Enhanced performance and selectivity of CO 2 methanation over g-C 3 N 4 assisted synthesis of Ni CeO 2 catalyst: Kinetics and DRIFTS studies

Abstract

The hydrogenation of CO2 on Ce–Ni catalyst modified with g-C3N4 (Ce–Ni–CN) as a sacrificial and protective template was studied by in-situ DRIFTS and Kinetics experiments to investigate the influence of modification on the catalytic activity and selectivity to gain mechanistic insight. After modification, the catalyst showed higher catalytic activity and selectivity. H2-TPR, CO2-TPD, TEM and XPS confirmed that this modification could enhance the interaction of nickel and ceria and decrease the particle size of nickel, which is in favor of the dissociation of H2 and adsorption of CO2. The in-situ DRIFTS experiments demonstrated that CO2 is adsorbed on ceria sites, forming carboxylate (CO2δ−), unidentate carbonate and bicarbonates, which, in turn, react with adsorbed and dissociated H on Ni to produce formate species. Furthermore, adsorbed methoxy species were observed, which are recognized to be intermediates in the methanation process. In-situ transient DRIFTS confirm that the adsorbed CO is not a reaction intermediate, but a by-product, which originates from the decomposition of weak-binding formate species on Ce³⁺ sites. The unmodified catalyst has more weak-binding formate species, which are more inclined to decompose into CO accounting for the low selectivity. Furthermore, the adsorbed CO on Ce³⁺ sites cannot react with the adsorbed hydrogen to produce methane. Kinetics studies are consistent with a Langmuir-Hinshelwood type mechanism in which the formation of bicarbonate is the rate-determining step (RDS).

Full text

Enhanced performance and selectivity of CO
2
methanation over g-C
3
N
4
assisted synthesis of
NieCeO
2
catalyst: Kinetics and DRIFTS studies
Yang Yu
a,b
, Yi Meng Chan
a
, Zhoufeng Bian
a
, Fujiao Song
c
,
Juan Wang
a,b
, Qin Zhong
b,*
, Sibudjing Kawi
a,**
a
Department of Chemical and Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4,
119260, Singapore
b
School of Chemical Engineering, Nanjing University of Science and Technology, Nanjing, 210094, PR China
c
School of Environmental Science and Engineering, Yancheng Institute of Technology, Yancheng, 224051, PR China
article info
Article history:
Received 9 April 2018
Received in revised form
24 May 2018
Accepted 14 June 2018
Available online 18 July 2018
Keywords:
CO
2
reduction
CeeNi catalyst
g-C
3
N
4
-modification
In-situ DRIFTS
Kinetics
abstract
The hydrogenation of CO
2
on CeeNi catalyst modified with g-C
3
N
4
(CeeNieCN) as a
sacrificial and protective template was studied by in-situ DRIFTS and Kinetics experiments
to investigate the influence of modification on the catalytic activity and selectivity to gain
mechanistic insight. After modification, the catalyst showed higher catalytic activity and
selectivity. H
2
-TPR, CO
2
-TPD, TEM and XPS confirmed that this modification could enhance
the interaction of nickel and ceria and decrease the particle size of nickel, which is in favor
of the dissociation of H
2
and adsorption of CO
2
. The in-situ DRIFTS experiments demon-
strated that CO
2
is adsorbed on ceria sites, forming carboxylate (CO
2
d
), unidentate car-
bonate and bicarbonates, which, in turn, react with adsorbed and dissociated H on Ni to
produce formate species. Furthermore, adsorbed methoxy species were observed, which
are recognized to be intermediates in the methanation process. In-situ transient DRIFTS
confirm that the adsorbed CO is not a reaction intermediate, but a by-product, which
originates from the decomposition of weak-binding formate species on Ce
3þ
sites. The
unmodified catalyst has more weak-binding formate species, which are more inclined to
decompose into CO accounting for the low selectivity. Furthermore, the adsorbed CO on
Ce
3þ
sites cannot react with the adsorbed hydrogen to produce methane. Kinetics studies
are consistent with a Langmuir-Hinshelwood type mechanism in which the formation of
bicarbonate is the rate-determining step (RDS).
©2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
Introduction
Carbon dioxide (CO
2
), which is generally generated from
combustion, is recognized as a significant factor responsible
for the greenhouse effect and a contributor to climate change
[1,2]. Hence, CO
2
fixation has gained much interest in recent
years [3e5]. Currently, hydrogenation of CO
2
is widely used in
the production of many useful chemicals, such as formic acid
*Corresponding author.
** Corresponding author.
E-mail addresses: zq304@mail.njust.edu.cn (Q. Zhong), chekawis@nus.edu.sg (S. Kawi).
Available online at www.sciencedirect.com
ScienceDirect
journal homepage: www.elsevier.com/locate/he
international journal of hydrogen energy 43 (2018) 15191e15204
https://doi.org/10.1016/j.ijhydene.2018.06.090
0360-3199/©2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

[6], methanol [7], and methane [8], of which methane may
account for the largest amount of hydrogenation product. CO
2
methanation is a strongly exothermic reaction, which can be
carried out under atmospheric pressure [9e11]. Many noble
metals (e.g., Ru, Rh, and Pd) have been proven to be effective
catalysts for CO
2
methanation at mild operating conditions
[12], but the high cost as well as their limited availability re-
stricts their practical application [13e16]. Therefore, the
transition metal catalysts (e.g., Ni, Co, and Cu) have aroused
much attention to obtain a feasible and economic catalytic
process [17,18].
Ni-based catalysts have been widely used for the metha-
nation of CO
2
because nickel is inexpensive and readily
available [19]. Because supports play an important role in
dispersing the active components, adsorption and catalytic
properties [20], preparation of highly dispersedly supported
metal catalysts has been the focus of abundant research. Many
materials have been used as support for the nickel catalysts
such as CeO
2
[17],Al
2
O
3
[19], TiO
2
[21], SiO
2
[22]. It is generally
reckoned that the high surface area of support can achieve a
high dispersion of nickel and thus improve the catalytic per-
formance [23]. In addition, the electronic structure, basic
properties, surface defects and the interaction between metal
and support would influence the CO2/H2 adsorption and
dissociation, and consequently impose significant influences
on the catalytic activity. CeO2 is an important support which
can enhance the catalytic performance via controlling the
dispersion of the active component and increasing the metal-
support interaction. According to the literature [24], the addi-
tion of CeO
2
to Ni catalysts enhances the catalytic perfor-
mance by forming a NieCeO
2
nanoparticle. The excellent
performance was ascribed to the large interface between Ni
and ceria surface. The results obtained from NieCeO
2
catalysts
for CO
2
methanation are indeed quite promising [25].
Graphitic carbon nitride (g-C
3
N
4
), with graphite-like sp
2
-
bonded CeN structure, is a promising material in the catalytic
field [26]. Although many endeavors have been devoted to
modify g-C
3
N
4
in order to improve their catalytic performance
[27,28], little attention has been paid to their use as sacrificial
templates. Fisher and co-authors synthesized ternary metal
nitride nanoparticles using g-C
3
N
4
as template and reactant
[29].
Several mechanisms have been proposed for CO
2
reduction
to methane (methanation or the Sabatier reaction). They are
mainly divided into two categories: (1) dissociation of CO
2
to
CO, followed by CO methanation [30e32], (2) direct hydroge-
nation of CO
2
to methane without forming a CO intermediate
[33,34]. However, the mechanism may vary according to the
operating pressure, as well as the nature of the catalysts [35].
Herein, CeeNi catalyst was fabricated elaborately via a g-
C
3
N
4
-modified sol-gel method. TG-DSC, BET, N
2
O pulse
decomposition, H
2
-TPR, XPS, TEM and CO
2
-TPD characteriza-
tions were used to analyze the physicochemical properties
and elucidate the interaction between nickel and ceria.
Combining kinetics experiments conducted over a wide range
temperatures and concentrations, with DRIFTS, the detailed
reaction mechanism involved in CO
2
methanation was pro-
posed. Further, the pathway of production of CO was revealed
by the transient in-situ DRIFTS to gain insights into the CO
2
hydrogenation reaction mechanism.
Experimental section
Preparation of the catalysts
All starting reagents were purchased from Sigma-Aldrich
Company (Singapore) and were used without further purifi-
cation. Deionized water was used throughout.
g-C
3
N
4
power was synthesized according to our previous
work [36].
NiO/CeO
2
modified with g-C
3
N
4
was synthesized by a cit-
rate sol-gel method. 3.6890 g Ce(NO
3
)
3
$6H
2
O, 0.4365 g
Ni(NO
3
)
2
$6H
2
O and 3.2664 g citric acid were dissolved into
deionized water with stirring for 0.5 h, then 2.5 g of the as-
prepared g-C
3
N
4
was added into the above-mentioned solu-
tion to continue stir for another 1 h. After that the mixture was
heated at 90 C under stirring until it became a viscous gel and
dried at 120 C for 12 h. The obtained solid was calcined in air
at 500 C for 4 h with a heating rate of 4 C$min
1
. The ob-
tained sample is labeled as CeeNieCN. For comparison, the
NiO/CeO
2
prepared in the same process but without adding g-
C
3
N
4
was denoted as CeeNi.
Sample characterization
The powder XRD patterns were recorded on a Shimadzu XRD-
6000 X-ray diffraction using Cu-K
a
radiation at 40 kV and
30 mA (2qfrom 5to 80).
An in-situ high temperature X-ray diffractometer (HT-XRD,
Philips, X'pert Pro) was utilized to measure the structural
evolution of precursor from 100 to 500 C in air with a heating
rate of 5 C$min
1
.
The thermal reduction capacity for oven dried sample was
comparatively studied by thermo-gravimetric analysis (TGA,
Netzsch, STA, 449 F3) in air and the temperature was ramped
from 50 to 700 C with a heating rate of 10 C$min
1
.
Specific surface areas of the different catalysts were
determined by N
2
adsorption-desorption measurements at
196 C and determined the Brunauer-Emmet-Teller (BET)
method using Micromeritics ASAP 2020 instrument.
X-ray photoelectron spectra (XPS) was performed on a
Thermo ESCALAB 250 (USA) apparatus with Al KaX-rays
(hv ¼1486.6 eV) radiation, calibrated by the C 1s peak at
284.6 eV, with an accuracy of 0.1 eV.
The micro morphology of the catalysts was examined on a
JEOL JEM-2100 transmission electron microscope (TEM) with
acceleration voltage of 200 kV.
Field-emission electron microscope (SEM) observations
were carried out using a FEI Quanta 250F.
H
2
-TPR was performed in a quartz U-tube reactor on
Thermal Scientific TPDRO 1100 apparatus equipped with a
thermal conductivity detector (TCD). Typically, 50 mg sample
was tested with a flow of 5% hydrogen in nitrogen (30 ml/min)
with a temperature ramping rate of 10 C/min after degassing
with helium for 10 min.
The metal dispersion measurement was carried out on
Thermal Scientific TPDRO 1100 as well with N
2
O decomposi-
tion method reported elsewhere [37]. 100 mg sample was
reduced at 450 C by 5% hydrogen in nitrogen for 1 h, and
cooled down to 80 C at the atmosphere of nitrogen. Purified
international journal of hydrogen energy 43 (2018) 15191e1520415192

N
2
O pulses were injected until the eluted peak area was con-
stant. Then the sample was reduced again in a TPR process.
The metal dispersion was estimated by comparing the
amount of H
2
consumed in first and second TPR process, ac-
cording to Eqs. (1) and (2).
NiSurface þN2O/NiO þN2(1)
NiO þH2/Ni þH2O (2)
The Ni active surface area was calculated from the amount
of H
2
consumed in the second reduction by assuming a stoi-
chiometry of 1 and a surface area of 6.5 10
20
m
2
per Ni atom.
Temperature-programmed desorption (TPD) was carried
out on the same instrument of TPR. 100 mg catalyst was
reduced at 450 C by 5% hydrogen in helium for 1 h and cooled
down to 50 C. At this temperature, CO
2
gas with a flow rate of
50 ml min
1
was purged for 1 h. After that, the flow gas was
switched to helium to purge the physically adsorbed CO
2
on
the catalysts. Finally, the temperature was increased to 800 C
with a ramping rate of 10 C$min
1
. TCD signal was recorded
continuously against temperature showing the desorption
temperature of CO
2
.
In-situ DRIFTS experiments were carried out by a Bruker V-
70 FTIR spectrometer, equipped with a liquid-nitrogen-cooled
MCT detector. 32 scans were averaged for each spectrum,
which were recorded at a resolution of 4 cm
1
. Prior to each
experiment, the samples were ex situ reduced in pure H
2
at
450 C for 1 h and purged with He for 30 min.
Catalytic evaluation
The catalytic performances were conducted at the atmo-
spheric pressure in a quartz reactor (4 mm i.d.) loaded with
300 mg catalyst. Prior to the test, the catalysts were reduced on
line for 1 h in a flow of 30 ml min
1
H
2
at 450 C, and then
cooled to 200 C in helium in a flow of 30 ml min
1
. Subse-
quently, 8 mol% CO
2
, 32 mol% H
2
, He balance, and the total
flow rate was of 50 ml min
1
with GHSV of about 10 L g
1
h
1
.
Effluent gases from the reactor were on-line by a gas chro-
matograph (HP 6980) with a Carboxen column and a thermal
conductivity detector (TCD). The CO
2
conversion and CH
4
selectivity are defined as follow:
ConversionCO2ð%Þ¼CinðCO2ÞCoutðCO2Þ
CinðCO2Þ
*100% (3)
SelecitivityCH4ð%Þ¼CðCH4Þ
CinðCO2ÞCoutðCO2Þ
*100% (4)
where CinðCO2Þand CoutðCO2Þare the inlet and outlet concentra-
tion, respectively. CðCH4Þis the methane concentration in the
product.
Kinetic modeling testing condition
The kinetic rate experiments were performed in a 4 mm ID
tubular reactor. 0.3 g of the NieCeeCN catalyst was mixed
with 0.1 g of silica to avoid temperature gradients and mass
transfer effects. The mixture was first reduced in H
2
for 1 h at
450 C. The temperature was controlled with an external oven
and the pressure of the reactants was at 1 atm. Flow rates of
reactants were regulated by mass flow controllers and com-
prises of a mixture of CO
2
,H
2
and He. The reactor outlet
composition was measured with an on-line Gas Chromatog-
raphy. The reactant gas composition and temperature was
then varied to investigate the dependence of methanation
rate on reactant partial pressure.
Result and discussion
Decomposition process of the precursor and g-C
3
N
4
To determine the calcination temperature and the decom-
position process of the precursors, the TG-DSC curves of the
precursors after drying at 120 C are shown in Fig. 1.Figs. S1
and S4 shows the XRD and FTIR patterns of g-C
3
N
4
,
respectively. The characterization results confirm the for-
mation of g-C
3
N
4
.InFig. 1a, the weight loss below 200 C
was attributed to the dehydration and the decomposition of
citric acid [31]. The DSC profiles of CeeNi precursor showed
only one exothermic peak at 295 C, which was due to the
decomposition of nickel and ceria citric complex decompo-
sition into oxides [38,39].Fig. 1b shows that when g-C
3
N
4
powder was added to the precursors, the compound
decomposed at a higher temperature of 346 C. This may be
induced by the enhanced interaction of ceria and nickel. The
TG-DSC curves of CeeNieCN are highly different from those
of CeeNi. There was one exothermic peak at 400 C. This
peak was due to the decomposition of g-C
3
N
4
for the strong
oxidation ability of CeO
2
under high temperature [40]. The
Fig. 1 eTG-DSC thermograms for heating the CeeNi (a) and CeeNieCN (b) precursors.
international journal of hydrogen energy 43 (2018) 15191e15204 15193

in-situ XRD pattern in Fig. S2 demonstrated the gradual
formation of CeO
2
at above 350 C, in consistent with the
results of TG-DSC. The TG-DSC profile of pure g-C
3
N
4
(Fig. S3) shows one exothermic peak appears at about 740 C.
This is much higher than the decomposition temperature
(400 C) of g-C
3
N
4
in the precursor of CeeNieCN. Thus, we
could infer the formation of CeO
2
could accelerate the
decomposition of g-C
3
N
4
. The results in Fig. 1 clearly show
that g-C
3
N
4
could serve as a sacrificial agent, which were
embedded in the precursor and were totally removed after
calcination at 500 C in air. The FTIR results (Fig. S4) and N 1s
spectra (Fig. S5) verified the complete removal of g-C
3
N
4
after calcinations process. This decomposition process may
probably prevent the agglomeration of nickel oxides during
the calcination process. Weight loss hardly occurred above
500 C, which indicated that the calcination temperature
could be higher than 500 C.
Textural analysis
To investigate the porosity and specific surface area of the
samples, N
2
adsorption-desorption measurements were car-
ried out at 196 C. Fig. 2a shows the adsorption-desorption
isotherms of the reduced CeeNi and CeeNieCN. The iso-
therms reveal that there is much less adsorption in the low
pressure region. Both samples show type IV isotherms. The
hysteresis loops are like a type of H3, indicating that these
samples contained mesopores with narrow silt-like shapes
[41].
In addition, the corresponding pore size distributions are
shown in Fig. 2b, which are calculated by the BJH method. The
pore size distribution curves suggested that the main peaks of
both samples are located in the range of 3e6 nm, indicating
the presence of large percentage of mesopores, along with
little amount of micro and macropores.
The textural properties of samples are summarized in
Table 1. It can be seen that the surface area of the g-C
3
N
4
modified sample was 19 m
2
g
1
, which is almost twice of the
original sample. Furthermore, the pore volume for CeeNi was
0.073 and 0.102 cm
3
g
1
for CeeNieCN. The higher surface
area and pore volume in these samples indicated that the
decomposition of g-C
3
N
4
played an important role in modi-
fying the textural properties, showing promise for their po-
tential application as catalyst template.
XRD
Fig. 3 depicts the XRD patterns of the calcined and reduced
catalysts. The formation of CeO
2
was revealed by the diffrac-
tion peaks at 2qvalues of 28.6, 33.1, 47.5, 56.4, 59.4, 69.6, 76.7
and 79.2, corresponding to (111), (200), (220), (311), (222), (400),
(331) and (420) crystal planes of CeO
2
, respectively. It could be
seen that all the diffraction peaks of these samples were
characteristic of typical cubic fluorite phase of ceria (JCPDS
Card No. 34-0394). Moreover, no diffraction peaks of nickel
species was observed on the XRD patterns, indicating they
were highly dispersed or/and existed as small particles that
were not sensitive to XRD [42]. In addition, the lattice pa-
rameters of the calcined and reduced samples are given in
Table 1 and Table S1. In principle, the variation of lattice
parameter can be predicted by comparing the relative ionic
radius when the oxidation of alien ions is steadily þ4[43];
however, with the charge of the doped ions unequal to that of
Ce
4þ
, the lattice volume will be influenced by some other
factors and not only by the ionic radius disparity [44]. Like the
conclusion of Deng et al. [45], the increase of lattice parameter
Fig. 2 eN
2
adsorption-desorption isotherms (a) and pore size distributions (b) of the reduced CeeNi and CeeNieCN.
Table 1 eTextural properties of the catalysts.
Sample BET (m
2
$g
1
) Pore volume
(cm
3
$g
1
)
Lattice parameter
a
(nm)
Ni dispersion
b
(%)
Ni metal surface
area (m
2
$g
cat
1
)
CeeNi-reduced 11 0.073 0.5408 17.8 0.85
CeeNieCN-reduced 19 0.102 0.5412 26.5 1.20
a
Calculated by Rietveld refinement of XRD.
b
Nickel metal dispersion determined by N
2
O decomposition.
international journal of hydrogen energy 43 (2018) 15191e1520415194

could be explained by a combined effect due to the formation
of the oxygen vacancies-related defects inside the ceria lat-
tice. The calcined CeeNieCN sample is found to have larger
lattice parameter compared to calcined CeeNi, indicating
more oxygen vacancies existing in CeeNieCN sample. After
reduction, more Ce
3þ
formed on the reduced sample. The
much greater ion radius of Ce
3þ
than Ce
4þ
can overcome,
offset, or compensate the diameter disparity between Ni
2þ
and Ce
4þ
, thereby inducing the lattice expansion. This phe-
nomenon could explain the increase of lattice parameter after
reducing the sample with H
2
.
TEM imagines of freshly reduced catalysts
TEM micrographs of the reduced catalysts reveal differences
in particle size of the supported Ni. Distinct nickel particles
are observed on the CeO
2
support as indicated by the red
arrow in Fig. 4. As shown in the red dotted area in Fig. 4a, large
clusters were observed. These clusters were composed of bulk
phase or un-interacted nickel species, which were unstable
and inclined to form secondary particles during the long term
Fig. 3 eXRD patterns for calcined and reduced CeeNi and
CeeNieCN catalysts.
Fig. 4 eTEM images of the catalysts upon reduction by pure H
2
at 450 C for 1 h (a &c) CeeNi (b &d) CeeNieCN, inset is the
Ni particle size distribution (c) CeeNi and (d) CeeNieCN.
international journal of hydrogen energy 43 (2018) 15191e15204 15195

