ni–Cu/Al
catalysts having similar Cu/Ni ratio and different total metal con-
tent. The pattern of the calcined 20Ni15Cu/Al does not appear
in Fig. 8 because it was already shown in Fig. 3. The two catalysts
with less metal content, 3Ni2Cu/Al and 12Ni8Cu/Al showed
mainly the reflections related to the Al support. This indicates a
2 3
O
2
4
2 3
O
2
O
3
2 3
O
2 3
O

8
6
I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
Table 6
Characteristics of the catalysts with constant nominal Cu/Ni ratio and different metal content.
3
Name/Al
2
O
3
Ni content (wt.%)
Theoretical (Real)
2.9
Cu Content (wt.%)
Ni + Cu (wt.%)
Ratio Cu/Ni
S
BET
2
Total pore volume (cm /g)
Average pore size (nm)
(
m /g)
Theoretical (Real)
Theoretical (Real)
Theoretical (Real)
3
1
2
2
3
Ni2Cu
2.1
(2.2)
5.0
(4.8)
0.72
(0.85)
271.9
207.9
229.6
142.0
95.1
0.201
0.194
0.247
0.217
0.112
2.72
3.40
3.66
5.43
5.79
(
2.6)
11.6
9.0)
2Ni8Cu
8.4
(7.5)
20.0
(16.5)
0.72
(0.83)
(
0Ni15Cu
9Ni21Cu
8Ni27Cu
20.3
14.7
(12.8)
35.0
(30.1)
0.72
(0.74)
(
(
(
17.3)
29.0
26.8)
21.0
(20.8)
50.0
(47.6)
0.72
(0.78)
37.8
37.6)
27.3
(26.7)
65.1
(64.3)
0.72
(0.71)
a
Chemical composition determined by ICP.
performed at 473 K (around 85%) were higher as compared to
the values recorded in the tests at 493 K (around 75%). This is in
accordance with the results from the tests with the first series of
catalysts. Therefore, the selectivity to 1,2-PDO depends more on
the Cu/Ni ratio of the catalysts and on the reaction temperature
than on the total metal content. Fig. 9C illustrates 1,2-PDO yield
obtained with each catalyst at 473 and 493 K. For both tempera-
tures, the maximum was observed when the catalyst having a
nominal metal content of 35 wt.% was used.
The acidity of 12Ni8Cu/Al
was evaluated by TPD-NH
. As expected, the acidity of the samples decreased as the nominal
metal content of the catalysts increased. This decrement on acidity
can be related to the Al content per gram of catalyst of the sam-
ples. In order to determine whether the decrement was also caused
by Ni and Cu particles occupying the acid sites of the Al support,
ammonia desorption results were normalized by mass of Al
see Table 7 last column). It can be observed that the increase in
the metal content reduced the support acidity, indicating that Ni
and Cu metal particles occupied acid sites of the Al [42].
2
O
3
, 20Ni15Cu/Al
. The results are displayed in Table
2 3
O , and 38Ni27Cu/
Al
7
2
O
3
3
2 3
O
2 3
O
2 3
O
(
Fig. 8. XRD patterns of calcined Ni–Cu/Al
metal ratio and different total metal content.
2
O
3
catalysts with constant nominal Cu/Ni
2 3
O
The role of acid sites in direct glycerol conversion mechanism to
,2-PDO through the intermediate formation of 1,3-dihydroxyiso-
high dispersion of the NiO and CuO phases on the support. In the
patterns of the two catalysts with the highest metal content
1
propoxide was firstly suggested for a system under H
and using Rh–Re/SiO [32] or Ir–Re/SiO [43] catalysts. The alkox-
ide was formed by adsorption on the ReO cluster. Next, the hy-
2
pressure
(
47.7 wt.% and 64.3 wt.%, respectively), diffraction peaks related
2
2
to the metal oxide phases were detected. Apart from CuO and
NiO phases, some mixed Cu–Ni oxides with different Cu/Ni ratio
were detected (Ni0.75Cu0.25O, Ni0.95Cu0.05O). Ni–Cu oxide phases
were not recorded in the XPS analysis of the catalysts with con-
stant nominal metal content of 35 wt.% (see Fig. 3). This seems to
indicate that for a significant formation of Cu–Ni oxide crystallite
phases high Ni and Cu contents are required.
x
dride coming from molecular hydrogen and activated on Ir or Rh
sites attacked the C–O bond. Chia et al. [44] and King et al. [45]
have recently suggested the formation of acid sites associated with
the presence of ReO clusters. Therefore, Re acting as an acid site is
x
responsible for the alkoxide adsorption, and Ir or Rh sites of the hy-
dride attack.
