Ruthenium Clusters on Carbon Nanofibers for Formic Acid Decomposition: Effect of Doping the Support with Nitrogen

Article abstract of DOI:10.1002/cctc.201500216

The catalytic properties of 1wt % Ru catalysts with the same mean Ru cluster size of 1.4-1.5nm supported on herringbone-type carbon nanofibers with different N contents were compared for H₂ production from formic acid decomposition. The Ru catalyst on the support with 6.8wt % N gave a 1.5-2times higher activity for the dehydrogenation reaction (CO₂, H₂) than the catalyst on the undoped support. The activity in the dehydration reaction (CO, H₂O) was the same. As a result, the selectivity to H₂ increased significantly from 83 to 92 % with N-doping, and the activation energies for both reactions were close (55-58kJ mol^-1). The improvement could be explained by the presence of Ru clusters stabilized by pyridinic N located on the open edges of the external surface of the carbon nanofibers. This N may activate formic acid by the formation of an adduct (>NH⁺HCOO^-) followed by its dehydrogenation on the adjacent Ru clusters. Nanoparticles on nanofibers: Ru catalysts on N-doped carbon nanofibers show a higher selectivity for hydrogen production from formic acid decomposition compared to undoped samples. This is attributed to the activation of formic acid on the pyridinic sites of the support followed by dehydrogenation on adjacent Ru clusters.

Full text of DOI:10.1002/cctc.201500216

DOI: 10.1002/cctc.201500216
Full Papers
Ruthenium Clusters on Carbon Nanofibers for Formic Acid
Decomposition: Effect of Doping the Support with
Nitrogen
Monika Zacharska,[a, b] Olga Yu. Podyacheva,[c, d] Lidiya S. Kibis,[c, d] Andrey I. Boronin,[c, d]
Boris V. Senkovskiy,[e, f] Evgeny Yu. Gerasimov,[c] Oxana P. Taran,[c] Artemiy B. Ayusheev,[c]
Valentin N. Parmon,[c, d] J. J. Leahy,[a, b] and Dmitri A. Bulushev*[a, c, d]
The catalytic properties of 1 wt% Ru catalysts with the same
mean Ru cluster size of 1.4–1.5 nm supported on herringbone-
type carbon nanofibers with different N contents were com-
pared for H2 production from formic acid decomposition. The
Ru catalyst on the support with 6.8 wt% N gave a 1.5–2 times
higher activity for the dehydrogenation reaction (CO2, H2) than
the catalyst on the undoped support. The activity in the dehy-
dration reaction (CO, H2O) was the same. As a result, the selec-
tivity to H2 increased significantly from 83 to 92% with N-
doping, and the activation energies for both reactions were
close (55–58 kJmolÀ1). The improvement could be explained
by the presence of Ru clusters stabilized by pyridinic N located
on the open edges of the external surface of the carbon nano-
fibers. This N may activate formic acid by the formation of an
adduct (>NH+HCOOÀ) followed by its dehydrogenation on the
adjacent Ru clusters.
Introduction
Formic acid can be produced from renewable biomass as a by-
product of the acid-catalyzed hydrolysis of cellulose to levulinic
acid.[1] It can also be produced as a primary product of the oxi-
dation of cellulose with oxygen using heteropoly acids and
their salts as catalysts.[2] As formic acid contains a sufficient
amount of hydrogen (4.4 wt%), it is considered as a potential
source of hydrogen for applications in fuel cells[3] or as a hydro-
gen donor for some hydrogenation reactions instead of molec-
ular hydrogen.[1b,4] Therefore, it is important to develop effec-
tive catalysts to direct the decomposition of formic acid to-
wards dehydrogenation to give H2 and CO2 as products, rather
than in the direction of dehydration to give H2O and CO, as
CO is a catalyst poison.
Often, a variety of carbon materials are used as catalyst sup-
ports for noble metals that decompose formic acid under mild
conditions. Such supports attract attention as inert and high-
surface-area materials able to stabilize metals in a highly dis-
persed state. Carbon nanofibers (CNFs) have attracted much
interest because of their exceptional structural and textural
properties and there is a possibility to control their properties
during synthesis.[5] Modification of the composition of CNFs
opens up new routes to improve the performance of a cata-
lyst.[5c,6] Recently, we showed that the doping of CNFs supports
with N led to an increase of the formic acid decomposition ac-
tivity of Pt catalysts by a factor of 10 if Pt is dispersed as 1 nm
clusters.[7] However, Pt is a very expensive metal, and cheaper
metallic catalysts are highly desirable. Ru can be considered as
such a substitute for Pt as it has demonstrated catalytic prop-
erties for formic acid decomposition comparable to that of
Pt.[8] We have also shown that Ru catalysts on N-doped CNFs
provide an improved activity and stability during the wet-air
oxidation of phenol compared to Ru catalysts on undoped
CNFs.[9] Moreover, Marco et al.[6b] discovered that Ru supported
on N-doped CNFs showed improved properties for the decom-
position of ammonia. Generally, the N-doping of carbons is
useful to create improved materials for a wide range of appli-
cations.[6a,10]
[a] M. Zacharska, Dr. J. J. Leahy, Dr. D. A. Bulushev
Chemical & Environmental Sciences Department
University of Limerick
(Ireland)
E-mail: dmitri.bulushev@catalysis.ru
[b] M. Zacharska, Dr. J. J. Leahy
Materials & Surface Science Institute
University of Limerick
(Ireland)
[c] Dr. O. Y. Podyacheva, Dr. L. S. Kibis, Prof. A. I. Boronin, Dr. E. Y. Gerasimov,
Dr. O. P. Taran, Dr. A. B. Ayusheev, Prof. V. N. Parmon, Dr. D. A. Bulushev
Boreskov Institute of Catalysis SB RAS
Novosibirsk 630090 (Russia)
[d] Dr. O. Y. Podyacheva, Dr. L. S. Kibis, Prof. A. I. Boronin, Prof. V. N. Parmon,
Dr. D. A. Bulushev
Novosibirsk State University
Novosibirsk 630090 (Russia)
[e] Dr. B. V. Senkovskiy
Dresden University of Technology
01062 Dresden (Germany)
[f] Dr. B. V. Senkovskiy
St. Petersburg State University
198504 St. Petersburg (Russia)
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201500216.