stability test. The enlarged image of CeeNi is shown in Fig. 4c.
The image also clearly indicates the aggregation of nickel
particles. By contrast, we could found that the nickel particles
are more dispersive on the support with the adding of g-C
3
N
4
.
The mean particle sizes decrease from 17 nm to 13 nm (inset
figures in Fig. 4c,d). These results demonstrate the greater
propensity for nickel species to aggregate on the undecorated
CeO
2
support due to the lack of strong metal-support inter-
action between nickel and ceria. On the other hand, the
thermal oxidizing decomposition of the g-C
3
N
4
could prevent
the aggregation of NiO that normally occurs in the process of
catalyst preparation as reported in the literature [23]. This
leads to stronger NieCe interaction that restricts the mobility
of surface nickel species and reduced assembly of nickel into
larger agglomerates, thus decreasing the particle size. This
result coincided with the results of TG-DSC.
H
2
-TPR
The reduction behavior of CeeNi and CeeNieCN under
reduction conditions were investigated by H
2
-TPR experi-
ments, with the obtained profiles plotted in Fig. 5. The
shoulder peak aat about 200 C was attributed to the reduc-
tion of the adsorbed oxygen [46]. It has been verified that Ni
2þ
is directly reduced to Ni
0
without any intermediate oxidation
state species [47]. Therefore, the different H
2
consumption
peaks appear according to various NiO species. From H
2
-TPR
profiles, both samples have two obvious reduction regions,
which were derived from the reduction of nickel species with
different degrees of interaction with the CeO
2
support. Firstly,
the bpeak was mostly ascribed to the bulk phase or un-
interacted NiO to Ni
0
species [22]. Secondly, the moderate-
temperature region gwas associated with the NiO species
which had relatively smaller particle diameter and enhanced
interaction with the support. A tiny and broad peak scentered
at about 550 C was due to the reduction of surface Ce
4þ
[48].
The high temperature zone reduction peak (above 700 C) was
caused by the reduction of bulk CeO
2
[49]. The intensities at
each peak appear to be different for both samples even though
both have the same nominal amount of Ni content. This is due
to the existence of different types of NiO species. After
modification, the area of peak bis much lower than the
counterpart of CeeNi, while the area of peak gbecomes
higher. This phenomenon indicates that the g-C
3
N
4
-modified
catalyst decreases the proportion of NiO with weak metal-
support interaction, and increases the proportion of nickel
species with moderate and strong interaction with the sup-
port. This finding corresponds with earlier results whereby a
higher dispersion of nickel species on the decorated sample
was measured (Table 1).
XPS
To identify the surface oxidation state and the surface species
present on the catalyst, XPS characterizations was carried out
for the freshly reduced catalysts. Fig. 6a displays the XPS
profiles of the hybrid, which is mainly composed of O, C, Ni
and Ce element. Fig. 6b shows the Ce 3d spectra of different
catalysts. Ce 3d core level spectra for both samples were
deconvoluted into eight contributions. The spin-orbit splitting
of Ce 3d
5/2
and Ce 3d
3/2
was 18.5 eV for all the samples, which
was in good agreement with literature [50]. The u''' satellite
peak was the fingerprint of Ce
4þ
state, and its high intensity
and area suggested that the main part of the cerium was in
Ce
4þ
oxidation state. After deconvolution, the appearance of
bands labeled as u'and v'were typical for Ce
3þ
ions, suggest-
ing that both Ce
4þ
and Ce
3þ
coexist on the surfaces of catalysts
[50].Table 2 presents the relative contributions of Ce
3þ
and
Ce
4þ
calculated based on the peak fitting and the area under
the fitted components. The Ce
3þ
/(Ce
3þ
þCe
4þ
) ratios were
found to be 23.3% and 31.8% for the reduced CeeNi and Cee
NieCN catalysts, respectively.
The O 1s spectra, which were made up of two compo-
nents, and displays in Fig. 6c. The peak of O 1s named as O
a
,
with a binding energy of ~529.3 eV, is assigned to the lattice
oxygen species in metal oxides [51]. The high binding
~531.5 eV energy was attributed to the surface chemisorbed
oxygen (O
b
)[51]. The adsorption of CO
2
in the reduced state
Ce
3þ
sites has a higher thermal stability than those in the
Ce
4þ
sites [52]. Thus, the adsorption oxygen was derived from
the carbonate species trapped by oxygen vacancies. The ra-
tios of oxygen vacancies could be obtained from the relative
content of adsorption oxygen and the results of O
b
/(O
b
þO
a
)
are shown in Table 2. The proportion of O
b
/(O
b
þO
a
)onthe
CeeNieCN catalyst is more than those on the CeeNi catalyst;
the sequence is in good agreement with the order of Ce
3þ
content.
Fig. 6d displays the binding energies (BEs) of Ni 2p
2/3
pre-
sent in the reduced CeeNi and CeeNieCN catalysts. The peaks
at ~854 and ~856.4 eV were characteristic of Ni
0
and NiO,
respectively [19,21,22]. The ratio of Ni
0
/(Ni
0
þNiO) was
calculated from the integrated area under the fitted compo-
nents and displayed in Table 2. It could be observed that more
NiO was reduced to Ni
0
over CeeNieCN catalyst as compared
to those on the CeeNi catalyst, demonstrating that the Cee
NieCN had moore reducible Ni species. In addition, the bulk
content of the catalysts was determined by SEM-EDS (Table
S2). The surface ratio of Ni/Ce on the CeeNieCN is close to
the bulk value, indicating the homogeneous dispersion of Ni
in the catalyst. This result is in agreement with the H
2
-TPR
and TEM result.
Fig. 5 eH
2
-TPR profiles of the calcined catalysts.
international journal of hydrogen energy 43 (2018) 15191e1520415196

CO
2
-TPD
The basic sites in the catalysts were characterized with CO
2
-
TPD measurements. It is commonly accepted that CO
2
is
adsorbed on the weaker basic sites in the lower temperature
regions, and the desorption peaks at 250 Ce500 C were
attributed to moderate basic sites. In addition, only those
adsorbed on stronger basic sites would be desorbed at higher
temperature [53]. The CO
2
-TPD profiles of both samples are
given in Fig. 7 up to 800 C. Both CeeNieCN and CeeNi cat-
alysts had one obvious CO
2
desorption peak at about 120 C.
This low-temperature CO
2
stripping peak centered at about
120 C were ascribed to low-strength basic sites, which
derived from the interaction between CO
2
and the weak
basic surface [53]. The broad desorption peaks from about
250 C to 500 C might be related to the moderate basic sites.
The strong desorption peaks presented at above 500 C were
ascribed to the existence of strong basic sites, which has
little contribution to CO
2
methanation [54]. The peak area of
low-strength and moderate basic sites is listed in Table 3.It
Fig. 6 eXPS spectra of (a) survey, (b) Ce 3d, (c) O 1s and (d) Ni 2p for reduced samples.
Table 2 eThe XPS analysis of the reduced catalysts.
Catalysts Ce
3þ
/(Ce
3þ
þCe
4þ
)Ni
0
/(Ni
0
þNiO) O
b
/(O
a
þO
b
) Ni/Ce
Surface
a
Bulk
b
CeeNi-reduced 0.233 0.363 0.289 0.169 0.184
CeeNieCN-reduced 0.318 0.405 0.338 0.181 0.177
a
The surface Ni contents were determined by XPS.
b
The bulk Ni contents were determined by SEM-EDX (Fig. S6) analysis.
Fig. 7 eCO
2
-TPD profiles of reduced CeeNi and CeeNieCN.
international journal of hydrogen energy 43 (2018) 15191e15204 15197

could be found that the peak area of 50e250 C is similar on
both samples. However, the peak area of moderate basic
sites is much higher on CeeNieCN sample. This indicated
the modification could enhance the moderate basic sites,
which is favor for the adsorption and activation of CO
2
.
Moreover, we could find that the slight shift of the weak and
moderate basic sites to higher temperature and the
increasing amount of the weak and moderate basic sites,
demonstrating the stronger binding and more sites forma-
tion on the modified sample.
Catalytic performance and selectivity
The catalytic performance of the obtained samples for the CO
2
methanation reaction was studied. Fig. 8a displays the CO
2
conversion as a function of temperature. The value of T
50
,
corresponding to the temperature at which 50% conversion is
obtained, was 305 and 330 C for CeeNieCN and CeeNi,
respectively. The sample of CeeNieCN thus clearly exhibits a
significantly enhanced low-temperature activity for CO
2
methanation. Moreover, it should be noted that the CeeNie
CN sample shows very satisfactory selectivity to CH
4
(above
99%) over the whole temperature region, while CeeNi has
lower selectivity at low temperature. The results demonstrate
that the g-C
3
N
4
modified sample displays significantly
enhanced low-temperature activity and selectivity. The T
50
are compared with literature values in Table 4. According to
Table 4, the T
50
shows a relative lower temperature than the
literatures value reported. In addition, we have tested these
catalysts for three times to study the reusability. The reus-
ability of the catalysts shown in Fig. S8 indicated that the CO
2
conversion and CH
4
selectivity did not change very much
during the repeating test. This confirms the good reusability of
these catalysts.
In-situ DRIFTS measurements
In-situ DRIFTS was performed to identify the species formed
on the surface of catalyst during the methanation reaction
and to understand the reaction mechanism. DRIFTS results
were recorded by introducing continuous flow of CO
2
followed
by H
2
to the reduced CeeNi (Fig. 9a,b) and CeeNieCN (Fig. 9c,d)
catalysts. Spectra were recorded every 15 min between 150
and 400 C at intervals of 25 C. Prior to the CO
2
methanation
experiments, the samples were reduced in H
2
at 450 C for 1 h
and then purged with He for 30 min. In both samples, the
bands from 3500 to 3750 cm
1
arises from two kinds of hy-
droxyl group, type I and type II-A [58]. As from Fig. 9a,c, the
intensity of these bands decreases with the increasing tem-
perature. After introducing CO
2
/H
2
(1:4 v/v), a band at ca.
2349 cm
1
came out immediately, indicating the presence of
adsorption of gas-phase CO
2
[59]. New intensity IR features
developed at 2850, 1592, 1393, and 1376 cm
1
, which can be
assigned to adsorbed formate species [60]. The feature at
1065 cm
1
was attributed to unidentate carbonate species [61],
which can react with surface OH groups to form bicarbonate.
The IR band at 1442 cm
1
represents adsorbed bicarbonates
[62]. It was previously reported that the formates are derived
from the hydrogenation of bicarbonate assisted by the
hydrogen on nickel particles [63]. The characteristic signal of
carboxylate (CO
2
d
, 1288 cm
1
)[64] exists on both samples,
which is prominent in the lower temperature. In addition,
Table 3 ePeak area of low-strength and moderate basic
sites.
Sample Peak area (mV$
o
C)
50e250 C 250e500 C
CeeNi 8506 5448
CeeNieCN 8291 6409
Fig. 8 e(a) CO
2
conversion vs temperature for CeeNi and CeeNieCN. (b) CH
4
selectivity vs temperature for CeeNi and CeeNie
CN.
Table 4 eComparison of T
50
of CO
2
methanation with
literatures value.
Catalyst Conditions T
50
(C) Reference
CeeNieCN 8%CO
2
,
32%H
2
, balance He
305 This work
Ni/Sm
2
O
3
10%CO
2
,
40%H
2
, balance N
2
~345 [55]
Ni/g-Al
2
O
3
10%CO
2
,
40%H
2
, balance N
2
~340 [55]
Ni/TiO
2
l%CO
2
,
50%H
2
, balance He
~345 [56]
20N/NB 15%CO
2
,
60%H
2
, balance N
2
~400 [57]
20N/B 15%CO
2
,
60%H
2
, balance N
2
~340 [57]
international journal of hydrogen energy 43 (2018) 15191e1520415198

carboxylates are not visible from the spectra on higher tem-
perature, inferring that these species convert into other in-
termediates with increasing temperature. According to
previous studies [64],CO
2
can convert to CO
2
d
on CeO
2
-based
catalysts. The adsorption of CO
2
on CeO
2
is mainly due to the
catalytic role of Ce
3þ
as Lewis sites. At the same time, the
characteristic band of methoxy (1008 cm
1
)[65] was observed
in the reaction process, with the concurrent decomposition of
formats to methoxy. Concomitantly to the progressive hy-
drogenation of methoxy with increasing temperature, the
formation of methane could be observed at around 3014 cm
1
and 1306 cm
1
[17]. This indicates that from this temperature
onwards, most CO
2
converted to CH
4
via methoxy interme-
diate species. Apparently, even at 250 C, the minor methane
peaks could be found on CeeNieCN (Fig. 8c,d). In 2127 cm
1
,a
tiny peak could be observed, which is attributed to CO adsorb
on Ce
3þ
sites [66,67].
At present, two CO formation routes have been proposed in
the literature [30e34]. The first one suggests the CO comes
from the decomposition of adsorbed formate species, which is
able to react with hydrogen to produce methane as an inter-
mediate. The other proposes that the dissociation of CO
2
into
CO
(ads)
and O
(ads)
species. The CO formation mechanism over
NieCe and NieCeeCN was investigated through a series of
DRIFTS measurements carried out under transient conditions
by switching between different gas streams (He, H
2
or CO
2
)at
300 C.
Fig. 10 displays a series of IR spectra recorded during CO
2
exposure (Fig. 10a,d), subsequent purging with He (Fig. 10b,e)
and finally H
2
(Fig. 10c,f) of both samples. Prior to spectral
acquisition at 300 C, the catalyst was reduced at 450 C for 1 h
and purged by He for 30 min at the same temperature. The IR
bands during CO
2
exposure are related to adsorbed species on
ceria support, as well as gas phase CO
2
(2349 and 1216 cm
1
)
[56]. Likewise, adsorbed formats, which have bands situated at
1557 and 1389 cm
1
, are present on both samples upon CO
2
introducing [57]. Bands situated at 1288 and 1049 cm
1
represent carboxylate (CO
2
d
) and unidentate carbonate spe-
cies, respectively [58,61]. From the spectra, it is notable that
the peak of 1049 cm
1
appears after introducing CO
2
for a long
time. This implies that unidentate carbonate species are
important intermediate products. When CO
2
was removed
from the gas feed by purging the cell with He (20e40 min,
Fig. 10a,e), the intensities of gas phase CO
2
decreased rapidly.
Introducing H
2
into the cell (t ¼40 min) resulted in fast in-
tensity decreases of the IR features associated with formats
and carboxlate/carbonate species. Meanwhile, in Fig. 10c,f, IR
bands of adsorbed CO (2184 and 2127 cm
1
) came out in the IR
spectrum, which was due to CO adsorbed on Ce
4þ
and Ce
3þ
,
respectively [63,64]. However, due to the reduction effect of H
2
on Ce
4þ
, the peaks of CO adsorbed on Ce
4þ
sites soon disap-
pear and the CO adsorb on Ce
3þ
bands increase at first and
then vanish gradually with time as the formate-related bands
simultaneously become weaker. In the meantime, minor
methane band at 3016 cm
1
exists even though the adsorbed
CO vanishes totally. Through these observations, we deduce
that production of CO does not occur in He but occurs under
H
2
, furthermore, the production of methane may not originate
from the hydrogenation of adsorbed CO but from the
decomposition of adsorbed formate species. As the in-situ
transient DRIFTS spectra shown in Fig. S7, the formation of
CH
4
ceased as soon as the IR bands of adsorbed formate dis-
appeared while the adsorbed CO species on Ce
3þ
sites stayed
the same, affirming that CH
4
is produced as a result of
Fig. 9 eCO
2
methanation DRIFTS experiments on the reduced (a, b) CeeNi and (c, d) CeeNieCN.
international journal of hydrogen energy 43 (2018) 15191e15204 15199

hydrogenation of adsorbed formate species. The main reason
is adsorbed CO on Ce
3þ
sites which are not adjacent to Ni sites
cannot get the dissociated H for further hydrogenation. While
the CO adsorbed on Ni
0
sites get the dissociated H and react
with it. Then the following process may follow the CO
methanation process.
To study which reaction route is the CO from, a series of
DRIFTS measurements were carried out under transit condi-
tions by switching between different gas streams (He, H
2
or
HCOOH) at 300 C. Prior to the transit experiments, the sam-
ples were reduced in H
2
at 450 C for 1 h and then purged with
He for 30 min.
Fig. 11a displays the DRIFTS spectra of HCOOH adsorption
for reduced CeeNi catalyst. We find four adsorbed formate
species on 2946, 2850, 1567 and 1359 cm
1
[57]. After intro-
duction of He, the intensity of formate species decreased a
little (Fig. 11b) and the CO
2
peak at 2349 cm
1
disappeared
completely. In Fig. 11c, when the flow is switched from He to
H
2
, the formation of adsorbed CO peak on Ce
3þ
(2127 cm
1
)
arises accompanied by simultaneous diminishment of the
adsorbed formate IR signals, clearly indicating the adsorbed
formate is converted to CO only in the presence of H
2
. The
series of IR spectra for CeeNieCN collected during the first
20 min adsorption of HCOOH is presented in Fig. 11d. Upon
exposure of the reduced CeeNieCN sample to HCOOH at
300 C, IR bands developed at 2946, 2850, 2720 and 1567 cm
1
,
which are characteristic of adsorbed formate species [57].
Besides, bicarbonate (1430 and 1260 cm
1
) and unidentate
carbonate (1342 and 1049 cm
1
) could also be detected [58,59].
The 2127 cm
1
peak can be assigned to chemisorbed CO on
Ce
3þ
sites [63,64]. When HCOOH was removed from the gas by
purging the gas with He (20e40 min, Fig. 11e), the intensities of
the above-mentioned formate, bicarbonate and unidentate
carbonate species remain almost unchanged. The intensity of
CO
2
(2127 cm
1
) became weaker. Introducing H
2
into the feed
gas (40e60 min) resulted in quick decreases of formate species
associated with unidentate carbonate and bicarbonate spe-
cies. Very strong IR band of gas phase methane (3016 cm
1
)
appeared in IR spectrum, whereas the intensities of adsorbed
CO remained much weaker over the entire 20 min even before
the gas-phase CO
2
signal reached stability (2349 cm
1
), which
also confirms that the adsorbed CO species are derived from
the decomposition of formate. The formation of gas-phase
CO
2
at 2349 cm
1
can be seen on both samples when
HCOOH was introduced. As reported by Rhodes et al. [68], the
dehydrogenation reaction is readily catalysed by metal and
basic metal oxides to CO
2
and H
2
. The produced H
2
could
reduce the formate to CO. So this can also explain the exis-
tence of small amount of CO when pure HCOOH was intro-
duced. The intensity of formate species adsorbed on CeeNie
CN is stronger than that in CeeNi, indicating that the modi-
fication provided more basic sites for HCOOH to adsorb.
Similarly, Bentiez and co-authors studied HCOOH hydroge-
nation on La-promoted Rh/Al
2
O
3
samples and found out the
higher the surface basicity, the higher amount of formate is
bonded to it [69]. This is consistent with the results of CO
2
-
TPD. When there are less basic sites, the adsorption ability of
formate species on the catalysts is weak. After modification,
the peak position slightly shifted to higher temperature and
the amount of the moderate basic sites became more. The
more basic sites only lead to more binding formats species.
The slight shift of the peak to higher temperature demon-
strates the stronger binding sites formation on the modified
sample. Thus we could infer that more strong-binding formats
exist on the catalysts. The weak-binding formate species are
more inclined to decompose into CO, while these strong-
Fig. 10 eDRIFS spectra collected at 300 C when the feed gas was switched (a, d) from He to CO
2
followed by a switch back to
(b, e) He and then to (c, f) H
2
over CeeNi (a, b and C) and CeeNieCN (d, e and f).
international journal of hydrogen energy 43 (2018) 15191e1520415200

binding formate species tend to convert to methoxy species
and then produce methane. This can explain why less amount
of CO produced when modified with g-C
3
N
4
.
Reaction kinetics of the CeeNieCN catalyst
Based on the in-situ DRIFTS results, a kinetic rate law can be
derived based on the proposed reaction mechanism. This
mechanism involving the formation of formate on catalysts
containing ceria is similar to the mechanism proposed by
others [70,77].
H2þ2M/2H M3.1
HMþO/OH*þ*M3.2
CO2þO/OCO2*3.3
OCO2*þOH /OCOOH*þO3.4
OCOOH*þH2/OCOH*þH2O 3.5
OCOH*þH2/OCH2OH 3.6
OCH2OH*þH2/OCH3*þH2O 3.7
OCH3*þH2/CH4þOH 3.8
*M refers to the Ni metal active site and O* refers to the surface
oxygen active site.
Based on the mechanism and Table 5, many different
kinetic rate expressions can be derived based on different
assumptions of the identity of the rate determining step and
Most Abundant Reaction Intermediate (MARI) species. The
kinetic study was conducted at the temperature range of
483e513 K. The form of the kinetic rate expressions
were determined from derivation of various Langmuir-
Hinshelwood and Eiley-Rideal based on the reaction mech-
anism or taken from previously reported CO
2
methanation
studies [71,73]. The full list of kinetic models tested is shown
in Table S2. The models were then fitted to the rate data
according to a least-squares method. Reaction expressions
were checked for physical feasibility, for example, expres-
sions with negative kinetic parameters were rejected. Other
expressions were rejected based on poor fit to the data. In
addition, an Arrhenius plot or a Van't Hoff plot for activation
energy (E
a
) or equilibrium constant (K
eq
) was conducted on
the remaining pool of models, and non-linear plots were
rejected. The computer fitting of the rate expression which
had the least square errors and a feasible rate parameters is
the model which assumes formation of hydrogen carbonate,
step (3.4), to be rate determining, with adsorbed CO
2
and
adsorbed hydroxyl to be MARI species, and steps (3.1e3.3) to
be in quasi-equilibrium. The derivation of this rate equation
as follows:
Determination of OH*. Based on steady state approxima-
tion of step (3.1):
Fig. 11 eDRIFS spectra collected at 300 C when the feed gas was switched (a, d) from He to HCOOH followed by a switch
back to (b, e) He and then to (c, f) H
2
over CeeNi (a, b and c) and CeeNieCN (d, e and f).
Table 5 eParameters for kinetic model.
Temperature/K k4K3K0:5
1K2K0:5
1K2K3
483 109.61 31.884 3.9603
493 215.28 23.159 3.0861
503 184.32 18.369 2.0052
513 345.62 17.909 2.0141
Activation energy (kJ/mol) 53.3 32.2 39.0
R
2
value 0.844 0.965 0.900
international journal of hydrogen energy 43 (2018) 15191e15204 15201

½HM2¼K1PH2½M2
½HM¼K0:5
1P0:5
H2½M
Based on steady state approximation of step (3.2):
½OH  ½M¼K2½HM½O
½OH ¼ K0:5
1K2P0:5
H2½O
Determination of OCO2*. Based on steady state approxi-
mation of step (3.3):
½OCO2 ¼ K3PCO2½O
Surface oxygen species balance with OCO2* and OH* as
MARI:
½O
t¼½Oþ½OCO2þ½OH 
½O
t¼½O
1þK3PCO2þK0:5
1K2P0:5
H2
Determination of rate determining step as step (3.4):
ro¼k4½OCO2½OH 
ro¼k4K3PCO2K0:5
1K2P0:5
H2½O
2
Substituting ½O:
ro
½O
2
t
¼k4K3K0:5
1K2PCO2P0:5
H2
1þK0:5
1K2P0:5
H2þK3PCO22
The kinetic parameters are shown in Table 5.
The apparent activation energy of the rate determining
step is 53.3 kJ/mol. This value is in line with the literature. A
comparison of this value for the energy of activation with
values previously reported is given in Table 6 and is seen to be
in good agreement with these data which cover a wide range
of conditions. The kinetic models reported for Ni/SiO
2
cata-
lysts [71e73] tested in this study had a large residual error
relative to the above kinetic model. The Ni/SiO
2
kinetic models
were all based on formation of CO as a reaction intermediary
followed by hydrogenation, whereas the IR studies indicate
that CO was not a key reaction intermediary for CeeNieCN.
The kinetic data is consistent with this finding; the poor fit of
the Ni/SiO
2
catalyst kinetic models is because CO
2
methana-
tion proceeds via a difference reaction path over Ni/CeO
2
. The
negative slope for the natural logarithm plots for K0:5
1K2and K3
terms are consistent with thermodynamic equilibrium ad-
sorptions of H
2
and CO
2
(Fig. 12).
Table 6 eKinetics of methanation of carbon dioxide.
Catalyts CeeNieCN Ni/KG Ni/Al
2
O
3
Ru/Al
2
O
3
Ni/SiO
2
Ru/Al
2
O
3
Ea (kJ/mol) 53.5 55e58 106 71 94 68
T (K) 483e513 555e673 473e503 477e633 500e600 408e473
Reference This work [74] [75] [75] [68] [76]
Fig. 12 eArrhenius plot or Van't Hoff plot of kinetic
parameters.
Fig. 13 eProposed mechanism for the CO
2
methanation.
international journal of hydrogen energy 43 (2018) 15191e1520415202

Based on the kinetic results, the IR spectra under transient
and steady-state conditions, a reaction mechanism that could
account for these observations is presented in Fig. 13. The first
step of the overall CO
2
methanation process is the adsorption
of CO
2
on Ce
3þ
sites to form carboxylate (CO
2
d
). In the subse-
quent steps, carboxylate can further convert to unidentate
carbonate species on surface oxygen sites adjacent to Ce
3þ
sites [77]. The bicarbonates are derived from reaction between
unidentate carbonate and OH on ceria support. In presence of
hydrogen, the-thus formed bicarbonates react with dissoci-
ated hydrogen on Ni
0
particles to form formats and then
methoxy species and to release CH
4
. In the overall CO
2
methanation process, CO does not participate in the main
reaction path-way and the CO adsorbed on Ce
3þ
sites cannot
react with the adsorbed H to form CH
4
. It only exists as a by-
product. Combined with the results of CO
2
-TPD and the
transit DRIFTS, weak-binding formates species are more in-
clined to decompose into adsorbed CO on Ce
3þ
sites only in
the presence of H
2
, while strong-binding formate species tend
to be further hydrogenation and then release methane. Ac-
cording to the proposed mechanism in Fig. 13, we can see that
both Ni and Ce were required in CO
2
methanation reaction
mechanism, so it is necessary to enhance the interaction be-
tween metal and support to accelerate the hydrogenation
process. The higher amount of Ni surface area and weak basic
sites observed for CeeNieCN can be explained the lower re-
action temperatures in enhancing CO
2
hydrogenation
performance.
Conclusions
In summary, a promising catalyst was prepared by the g-C
3
N
4
-
modified sol-gel method. The catalyst shows excellent activity
and selectivity for CO
2
methanation. It was found that with
the introduction of g-C
3
N
4
, the catalysts were improved
distinctly on the physicochemical properties, which is favor
for the dissociation of H
2
and adsorption of CO
2
. Moreover, g-
C
3
N
4
could serve as a sacrificial template, which could anchor
and prevent the nickel particles from migrating and sintering.
In-situ DRIFTS measurements suggest that the CO
2
hydroge-
nation reaction does not need CO as an intermediate species,
but through the formation of formate species. In presence of
H
2
, the weak-binding formate species dissociate directly to
adsorbed CO on Ce
3þ
sites, which can not react with the
adsorbed H to produce CH
4
. The kinetics results are in good
agreement with Langmuir-Hinshelwood type mechanism in
which the formation of bicarbonate is the rate-determining
step (RDS). The method reported here can be extended to
fabricate other types of heterogeneous catalyst with signifi-
cantly improved activity and selectivity.
Acknowledgments
This work was financially supported by the National Uni-
versity of Singapore and NEA Project (ETRP 1501103, R279-
000-491-279), A*STAR for the funding from AME IRG 2017
Project R279-000-509-305, Industry-Academia Cooperation
Innovation Fund Projects of Jiangsu Province (BY2016004-09),
Jiangsu Province Scientific and Technological Achievements
into a Special Fund Project (BA2015062, BA2016055 and
BA2017095), A Project Funded by the Priority Academic Pro-
gram Development of Jiangsu Higher Education of Jiangsu
Higher Education Institutions. Yang Yu would like to thank
the China Scholarship Council for financially supporting his
Ph.D. work.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.ijhydene.2018.06.090.
references
[1] Aziz MAA, Jalil AA, Triwahyono S, Ahmad A. Green Chem
2015;17:2647e63.
[2] Najafpour MM, Renger G, Hoły
nska M, Moghaddam AN,
Aro EM, Carpentier R, et al. Chem Rev 2016;116:2886e936.
[3] Schrag DP. Science 2007;315:812e3.
[4] Zhang LZ, Lu YQ, Rostam-Abadi M. Phys Chem Chem Phys
2012;14:16633e43.
[5] Kwak BS, Vignesh K, Park NK, Ryu HJ, Baek JI, Kang M. Fuel
2015;143:570e6.
[6] Zhang ZF, Hu SQ, Song JL, Li WJ, Yang GY, Han BX.
ChemSusChem 2009;2:234e8.
[7] Wang W, Wang S, Ma X, Gong J. Chem Soc Rev
2011;40:3703e27.
[8] Song CS. Catal Today 2006;115:2e32.
[9] Abe T, Tanizawa M, Watanabe K, Taguchi A. Energy Environ
Sci 2009;2:315e21.
[10] Ku
smierz M. Catal Today 2008;137:429e32.
[11] Kowalczyk Z, Stolecki K, Rarog-Pilecka W, Mi
skiewicz E,
Wilczkowska E, Karpi
nski Z. Appl Catal A Gen 2008;342:35e9.
[12] Du GA, Lim SY, Yang YH, Wang C, Pfefferle L, Haller GL. J
Catal 2007;249:370e9.
[13] Hetterley RD, Mackey R, Jones JTA, Khimyak YZ, Fogg AM,
Kozhevnikov IV. J Catal 2008;258:250e5.
[14] Lang YQ, Wang QQ, Xing JM, Zhang B, Liu HZ. AIChE J
2008;54:2303e9.
[15] Wang Q, O'Hare D. Chem Rev 2012;112:4124e55.
[16] Wei M, Zhang X, Evans DG, Duan X, Li XJ, Chen H. AIChE J
2010;53:2916e24.
[17] Ang ML, Oemar U, Saw ET, Mo L, Kathiraser Y, Chia BH, et al.
ACS Catal 2014;4:3237e48.
[18] Bian ZF, Suryawinata IY, Kawi S. Appl Catal B Environ
2016;195:1e8.
[19] He S, Li CM, Chen H, Su DS, Zhang BS, Cao XZ, et al. Chem
Mater 2013;25:1040e6.
[20] Vance CK, Bartholomew CH. Appl Catal 1983;7:169e77.
[21] Liu J, Li CM, Wang F, He S, Chen H, Zhao YF, et al. Catal Sci
Technol 2013;3:2627e33.
[22] Ashok J, Ang ML, Terence PZL, Kawi S. ChemCatChem
2016;8:1308e18.
[23] Zhen WL, Li B, Lu GX, Ma JT. Chem Commun
2015;51:1728e31.
[24] Miyazawa T, Kimura T, Nishikawa J, Kado S, Kunimori K,
Tomishige K. Catal Today 2006;115:254e62.
[25] Ocampo F, Louis B, Roger AC. Appl Catal A Gen
2009;369:90e6.
international journal of hydrogen energy 43 (2018) 15191e15204 15203

[26] Huang LY, Xu H, Li YP, Li HM, Cheng XN, Xia JX, et al. Dalton
Trans 2013;42:8606e16.
[27] Wang NN, Zhou Y, Chen CH, Cheng LY, Ding HM. Catal
Commun 2016;73:74e9.
[28] Zhou XJ, Dong YY, Li XY, Zhao QD, Cui YB, Lu G. Catal
Commun 2015;69:109e13.
[29] Fischer A, Mu
¨ller JO, Antonietti M, Thomas A. ACS Nano
2008;2:2489e96.
[30] Peebles DE, Goodman DW, White JM. J Phys Chem
1983;87:4378e87.
[31] Marwood M, Doepper R, Renken A. Appl Catal A
1997;151:223e46.
[32] Falconer JL, Zagli AE. J Catal 1980;62:280e5.
[33] Fujita S, Terunuma H, Kobayashi H, Takezawa N. React Kinet
Catal Lett 1987;33:179e84.
[34] Schild C, Wokaun A, Baiker A. J Mol Catal 1990;63:243e54.
[35] Wang W, Wang SP, Ma XB, Gong JL. Chem Soc Rev
2011;40:3703e27.
[36] Yu Y, Zhong Q, Cai W, Ding J. J Mol Catal A Chem
2015;398:344e52.
[37] Tada S, Yokoyama M, Kikuchi R, Haneda T, Kameyama H. J
Phy Chem C 2013;117:14652e8.
[38] Wei YG, Wang H, Li KZ, Zhu X, Du YP. J Rare Earth
2010;28:357e61.
[39] Tai LW, Lessing PA. J Mater Res 1992;7:502e10.
[40] Huang LY, Li YP, Xu H, Xu YG, Xia JX, Wang K, et al. RSC Adv
2013;3:10968e74.
[41] Luo JY, Meng M, Li X, Li XG, Zha YQ, Hu TD, et al. J Catal
2008;254:310e24.
[42] Shi F, Tse MK, Pohl MM, Bru
¨ckner A, Zhang SM, Beller M.
Angew Chem Int Ed 2007;46:8866e8.
[43] Gupta A, Waghmare U, Hegde MS. Chem Mater
2010;22:5184e98.
[44] Wang XQ, Rodriguez JA, Hanson JC, Gamarra D, Martinez-
Arias A, Fernandez-Garcia M, et al. J Phys Chem B
2005;109:19595e603.
[45] Deng J, Chu W, Wang B, Xu ZX, Yang W, Zhao XS. Catal Sci
Technol 2016;6:4565e76.
[46] Xue L, Zhang CB, He H, Teraoka Y. Appl Catal B Environ
2007;75:167e74.
[47] Wang N, Shen K, Yu XP, Qian WZ, Chu W. Catal Sci Technol
2013;3:2278e87.
[48] Perez-Alonso FJ, Meli
an-Cabrera I, Granados ML, Kapteijn F,
Fierro JLG. J Catal 2006;239:340e6.
[49] Damyanova S, Bueno JMC. Appl Catal A 2003;253:135e50.
[50] B
eche E, Charvin P, Peranau D, Abanades S, Flamant G. Surf
Interface Anal 2008;40:264e7.
[51] Najafpour MM, Mehrabani S, Bagheri R, Song ZL, Shen JR,
Allakhverdiev SI. Int J Hydrogen Energy 2018;43:2083e90.
[52] Boaro M, Giordano F, Recchia S, Santo VD, Giona M,
Trovarelli A. Appl Catal B Environ 2004;52:225e37.
[53] Pino L, Vita A, Cipiti F, Lagan
a M, Recupero V. Appl Catal B
Environ 2011;104:64e73.
[54] Yan Y, Dai YH, He H, Yu YB, Yang YH. Appl Catal B Environ
2016;196:108e16.
[55] Muroyama H, Tsuda Y, Asakoshi T, Masitah H, Okanishi T,
Eguchi TMK. J Catal 2016;343:178e84.
[56] Le TA, Kim MS, Lee SH, Kim TW, Park ED. Catal Today
2017;293e294:89e96.
[57] Lu XP, Gu FN, Liu Q, Gao JJ, Liu YJ, Li HF, et al. Fuel Process
Technol 2015;135:34e46.
[58] Saw ET, Oemar U, Tan XR, Du Y, Borgna A, Hidajat K, et al. J
Catal 2014;314:32e46.
[59] Gregoire G, Brinkmann NR, van Heijnsbergen D, Schaefer HF,
Duncan MA. J Phys Chem A 2003;107:218e27.
[60] Holmgren A, Andersson B, Duprez D. Appl Catal B Environ
1999;22:215e30.
[61] Li C, Sakata Y, Arai T, Domen K, Maruya KI, Onishi T. J Chem
Soc Faraday Trans 1989;85:1451e61.
[62] Vimont A, Lavalley JC, Sahibed-Dine A, Arean CO,
Delgado MR, Daturi M. J Phys Chem B 2005;109:9656e64.
[63] Zhu F, Zhu GL, Zhou S. Chemsuschem 2015;8(4):609e12.
[64] Graciani J, Mudiyanselage K, Xu F, Baber AE, Evans J,
Senanayake SD, et al. Science 2014;345:546e50.
[65] Wang F, He S, Chen H, Wang B, Zheng LR, Wei M, et al. J Am
Chem Soc 2016;138:6298e305.
[66] Resini C, Venkov T, Hadjiivanov K, Presto S, Riani P,
Marazza R, et al. Appl Catal A 2009;353:137e43.
[67] Gamarra D, Belver C, Fern
andez-Garcia M, Martı´nez-Arias A.
J Am Chem Soc 2011;129:12064e5.
[68] Rhodes C, Hutchings GJ, Ward AM. Catal Today
1995;23:43e58.
[69] Benitez JJ, Carrizosa I, Odriozola JA. Appl Surf Sci
1993;68:565e73.
[70] Aldana PAU, Ocampo F, Kobl K, Louis B, Thibault-Starzyk F,
Daturi M, et al. Catal Today 2013;215:201e7.
[71] Weatherbee GD, Bartholomew CH. J Catal 1982;77:460e72.
[72] Fujita S, Terunuma H, Nakamura M, Takezawa N. Ind Eng
Chem Res 1991;30(6):1146e51.
[73] Inoue H, Funakoshi M. J Chem Eng Jpn 1984;17(6):602e10.
[74] Dew JN, White RR, Sliepcevich CM. Ind Eng Chem
1955;1:140e6.
[75] Herwijnen TV, Doesburg HV, de Jong WA. J Catal
1973;1:391e402.
[76] Solymosi F, Erd€
ohelyi A, B
ans
agi T. J Catal 1981;68:371e92.
[77] Pan QS, Peng JX, Wang S, Wang SD. Catal Sci Technol
2014;4:502e9.
international journal of hydrogen energy 43 (2018) 15191e1520415204