In our system, acid sites are provided by Al
2 3
O support instead
of ReO , while hydrides are formed from formic acid decomposi-
x
3.2.3. Activity test results
tion and not from dissolved molecular hydrogen dissociation.
Therefore, not only the glycerol hydrogenolysis mechanism, but
also the formic acid decomposition mechanism needs to be taken
into account. Concerning alkoxide formation, 1,3-dihydroxyiso-
propoxide is formed by glycerol adsorption on an acid site of the
alumina support. Alkoxide formation on the acid sites of an alu-
mina support has already been reported [46]. Regarding hydride
formation, it is likely that formic acid adsorbs dissociatively to give
a formate species and an adsorbed hydrogen atom, and that the
Fig. 9 shows the main activity test results obtained at 473 and
93 K, with the series of Ni–Cu/Al O catalysts with nearly con-
2 3
4
stant Cu/Ni ratio and different total metal content. For both tem-
peratures, glycerol conversion increased with total metal loading,
until a maximum was reached for the catalyst having a nominal
metal content of 35 wt.%. For higher nominal metal contents, glyc-
erol conversion decreased with additional increases of the total
amount of metals in the catalyst. Hence, there is an optimum me-
tal-Al
is coherent with the fact that – as it will be discussed below – both
metal and acid sites of the Al support are supposed to play a
2 3
O proportion that enhances glycerol conversion. This result
2
formate species then dissociate to give gaseous CO and another
adsorbed hydrogen atom [47,48]. If 1,3-dihydroxyisopropoxide is
adsorbed in an acid site near a metal site having an adsorbed
hydrogen atom, the hydride attacks the CO bond of the alkoxide.
Finally, the hydrolysis of the reduced alkoxide gives 1,2-PDO (see
Fig. 10).
2 3
O
role in the glycerol hydrogenolysis reaction.
Regarding the effect of total metal content of the catalyst on the
selectivity to 1,2-PDO, it can be observed in Fig. 9A and B that there
was no significant influence. The selectivities in the tests

I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
87
Fig. 9. Glycerol conversion, 1,2-PDO selectivity and yield obtained at 473 and 493 K reaction temperature for each Ni–Cu/Al
2
O
3
catalyst with constant Cu/Ni ratio and
À1
À1
different metal loading. 16-h reaction time, 45 bar N
2
pressure, 135 mL glycerol aqueous solution (4 wt.%), 0.9 g of catalyst, 2.1 or 3:6 mmol gcatal
h
formic acid feed rate.
Table 7
493 K and the same operating conditions. The yield to 1,2-PDO ob-
tained (10.1%) was lower than that obtained with the catalyst hav-
Acidity of the reduced samples from temperature-programed desorption of ammonia.
Sample/Al NH desorption NH desorption
2
O
3
3
3
ing the higher metal content, 38Ni27Cu/Al
times lower than the one obtained with the optimized 20Ni15Cu/
Al catalyst (36.4%).
2 3
O (22.3%), and 3.6
mmol NH
3
/g catal
3 2 3
mmol NH /g Al O
2 3
O
1
2
3
2Ni8Cu
0Ni15Cu
8Ni27Cu
0.33
0.16
0.06
0.39
0.23
0.17
3.3. Kinetic study
This mechanism allows explaining why there must be an opti-
mum proportion of acid and metal sites. For the catalysts present-
ing low metal content, 1,3-dihydroxyisopropoxide is formed on the
highly available acid sites. However, due to the low relative
amount of metal sites, only a low proportion of these alkoxides
are adsorbed near a metal site having an adsorbed hydrogen atom
and are therefore desorbed without reacting. In the opposite case,
which means a catalyst with high metal content, there are not en-
ough acid sites available for glycerol adsorption. For the Ni–Cu/
So far, no information about the evolution of glycerol conver-
sion with the reaction time has been shown, as in the above pre-
sented test results, liquid and gas phase were only analyzed at
the end on the experiments after 16 h. In order to gain a deeper
knowledge on the process, further tests were carried out with
2 3
the optimized catalyst, 20Ni15Cu/Al O , but taking five liquid sam-
ples in regular time intervals during the activity test.
As observed in Fig. 11A, the activity of the 20Ni15Cu/Al
2 3
O cat-
alyst decreased markedly during reaction time. In the first 2 h, an
2
Al O
3
catalytic system, the optimum was obtained for a catalyst
À3
À1
À1
average glycerol TON of 4:7 Â 10 molglyc
g
h
was obtained,
catal
having a nominal metal content of 35 wt.%.
However, the formation of 1,3-dihydroxyisopropoxide also on
metal sites cannot be excluded. A test was run using Raney nickel,
while for the last 16 h the average TON was only
In principle, complete conversion
should be possible if hydrogen donor was fed in excess. Three main
À3
À1
À1
0
:5 Â 10 molglyc
g
h
.
catal
Fig. 10. Model structure of the transition states of the hydride attack to the adsorbed 1,3-dihydroxypropoxide, under inert atmosphere, with formic acid as hydrogen donor
molecule and over Ni–Cu/Al catalyst.
2 3
O

8
8
I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
Fig. 11. Evolution of glycerol conversion with reaction time. 20Ni15Cu/Al
2
O
3
catalyst, 493 K, 24-h reaction time, 45 bar N
2
pressure, 135 mL glycerol aqueous solution
À1
À1
(
4 wt.%), 3:6 mmol gcatal
h
formic acid feed rate. (A) Effect of reusing the catalyst. (B) Effect of the presence of 1,2-PDO in the feed (0.5 mol of 1,2-PDO/mol of glycerol). (C)
Effect of the amount of catalyst placed in the reactor. (D) Evolution of 1,2-PDO and 1-propanol selectivity in the test with high catalyst loading.
factors were selected as the possible causes for this decrease of the
activity: (i) deactivation of the catalyst, (ii) insufficient hydrogen
supply, and (iii) competitive adsorption for active sites between
glycerol and 1,2-PDO. Thermodynamic equilibrium between glyc-
erol and formic acid was not considered as a possible factor, be-
cause under similar operating conditions but using hydrogen
pressure higher glycerol conversion and 1,2-PDO yields have been
reported.
In order to determine whether there is competition for active
sites between 1,2-PDO and glycerol, 1,2-PDO was introduced to-
gether with glycerol in the aqueous feed solution (0.5 mol of 1,2-
PDO/mol of glycerol). As Fig. 11B shows, glycerol conversion was
significantly affected by the presence of 1,2-PDO in the feed. In-
deed, final glycerol conversion obtained after 24 h decreased from
52.7% to 35.4%. It is therefore clear that 1,2-PDO affects glycerol
reaction rate; there is product inhibition.
In order to check whether the catalyst was deactivated during
the reaction, it was decided to reuse the 20Ni15Cu/Al O catalyst
2 3
Another activity test was carried out without 1,2-PDO in the
feed (only glycerol aqueous solution), keeping the operating condi-
tions as in the previous tests, and using three times more of the
in a second run, under the same experimental conditions. It can
be noticed from Fig. 11A that similar glycerol conversions with
reaction time were achieved in the first and second run. Therefore,
in the short term, the catalyst was not deactivated, as it was ex-
pected from the low carbon content detected in the spent catalysts
after the activity tests, and hence another factor must be the rea-
son of the decrease in the catalyst activity.
2 3
20Ni15Cu/Al O catalyst (2.7 g). Hence, there were three times
more available active sites in the reactor for the same molar
amount of glycerol. It can be noticed in Fig. 11C that glycerol con-
version significantly increased when a higher amount of catalyst
was used. Indeed, after 24 h, a glycerol conversion of 89.9% and a
selectivity to 1,2-PDO of 81.6% were measured. These results
clearly indicate that the competition for active sites between the
OH groups of glycerol and 1,2-PDO is the main factor that limits
the activity of the catalyst during the reaction time.
Formic acid was continuously fed during 24 h at a constant rate
À1 À1
of 3:6 mmol h
gcatal. As written above, it can be assumed that for-
mic acid reacts instantaneously when it enters the reactor, releas-
ing CO and two active hydrogen species. As a consequence, the
2
The competition for active sites between the OH groups of 2-
supply of active hydrogen was constant during the entire activity
test: the same active hydrogen supply rate when high TON (after
2 3
propanol and glycerol in Ni–Cu/Al O catalyst was previously re-
ported [29]. 1,2-PDO competes with glycerol to be adsorbed in
the acid sites of the support to form 2-hydroxypropoxide. Part of
this adsorbed alkoxide reacts with the active hydrogen atoms com-
ing from formic acid, being activated in the Ni–Cu metal sites, to
yield 1-propanol. In Fig. 11D, it can be observed that after 10-h
reaction time, there is a decrease in the selectivity to 1,2-PDO
and an increase in the selectivity to 1-propanol. The fact that high
2
h) were registered and when low TON (in last 16 h) were regis-
tered. In other words, hydrogen donor and therefore active hydro-
gen species were always present in excess, and therefore reaction
rate was not limited by the concentration of those. Hence, the
hydrogen supply is not the main factor behind the decrease in
the activity of the catalyst with the reaction time.

I. Gandarias et al. / Journal of Catalysis 290 (2012) 79–89
89
1
,2-PDO selectivity was obtained after 24 h indicates that the Ni–
Cu/Al catalytic system is more active in the glycerol hydrogen-
olysis to yield 1,2-PDO than in the 1,2-PDO hydrogenolysis to yield
[2] D.C. Elliott, D. Beckman, A.V. Bridgwater, J.P. Diebold, S.B.S.B. Gevert, Energ.
Fuel 5 (1991) 399.
2 3
O
[
3] D.L. Klass, Biomass for Renewable Energy, Fuels, and Chemicals, Academic
Press, San Diego, 1998.
1
-propanol.
[4] G.W. Huber, R.D. Cortright, J.A. Dumesic, Angew. Chem. Int. Ed. 43 (2004) 1549.
[
5] C. Montassier, J.C. Ménézo, L.C. Hoang, C. Renaud, J. Barbier, J. Mol. Catal. 70
1991) 99.
It is worth to highlight the high glycerol conversion, 90%, and
(
selectivity values, 82%, obtained in a system with in situ generation
of hydrogen. Actually, these results are comparable to the best re-
sults reported under the traditional liquid phase process using
hydrogen pressure (see Table 1). Further research should be
accomplished to determine the effect of the pressure in glycerol
hydrogenolysis through transfer hydrogenation, as less severe con-
ditions can make the process even more attractive. Moreover, the
performance of formic acid as hydrogen donor should be compared
to molecular hydrogen under the same operating conditions.
[
6] S. Sato, M. Akiyama, R. Takahashi, T. Hara, K. Inui, M. Yokota, Appl. Catal. A:
Gen. 347 (2008) 186.
7] S. Wang, H. Liu, Catal. Lett. 117 (2007) 62.
8] T. Miyazawa, Y. Kusunoki, K. Kunimori, K. Tomishige, J. Catal. 240 (2006) 213.
9] E.P. Maris, R.J. Davis, J. Catal. 249 (2007) 328.
[
[
[
[10] D.G. Lahr, B.H. Shanks, Ind. Eng. Chem. Res. 42 (2003) 5467.
11] Y. Kusunoki, T. Miyazawa, K. Kunimori, K. Tomishige, Catal. Commun. 6 (2005)
45.
12] M.A. Dasari, P. Kiatsimkul, W.R. Sutterlin, G.J. Suppes, Appl. Catal. A: Gen. 281
(2005) 225.
13] J. Chaminand, L. Djakovitch, P. Gallezot, P. Marion, C. Pinel, C. Rosier, Green
Chem. 6 (2004) 359.
[
[
[
[
[
6
14] M. Akiyama, S. Sato, R. Takahashi, K. Inui, M. Yokota, Appl. Catal. A: Gen. 371
(
2009) 60.
4
. Conclusions
15] Z. Yuan, L. Wang, J. Wang, S. Xia, P. Chen, Z. Hou, X. Zheng, Appl. Catal. B:
Environ. 101 (2011) 431.
In the present work, the performance of Ni–Cu/Al
2
O
3
bimetallic
[16] A. Wawrzetz, B. Peng, A. Hrabar, A. Jentys, A.A. Lemonidou, J.A. Lercher, J. Catal.
69 (2010) 411.
2
catalysts on the glycerol hydrogenolysis to 1,2-PDO under inert
atmosphere and using formic acid as the source of hydrogen was
studied. It was observed that there is an optimum Cu/Ni ratio
and also an optimum in the ratio and distribution of metal and acid
sites that maximize the yield of 1,2-PDO. The glycerol hydrogenol-
[
[
17] E.S. Vasiliadou, A.A. Lemonidou, Appl. Catal. A: Gen. 396 (2011) 177.
18] R.A.W. Johnstone, A.H. Wilby, I.D. Entwistle, Chem. Rev. 85 (1985) 129.
[19] E. D´ Hont, S. Van de Vyver, B.F. Sels, P.A. Jacobs, Chem. Commun. (2008) 6011.
[
[
20] D. Roy, B. Subramaniam, R.V. Chaudhari, Catal. Today 156 (2010) 31.
21] F. Auneau, S. Noël, G. Aubert, M. Besson, L. Djakovitch, C. Pinel, Catal. Commun.
1
6 (2011) 144.
[22] M.G. Musolino, L.A. Scarpino, F. Mauriello, R. Pietropaolo, Green Chem. 11
2009) 1511.
23] M.G. Musolino, L.A. Scarpino, F. Mauriello, R. Pietropaolo, ChemSusChem 4
2011) 1143.
2 3
ysis occurs when glycerol is adsorbed on an acid site of Al O to
(
form a secondary alkoxide, and when this alkoxide interacts with
an adsorbed hydrogen atom coming from formic acid decomposi-
tion and being activated on a metal site. The optimum balance be-
tween the alumina acid sites and the metal sites was obtained for
the catalyst with a nominal metal content of 35 wt.%. Regarding
the role of the metallic sites, Ni provides comparatively high activ-
ity in the C–C and C–O bond hydrogenolysis, whereas Cu provides
some activity in the C–O bond hydrogenolysis but not in the C–C
bond hydrogenolysis. The formation of a Cu–Ni alloy during the
[
(
[24] T. Miyazawa, S. Koso, K. Kunimori, K. Tomishige, Appl. Catal. A: Gen. 318
(2007) 244.
[
25] I. Gandarias, P.L. Arias, J. Requies, M. El Doukkali, M.B. Güemez, J. Catal. 282
2011) 237.
(
[
26] D.J. Hayes, S. Fitzpatrick, M.H.B. Hayes, J.R.H. Ross, Biorefineries-Industrial
Processes and Products, Wiley-VCH, Weinheim, 2006.
[
[
27] B. Loges, A. Boddien, F. Gaertner, H. Junge, M. Beller, Top. Catal. 53 (2010) 902.
28] A. Boddien, B. Loges, F. Gaertner, C. Torborg, K. Fumino, H. Junge, R. Ludwig, M.
Beller, J. Am. Chem. Soc. 132 (2010) 8924.
reduction of the Ni–Cu/Al
2
O
3
catalysts reduces the active Ni
[29] D.A. Bulushev, J.R.H. Ross, Catal. Today 163 (2011) 42.
[
[
[
30] T. Kim, M. Song, H. Koh, K. Kim, Appl. Catal. A: Gen. 210 (2001) 35.
31] J. Lee, E. Lee, O. Joo, K. Jung, Appl. Catal. A: Gen. 269 (2004) 1.
32] X. Wang, X. Pan, R. Lin, S. Kou, W. Zou, J. Ma, Int. J. Hydrog. Energ. 35 (2010)
4060.
ensemble size. As a result, the activity of the bimetallic catalyst
for C–C bond cleavage is diminished while the activity for C–O
bond cleavage is promoted. The optimum Cu/Ni ratio that maxi-
mized the yield to 1,2-PDO was 0.72.
[
[
33] M. Turco, G. Bagnasco, C. Cammarano, P. Senese, U. Costantino, M. Sisani, Appl.
Catal. B: Environ. 77 (2007) 46.
34] J. Shen, C. Song, Catal. Today 77 (2002) 89.
The kinetic study revealed that the OH groups of glycerol and of
1
,2-PDO competed for the same acid sites to form alkoxides. In or-
[35] F.E. López-Suárez, A. Bueno-López, M.J. Illán-Gómez, Appl. Catal. B: Environ. 84
2008) 651.
36] L. De Rogatis, T. Montini, A. Cognigni, L. Olivi, P. Fornasiero, Catal. Today 145
2009) 176.
[37] V. Ponec, W.M.H. Sachtler, J. Catal. 24 (1972) 250.
(
der to minimize this factor, high catalyst loading was placed in the
reactor, and a glycerol conversion of 90% and a 1,2-PDO selectivity
of 82% were achieved after 24 h operating at 493 K.
[
(
[
[
[
[
38] J.H. Sinfelt, J.L. Carter, D.J.C. Yates, J. Catal. 24 (1972) 283.
39] V. Ponec, Appl. Catal. A: Gen. 222 (2001) 31.
40] B. Coq, F. Figueras, Coord. Chem. Rev. 178–180 (1998) 1753.
41] E. Asedegbega-Nieto, B. Bachiller-Baeza, A. Guerrero-Ruíz, I. Rodríguez-Ramos,
Appl. Catal. A: Gen. 300 (2006) 120.
Acknowledgments
This work was supported by funds from the Spanish Ministry of
Science and Innovation ENE2009-12743-C04-04, and from the Bas-
que Government (Researcher Training Programme of the Depart-
ment of Education, Universities and Research). The authors
greatly acknowledge Drs Schneider, Pohl, Radnik, Mr. Eckelt and
Ms Evert for the work done in the characterization of the catalysts,
and the Inorganic Chemistry Department at the University of Ma-
laga for their technical support.
[
42] V. Ponec, G.C. Bond, Catalysis by Metals and Alloys, Academic Press,
Amsterdam, 1995.
[43] Y. Shinmi, S. Koso, T. Kubota, Y. Nakagawa, K. Tomishige, Appl. Catal. B:
Environ. 94 (2010) 318.
[
44] Y. Amada, Y. Shinmi, S. Koso, T. Kubota, Y. Nakagawa, K. Tomishige, Appl. Catal.
B: Environ. 105 (2011) 117.
[45] M. Chia, Y.J. Pagán-Torres, D. Hibbitts, Q. Tan, H.N. Pham, A.K. Datye, M.
Neurock, R.J. Davis, J.A. Dumesic, J. Am. Chem. Soc. 133 (2011) 12675.
[
46] D.L. King, L. Zhang, G. Xia, A.M. Karim, D.J. Heldebrant, X. Wang, T. Peterson, Y.
Wang, Appl. Catal. B: Environ. 99 (2010) 206.
[
[
47] M.I. Zaki, M.A. Hasan, L. Pasupulety, Langmuir 17 (2001) 4025.
48] J.A. Wang, X. Bokhimi, O. Novaro, T. López, F. Tzompantzi, R. Gómez, J.
Navarrete, M.E. Llanos, E. López-Salinas, J. Molec. Catal. A: Chem. 137 (1999)
References
239.
[
1] T.E. Bull, Science 285 (1999) 1209.