In contrast to Pt/C catalysts, Ru/C catalysts are applied
widely in reactions related to biomass conversion, especially
hydrogenation, that could use hydrogen donor molecules such
Part of a Special Issue on Carbon in the Catalysis Community. A link to the
Table of Contents will appear here once the Special Issue is assembled.
ChemCatChem 2015, 7, 2910 – 2917
2910
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
as formic acid or alcohols. Thus, Panagiotopoulou and Vla-
chos[11] showed high yields of methylfuran from hemicellulose-
derived furfural over a Ru/C catalyst at ꢀ453 K. In this case,
methylfuran was produced by the transfer hydrogenation reac-
tion using 2-propanol. Earlier, West et al.[12] applied Ru/C cata-
lysts for the hydrogenation of biomass-derived ketones and
acids to alcohols. The same group[1b] reported that the use of
Ru/C catalysts can be promising for the conversion of levulinic
acid to g-valerolactone by transfer hydrogenation with formic
acid. Similarly, Heeres et al.[4b] produced g-valerolactone from
C6 sugars by dehydration/transfer hydrogenation with formic
acid on Ru/C catalysts. Notably, methylfuran, g-valerolactone,
and alcohols can be used as motor fuel additives or as valuable
chemical intermediates. For efficient transfer hydrogenation
with formic acid, metal-based catalysts should facilitate the de-
composition of formic acid to give hydrogen at low tempera-
tures.
0.88 cm3 gÀ1. The contribution of micropores to the total pore
volume was negligibly small (<2%). These data indicate that
all the N-CNFs supports were mesoporous materials. The tex-
tural characteristics of the materials almost did not change
after the Ru deposition.
TEM characterization
TEM was used to determine the mean Ru particle size and par-
ticle size distributions for the Ru catalysts, and typical images
of the catalysts are shown in Figure 1. The catalyst support
used displays approximately the same diameter of carbon
nanofibers of ꢀ30 nm. The morphology of the fibers allows
their assignment as a herringbone-type structure that contains
a lot of open edge sites of the graphene layers.[5b,c] The distri-
butions of Ru particle sizes obtained for the catalysts with 0, 2,
and 6.8 wt% N are monomodal and narrow with mean particle
sizes of 1.4, 1.4, and 1.5 nm with almost the same standard de-
viations of 0.3, 0.2, and 0.2 nm, respectively (Table 1). The
mean Ru particle sizes correspond to 88–94% dispersion,
Although transfer hydrogenation with formic acid looks
promising for a wide set of reactions, the decomposition of
formic acid over Ru/C catalysts has been studied to a much
lesser extent than that over Pd, Pt, Au, and Ni catalysts. Soly-
mosi et al.[13] showed that the activity of a Ru/C catalyst was
higher than that of a Rh/C catalyst but lower than the activities
of Ir, Pt, and Pd catalysts supported on carbon. O’Neill et al.[14]
studied the decomposition of formic acid in the liquid phase
in the presence of water and showed that the activity of a Ru/
C catalyst was lower than that of a Pd/C catalyst. Unfortunate-
ly, all these comparisons were performed for catalysts with dif-
ferent metal dispersions.
In the present study, formic acid decomposition was investi-
gated over 1 wt% Ru catalysts supported on CNFs doped with
different amounts of N. The advantage of these catalysts was
that the Ru clusters formed possessed approximately the same
mean size of ꢀ1.4–1.5 nm on undoped and N-doped supports.
The size of the metal particles may decrease for the N-doped
carbon supports compared to that on the undoped sup-
ports.[10,15] This can complicate the understanding of the real
reasons for the N-doping effect, as structure-sensitivity effects
related to the metal particles size can mask it. In the present
paper, we show clearly that the selectivity to hydrogen in-
creased noticeably with the amount of N in the support. This
was achieved by an increase in the rate of dehydrogenation,
whereas the dehydration rate did not change. Additionally, cer-
tain differences were observed for the state of the Ru clusters
in the N-doped samples as compared to the undoped samples,
which arise from the participation of the support N sites in the
coordination of these clusters.
Results and Discussion
Characterization of the Ru/N-CNFs samples
Characterization of the supports by N2 adsorption
The BET surface areas for the N-CNFs supports with 0, 2, and
6.8 wt%
N
were 229, 303, and 209 m2 gÀ1, respectively
(Table S1). The average pore diameters were in the range of
8.2–11.7 nm, and the pore volumes were in the range of 0.47–
Figure 1. TEM images of the Ru/N-CNFs catalysts and Ru particle size distri-
butions.
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2911
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 1. N and Ru contents in the samples, mean Ru particle sizes determined by TEM, TOF for the total formic acid decomposition, and apparent activa-
tion energies (Ea) for the dehydration and dehydrogenation reactions.
Catalysts
N [wt%] Ru [wt%] Mean Ru particle size [nm] TOF at 373 K [sÀ1
]
Ea for dehydration [kJmolÀ1
]
Ea for dehydrogenation [kJmolÀ1
]
1.3Ru/0N-CNFs
1.3Ru/2N-CNFs
1Ru/6.8N-CNFs 6.8
0
2
1.3
1.3
1.0
1.4Æ0.3
1.4Æ0.2
1.5Æ0.2
0.009
0.011
0.016
57
56
56
55
57
58
which implies that almost all the Ru in the particles on the
studied supports was available for the catalytic reaction.
respectively. The contribution of this peak to the overall N1s
spectrum became stronger after reduction (45%). The impact
of the pyrrolic N peak decreased from 37 to 26% with the in-
crease of N content from 2 to 6.8 wt%. The relative content of
the graphite-like (quaternary) N was in the range of 27–29%
for all samples. The contribution of the oxidized N species de-
creased slightly from 8 to 5% after reduction.
Photoelectron spectroscopy
Photoelectron spectroscopy was used to analyze the chemical
states of N and Ru in the samples. Apart from the collection of
the spectra for the original samples, the spectra of the 1Ru/
6.8N-CNFs catalyst reduced in H2 at 623 K in the preparation
chamber of the spectrometer were used to clarify the Ru
states. N1s spectra (Figure 2) of the 1.3Ru/2N-CNFs, 1Ru/6.8N-
CNFs, and 1Ru/6.8-CNFs catalysts after reduction were fitted
Figure 2. N1s spectra of 1.3Ru/2N-CNFs (1), 1Ru/6.8N-CNFs (2), and 1Ru/
6.8N-CNFs (3) after reduction.
with four peaks with intensity maxima at binding energies of
398.6, 399.8–400.0, 400.7–400.9, and 402.0 eV that can be as-
signed to pyridinic, pyrrolic, graphite-like (quaternary), and oxi-
dized N species, respectively.[10,16] The intensity of the pyridinic
N relative to the total intensity of the N1s spectrum was 27
and 37% for the 1.3Ru/2N-CNFs and 1Ru/6.8N-CNFs samples,
Figure 3. Ru3d5/2 spectra of a) 1.3Ru/0N-CNFs (1) and 1.3Ru/2N-CNFs (2);
b) 1Ru/6.8N-CNFs (1) before and (2) after reduction (the spectra were record-
ed at hn=600 eV).
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2912
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
The Ru3d5/2 spectra for all samples are shown in Figure 3.
Following the spectral fitting, two peaks are observed at bind-
ing energies of 280.6–280.9 and 281.2–281.4 eV (Figure 3a and
b curve (1)). According to the literature data, the latter peak
can be assigned to an oxidized Ru species similar to that in
RuO2.[17] The Ru3d5/2 binding energy value obtained for the
former component is higher than the value typical for the bulk
metallic state of Ru (~280.0 eV).[9,17a,18] It is well known that the
size of particles and their shape can influence the binding
energy because of the initial and final state effects.[19] A posi-
tive shift in the binding energy can be the result of the charg-
ing of Ru clusters through their interaction with the support or
the adsorption of some carbon- or oxygen-containing species
on their surface. Thus, we can assign the peak at a binding
energy of 280.6–280.9 eV to very small metallic Ru0 clusters or
Figure 4. Dependence of I(Ru3d5/2)/I(C1s) as a function of KE.
to a slightly charged species Rud+
.
It is interesting that the highest content of these reduced Ru
species was observed for the sample with the maximum N
content (6.8 wt%). The ratio of the intensities of these species
to the oxidized species increased by ꢀ1.5 times as compared
to that of the undoped sample. A similar result was obtained
by Marco et al.[6b,20] for samples with a higher concentration of
Ru (3.5–5 wt%). These authors also observed an increase in the
content of metallic Ru species with the introduction of N into
the CNFs. We propose that the N-doped support facilitates the
formation of Ru with a higher contribution of the metallic clus-
ters or slightly charged Rud+ species in agreement with the lit-
erature data that reports the stabilization of Ru particles in a re-
duced state through electron donation from the N-modified
carbon support.[20]
A dependence of the I(Ru3d5/2)/I(C1s) ratios on the KE is
shown in Figure 4. This ratio passes through a maximum for all
samples, which indicates that the Ru clusters might be covered
with a thin layer of carbonaceous species that shields the Ru
signal. It is seen that the relative intensity of the Ru3d5/2 line
increases with the increase in N content. The I(Ru3d5/2)/I(C1s)
ratio for the Ru/6.8N-CNFs sample is substantially higher in
comparison with other samples. To explain these results, it
should be mentioned again that both the total Ru concentra-
tion and mean particle size of Ru (dispersion) on the support
obtained by TEM are very close for all samples (Table 1). An
analysis of the data indicated that there is no correlation of
the intensity of the Ru signal with the porous structure of the
N-CNFs supports (Table S1). As mentioned above, all used sup-
ports are mesoporous materials. Hence, the considerable in-
crease of the Ru content in the surface layer of the samples
with N can be explained by the stronger interaction of Ru with
N sites on the external surface of the N-CNFs bundles that pre-
vents Ru diffusion to the internal volume of the bundles and
prevents the incorporation of Ru between graphene layers,
which might take place in the undoped sample similar to that
found for Pd/CNFs catalysts.[23] Therefore, a more uniform dis-
tribution of Ru through the support in the undoped sample
should be expected. Such a binding of noble-metal clusters
with N sites is in line with recent multiple results obtained by
electron energy loss spectroscopy combined with scanning
TEM.[10,16,24]
The additional treatment of 1Ru/6.8N-CNFs with hydrogen
results in the complete reduction of the oxidized Ru species.
Only one peak was observed in the Ru3d5/2 spectrum (Fig-
ure 3b curve (2)) at a binding energy of 280.3 eV close to the
peak for the bulk metallic Ru. The shift to low binding energy
values indicates the sintering of small Ru clusters under hydro-
gen exposure. The ruthenium oxide, responsible for the peak
at a binding energy of 281.2–281.4 eV seen for the samples
that were not reduced in the spectrometer, could be formed
because of the oxidation of metallic Ru species by ambient air.
To analyze the Ru localization in the support layers, a depth
analysis with a variation of the incident photon energy was
performed. The change of hn from 400 to 1000 eV results in
a variation of the kinetic energy of photoelectrons (KE) from
100 to 700 eV. According to the literature data, such a change
of the KE leads to an increase of an inelastic mean free path
(l) from ~4 to ~11 Š,[21] which defines the surface sensitivity of
photoelectron spectroscopy.[22] The relative intensity of the
Ru3d5/2 line normalized to the intensity of the C1s line
(I(Ru3d5/2)/I(C1s) ratio) might provide information on the parti-
cle localization. If the particles are localized inside the subsur-
face layers of the support, one might expect a decreased
I(Ru3d5/2)/I(C1s) ratio. However, the localization of the particles
on the external surface of support should result in an in-
creased I(Ru3d5/2)/I(C1s) ratio. It is clear that at low KE values
(100–300 eV) a decrease of the Ru3d5/2 line intensity caused by
the screening by the support would be more pronounced.
Formic acid decomposition
The conversion of the formic acid decomposition over the Ru
catalysts with very similar metal charges, mean Ru particle
sizes, but different contents of N in the support of 0, 2 and
6.8 wt% is shown in Figure 5a. The data demonstrate that the
catalytic activity of Ru catalysts doped with N improves with
the increased N-doping of the support. Thus, it is seen at
423 K that the Ru catalyst with the highest amount of N gives
a conversion 1.5 times higher than that of the sample without
N. Similarly, Marco et al.[6b] observed an approximately two
times higher H2 production for NH3 decomposition over Ru
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2913
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
The turnover frequencies (TOFs) for the total formic acid de-
composition over the Ru catalysts at 373 K are shown in
Table 1. At 373 K and under the same conditions, the Ru/CNFs
catalyst without N demonstrated the same TOF (0.009 sÀ1) as
the undoped Pt/CNFs catalyst with a mean particle size of
1 nm.[7] Solymosi et al.[13] obtained
a
comparable TOF
(0.013 sÀ1) under similar conditions for a 2 wt% Ru/C catalyst
with a low metal dispersion (5.6%). However, they obtained
a much higher TOF (0.064 sÀ1) for a 2 wt% Pt/C catalyst with
a dispersion of 23.5%. If our Ru samples were doped with
6.8 wt% N, the TOF almost doubled, whereas that of our Pt
samples increased by an order of magnitude. The mean parti-
cle size of the Pt catalysts (ꢀ1 nm) is smaller than that of the
Ru catalysts (ꢀ1.5 nm). Moreover, the Pt catalysts contained
a substantial amount of atomically dispersed Pt and/or Pt clus-
ters with a few atoms,[7] which can be active for formic acid de-
composition in the presence of N on the support. This could
explain the more pronounced effect of the N-doping for the Pt
catalysts.
The selectivities to hydrogen do not change much with tem-
perature (Figure 5b). At 50% conversion for the 1.3Ru/0N-
CNFs, 1.3 Ru/2N-CNFs, and 1Ru/6.8N-CNFs catalysts, the selec-
tivities were 82.9, 87, and 91.5%, respectively. Hence, N incor-
poration in the N-CNFs structure improves the selectivity great-
ly. However, the selectivity of the Ru catalysts was lower than
that of the Pt catalysts on the similar supports.[7] For the un-
doped Pt catalyst and the Pt catalyst with 7.5 wt% N, the se-
lectivities at 50% conversion were 96.2 and 99.5%, respective-
ly.
Figure 5. Temperature dependences of a) formic acid conversion over the
Ru/N-CNFs catalysts and N-CNFs supports with different N contents and
b) the selectivity of H2 production.
To understand whether the water gas shift (WGS) reaction
contributes to the formic acid decomposition, this reaction
was investigated for the 1.3Ru-6.8N/CNFs catalyst. It is impor-
tant that no conversion of CO was observed over the tempera-
ture range of 473–593 K. Similarly, Basinska et al.[25] demon-
strated that a 2 wt% Ru/C catalyst with a mean Ru particle size
of 3.4 nm showed some activity but only at very high tempera-
tures (623 K). This proves that there is no significant contribu-
tion from the WGS reaction to the formic acid decomposition
over the Ru catalysts under the reaction conditions studied.
Hence, the formation of CO2 and H2 from formic acid decom-
position takes place in parallel with the formation of CO and
H2O.
catalysts on monoliths covered with N-doped CNFs than on
undoped samples.
A special long-term stability run was performed with the
1Ru/6.8N-CNFs catalyst for 25 h at 448 K (Figure S1). The
change of the conversion during this run was small (from 33
to 29%). This experiment showed clearly the very high stability
of the catalyst under the reaction conditions and confirmed
that the binding of the Ru clusters to the N-CNFs support is
strong and that Ru sintering is complicated.
Two supports without Ru were also studied for the formic
acid decomposition to estimate whether there is a strong con-
tribution from the support during the conversion of formic
acid (Figure 5a). The support without N (0N-CNFs) was almost
inactive for this reaction and showed less than 5% conversion
at 573 K. The support with 6.8 wt% N showed a much higher
activity that corresponds to 58% conversion at 573 K. The in-
creased conversion over the support with N compared to that
of the undoped support strongly indicates an improved activa-
tion of formic acid on the N sites. The obtained results allowed
us to conclude that the conversion of formic acid with the Ru
catalysts at 448 K was ꢀ30–50 times higher than that of the
supports without Ru. Hence, the activities of the supports are
very low compared to the activities of the Ru catalysts studied.
However, the activation of formic acid on the N sites may also
take place on the samples with Ru and has a positive influence
on the catalytic activity.
Arrhenius plots of the TOF values for the dehydrogenation
and dehydration reactions calculated from the conversion and
selectivity values are shown in Figure 6. For the dehydrogena-
tion reaction, the TOFCO values for 1.3Ru/0N-CNFs, 1.3Ru/2N-
2
CNFs, and 1Ru/6.8N-CNFs catalysts at 373 K were 0.007, 0.009,
and 0.014 cmÀ1, respectively, whereas for the dehydration reac-
tion there was no significant difference in the TOFCO values
(ꢀ0.001 sÀ1). These data show that N-doping promotes the
formic acid decomposition reaction to the direction of dehy-
drogenation and does not affect the dehydration reaction. The
apparent activation energies for the dehydrogenation and de-
hydration reactions over all the Ru catalysts were also calculat-
ed from the Arrhenius plots (Figure 6). The apparent activation
energies are the same for both reactions in the range of 55–
58 kJmolÀ1 (Table 1).
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2914
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
ble by regular TEM can be also proposed. These Ru species
may dehydrogenate formic acid only if they are coordinated
by the N sites of the support.
Conclusions
We have demonstrated that N-doped carbon nanofibers are
promising catalyst supports for active Ru catalysts for the pro-
duction of hydrogen from vapor-phase formic acid decomposi-
tion. The synthesized N-doped and undoped samples pos-
sessed approximately the same Ru content, mean particle size,
and particle sizes distributions. Photoelectron spectroscopy
data for the original samples showed that there are two Ru
states in the samples: slightly charged or metallic Ru clusters
and oxidized Ru species similar to those in RuO2. For the
sample with the highest N content, the former Ru clusters are
the main species. The Ru content on the external surface of
the N-doped support is considerably higher than that of the
undoped support. The increase of the surface metal concentra-
tion can occur as a result of the coordination of the Ru clusters
by pyridinic N located on the open edges of graphene layers
of the herringbone-type carbon nanofibers used. The strong
binding of the Ru species with pyridinic N could prevent their
diffusion into the internal volume. This allows the high accessi-
bility of these clusters to formic acid as well as the possibility
of an additional route for dehydrogenation by the activation
of formic acid through the formation of adducts on the interfa-
cial pyridinic N adjacent to the Ru clusters. As a result, the rate
of the dehydrogenation reaction increases, whereas the rate of
the dehydration reaction does not change. This leads to a con-
siderable increase of the selectivity of the hydrogen produc-
tion from 83 to 92%. The results obtained could be useful for
the development of Ru/C catalysts for the transfer hydrogena-
tion of biomass-related products with formic acid or alcohols.
Figure 6. Arrhenius plots for TOFs in formic acid dehydrogenation and dehy-
dration reactions over the Ru/N-CNFs catalysts with different N contents.
We assigned the dependence of the TOFCO and TOFCO
2
values on the N content to the appearance of a new route for
the formic acid dehydrogenation that involves N sites on the
support in the neighborhood of the metal clusters. Basic pyri-
dinic N[15b] in the vicinity of the Ru clusters can form an adduct
with formic acid [Eq. (1)]:
> NþHCOOH !> NHþHCOOÀ
ð1Þ
The higher activity observed for the formic acid decomposi-
tion on the N-doped CNFs without Ru as compared to the un-
doped CNFs (Figure 5a) may indirectly support the presence of
such activation of formic acid on N-containing sites. The inter-
mediate formed can interact with the Ru clusters to give H2
and CO2, which thus provides an additional selective route to
H2 for formic acid conversion [Eq. (2)]:
Ru
> NHþHCOOÀ
> NþH2þCO2
ð2Þ
ƒ!
The apparent activation energy for this reaction could be
close to that for the reaction that takes place on metallic Ru.
The rate of the dehydration reaction does not change in this
case as it is determined by the surface sites on Ru clusters,
which are same size for all the catalysts studied. The appear-
ance of the new route that is selective to H2 leads to the ob-
served increase of selectivity (Figure 5b). As the interface of
the Ru clusters and N sites of the support is not as contiguous
as it was for the Pt catalysts with smaller clusters and the pos-
sible presence of isolated Pt atoms, the contribution of this
route to the total conversion of formic acid for the Ru catalysts
is smaller. In accordance, we observed a decrease of the effect
of N for the Pt/N-CNFs catalysts with an increase of the mean
size of Pt clusters from 1.0 to 2.0 nm.[7] Importantly, the pro-
posed explanation based on the appearance of the new dehy-
drogenation route is also in line with the literature results for
the liquid-phase formic acid decomposition with some homo-
geneous catalysts, which include Ru complexes.[3,26] This shows
that the addition of amines to the liquid accelerated the
formic acid decomposition considerably. An alternative explan-
ation based on the presence of sub-nm sized Ru species invisi-
Experimental Section
Catalyst preparation
N-doped CNFs were synthesized by the decomposition of a mixture
of ethylene and ammonia on a 65 wt% Ni/25 wt% Cu/10 wt%
Al2O3 catalyst at 823 K for 1 h as described previously.[9,15a] The cat-
alysts were prepared by coprecipitation from nitrate salt solutions.
To obtain 0, 2, and 6.8 wt% N in the supports, the concentration of
ammonia in the mixture was increased, which corresponded to 0,
50, and 75 vol%. To obtain the catalyst, all the obtained materials
were treated with concentrated HCl several times over several
weeks at RT and then with boiling 2m HCl for 30 min. After this,
the materials were washed with distilled water until no chloride
ions were detected in the rinsing liquid. In spite of the harsh treat-
ment in HCl, the presence of some Ni was observed in the catalysts
by photoelectron spectroscopy and energy-dispersive spectrosco-
py (EDS) combined with TEM. Based on these studies and studies
of the catalytic properties of the CNFs supports and Ru/N-CNFs
catalysts, this Ni was assigned as encapsulated by carbon nickel
carbide particles of a relatively large size (>10 nm) that are not ac-
cessible for interaction with the Ru precursor and formic acid. The
observed encapsulation indicates that this Ni can have hardly any
effect on the decomposition of formic acid.
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2915
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Ru catalysts on the N-doped CNFs were prepared by deposition of
Ru from aqueous solutions of Ru(NO)(NO3)3 by the incipient wet-
ness impregnation method.[9] After the impregnation, the samples
were dried and reduced in flowing H2 at 573 K for 2 h.
concentrations of CO and CO2 formed in each run and related to
the initial concentration of formic acid.
Apparent activation energies and TOFs were calculated at low con-
versions up to 15%. The total TOFs were determined as the rate of
the sum of CO and CO2 formation related to the number of surface
metal sites calculated using the mean particle size values (TEM)
Catalyst characterization
with the Ru atomic area as 0.064 nmatomÀ1
.
The content of Ru in the samples was determined by X-ray fluores-
cence analysis by using an ARL Advant’x (Thermo Scientific) spec-
trometer. It was 1.3, 1.3, and 1 wt% for the samples that contained
0, 2, and 6.8 wt% N, respectively (Table 1). TEM measurements
were performed by using a JEM-2010 (JEOL) instrument with an ac-
celerating voltage of 200 kV. The size distributions of the Ru parti-
cles were calculated from the measurements of at least 300 parti-
cles.
Acknowledgements
The authors would like to acknowledge the support of the Earth
and Natural Sciences (ENS) Doctoral Studies Programme, funded
by the Higher Education Authority (HEA) through the Programme
of Research at Third Level Institutions, Cycle 5 (PRTLI-5), co-
funded by the European Regional Development Fund (ERDF). This
publication has also emanated from research conducted with the
financial support of Science Foundation Ireland under Grant
Number 06/CP/E007. We thank Helmholtz-Zentrum (Berlin) for
the allocation of synchrotron radiation beamtime. The presented
photoelectron spectroscopy experiments have been supported by
the Russian-German Laboratory at BESSY II (Project No.
14201409). The Ministry of Education and Science of the Russian
Federation is gratefully acknowledged for the support.
High-resolution photoelectron spectra were recorded with synchro-
tron radiation at the BESSY II RGL station at Helmholtz-Zentrum in
Berlin. The survey spectra were collected at an incident photon
energy (hn) of 850 eV. The core-level spectra of C1s and Ru3d
were recorded with a variation of the incident photon energy from
400 to 1000 eV. An incident photon energy of 800 eV was used to
acquire the N1s spectra. The application of synchrotron radiation
allowed the high-resolution acquisition of data in the 280–300 eV
spectral region typical of Ru3d and C1s core-level lines, which
thus allows the possibility to analyze the Ru state using the
Ru3d5/2 spectra. For the scale calibration, the core-level line of met-
allic gold (Au4f7/2) with a binding energy of 84.0 eV was used as
an external reference in all experiments. The spectra processing
and analysis was performed using the XPS-Calc program after sub-
traction of the Shirley background.[15a,27] To shed light on the chem-
ical state of Ru, additional measurements were performed after re-
duction of the Ru sample with the highest N content in the prepa-
ration chamber of the spectrometer. For this, the reduction in hy-
drogen at 2”10À2 mbar for 2 h at 623 K was undertaken. After the
reduction, the preparation chamber was evacuated and the
sample was transferred to the analysis chamber without contact
with air.
Keywords: doping
·
dehydrogenation
·
hydrogen
·
ruthenium · supported catalysts
[1] a) D. J. Hayes, S. Fitzpatrick, M. H. B. Hayes, J. R. H. Ross, in Biorefineries
– Industrial Processes and Products, Vol. 1 (Eds.: B. Kamm, P. R. Gruber,
M. Kamm), Wiley-VCH, Weinheim, 2006, pp. 139–164; b) J. C. Serrano-
Ruiz, D. J. Braden, R. M. West, J. A. Dumesic, Appl. Catal. B 2010, 100,
184–189.
[2] a) J. Li, D. J. Ding, L. Deng, Q. X. Guo, Y. Fu, ChemSusChem 2012, 5,
1313–1318; b) J. Zhang, M. Sun, X. Liu, Y. Han, Catal. Today 2014, 233,
77–82.
The surface areas of the materials were calculated using the BET
method from the data obtained by N2 adsorption at 77 K by using
an ASAP-2400 instrument (Micromeritics).
[3] M. Grasemann, G. Laurenczy, Energy Environ. Sci. 2012, 5, 8171–8181.
[4] a) D. A. Bulushev, J. R. H. Ross, Catal. Today 2011, 163, 42–46; b) H.
Heeres, R. Handana, D. Chunai, C. B. Rasrendra, B. Girisuta, H. J. Heeres,
Green Chem. 2009, 11, 1247–1255.
[5] a) D. A. Bulushev, I. Yuranov in Dekker Encyclopedia of Nanoscience and
Nanotechnology, Second ed. (Eds.: C. I. Contescu, K. Putiera), CRC Press,
Boca Raton, 2008, pp. 3195–3205; b) A. E. Shalagina, Z. R. Ismagilov,
O. Y. Podyacheva, R. I. Kvon, V. A. Ushakov, Carbon 2007, 45, 1808–
1820; c) O. Y. Podyacheva, Z. R. Ismagilov, Catal. Today 2015, 249, 12–
22.
[6] a) K. N. Wood, R. O’Hayre, S. Pylypenko, Energy Environ. Sci. 2014, 7,
1212–1249; b) Y. Marco, L. Roldan, S. Armenise, E. Garcia-Bordeje,
ChemCatChem 2013, 5, 3829–3834.
[7] L. J. Jia, D. A. Bulushev, O. Y. Podyacheva, A. I. Boronin, L. S. Kibis, E. Y.
Gerasimov, S. Beloshapkin, I. A. Seryak, Z. R. Ismagilov, J. R. H. Ross, J.
Catal. 2013, 307, 94–102.
[8] a) J. Fahrenfort, L. L. van Reyen, W. M. H. Sachtler in Proceedings of the
Symposium on the Mechanism of Heterogeneous Catalysis (Ed.: J. H.
de Boer), Elsevier, Amsterdam, 1960, p. 23; b) K. Tedsree, T. Li, S. Jones,
C. W. A. Chan, K. M. K. Yu, P. A. J. Bagot, E. A. Marquis, G. D. W. Smith,
S. C. E. Tsang, Nat. Nanotechnol. 2011, 6, 302–307.
Catalytic activity measurements
The vapor-phase formic acid decomposition was performed in
a fixed-bed 4 mm id quartz reactor as described earlier.[28] Activity
tests were performed at atmospheric pressure with 7 mg of cata-
lyst placed between two layers of quartz wool. The catalysts were
reduced in a 1 vol% H2/Ar mixture for 1 h at 573 K and cooled in
He to the reaction temperature (333 K). To obtain a temperature
dependence of the acid conversion, the composition of the outlet
gas mixture was measured several times at each temperature (at
least for 30 min) to ensure that steady-state was reached.
A controlled amount of formic acid (Sigma–Aldrich, 98–100%) was
introduced into the system by using a syringe-pump (Sage) to
obtain 1.9 vol% of formic acid vapor in He. The WGS reaction was
performed using a 2.5 vol% CO/2.3 vol% H2O/He mixture. A total
gas flow rate of 51 cm3 minÀ1 was used in all experiments.
[9] A. B. Ayusheev, O. P. Taran, I. A. Seryak, O. Y. Podyacheva, C. Descorme,
M. Besson, L. S. Kibis, A. I. Boronin, A. I. Romanenko, Z. R. Ismagilov, V.
Parmon, Appl. Catal. B 2014, 146, 177–185.
The products were analyzed by using an HP-5890 gas chromato-
graph equipped with a thermal conductivity detector and Porapak-
Q column. As no other carbon-containing products were detected,
the conversion of formic acid was determined as the sum of the
[10] S. Pylypenko, A. Borisevich, K. L. More, A. R. Corpuz, T. Holme, A. A.
Dameron, T. S. Olson, H. N. Dinh, T. Gennett, R. O’Hayre, Energy Environ.
Sci. 2013, 6, 2957–2964.
[11] P. Panagiotopoulou, D. G. Vlachos, Appl. Catal. A 2014, 480, 17–24.
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2916
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[12] R. M. West, E. L. Kunkes, D. A. Simonetti, J. A. Dumesic, Catal. Today
2009, 147, 115–121.
[21] S. Tanuma, C. J. Powell, D. R. Penn, Surf. Interface Anal. 2011, 43, 689–
713.
[22] Practical Surface Analysis by Auger and X-ray Photoelectron Spectroscopy
(Eds.: D. Briggs, M. P. Seah), Wiley, New York, 1983.
[23] D. I. Kochubey, V. V. Chesnokov, S. E. Malykhin, Carbon 2012, 50, 2782–
2787.
[24] a) G. Bae, D. H. Youn, S. Han, J. S. Lee, Carbon 2013, 51, 274–281; b) S.
Stambula, N. Gauquelin, M. Bugnet, S. Gorantla, S. Turner, S. H. Sun, J.
Liu, G. X. Zhang, X. L. Sun, G. A. Botton, J. Phys. Chem. C 2014, 118,
3890–3900.
[25] A. Basinska, L. Kepinski, F. Domka, Appl. Catal. A 1999, 183, 143–153.
[26] B. Loges, A. Boddien, F. Gartner, H. Junge, M. Beller, Top. Catal. 2010, 53,
902–914.
[27] a) I. Mazov, V. L. Kuznetsov, I. A. Simonova, A. I. Stadnichenko, A. V. Ish-
chenko, A. I. Romanenko, E. N. Tkachev, O. B. Anikeeva, Appl. Surf. Sci.
2012, 258, 6272–6280; b) E. D. Grayfer, V. G. Makotchenko, L. S. Kibis,
A. I. Boronin, E. M. Pazhetnov, V. I. Zaikovskii, V. E. Fedorov, Chem. Asian
J. 2013, 8, 2015–2022.
[13] F. Solymosi, A. Koos, N. Liliom, I. Ugrai, J. Catal. 2011, 279, 213–219.
[14] B. J. O’Neill, E. I. Gürbüz, J. A. Dumesic, J. Catal. 2012, 290, 193–201.
[15] a) O. Y. Podyacheva, Z. R. Ismagilov, A. I. Boronin, L. S. Kibis, E. M. Slavin-
skaya, A. S. Noskov, N. V. Shikina, V. A. Ushakov, A. V. Ischenko, Catal.
Today 2012, 186, 42–47; b) A. Benyounes, M. Kacimi, M. Ziyad, P. Serp,
Chin. J. Catal. 2014, 35, 970–978.
[16] O. A. Stonkus, L. S. Kibis, O. Y. Podyacheva, E. M. Slavinskaya, V. I. Zaikov-
skii, A. H. Hassan, S. Hampel, A. Leonhardt, Z. R. Ismagilov, A. S. Noskov,
A. I. Boronin, ChemCatChem 2014, 6, 2115–2128.
[17] a) J. L. Gomez de La Fuente, F. J. Perez-Alonso, M. V. Martinez-Huerta,
M. A. Pena, J. L. G. Fierro, S. Rojas, Catal. Today 2009, 143, 69–75; b) T. F.
Hsieh, C. C. Chuang, W. J. Chen, J. H. Huang, W. T. Chen, C. M. Shu,
Carbon 2012, 50, 1740–1747.
[18] a) B. Folkesson, Acta Chem. Scand. 1973, 27, 287–302; b) Y. C. Wei, C. W.
Liu, W. J. Chang, K. W. Wang, J. Alloys Compd. 2011, 509, 535–541.
[19] a) M. G. Mason, L. J. Gerenser, S. T. Lee, Phys. Rev. Lett. 1977, 39, 288–
291; b) G. K. Wertheim, S. B. DiCenzo, Phys. Rev. B 1988, 37, 844; c) T. P.
Wu, W. E. Kaden, W. A. Kunkel, S. L. Anderson, Surf. Sci. 2009, 603,
2764–2770.
[28] D. A. Bulushev, S. Beloshapkin, J. R. H. Ross, Catal. Today 2010, 154, 7–
12.
[20] S. Armenise, L. Roldan, Y. Marco, A. Monzon, E. Garcia-Bordeje, J. Phys.
Chem. C 2012, 116, 26385–26395.
Received: March 3, 2015
Published online on June 18, 2015
ChemCatChem 2015, 7, 2910 – 2917
www.chemcatchem.org
2917
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim