Function of the support and metal loading on catalytic carbon dioxide reduction using ruthenium nanoparticles supported on carbon nanofibers

Article abstract of DOI:10.1002/cctc.201500016

Catalytic CO₂ reduction has been performed using carbon nanofibers or nitrogen-doped carbon nanofibers as a support for Ru nanoparticles. The catalyst that consists of 5 wt % Ru on nitrogen-doped carbon nanofibers exhibited the highest conversi

Full text of DOI:10.1002/cctc.201500016

DOI: 10.1002/cctc.201500016
Full Papers
Function of the Support and Metal Loading on Catalytic
Carbon Dioxide Reduction Using Ruthenium Nanoparticles
Supported on Carbon Nanofibers
[
a]
Laura Roldµn, Yanila Marco, and Enrique García-BordejØ*
Catalytic CO2 reduction has been performed using carbon
nanofibers or nitrogen-doped carbon nanofibers as a support
for Ru nanoparticles. The catalyst that consists of 5 wt% Ru on
nitrogen-doped carbon nanofibers exhibited the highest con-
version at a relatively low temperature, complete selectivity to
tion, and transient response to CO removal) revealed that the
2
catalyst support participates actively in the reaction. The Ru
content governed the selectivity, which either favored CO for-
mation for lower Ru contents (0.5–2 wt%) or CH formation for
4
5 wt% Ru loading. The mean Ru particle size determined by
TEM was similar for each of the metal loadings. Therefore, the
substantially different selectivity patterns cannot be attributed
CH , and stable catalytic performance. The catalytic per-
4
formance was substantially superior to catalysts supported on
carbon nanotubes and akin to the best metal-oxide-supported
catalyst in the literature. The characterization of the prepared
catalyst by transient experiments (CO2 temperature-pro-
grammed desorption, temperature-programmed surface reac-
to structure sensitivity. The higher selectivity to CH can be ex-
4
plained by the enhanced supply of adsorbed hydrogen to the
activated adsorbed CO intermediate, which was demonstrated
to be the rate-determining step.
Introduction
[
3]
The perceived risk of running out of conventional fossil fuels
and the pollution risks associated with burning fuels has led
the boom in programs for the development of renewable
energy. The efficient utilization of renewable energy sources
from wind turbines and solar panels is still an ongoing chal-
lenge. Naturally, power generated by renewable energies
sible to convert H with CO into synthetic natural gas (CH ).
2 2 4
Synthetic natural gas can be stored and used in the existing
gas infrastructure without limitation. An additional advantage
is that the gas infrastructure (transport, storage, and distribu-
tion) is immediately available and there is no investment
needed. This will also allow the transportation of energy from
areas with high renewable power (sun or wind rich) to remote
areas with less renewable energy. A cost-estimation study on
a CO recycling system that produces CH using photovoltaic
(wind, photovoltaics) is intermittent, highly volatile, and not in
line with the demand for electricity. Electricity from wind and
sun is generated at irregular intervals, and the energy is pro-
duced at locations where it is not consumed directly. Only the
large-scale storage of energy can secure an economic supply
from renewable sources reliably. Thus, solutions for long-range
transportation and the storage of renewable energies are cur-
2
4
energy in the Middle East and transports it to Japan showed
that the cost of the energy input of CO recycling is similar to
2
that to obtain liquefied natural gas (LNG) from wells (mining,
[4]
liquefaction, and purification). Moreover, 79% of CO emis-
2
[
1]
rently of great interest.
sions from a 1 GW CH -combustion power plant can be re-
4
Recently, much interest has been devoted to energy conver-
sion based on the hydrogen cycle (“hydrogen economy”) such
as water splitting. It is easy and affordable to convert electricity
to hydrogen by electrolysis, a proven technology in the chemi-
duced by CO conversion to CH if CO is recovered at the
2
4
2
power plant. The conversion of CO2 into fuels and useful
chemicals using renewable H2 will permit that the carbon
cycle, which has generally an open end because of the burning
of fossil fuels and release of CO2 into atmosphere, can be
[
2]
cal industry. The main unsolved problem of the use of H as
2
an energy carrier is its storage. Hydrogen can be stored and
used in the existing gas infrastructure but the discharge time
is typically lower than 100 h, depending on technical regula-
tions. Research is underway to widen the use of hydrogen in
the existing gas infrastructure. For long-term storage it is pos-
closed by the capture of CO in the power station and recy-
2
cling it to fuel. Thus, this approach allows us to kill two birds
with one stone, the storage of renewable H and the reduction
2
of CO2 emissions. Several projects are currently underway
worldwide to exploit this concept of green natural gas such as
“
e-gas” by Audi and “Power2gas” by E.ON.
[
a] Dr. L. Roldµn, Y. Marco, Dr. E. García-BordejØ
Instituto de Carboquímica (ICB-CSIC)
Miguel Luesma Castµn 4, E-50018 Zaragoza (Spain)
Fax: (+34)976733318
The reaction of CO and H was first reported a century ago
2
2
[5]
and it is known as the Sabatier reaction. This has been inves-
tigated more intensively recently because of its fundamental
and practical significance in the context of catalysis, surface
science, biology, nanotechnology, and environmental sci-
E-mail: jegarcia@icb.csic.es
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201500016.
[6,7]
[7]
ence.
Photocatalysis, electrocatalysis, and thermal homo-
ChemCatChem 2015, 7, 1347 – 1356
1347
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[
6,8]
geneous
and heterogeneous catalysis have been used to
Results
hydrogenate CO . State-of-the-art photocatalysts used in artifi-
2
cial photosynthesis produce low CO conversions that render
2
Catalytic testing
[
9]
the process inefficient thus far. Homogeneous catalysts show
satisfactory activity and selectivity, but recovery and regenera-
tion are problematic. Alternatively, heterogeneous catalysts are
preferable in terms of stability, separation, handling, reuse, and
reactor design, which reduces the cost of large-scale produc-
tion. With regard to heterogeneous catalysis, supported transi-
The CO conversion and CH and CO yields for three different
2
4
Ru loadings on CNF and N-CNF are shown in Figure 1A–C, re-
spectively. There are very significant differences in terms of
conversion and selectivity as a function of the loading, irre-
spective of the support. If we compare the performance of 0.5
[
10–13]
[14,15]
tion metals, Ni
and Co,
and noble metals, which in-
and 2 wt% Ru loaded catalysts, the differences in CO conver-
2
[
16–18]
[19]
[19,20]
[19,21–23]
clude Ru,
Pd, Pt,
and Rh,
are active for this re-
sions are very modest or even nonexistent for the CNF- and N-
action. The support plays an important role in the catalytic re-
action as it may interact with the reactant(s), stabilize inter-
mediates or reaction products, or create special interfacial sites
at which reactions can proceed. The support can have elec-
CNF-supported catalysts. The CO conversion increases signifi-
2
cantly with the loading from 2 to 5 wt% Ru. The catalysts that
contain 5 wt% Ru achieve thermodynamic equilibrium at
[
24]
tronic effects on the catalytic nanoparticles or it can adsorb
some of the reactants to increase the local concentration of
[
25]
the reactant. Accordingly, some reactions have been report-
ed to take place in the interphase between the metal and sup-
[
26,27]
port.
Several support materials have been used in CO hy-
2
[
28–31]
[27,32,33]
drogenation such as alumina,
titania,
titanium car-
[
34]
[15]
[12,18,35]
[24]
bide,
silica,
ceria,
zeolite,
and carbon materials
[
36]
[37,38]
such as activated carbon and carbon nanotubes (CNT).
CO reduction is reported to require a bifunctional catalyst
2
that has one function to activate CO and another to activate
2
[
19]
H . Park and MacFarland observed a selectivity shift from CO
2
to CH by modifying Pd on SiO with MgO, whereas MgO/SiO
2
4
2
showed no measurable activity. They rationalized their results
by suggesting a bifunctional mechanism in which CO first ad-
2
sorbs strongly onto MgO to inhibit CO desorption, and Pd dis-
sociates H . There is also clear evidence that CO interacts with
2
2
the Al O support to produce alumina-bound carbonates/bicar-
2
3
[
31,39]
bonates.
Therefore, metal oxide supports are clearly not in-
nocent in this reaction. However, CNT is usually considered as
an inert support material. The activity was negligible if CNT
was used as a Ru support for CO hydrogenation, which was
2
[
40]
attributed to the lack of the ability to activate CO2. To the
best of our knowledge, Ru on carbon nanofibers (CNF) has not
been studied for CO methanation.
2
Herein, we used CNF and nitrogen-doped CNF (N-CNF) as
supports for several Ru loadings. The catalysts showed remark-
able activity in CO2 hydrogenation, in contrast to previous
studies in which CNT was used as a support. This prompted us
to study the reaction mechanism. As in situ spectroscopic char-
acterization is not usually performed for carbon-based catalysts
because it is hampered by the high absorbance of carbon, the
reaction mechanism was studied here by catalytic transient re-
sponse. This characterization enabled us to explain the effect
of the support and metal loading on the excellent selectivity
to CH . It was revealed that the support played an active role
4
in the reaction, which gives the catalyst an outstanding per-
formance compared to those supported on metal oxides.
Figure 1. A) CO
loaded catalysts on CNF and N-CNF: (~) 0.5%Ru/CNF, (
%Ru/CNF, (~) 0.5%Ru/N-CNF, (
) 2%Ru/N-CNF, (*) 5%Ru/N-CNF. The
dotted line represents values at thermodynamic equilibrium.
2
conversion, B) CH
4
yield, and C) CO yield for different Ru-
&
) 2%Ru/CNF, (*)
5
&
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1348
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
around 620 K. If we compare the two supports, 5%Ru/N-CNF
provided a higher conversion than 5%Ru/CNF. The increase of
conversion as the metal loading increases can be rationalized
easily based on the larger number of sites to activate CO2.
However, the effect of loading on selectivity is more noticeable
and not that straightforward to explain. The selectivity to CH4
increases substantially as the metal loading increases (Fig-
of CO2 at room temperature on the catalyst prereduced at
723 K and flushed with Ar at room temperature to remove spe-
cies physisorbed weakly. All the catalysts showed similar TPD
profiles. Representative TPD profiles that correspond to 5%Ru/
N-CNF are shown in Figure 2. The formation of H O peaked at
2
ure 1B). The highest selectivity to CH is achieved with 5%Ru/
4
N-CNF, which exhibited a selectivity to CH between 90–97%
4
in all the range of studied temperatures (up to 670 K). The
onset of CO occurs at around the same temperature (~520 K)
for all catalysts (Figure 1C). The CO yield increases as the tem-
perature increases, and a peak maximum is observed at tem-
peratures around that of the onset of CH4 evolution (Fig-
ure 1B). At temperatures above the onset of CH , the CO yield
4
decreases or remains stable. For this reason, with the catalyst
for which the onset in CH evolution occurs at the lowest tem-
4
perature, that is, 5%Ru/N-CNF, the CO yield remains at negligi-
ble values in the full range of temperatures. For the highest
tested temperatures, some catalysts exhibited a slight increase
2
Figure 2. TPD of preadsorbed CO at 298 K on 5%Ru/N-CNF.
of CO and a decrease of CH , which may be attributed to the
4
thermodynamic equilibrium. As the hydrogenation of CO is
2
À1
highly exothermic (1H=À164 kJmol ), at higher temperatures
400 K, which indicates that hydrogen and oxygen were ad-
the thermodynamic equilibrium favors the steam reforming of
sorbed on the catalyst surface. H should come from the disso-
2
CH and the reverse water gas shift, which are endothermic
ciative adsorption of H on the metal surface during the reduc-
4
2
[
29]
processes and lead to the production of CO. Thus, these re-
actions are responsible for the small increase of the CO yield at
the highest temperatures. We decided to keep reaction tem-
peratures below 673 K to avoid these undesired reactions and
the gasification of the support that was found for tempera-
tures >650 K in the temperature-programmed surface reaction
tion step that was not removed on flushing with Ar. Oxygen
should come from the dissociative adsorption of CO into ad-
2
sorbed oxygen (O ) and adsorbed carbon monoxide (CO ) on
ad
ad
the catalyst surface. A very small amount of CH seems to be
4
formed from the reaction of chemisorbed hydrogen (H ) and
ad
COad species. The CO2 profile shows desorption peaks at
around 350 K and at 700 K, which may be attributed to physi-
sorbed CO and some CO -desorbing oxygenated groups, re-
(TPSR) experiment shown below.
Turnover rates, CH productivity, and space time yields (STY)
4
2
2
at 673 K for all the tested catalysts are summarized in Table 1.
spectively, formed in the CO adsorption stage. There is no CO
2
desorption in the range of temperatures, which indicates that
COad species formed by CO2 dissociation are adsorbed very
strongly on the catalyst/support.
Table 1. Turnover rates at two temperatures and space time yield at
673 K
To shed more light onto the CO reduction mechanism, we
2
performed TPSR experiments (Figure 3). The formation of H O
2
Catalyst
Turnover rate
at 673 K
CH
at 673 K
[molCH4 mol
4
productivity
STY
at 673 K
catÀ1 À1
[kgCH4 kg h ]
at low temperatures (peak around 370 K in Figure 3B and D)
indicates that CO flushed before TPSR dissociated as O and
RuÀ1 À1
RuÀ1 À1
[molconv mol
s
]
s
]
2
ad
0
2
5
0
2
5
.5%Ru/CNF
%Ru/CNF
%Ru/CNF
.5%Ru/N-CNF 0.26
%Ru/N-CNF
%Ru/N-CNF
0.30
0.07
0.05
0.04
0.03
0.05
0.06
0.02
0.05
0.13
0.37
1.42
0.20
0.23
1.54
COad species on the Ru nanoparticle surface in agreement with
TPD. In TPSR, Oad species generate H O readily by reaction with
2
Had species that come from the dissociation of H present in
2
the gas phase.
0.06
0.05
A peak that corresponds to CH formation is observed in the
4
traces shown in Figure 3A and C. The calibration of the CH4
signal and the integration of the peak after the baseline is sub-
The catalysts that contained 0.5% Ru exhibited the highest
tracted enabled the quantification of desorbed CH (Table 2).
4
turnover, but the CH productivity per mole of Ru is in the
The amount of desorbed CH gives an indication of the CO
2
4
4
same range as that for the other Ru loadings. The STY per cata-
lyst weight increases proportionally as the loading increases.
Thus, the highest STY value at 673 K was achieved using
that has been dissociated and retained on the catalyst surface
during CO saturation at room temperature before the TPSR
2
experiment. For catalysts that evolve CH at lower tempera-
4
5
%Ru/N-CNF as the catalyst.
tures, that is, those with higher Ru loadings, the quantified CH4
can be ascribed unambiguously to the hydrogenation of COad
species, which were adsorbed previously. For catalysts that
To gain an insight into the reaction mechanism, transient ex-
periments were performed. First, we performed temperature-
programmed desorption (TPD) in Ar flow after the adsorption
contain the lowest loading (0.5 wt%), as the CH peak occurs
4
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1349
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 3. Signals of A, C) CH
4
and B, D) H
2
O in TPSR experiments on A, B) Ru on CNF and C, D) Ru on N-CNF with different metal loadings.
a H O desorption peak, which in-
2
Table 2. Quantification of TPSR and catalytic parameters.
dicates that the formation of at
least part of CH occurs via an
4
Catalyst
Temperature
of CH peak
K]
Molar Ru
loading
[mmolg
CH
4
per catalyst
4
Molar CH /Ru ratio
oxygenated intermediate such
as COad.
4
weight in TPSR
[mmolg
À1
À1
[
]
]
The transient behavior after
0
2
5
0
2
5
.5%Ru/CNF
%Ru/CNF
%Ru/CNF
.5%Ru/N-CNF
%Ru/N-CNF
%Ru/N-CNF
606
557
500
575
559
481
0.049
0.20
0.50
0.05
0.19
0.49
3.75
2.37
4.08
1.93
1.16
3.14
76
12
8.2
39
5.8
6.3
the sudden removal of CO from
2
the gas feed at three reaction
temperatures for some selected
catalysts is shown in Figure 4.
Similar results for the rest of the
catalysts are provided in Figur-
at high temperatures, it cannot be completely ruled out that
es S2 and S3. It is a general behavior of the catalysts that both
CO and CO signals decay as soon as CO is removed from the
part of the quantified CH stems from the gasification of the
4
2
2
support. If we compare the CH produced in TPSR with the
gas feed. This suggests that the produced CO is released to
4
amount of Ru present on the catalyst (Table 2), it is apparent
that the amount of CO2 retained on the catalyst surface is
almost one order of magnitude higher than the amount of Ru,
which suggests that the support participates in the storage of
the gas phase directly by the decomposition of CO on the
2
metal catalyst surface. However, CH desorption shows a tail,
4
which indicates that it is formed by a multistep process in
which adsorbed intermediates are further hydrogenated until
CH4 is released. Notably, the CH4 concentration in the gas
COad that comes from CO dissociation. The TPSR of the pris-
2
tine supports after CO preadsorption did not show desorption
phase increases if the CO and CO concentrations decay. The
2
2
of any gas at temperatures below 673 K (Figure S1), which indi-
cates that the support needs the participation of Ru nanoparti-
CH concentration reaches a maximum and then it declines be-
4
cause adsorbed CH -generating intermediates are depleted
4
cles to dissociate CO into species that are subsequently trans-
from the catalyst surface. The CH4 peak is less intense at
higher reaction temperatures. At the same reaction tempera-
tures, the maximum is more pronounced for 5%Ru/CNF (Fig-
ure 4B) than for 5%Ru/N-CNF (Figure 4A), and at a lower load-
ing (2 wt% Ru, Figure 4C) the maximum is more intense than
that of the higher loading (5 wt% Ru, Figure 4B). The appear-
2
ferred to the support.
The peak of CH4 formation in the TPSR trace (Table 2 and
Figure 3A and C) occurs at decreasing temperatures as the Ru
loading increases, which is in agreement with the increasing
selectivity to CH . Moreover, the CH peak is concomitant with
4
4
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1350
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
tent with the fact that the coverage of the metal by adsorbed
species is lower at higher reaction temperatures, which leaves
more sites for H chemisorption.
2
The H O signal shown in Figure 4 (dashed line) continues at
2
the same concentration long after the removal of CO2 from
the gas feed. Afterwards, the H O concentration decays to zero
2
(
Figure 4C and Figure S3A and B). The continuation of the H O
2
formation after CO removal could be because of the reaction
2
of accumulated Oad species with Had species that are fed con-
tinuously. The duration of H O formation thus has a direct rela-
2
tionship with the amount of reactive Oad species on the cata-
lyst/support surface generated during the previous CO reduc-
2
tion reaction. The duration of H O formation increases as the
2
reaction temperature and Ru loading increase. These two fac-
tors favor higher CO conversions and hence the build-up of
2
a higher amount of Oad species. The duration is also longer for
CNF-supported catalysts than for N-CNF-supported ones (Fig-
ures S2 and S3), which may suggest that CNF can lodge more
adsorbed Oad species than N-CNF. A reaction pathway is sug-
gested that is consistent with the results gathered during the
transient experiments.
To assess the stability, the catalyst with the highest activity,
5
%Ru/N-CNF, was tested at the temperature of maximum CH4
productivity, 623 K, over 20 h (Figure 5). The conversion in-
Figure 5. Long-term reaction stability tests with two catalysts that contain
different Ru loadings (0.5 and 5 wt% Ru on N-CNF) and different selectivities
Figure 4. Experiments of transient response to CO
tion mixture at three reaction temperatures for three selected catalysts:
A) 5%Ru/N-CNF, B) 5%Ru/CNF, C) 2%Ru/CNF. CO concentration, thin solid
line; CH concentration, thick solid line; CO concentration, dotted line; H
2
removal from the reac-
at 623 K reaction temperature. ( ) CH
4
yield [%], (~) CO yield [%].
&
2
creased slightly with time on stream, and the selectivity to CH4
remained constant (>95%). Additionally, a long-term test was
performed using a catalyst with the lowest Ru content,
4
2
O
concentration, dashed line. Feed gas composition: 5% CO
balance.
2
, 15% H
2
, Ar to
0.5%Ru/N-CNF. This catalyst also exhibited an increase in con-
ance of the CH peak seems to indicate that the rate-determin-
version, and the initial high selectivity to CO even increased
further. Therefore, the catalysts are highly stable, and the dif-
ference between their selectivities was accentuated after the
long-term test.
4
ing step for CH4 formation is the activation of H2 and the
supply of four Had species to reduce the COad intermediate. If
CO is removed from the feed gas, more Ru adsorption sites
2
become available for the dissociative adsorption of H . This
2
favors the supply of Had chemisorbed species to the COad inter-
mediate, which enhances CH4 formation. This is consistent
with the inverse relationship between the intensity of the CH4
peak and the conversion at steady state, that is, the peak is
less intense for 5%Ru/N-CNF than for 5%Ru/CNF, for higher
loadings, and for higher reaction temperatures. This is consis-
TEM characterization
We characterized the Ru particle size of the prepared catalysts
after a reduction step at 723 K by scanning transmission elec-
tron microscopy (STEM; Figure 6). For all the catalysts, more
than 90% of the observed Ru particles had diameters smaller
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1351
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 6. STEM images and particle size distributions of the different catalysts after a reduction step at 723 K. A) 0.5%Ru/CNF, B) 0.5%Ru/N-CNF, C) 2%Ru/CNF,
D) 2%Ru/N-CNF, E) 5%Ru/CNF, F) 5%Ru/N-CNF.
than 1.5 nm, and no particles larger than 3.5 nm were found. Discussion
Metal particles of sizes that ranged from 2–3.5 nm were absent
in the catalysts that contained 0.5 wt% Ru and were only be-
tween 4–9% of the total number of particles in the catalysts
that contained 2 and 5 wt% Ru loadings. These two latter cata-
The mechanism of CO reduction to CH on oxide-supported
2 4
metal catalysts has been debated widely in the literature.
Some authors propose that COad is a key intermediate in the
lysts showed very different selectivities to CH , despite their
CO methanation reaction, which is subsequently hydrogenat-
4
2
very similar particle size distribution. The difference between
these two catalysts is the spatial distribution of the nanoparti-
cles observed in the STEM images. For 2 wt% Ru loading, the
nanoparticles are distributed sparsely, whereas the Ru nanopar-
ticles are closer to each other for 5 wt% Ru loading, which
leaves less support space between them.
ed by the mechanism suggested for CO methanation. COad can
be formed by the reverse water gas shift via a formate inter-
[17]
mediate
or by dissociative CO2 adsorption in a redox
The formation of CH from CO was proposed to
[41–43]
center.
4
ad
[44–47]
proceed either by the initial CÀO bond breaking
or with
the association of H with CO to form an intermediate and
2
ad
48,49]
[
The catalysts recovered after the stability test for 20 h
shown in Figure 5 were also characterized by STEM (Figure S4).
The particle size distribution did not show significant change
after long-term operation at 623 K, which is in line with the
stable catalytic performance.
subsequent bond breaking.
Based on the results of the transient experiments conducted
here, the following reaction mechanism is proposed. CO is dis-
2
sociated on reduced Ru nanoparticles even at room tempera-
ture, and Oad species and COad species spill over to the CNF
support close to the interphase of metal nanoparticles. The dis-
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1352
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
sociative chemisorption energies calculated in Ref. [50] for CO,
port. The concentration of Ru on the perimeter was estimated
CO , and H on Ru are À1.62, À1.09, and À0.77 eV, respectively,
from the catalyst dispersion (Table S1). The CH yield exhibits
2
2
4
with respect to molecules in a vacuum. Therefore, CO is chemi-
sorbed less strongly than CO and H on Ru nanoparticles and
a quasilinear relationship with the concentration of Ru on the
perimeter, especially for the catalyst supported on CNF
(Figure 7). The similar importance of the metal catalyst perime-
2
2
it can be displaced by the reactants towards the metal–sup-
port interphase, which favors the proposed spill-over mecha-
nism. The COad species might be stored on the support be-
cause of some interaction with the edges of graphitic basal
planes of CNF. The accumulation of Oad species on the support
is evidenced by the formation of water at around 370 K in the
TPSR experiments (Figure 3) and by the sustained formation of
H O in the transient response experiments after CO2 is re-
2
moved from the gas feed (Figure 4). The quantified amount of
CH produced in TPSR, which is one order of magnitude larger
4
than the amount of Ru, indicates that COad intermediate spe-
cies accumulated on the carbon support as well. The presence
of Ru nanoparticles is necessary to dissociate CO as no species
2
were desorbed in TPSR using pristine supports (Figure S1).
Some of the CH -generating intermediates are partially oxi-
4
dized as inferred from the synchronous evolution of CH and
4
Figure 7. CH
perimeter for different loadings of Ru nanoparticles supported either on
CNF ( ) or N-CNF ( ).
4
yield at 673 K as a function of the concentration of Ru on the
H O in TPSR. The nature of this CH -generating intermediate
2
4
species is not clarified yet. Some authors have proposed the
&
&
formation of reversible bicarbonates by the reaction of CO
2
[
31,39]
with the OHÀ groups of the Al O support.
Similarly, bicar-
2
3
bonate species may be formed on OHÀ groups present on the
edges of the CNF basal planes. Conversely, the mechanism of
bicarbonate formation would hardly occur in the case of the
CNT support because exposed basal planes of CNT have fewer
defects for COad chemisorption. This would explain the negligi-
ter has been reported previously for CO /CH reforming to
2 4
[26]
syngas using a Pt/ZrO catalyst. It is claimed that CO is acti-
2
2
vated by carbonate species on the support that must be in the
proximity of the Pt particles to react with the methane activat-
ed on the metal.
[40]
ble activity if CNTs are used as a Pd catalyst support.
According to the reaction mechanism proposed based on
the results of the transient experiments, the support plays
a crucial role in the reaction. Therefore, a longer interphase pe-
rimeter between the metal nanoparticles and the support
The catalytic results showed that the selectivity pattern de-
pends strongly on the metal loading. The selectivity is steered
either towards CO for low Ru loadings or towards CH4 for
higher Ru loadings. Several authors found similar selectivity
patterns as a function of the metal loading for different cata-
could be the reason for the different CH selectivities. As the
4
supply of Had to the adsorbed COad intermediate was found to
be the rate-determining step in transient response experi-
ments (Figure 4), the closer proximity of the Ru nanoparticles
and larger interphase perimeter may also favor the supply of
Had to the COad intermediate and hence boost the formation of
CH . The CH peak upon CO removal in transient experiments
[
30]
[40]
lysts such as Ru on alumina,
Pd on alumina,
or Ni on
[
10]
MCM-41. Some of these authors attributed it to the different
metal particle size, nano-sized metal clusters (2–5 nm) in 10%
Pd on Al O and dispersed atomically for 0.5% Pd on
2
3
[30,40]
[10]
Al O .
In another case, sub-nanometer Ni clusters were
2
3
4
4
2
reported irrespective of the metal loading, and the size did not
change after reaction. Conversely, other authors observed an
almost vanished for the catalyst that exhibited the highest ac-
tivity at steady state, which indicates that H activation and H
2
ad
increase of CH selectivity for smaller particles using Pd nano-
supply is not that rate-limiting in these cases. The overall reac-
tion is governed by a subtle balance of adsorbed molecules,
dissociated species, and the reaction between them. Therefore,
it is not straightforward to gather the whole process into one
picture. For clarity, the simplified scheme shown in Figure 8 il-
lustrates how different Ru loadings can affect the selectivity
pattern. For the lower loadings (A), the metal particles are
more separated and the supply of four Had atoms to the acti-
vated COad intermediate on the support is hindered. In con-
trast, for the highest loading, the separation between nanopar-
ticles is smaller and the supply of four Had atoms is enhanced.
The 5 wt% Ru catalyst supported on N-CNF outperforms its
counterpart supported on CNF. As a result of the different fac-
tors involved in this reaction, further research is needed to un-
ravel the exact reason of the enhanced performance of the N-
4
particles embedded in porous silica and ascribed this behavior
[51]
to the increased amount of corner and edge atoms.
No
effect of particle size on the selectivity to CH was observed if
4
[
52]
a nanoparticle model Co catalyst on silica was used. There-
fore, it seems that there is not a particle size effect that can be
generalized for all catalytic systems. In our case, we cannot at-
tribute the different selectivity pattern to different Ru particle
sizes because the differences in the particle size distribution
measured by STEM are inappreciable, especially between the 2
and 5 wt% Ru loaded catalysts. The main difference between
these catalysts observed by STEM is that the Ru nanoparticles
are closer for the highest Ru loading. As the apparent particle
size is very similar for all samples, those with higher Ru load-
ings also have a longer perimeter of interphase with the sup-
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1353
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 3. Comparison of CH
4
productivity of this catalyst with some of the
best catalysts reported in the literature tested under similar conditions.
Entry Reference Catalyst
Loading
T
[K]
CH
À1
metal
[molh g ]
4
productivity
À1
[
wt%]
1
2
3
4
5
6
7
this work
[19]
[30]
[54]
[23]
Ru on N/CNF
Pd-Mg on SiO
5
6.2
5
23
1
5
643 1.9
723 0.5
643 1.7
523 0.8
423 0.4
723 0.05
673 0.05
2
Ru on Al
2 3
O
Ni-Fe on Al
2
O
3
Rh on Al
2
O
3
[18]
[55]
Ru on CeO
Ni on USY
2
14
and selectivity reported for catalysts supported on carbon
nanotubes. Moreover, 5%Ru/N-CNF showed a CH productivity
4
comparable to the best catalysts supported on metal oxides
reported in the literature.
Thus, the catalyst can be used efficiently in a CO recycling
2
process situated near to where renewable or byproduct H can
2
Figure 8. Different selectivity patterns depending on the different spatial
separation of the Ru nanoparticles for the several Ru loadings. A) Hindered
supply of Had for 0.5 and 2 wt% loaded catalyst. B) Enhanced supply of Had
to 5 wt% loaded catalyst.
be utilized.
To gain insights into the reasons of this outstanding per-
formance, catalytic transient response experiments were per-
formed. Transient experiments underscore the active participa-
tion of the CNF and N-CNF support on the reaction to store
the reaction intermediate species. Thus the reaction occurs
likely close to the interphase between Ru metal nanoparticles
and the support. The Ru nanoparticles assist the dissociation
of H to H , CO to O , and a reduced intermediate CO spe-
CNF-supported catalyst. Transient response experiments (Fig-
ure 4A and B) seem to indicate that the supply of Had active
species is favored for the N-CNF-supported catalyst. Additional-
ly, the CH desorption peak in TPSR of 5%Ru/N-CNF occurs at
4
2
ad
2
ad
ad
approximately 20 K lower than that of 5%Ru/CNF, which could
be a result of the enhanced supply of Had or the formation of
cies. The dissociated species spill over to the carbon support,
which stores the Oad and COad species. Transient experiments
revealed that the CH4 formation rate is controlled by the
supply of four Had atoms to the COad intermediate. The selec-
tivity depends strongly on the metal loading. For 0.5 and
2 wt% Ru loadings, the reduction is steered mainly to CO for-
more reactive COad intermediates upon CO dissociation.
2
The 5 wt% Ru on N-CNF catalyst showed an activity compa-
rable to that supported on metal oxide supports. The reaction
reaches thermodynamic equilibrium conversion at 623 K and
complete selectivity to CH4 at
a
high space velocity
mation, whereas for 5 wt% Ru loading, the selectivity to CH is
4
À1
(
19000 h ). The catalyst is stable during long-term operation,
97%. As the Ru particle sizes did not differ significantly for the
catalysts that contain different loadings, the different selectivi-
ty pattern cannot be ascribed to structure sensitivity. The
and the Ru particle size distribution remains constant after
long-term testing because of the strong attachment of Ru
nanoparticles to the carbon support. It has been reported re-
cently that Ru nanoparticles on CNT are thermally stable and
maintain a high dispersion because of their anchoring by ace-
tate ligands, which are reconstructed after thermal treatment
higher selectivity to CH is most likely explained by the en-
4
hanced supply of the four Had atoms to the activated COad in-
termediate, which is favored by the proximity between nano-
particles or to the longer interphase perimeter between the
metal nanoparticles and support.
[
53]
in the presence of oxygen. Not all the metals supported on
CNT showed stability in CO methanation. For instance, a Ni
2
supported on CNT catalyst tested in CO methanation at 623 K
2
for 6 h showed a decrease of yield from 75 to 62%, which was Experimental Section
[
37]
attributed to Ni sintering.
Catalyst preparation
The catalyst here prepared not only outperforms other cata-
lysts supported on carbon materials but also exhibits superior
CNF were grown on a 20 wt% Ni on alumina catalyst prepared by
incipient wetness impregnation of nickel nitrate on alumina (Pural,
Sasol) and calcined at 873 K in N2 flow. The powder catalyst was
placed on a porous frit of a vertical reactor. After the catalyst was
CH productivity than some of the best metal-oxide-supported
4
catalysts tested under similar conditions (Table 3).
À1
reduced at 823 K under 100 mLmin of a H /N mixture for 1 h,
2
2
À1
Conclusions
the CNF growth was performed at 873 K under 100 mLmin of
a C H /H mixture (50:50). N-CNF were grown on a 20 wt% Fe on
2
6
2
Ru supported on carbon nanofibers (CNF) or nitrogen-doped
carbon nanofibers (N-CNF) showed remarkable activity and sta-
bility in CO reduction to CH , in contrast with the poor activity
alumina catalyst prepared by incipient wetness impregnation of
iron nitrate on alumina (Pural, Sasol) and calcined at 873 K in N
2
À1
flow. The catalyst was reduced at 823 K under 100 mLmin of
2
4
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1354
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
a H /N (50:50) mixture for 1 h. Subsequently, the reactor was
heated to the growth temperature (1023 K) under Ar. After the
all the changes in the MS signals reflected the changes of the
gases in contact with the catalyst accurately.
2
2
temperature was reached, the reaction mixture was admitted. The
À1
TPD and TPSR experiments were conducted as follows. The catalyst
reaction mixture consisted of 100 mLmin
Ar mixed with
À1
À1
was heated to 723 K at a heating rate of 10 Kmin in inert gas. At
0
.15 mLmin of ethylendiamine (10 mL in total) fed using a sy-
À1
this temperature, the catalyst was reduced with 100 mLmin of
ringe. The ethylenediamine/argon mixture passed through an
evaporator at 473 K and a heated line to the reactor that contained
the growth catalyst.
a H /N mixture for 0.5 h. The reactor was allowed to cool to 298 K
2
2
and CO2 was flushed for 0.5 h. The gas was switched to
À1
1
00 mLmin Ar for 1 h to remove CO physisorbed weakly. Then
2
À1
Both CNFs and N-CNFs supports were purified from the growth
catalyst under reflux of NaOH (5m) for 4 h at 353 K and later under
reflux of HCl at 373 K for another 4 h. After HCl treatment, the ma-
terial was rinsed thoroughly with distilled water until the filtrate
was neutral. After this purification process, the residual catalyst
was less than 1 wt% as determined by oxidation by using a ther-
mobalance and no HCl traces were detected by X-ray photoelec-
tron spectroscopy (XPS).
the gas was adjusted to 60 mLmin of Ar for TPD experiments or
À1
to 60 mLmin 15% H in Ar for TPSR experiments. If the signal in
2
the mass spectrometer was stabilized, the temperature was in-
À1
creased to 723 K at a rate of 10 Kmin and the desorbed gases
were monitored.
Ru metal nanoparticle size on CNF was measured by STEM by
using a FEI TECNAI F30 electron microscope equipped with Gatan
energy filter and cold field-emission gun (FEG) operated at 300 kV
with a 1.5 Š lattice resolution. TEM specimens were prepared by ul-
trasonic dispersion in ethanol of powder catalyst. A drop of the
suspension was applied to a holey carbon support grid. The parti-
cle size distribution was calculated by statistical analysis of 400 par-
ticles in ~20 images on CNFs. The mean Ru particle size evaluated
The preparation of Ru catalysts was performed by incipient wet-
ness impregnation. The CNF or N-CNF supports were crushed to
powders of a particle size smaller than 200 mm. The desired
amount of Ru(NO)(NO ) was dissolved in an ethanol/water mixture
3
(
4:1) and impregnated in the catalysts to achieve different Ru load-
¯
ings (0.5, 2, 5 wt%) with respect to the total sample. After drying,
as the surface area weighted diameter (d ) was computed accord-
Ru
the catalyst was first calcined in N at 723 K at a heating rate of
ing to Equation (1):
2
À1
1
Kmin and subsequently reduced in H at the same temperature
2
and heating rate. The reduced catalysts are denoted as loading%
Ru/support. The actual Ru content on the catalyst was analyzed by
inductively coupled plasma optical emission spectroscopy (ICP-
OES).
in which n represents the number of particles with diameter
i
d(Sn ꢀ400)
i
i i
Catalytic tests and characterization
Catalytic testing was performed by using a continuous-flow 4 mm
i.d. quartz reactor inside a vertical furnace with a temperature con-
troller (Eurotherm). Catalyst (50 mg) diluted in SiC was placed on
quartz wool inside the reactor. Before the catalytic test, the catalyst
was heated to 723 K in a N2 flow using a heating ramp of
Acknowledgements
The financial support of the European Commission (FREECATS
project, FP7 Grant agreement number 280658), the Spanish Min-
istry MINECO and the European Regional Development Fund
À1
À1
1
7
0 Kmin and reduced with 100 mLmin of H /N2 (50:50) at
23 K for 0.5 h. The reaction temperature was controlled by using
2
(
project ENE2013-48816-C5-5-R), and the Regional Government of
a thermocouple inside the catalytic bed. The reaction conversion
Aragon (DGA-ESF-T66 Grupo Consolidado) is gratefully acknowl-
edged.
À1
and selectivities were recorded at steady state using 60 mLmin
of a reaction mixture consisting of 5% CO , 15% H , and Ar to bal-
2
2
À1
ance. This flow rate gives rise to a space velocity of 19000 h . Gas
analysis was performed by using a Pfeiffer vacuum mass spectrom-
eter. The following m/z signals were recorded: 2, 16, 18, 28, 40, 44.
The signals of the gases were calibrated by taking into account the
baseline of Ar and the fragmentation pattern of each mass. The
Keywords: hydrogenation · doping · nanoparticles · reaction
mechanisms · ruthenium
main m/z signals used for each gas were m/z=2 (H ), 16 (CH ), 18
2
4
[1] G. Centi, S. Perathoner, Greenhouse Gases Sci. Technol. 2011, 1, 21–35.
[2] Energy Alliance for Hydrogen from Wind, http://www.performing-ener-
gy.de, 2014.
(
H O), 28 (CO), 40 (Ar), and 44 (CO ). The concentration of CO was
2 2
calculated by subtracting the contribution of CO to m/z=28. The
2
concentration of CH was calculated by subtracting the contribu-
[3] G. A. Olah, Angew. Chem. Int. Ed. 2005, 44, 2636–2639; Angew. Chem.
4
2
005, 117, 2692–2696.
tion of CO and CO to m/z=16. The correct calibration of the mass
2
[
4] K. Hashimoto, M. Yamasaki, K. Fujimura, T. Matsui, K. Izumiya, M.
Komori, A. A. El-Moneim, E. Akiyama, H. Habazaki, N. Kumagai, A. Kawa-
shima, K. Asami, Mater. Sci. Eng. A 1999, 267, 200–206.
5] P. Sabatier, Compt. Rend. Heb. Acad. Sci. 1902, 134, 689–691.
6] G. Centi, E. A. Quadrelli, S. Perathoner, Energy Environ. Sci. 2013, 6,
spectrometer was confirmed by analyzing the gases by using a cali-
brated Agilent Micro GC 3000 A. Stability tests were conducted
under the same conditions but leaving the reaction overnight.
[
[
The transient response behavior of the reactant and products to
1711–1731.
the sudden removal of CO from the reactant gas under isothermal
2
[
[
[
7] E. V. Kondratenko, G. Mul, J. Baltrusaitis, G. O. Larrazabal, J. Perez-Ramir-
ez, Energy Environ. Sci. 2013, 6, 3112–3135.
8] F. J. Fernµndez-Alvarez, A. M. Aitani, L. A. Oro, Catal. Sci. Technol. 2014,
reaction conditions was studied. To this end, after the reaction at
steady state using a feed gas consisting of 5% CO , 15% H , and
2
2
Ar to balance was recorded, CO was removed suddenly from the
2
4, 611 –624.
reaction mixture and the total flow rate was kept constant by com-
pleting the balance with Ar. The dead volume at the outlet line to
the mass spectrometer was determined to be negligible. Therefore,
9] J. M. Thomas, ChemSusChem 2014, 7, 1801–1832.
[10] G. Du, S. Lim, Y. Yang, C. Wang, L. Pfefferle, G. L. Haller, J. Catal. 2007,
249, 370–379.
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1355
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[
11] F. Studt, I. Sharafutdinov, F. Abild-Pedersen, C. F. Elkjær, J. S. Hummel-
[35] C. de Leitenburg, A. Trovarelli, J. Ka sˇ parb, J. Catal. 1997, 166, 98–107.
[36] M. C. Romµn-Martínez, D. Cazorla-Amorós, A. Linares-Solano, C. S.-M.
de Lecea, Appl. Catal. A 1994, 116, 187–204.
shøj, S. Dahl, I. Chorkendorff, J. K. Nørskov, Nat. Chem. 2014, 6, 320–
324.
[
[
[
[
12] S. Tada, T. Shimizu, H. Kameyama, T. Haneda, R. Kikuchi, Int. J. Hydrogen
Energy 2012, 37, 5527–5531.
13] M. Tsuji, T. Kodama, T. Yoshida, Y. Kitayama, Y. Tamaura, J. Catal. 1996,
[37] Y. Feng, W. Yang, S. Chen, W. Chu, Integr. Ferroelectr. 2014, 151, 116–
125.
[38] J. P. O’Byrne, R. E. Owen, D. R. Minett, S. I. Pascu, P. K. Plucinski, M. D.
Jones, D. Mattia, Catal. Sci. Technol. 2013, 3, 1202–1207.
[39] C. J. Keturakis, F. Ni, M. Spicer, M. G. Beaver, H. S. Caram, I. E. Wachs,
ChemSusChem 2014, 7, 3459–3466.
[40] J. H. Kwak, L. Kovarik, J. Szanyi, ACS Catal. 2013, 3, 2094–2100.
[41] K. H. Ernst, C. T. Campbell, G. Moretti, J. Catal. 1992, 134, 66–74.
[42] L. F. Liotta, G. A. Martin, G. Deganello, J. Catal. 1996, 164, 322–333.
[43] W. Wang, S. Wang, X. Ma, J. Gong, Chem. Soc. Rev. 2011, 40, 3703–3727.
[44] A. Bell, Catal. Rev. Sci. Eng. 1981, 23, 203–232.
1
64, 315–321.
14] G. Melaet, W. T. Ralston, C. S. Li, S. Alayoglu, K. An, N. Musselwhite, B.
Kalkan, G. A. Somorjai, J. Am. Chem. Soc. 2014, 136, 2260–2263.
15] R. E. Owen, J. P. O’Byrne, D. Mattia, P. Plucinski, S. I. Pascu, M. D. Jones,
Chem. Commun. 2013, 49, 11683–11685.
[
16] N. M. Gupta, V. S. Kamble, K. A. Rao, R. M. Iyer, J. Catal. 1979, 60, 57–67.
17] M. R. Prairie, A. Renken, J. G. Highfield, K. Ravindranathan Thampi, M.
Grätzel, J. Catal. 1991, 129, 130–144.
[
[
18] S. Sharma, Z. Hu, P. Zhang, E. W. McFarland, H. Metiu, J. Catal. 2011,
[45] D. W. Goodman, R. D. Kelley, T. E. Madey, J. Yates, J. Catal. 1980, 63,
226–234.
278, 297–309.
[
[
[
[
[
[
19] J. N. Park, E. W. McFarland, J. Catal. 2009, 266, 92–97.
[46] N. M. Gupta, V. P. Londhe, V. S. Kamble, J. Catal. 1997, 169, 423–437.
[47] H. H. Nijs, P. A. Jacobs, J. Catal. 1980, 66, 401–411.
[48] M. P. Andersson, F. Abild-Pedersen, I. N. Remediakis, T. Bligaard, G.
Jones, J. Engbæk, O. Lytken, S. Horch, J. H. Nielsen, J. Sehested, J. R.
Rostrup-Nielsen, J. K. Nørskov, I. Chorkendorff, J. Catal. 2008, 255, 6–19.
[49] I. A. Fisher, A. T. Bell, J. Catal. 1996, 162, 54–65.
20] A. V. Boix, M. A. Ulla, J. O. Petunchi, J. Catal. 1996, 162, 239–249.
21] A. Boffa, C. Lin, A. T. Bell, G. A. Somorjai, J. Catal. 1994, 149, 149–158.
22] C. Deleitenburg, A. Trovarelli, J. Catal. 1995, 156, 171–174.
23] M. Jacquemin, A. Beuls, P. Ruiz, Catal. Today 2010, 157, 462–466.
24] S. Scir›, C. Crisafulli, R. Maggiore, S. Minicò, S. Galvagno, Catal. Lett.
1
998, 51, 41–45.
[50] T. Bligaard, J. K. Nørskov, S. Dahl, J. Matthiesen, C. H. Christensen, J. Se-
hested, J. Catal. 2004, 224, 206–217.
[
25] E. E. Santiso, A. M. George, C. H. Turner, M. K. Kostov, K. E. Gubbins, M.
Buongiorno-Nardelli, M. Sliwinska-Bartkowiak, Appl. Surf. Sci. 2005, 252,
[51] J. Martins, N. Batail, S. Silva, S. Rafik-Clement, A. Karelovic, D. P. Debeck-
er, A. Chaumonnot, D. Uzio, Catal. Commun. 2015, 58, 11–15.
[52] G. Melaet, A. Lindeman, G. Somorjai, Top. Catal. 2014, 57, 500–507.
[53] B. F. Machado, M. Oubenali, M. Rosa Axet, T. Trang Nguyen, M. Tunckol,
M. Girleanu, O. Ersen, I. C. Gerber, P. Serp, J. Catal. 2014, 309, 185–198.
[54] J. Sehested, K. Larsen, A. Kustov, A. Frey, T. Johannessen, T. Bligaard, M.
Andersson, J. Nørskov, C. Christensen, Top. Catal. 2007, 45, 9–13.
[55] I. GraÅa, L. V. Gonzµlez, M. C. Bacariza, A. Fernandes, C. Henriques, J. M.
Lopes, M. F. Ribeiro, Appl. Catal. B 2014, 147, 101–110.
766–777.
[
[
26] J. H. Bitter, K. Seshan, J. A. Lercher, J. Catal. 1997, 171, 279–286.
27] D. Li, N. Ichikuni, S. Shimazu, T. Uematsu, Appl. Catal. A 1999, 180, 227–
235.
[
[
28] R. Büchel, A. Baiker, S. E. Pratsinis, Appl. Catal. A 2014, 477, 93–101.
29] C. Janke, M. S. Duyar, M. Hoskins, R. Farrauto, Appl. Catal. B 2014, 152–
153, 184–191.
[
[
[
30] J. H. Kwak, L. Kovarik, J. Szanyi, ACS Catal. 2013, 3, 2449–2455.
31] J. Szanyi, J. H. Kwak, Phys. Chem. Chem. Phys. 2014, 16, 15126–15138.
32] M. R. Prairie, J. G. Highfield, A. Renken, Chem. Eng. Sci. 1991, 46, 113–
121.
[
[
33] Y. Traa, Chem. Eng. Technol. 1999, 22, 291–293.
34] J. A. Rodriguez, J. Evans, L. Feria, A. B. Vidal, P. Liu, K. Nakamura, F. Illas,
J. Catal. 2013, 307, 162–169.
Received: January 8, 2015
Revised: January 30, 2015
Published online on April 9, 2015
ChemCatChem 2015, 7, 1347 – 1356
www.chemcatchem.org
1356
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim