Structural tuning and catalysis of tungsten carbides for the regioselective cleavage of C–O bonds

Article abstract of DOI:10.1016/j.jcat.2018.11.020

Tungsten carbides exhibit excellent performance in many heterogeneous processes because of their distinctive catalytic properties. Preparation of tungsten carbides with controllable phase composition relevant to their catalytic behavior is essential yet challenging. In this study, tungsten carbides embedded in carbon spheres (W_xC@CS) were fabricated through carburization of organic–inorganic hybrid precursors. W_1.25C@CS with rational structure-tuning properties exhibits promising regioselectivity (reaching 91.5%) toward aryl C–O bond cleavage, specifically during hydrogenolysis of guaiacol to phenol. A structure reconstruction strategy was adopted to elucidate structure–performance relationship by transforming commercially available bulk WC from inert phase to composition-dependent active catalysts. Combined catalytic and characteristic analyses illustrate that the catalyst performance is dependent on the C-defect structure. The intimate connection between the phenol space time yield and the C/W atomic ratio on the exterior interface of the catalyst was verified. The C/W atomic ratio of 7.2 leads to the optimal catalytic performance. Density functional theory calculations were performed to define the catalytic mechanism at the atomic level. The theoretical analysis suggests an appropriate configuration of surface W and C atoms for activation of hydrogen and guaiacol molecules, rendering the intrinsic active sites for phenol production. This work provides insights into controlling the surface compositions of tungsten carbides to develop efficient C–O bond cleavage catalysts, which verifies the importance of hydrogenolysis catalysis in lignin-derived compounds involving complex O-containing guaiacols and phenolics.

Full text of DOI:10.1016/j.jcat.2018.11.020

Journal of Catalysis 369 (2019) 283–295
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Structural tuning and catalysis of tungsten carbides for the regioselective
cleavage of CAO bonds
a
b,
⇑
a
a
a
a
Huihuang Fang , Alberto Roldan , Chenchen Tian , Yanping Zheng , Xinping Duan , Kun Chen ,
a
b
a,
⇑
Linmin Ye , Stefano Leoni , Youzhu Yuan
a
State Key Laboratory of Physical Chemistry of Solid Surfaces, National Engineering Laboratory for Green Chemical Productions of Alcohols-Ethers-Esters and iChEM, College
of Chemistry and Chemical Engineering, Xiamen University, Xiamen 361005, China
School of Chemistry, Cardiff University, Main Building, Park Place, Cardiff CF10 3AT, United Kingdom
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Tungsten carbides exhibit excellent performance in many heterogeneous processes because of their dis-
tinctive catalytic properties. Preparation of tungsten carbides with controllable phase composition rele-
vant to their catalytic behavior is essential yet challenging. In this study, tungsten carbides embedded in
Received 25 August 2018
Revised 12 November 2018
Accepted 19 November 2018
x
carbon spheres (W C@CS) were fabricated through carburization of organic–inorganic hybrid precursors.
W1.25C@CS with rational structure-tuning properties exhibits promising regioselectivity (reaching 91.5%)
toward aryl CAO bond cleavage, specifically during hydrogenolysis of guaiacol to phenol. A structure
reconstruction strategy was adopted to elucidate structure–performance relationship by transforming
commercially available bulk WC from inert phase to composition-dependent active catalysts.
Combined catalytic and characteristic analyses illustrate that the catalyst performance is dependent on
the C-defect structure. The intimate connection between the phenol space time yield and the C/W atomic
ratio on the exterior interface of the catalyst was verified. The C/W atomic ratio of 7.2 leads to the optimal
catalytic performance. Density functional theory calculations were performed to define the catalytic
mechanism at the atomic level. The theoretical analysis suggests an appropriate configuration of surface
W and C atoms for activation of hydrogen and guaiacol molecules, rendering the intrinsic active sites for
phenol production. This work provides insights into controlling the surface compositions of tungsten car-
bides to develop efficient CAO bond cleavage catalysts, which verifies the importance of hydrogenolysis
catalysis in lignin-derived compounds involving complex O-containing guaiacols and phenolics.
Ó 2018 Elsevier Inc. All rights reserved.
Keywords:
Tungsten carbide
Guaiacol
Hydrogenolysis
CAO bond cleavage
Phenol
Regioselectivity
1
. Introduction
Transition metal carbides (TMCs) have gained substantial atten-
higher d-band electronic density of states at the Fermi level,
thereby affecting their catalytic performance [6,11,12]. TMC for-
mation, carbon diffusion, and carburization control are important
in designing efficient TMC catalysts with dominant catalytic
behavior that distinguishes them from their parent metal [13].
As typical TMCs, tungsten carbides possess excellent catalytic
performance in hydrogenolysis, hydrogenation, isomerization,
tion as potential catalytic materials in heterogeneous catalysis due
to their distinctive catalytic behavior in various significant reac-
tions [1–8]. The insertion of carbon atoms into the parent transi-
tion metals induces structural rearrangement, which then alters
steric and electronic properties. These changes contribute to the
formation of bonds with certain covalent characteristics and
charge transfer between the metal and the carbon atoms, leading
to the formation of TMCs (e.g., tungsten carbides and molybdenum
carbides) [2,9,10]. Theoretical studies have indicated that the
presence of carbon in the lattice of the parent metal promotes a
significant redistribution of the density of states and leads to
deoxygenation, and H
ter and co-workers [3,4,7,16] developed W
2
evolution reactions (HER) [3–7,14–17]. Bit-
C-based catalysts for
2
efficient deoxygenation of oxygen-rich biomass-derived feeds,
including stearic acid, oleic acid, and guaiacol, to produce high
yields of hydrocarbons and phenolics. Previous reports showed
that WC monocarbides possess HER activity and are thus promis-
ing candidates for catalysis of such reaction [18,19]. Gong et al.
[
2
6] synthesized phase-pure W C nanoparticles supported on
carbon nanotubes that displayed superior HER performance to that
of WC materials. The difference in their activities relies on the
carburizing level and structure tuning. However, the structure–
⇑
Corresponding authors.
E-mail addresses: RoldanMartinezA@cardiff.ac.uk (A. Roldan), yzyuan@xmu.edu.
cn (Y. Yuan).
https://doi.org/10.1016/j.jcat.2018.11.020
0
021-9517/Ó 2018 Elsevier Inc. All rights reserved.

2
84
H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
activity relationship in these nanoparticles has been rarely studied
due to the complexity of carbide phases involved in phase purity,
surface defects, surface carbon, surface termination, and surface
oxides [20,21]. Therefore, controlling carbide phase formulation
and understanding the origin of catalytic activity are important
yet challenging to be investigated.
A critical issue in facilitating the identification of phase formula-
tion and surface compositions is the accurate synthesis of TMCs by
controlling carbon diffusion. Previous reports showed that the typ-
ical protocol for synthesis of TMCs is temperature-programmed
yellow powder precursors. The precursors were washed, dried,
and carburized at desired temperatures under H . Typically,
W@CS was prepared by carburizing at 700 °C for 2 h at a heating
rate of 5 °C/min. W C@CS was carefully carburized at 800 °C at a
heating rate of 1 °C/min for 30 min. W1.25C@CS and WC@CS were
carburized at 850 °C at a heating rate of 3 °C/min for 3 and 6 h,
respectively. Prior to exposure to air, the as-prepared samples were
2
2
2 2
passivated by 1% O /99% N for 30 min.
Commercially available WC (denoted as Com-WC) samples with
different surface reconstructions were prepared. Com-WC was
obtained from Aladdin Co. Ltd. In brief, 1 g of Com-WC and
4 3 8
reduction using gaseous carbon precursors, such as CH , C H , or
CO [22–25]. In this method, carbon diffusion rapidly occurs through
the gas–solid interface and is barely modulable, which results in the
uncontrollability of the phase composition (i.e., C-defects on the
carbide catalysts). To solve this problem, researchers have devel-
oped and applied carbothermal hydrogen reduction by using solid
carbon as a controllable source during pyrolysis [3,6,14,15,26–
6
240 mg of WCl were dispersed in 100 mL of ethanol under contin-
uous stirring for 30 min. The solution was slowly added with 4 mL
of water and heated to 80 °C overnight. The obtained solid was fil-
tered, washed several times, and vacuum dried. W-Com-WC was
pretreated at 450 °C for 4 h at a rate of 5 °C/min under H
mal W1+xC-Com-WC was first pretreated from 20 °C to 450 °C at a
rate of 5 °C/min under H and from 450 °C to 750 °C at a rate of 1 °C
/min for 1 h under 15% CH /H2. C-defect W C-Com-WC samples
with different surface C/W atomic ratios were pretreated with sim-
ilar methods under H , followed by 450 °C to 750 °C at a rate of
1 °C/min for 0–6 h under 15% CH /H . Surface C/W atomic ratios
were calculated from the HS-LEIS profiles. WO -Com-WC was pre-
treated at 450 °C for 4 h at a rate of 5 °C/min under Ar. Similar pas-
2
. The opti-
2
8]. This method slows down carbon diffusion through the solid–
2
solid interface to avoid excessive carbon deposition and simplify
the tuning of the phase composition. Various carbon materials, such
as activated carbon, carbon nanotubes, and carbon fibers, are used
to produce metal carbides [3,6,15]. In addition, Xu and Wu et al.
4
x
2
4
2
[
27,28] used metal–organic frameworks as precursors to synthesize
metal carbide nanoparticles by pyrolysis and carbon diffusion. In
our previous work, we prepared tungsten carbides (W C@CS) with
3
x
sivated treatments were conducted to prepare W
x
C@CS prior to
well-defined phase composition by controlling the carburizing level
of phenolic polymers in the presence of tungsten precursors. The
optimal catalyst exhibited promising performance for selective
hydrogenolysis of aryl ether CAO bonds [14] over O-containing
lignin derived compounds including guaiacol, diphenyl ether,
phenetole, veratrole, dimethoxyphenol and anisole, of particular
importance connecting the conversion and valorization of
oxygen-rich lignocellulosic biomass [14,29–31]. This method for
carbide synthesis displays significant advances in fabrication and
control of structural phases. Nevertheless, the identity of reactive
phase, the role of C defects on metal surroundings, and the struc-
ture–activity relationship remain unclear.
exposure to air.
2.2. Activity test
The catalytic hydrogenolysis for guaiacol was performed on a
fixed bed reactor with a computer-controlled autosampling system.
In brief, 200 mg of fresh catalyst was loaded in the center of the
quartz tubular reactor sandwiched with quartz power. Before test-
ing, the catalysts were pretreated at 450 °C for 4 h under 5% H
at a heating rate of 3 °C/min. The catalyst bed was cooled naturally
to target the desired temperature, and pure H gas was fed into the
reactor at 3.0 MPa. Liquid guaiacol was pumped into the reactor by
a Series III digital HPLC pump (Scientific Systems, Inc.) with
required weight liquid hourly space velocity (WLHSV). The prod-
ucts were analyzed online using an Agilent 7890A gas chro-
matograph (GC) equipped with an autosampling value, flame
ionization detector, and a DB-Wax capillary column. A GC 2060
with thermal conductivity detector and a TDX column were used
2 2
/N
2
x
In this work, W C@CS catalysts with different phase formula-
tions were fabricated and applied to catalyze the hydrogenolysis
of guaiacol. This study aimed to determine the capability of the
catalysts to cleave CAO bonds and elucidate the role of active C-
defect sites. The catalysts were also characterized in detail through
X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS),
high-sensitivity low-energy ion scattering spectroscopy (HS-LEIS),
transmission electron microscopy (TEM), and temperature pro-
grammed desorption (TPD) analyses and first-density functional
4 2 2
to analyze gas products (CH , CO, CO , and H O). The absence of
heat and mass transfer limitations were confirmed by the estimated
Thiele modulus and Mears criterion [32–34]; see supporting infor-
mation for details. Detailed guaiacol conversion, product selectivity
and carbon balance were calculated by Eqs. (1)–(4) as follows.
(
DFT) calculations. The optimal W1.25C@CS displays high activity
for phenol production because of the surface configuration and
variation in electronic properties through appropriate carburiza-
tion control. A strategy was designed by reconstructing commer-
cially available WC and used to understand the origin of catalytic
activity. A correlation between phenol space time yield (STY) and
the surface C/W atomic ratio was established and the plausible
scheme of guaiacol hydrogenolysis was proposed.
ðmoles of GUAÞin ꢀ ðmoles of GUAÞout
Conversion ¼
ðmoles of GUAÞin
ꢁ 100%
ð1Þ
ð2Þ
moles of ring product i
S
C6i
C1i
¼
¼
ꢁ 100%
the sum moles of guaiacol consumed
2
. Experimental
moles of produced methane or methanol
the sum moles of methane and methanol
S
ꢁ 100%
ð3Þ
2.1. Synthesis of materials
moles of C measured from the reactor effluent
Carbon balance ¼
moles of C in the guaiacol feed
Tungsten carbides (W
x
C@CS, CS: carbon spheres) were synthe-
sized by carburization of organic–inorganic hybrid precursors
14]. In brief, 5.0 g of meta-tungstate and 1.2 g of resorcinol were
dissolved in deionized water under continuous stirring for
0 min. The solution was then slowly added with 2.3 mL of
formaldehyde and heated to reflux at 85 °C for 24 h to obtain
ð4Þ
[
Unless otherwise noted, the carbon balance was about 95% ± 2%.
SC6i was obtained based on the number of C6 rings except for C1
3
4 3
products (CH and CH OH); SC1i was calculated based on the

H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
285
number of methoxy groups from the reacted substrate. In some
cases, the total selectivity to C1 products is less than 100% because
of the formation of coke or trace transalkylation reaction.
The rate of guaiacol conversion was calculated as k = (F/m)ln
of 550 eV to expand the plane-waves of the Kohn–Sham valence
states [38]. The inner electrons were represented by the
projector-augmented wave (PAW) pseudopotentials by consider-
ing also the nonspherical contributions from the gradient correc-
tions [39]. All the calculations were performed using long-range
dispersion correction approach by Grimme (D3) [40]; this
approach is an improved version of pure DFT when considering
large polarizable atoms [41–43]. The optimization thresholds were
(
1/1 ꢀ X), where k is the rate constant expressed as moles of gua-
iacol consumed per second and per gram of catalyst, F is the molar
flow rate of guaiacol, m represents the mass of catalyst, and X is the
total conversion of guaiacol. All the catalysts were evaluated twice
to ensure the accuracy of measurements (the errors are below 5%).
The Arrhenius plots and activation energies were estimated based
on the activity at different temperatures on the condition that the
conversion was below 40%. The turnover frequency (TOF) values
were based on the CO chemisorption uptake, indicating the moles
of guaiacol converted by per active site at the catalyst surface per
ꢀ5
10 eV and 0.01 eV/Å for electronic and ionic relaxation, respec-
tively. The Brillouin zone was sampled by -centered k-point mesh
generated through a Monkhorst–Pack grid [44] with a maximum
C
ꢀ1
separation of 0.20 Å between k-points to ensure electronic and
ionic convergence. Electronic partial occupancies were determined
using tetrahedron method with Blöchl corrections to increase the
integration efficiency for all the calculations [45].
ꢀ1
ꢀ1
ꢀ1
second (mol-guaiacol mol-site
s , for short, s ). The guaiacol
conversion for the TOF calculation was lower than 10%.
The volume, shape, and internal coordinates of the W
2
C bulk
(
crystal symmetries P-31m) were optimized, leading to cell param-
2
.3. Material characterization
eters a = b = 5.190 Å, c = 4.724 Å, and formation angle of 120.0°,
which are in full agreement with the experimental benchmark
(a = b = 5.208 Å and c = 4.737 Å) [46]. The bulk of WC has a crystal
symmetry P-6m2; after relaxation within the computational setup,
the bulk remained in the same symmetry with cell parameters of
a = b = 2.906 Å and c = 2.838 Å, which are also in excellent agree-
ment with crystallographic data (a = b = 2.904 Å and c = 2.383 Å)
[47].
XRD analysis was conducted using a Rigaku Ultima IV X-ray
a
diffractometer equipped with Cu–K radiation (35 kV and 15 mA)
at scanning 2h from 10° to 90°. The obtained diffraction data were
analyzed using the JCPDS database. The XRD Rietveld refinement
was carried out using Topas software to calculate W/W
2
C/WC
weight ratio.
TEM, HR-TEM, and HAADF-STEM images were obtained using a
Philips Analytical FEI Tecnai 30 electron microscope operated at an
acceleration voltage of 300 kV. Fresh samples were dispersed ultra-
sonically, dropped, and dried on copper grid with lacey support
films.
Scanning electron microscopy (SEM) images were recorded
using Hitachi S4800 electron microscope to observe the morphol-
ogy of the samples.
We have simulated the low-Miller index surfaces considering
different terminations, i.e., carbon and tungsten, similar to those
in slab models. The slab models were built following dipole
method that is based on the condition that the net dipole perpen-
dicular to the surface must be zero, as determined by Tasker [48].
All the surfaces studied were generated by cutting the optimized
bulk of the iron minerals by using the METADISE code [49]. We
added a vacuum width of 15 Å between vertically repeated slabs
to avoid spurious interaction between them. These slabs models
contain sufficient atomic layers to allow the uppermost five layers
to be completely relaxed without symmetry restrictions, and the
bottom ones were frozen at the bulk lattice parameter. Table S1
contains a summary of areas and surface energies calculated
according to Eq. (5) for each slab investigated. The surface energy
XPS measurements were conducted on an Omicron Sphera II
photeoelectron spectrometer equipped with an Al–K
tion source (h = 1486.6 eV). The binding energy was calibrated
using the C 1s peak at 284.5 eV. Before XPS experiments, the sam-
ples were pretreated with 5%H /N at 450 °C for 4 h to ensure these
a
X-ray radia-
m
2
2
samples are in similar states with fresh ones before the activity
test. HS-LEIS profiles were obtained by an Ion-TOF Qtac100 instru-
ment. He was selected as the ion source at a kinetic energy of 3 keV
(c), which measures the excess of energy of the surface with
respect to the bulk, has been calculated as Eq. (5).
ꢀ2
with ion flux of 6000 pA m and a spot size of 2000 mm ꢁ 2000 mm
to obtain surface information and minimize surface damage. The C/
W atomic ratio on the surface was calculated by integrating the
relative intensity of W and C from both HS-LEIS and XPS measure-
ments. To eliminate the interference of surface oxidation as much
as possible, the data were collected based on the deconvoluted
spectra by excluding the influence of oxide species.
unrelaxed
slab
E
slab ꢀ nEbulk
E
ꢀ nEbulk
c
¼
ꢀ
ð5Þ
A
2A
unrelaxed
where Eslab is the absolute energy of the relaxed slab; Eslab
is the
energy of surface slab where the atoms are frozen at their bulk posi-
tions; Ebulk is the energy of bulk carbide; n is the number of stoichio-
metric units in the surface cell; and A is the surface area.
CO and
Micromeritics AutoChem II 2920 chemisorption analyzer. In brief,
00 mg of the sample was used in each test. Before test, the sample
2
H -TPD measurements were carried out with a
2
3. Results and discussion
was in situ pretreated under the condition similar to that of the
activity evaluation of the catalysts. Ar gas was introduced to clean
the surface of the sample for 1 h at 400 °C. The sample was then
3.1. Preparation of catalysts with controllable phase composition
cooled down at room temperature and subjected to switching H
2
x
The W C@CS catalysts were synthesized by carburization of
or CO stream for adsorption for 1 h. The sample was flushed with
Ar until a stable baseline was achieved. TPD measurements were
conducted from 30 °C to 500 °C at a ramping rate of 5 °C/min.
The desorbed exit gases were monitored with a thermal conductiv-
ity detector.
organic–inorganic hybrid precursors over specific heating rate,
carrier-gas flow, and carburizing time. Scheme 1 shows the evolu-
tion of crystalline phase under the carburization treatment [14].
Metal W nanoparticles were generated through the carbothermal
hydrogen reduction at the first stage; then the diffusion of intersti-
tial carbon atoms into the reduced metal W promoted the forma-
2.4. Computational details
tion of the W
2
C phase. Furthermore, deeper carburization formed
C content.
the WC crystalline phase with gradually declining W
2
Periodic plane-wave DFT calculations were performed using the
Vienna ab-initio simulation package (VASP) [35,36] and the open-
shell Perdew–Burke–Ernzerhof functional [37] at a kinetic energy
We identified the crystallinity and phase purity using the XRD pat-
terns (Fig. S1). The formation of various crystalline phases depends
on the control of carburizing level. The sharp diffraction peaks cen-

2
86
H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
Scheme 1. Evolution of crystalline phase under carburization of precursors.
tred at 2h of 40.42°, 58.36°, and 73.33° display the high crystallinity
of metal W (PDF#04-0806), as depicted in W@CS. By contrast,
such as Pt and Ru (dash line) [50,51]; the other mechanisms
encompass the direct hydrogenolysis of CAO bonds to give phe-
nols, which may be then overhydrogenated into corresponding
cyclohexanol and cyclohexane (dark line) [7,14,29]. As an interme-
diate product, phenol could be generated by direct cleavage of aryl
W
3
2
C@CS exhibits distinctive peaks located at 34.47°, 38.10°, and
9.67°, which are indexed to the (1 0 0), (0 0 2), and (1 0 1) of the
facets of W C (PDF#65-3896), respectively. W C has a hexagonal
2
2
structure (P-31m) with lower C composition, while WC contains
a hexagonal structure of P-6m2; whereas the C planes are sand-
wiched between tungsten layers. With deeper carburization, typi-
cal peaks centered at 31.70°, 35.89°, and 48.65° are visible in
CAO bonds (producing PhOH and CH
aliphatic CAO bonds firstly (producing catechol and CH
followed by dehydroxylation of catechol to phenol. In our case, no
catechol was observed over optimal W1.25C@CS catalyst with dif-
ferent conversions, which confirms the direct hydrogenolysis of
3
OH) or the cleavage of weak
), and then
4
W
1.25C@CS, revealing the generation of WC (PDF#65-4539). The
XRD pattern in WC@CS obtained well-crystallized WC phase with-
out any W C or metal W phase, indicating complete carburization.
3
CarylAOCH bond of guaiacol to phenol; this is in agreement with
2
previous reports [7,14].
For further analysis, Table S2 shows the identification of phase
composition by XRD Rietveld refinement and other physicochemi-
cal properties (see Section 2 for details on the preparation and
characterization process).
As shown in Fig. 1, the W-based catalysts show preferences on
the hydrogenolysis of CAO bonds without arene hydrogenation.
W@CS presents a 28.2% guaiacol conversion through cleavage of
aliphatic CH
3
AO bonds, producing 56.7% of catechol selectivity.
C@CS exhibits a high conversion of 83.6%
2
In order, W@CS, W C@CS, and WC@CS, with an optimal
In contrast, the W
2
W
1.25C@CS, were readily fabricated through the method state
towards phenol, anisole, and benzene, indicating a great capability
for hydrogenolysis of aryl CAO bonds with poor selectivity. The
optimal W1.25C@CS catalyst exhibits the highest conversion of
98.6% with 91.5% phenol selectivity, exhibiting a regioselective
above. Notably, the phase evolution of tungsten carbides affects
the structural and electronic properties and hence, the catalytic
performance. In this case, the phase composition can be designed
rationally by optimizing the carburizing level, which may facilitate
seeking promising active phases for a wider catalytic application.
For theoretical calculation, we modeled the catalyst surfaces
3
cleavage of CarylAOCH bond. Absolute carburization over the
WC@CS catalyst caused the rapid decline of activity, obtaining only
2
with the lowest surface energies, i.e., W C{1 0 1} and WC{1 1 0},
in agreement with the XRD data. We built several structures,
decreasing the C/W ratio from the pristine WC{1 1 0} to 0.88, con-
sidering different carbon vacancies, i.e., top and subsurface car-
bons, as well as scattered and contiguous positions. Their relative
Phenol
Cresols
Anisole
Benzene
Others
Catechol
C1 product select. / %
14.1
81.7
CH4
MeOH 37.8
57.2
16.1
80.6
19.2
70.1
17.3
75.8
stabilities are shown in Fig. S2, where the relative energy (
DE) con-
siders the species’ chemical potential as unity. Hence, E is defined
D
1
00
as the energy difference between the vacancy system plus the
number of carbons (as in graphite) and the pristine WC{1 1 0}. It
shows that any C-vacancy above a C/W ratio of 0.918 is more stable
at the top atomic layer, and below that ratio, the vacancy will be at
80
the subsurface. We simulated a W
x
C structure with C/W ratio of
6
4
2
0
0
0
0
6
2.5%, in agreement with the reconstructed W1+xC surface ratio.
3.2. Catalytic performance
The catalytic conversion of guaiacol involves the cleavage of
diverse CAO bonds and the hydrogenation of arene species. We
selected this reaction as a model to elucidate the active sites of
the catalysts and understand the structure–performance relation-
ship. Scheme S1 shows two competitive pathways on guaiacol con-
version: one involves the hydrogenation of the aromatic ring to
alkylated cyclohexanols, followed by the cleavage of aliphatic
CAO bonds to cyclohexane, which usually occur on noble metals,
a
b
c
d
e
Fig. 1. Catalytic performance of GUA conversion (black dot) over tungsten carbide
catalysts with different phase composition: (a) W@CS, (b) W C@CS, (c) W1.25C@CS,
d) WC@CS, (e) Physically mixed W C@CS and WC@CS with the similar phase
composition with W1.25C@CS. Reaction conditions: weight liquid hourly space
2
(
2
ꢀ
1
2 2
velocity (WLHSV) = 3.0 h , P(H ) = 3.0 MPa, H /GUA molar ratio = 50, T = 300 °C.

H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
287
3
0.1% guaiacol conversion and 70.7% phenol selectivity. As shown
atomic formulations on the kinetics of guaiacol conversion, as
shown in Table 1. The reaction rate of C@CS is about
, which is nearly 2 fold higher than that of
1.25C@CS. With further carburization, WC@CS presents a rate of
in Table S2, the molar ratios of Ccarbides/CMatrix were calculated
based on a set of characterizations to correlate the hydrogenolysis
properties. With the increase of carburization, the Ccarbides/CMatrix
molar ratios displayed an increasing trend, giving values of 3.1%,
W
2
ꢀ
5
ꢀ1 ꢀ1
2.1 ꢁ 10 mol g
s
W
ꢀ6
ꢀ1 ꢀ1
7.0 ꢁ 10 mol g
s . In addition, W@CS displays the lowest rate
5
.4%, and 7.7% for W
2
C@CS, W1.25C@CS and WC@CS catalysts,
although it has the highest CO chemisorption uptake. The TOF val-
ues were further confirmed the capability for guaiacol conversion
over these samples with different carbidic or metal species.
respectively. Interestingly, the catalytic performance of these cata-
lysts showed significant difference. The guaiacol conversion exhib-
ited a volcanic trend and the product selectivities toward phenol,
anisole, and benzene were varied with increase of Ccarbides content.
W@CS with nearly no Ccarbides species showed a fairly poor guaiacol
conversion with high catechol selectivity. That is to say, the trans-
formation of carbides could lead to different electronic properties
in bulk and surface atomic formulations, significantly altering the
hydrogenolysis of guaiacol. To further confirm the effect of struc-
ture transformation on the catalytic performance, controlled
ꢀ
2 ꢀ1
W
2
C@CS possesses the highest TOF of 6.1 ꢁ 10
s , and with fur-
ther C insertion to the matrix carbides, the TOF decrease obviously,
ꢀ2
ꢀ1
ꢀ2 ꢀ1
which is 4.8 ꢁ 10
s
and 1.7 ꢁ 10
s
on W1.25C@CS and
WC@CS, respectively. In contrast, W@CS with almost W terminals
on the surface exhibits the lowest TOF value. This indicates that the
active sites contributed by formulation C and W atom have differ-
ent catalytic ability for guaiacol conversion.
We explored different adsorption sites following standard sim-
ulation methods and found the different interaction strength
between molecules and catalysts. The slabs models were based
2
experiments were conducted by physical mixture of W C@CS and
WC@CS as a contrastive catalyst. The mixed catalysts showed
poorer performance (11.5% and 73.7% selectivities of anisole and
phenol, respectively) even when the catalyst contained a similar
2
on supercell with areas 136.27, 109.75 and 114.21 Å respectively
2
for W C, W1.25C and WC. These areas ensure minimal steric effect
amount of W
2
C and WC with the original W1.25C@CS catalyst.
between molecules as we adsorbed a single species at a time. Over-
all, the adsorption energies of the different species are stronger on
W C, followed by WC and W1.25C (Table 2). On W C, the adsorption
2 2
The apparent activation energy (Ea) is a critical parameter
directly related to the performance of the catalysts. We used the
Arrhenius plots to derive the Ea for guaiacol conversion (Fig. 2).
Ea was 189.7 kJ/mol for W@CS, which is considerably higher than
of phenol is stronger than any other species investigated, which
may explain yields on further hydrogenated species, e.g. benzene.
This is also confirmed by the experimental data as shown in
Table S3. By contrast, the adsorption energies of W1.25C and WC
become weaker with the degree of hydrogenation. The strength
of these interactions is expressed also on the distances between
the molecule and the surface (Table 2). Furthermore, the adsorp-
tion energies and distances relate well with the charge transfer
2
those found for tungsten carbides. Notably, W C@CS displays a
quite low Ea value of 111.1 kJ/mol compared with the other W-
based catalysts considered here. In fact, the Ea increases with the
catalyst carburization up to a constant of approximately 150.0 kJ/-
mol. Thus, W C performs the cleavage of CAO bonds faster than WC
2
but with poor regioselectivity. W1.25C@CS has a moderate Ea value
of 145.5 kJ/mol and displays the highest phenol selectivity. To con-
2
from the surface to the molecule. The charge transferred on W C
firm the capabilities of CarylAOCH
species, anisole, with only one similar aryl methoxyl group, was
evaluated at low conversions over the W C@CS and WC@CS
Table S3). The Ea for anisole conversion over W C@CS is about
5.8 kJ/mol, which is similar with the report by Chen and Bhan
3
cleavage over different carbidic
(0.7 e) is more than twice that on WC (0.3 e) and considerably lar-
ger than that on W1.25C (0.1 e). The charge transferred does not
directly correlates with the work function of surface, i.e., 4.2,
2
(
8
2
2
4.13, and 4.53 eV, respectively, for W C, W1.25C, and WC, which
indicates that the redox properties of the catalysts does not control
the process. Apart from a relatively low adsorption energy with
limited electron transfer to the molecule, we cannot identify any
other clear trend between the electronic and geometric structure
with the catalytic activity.
[
33]. For WC@CS, the Ea is 112.3 kJ/mol, which is much higher than
that of W C@CS. Both guaiacol and anisole show a significant acti-
vation barrier over the W C@CS and WC@CS catalysts. The differ-
2
2
ence in Ea is most likely caused by the structure transformation
and appropriate carbon diffusion to parent tungsten, which is
essential for the regioselective hydrogenolysis of CarylAOCH
The TOF values and reaction rates were further calculated to
investigate the effect of carburization with different C and W
3
bond.
3.3. Texture feature, electronic property, and chemisorption behavior
of the catalysts
TEM measurements were used to gain more detailed informa-
tion about the textural structure. The samples with different car-
burization show similar morphologies (Fig. S3). The carbide
nanoparticles are formed and evenly embedded in carbon spheres
with average diameters of 4–10 nm. Taking W1.25C@CS nanoparti-
cles as representatives, the HR-TEM analyses are shown in Fig. 3.
Fig. 3a exhibits clear lattice fringes with interplanar distances of
-
-
-
-
-
-
3
4
5
6
7
1
51.9 kJ/mol
1
11.1 kJ/mol
0
.227, 0.259, 0.188, and 0.252 nm, corresponding to the (1 0 1),
(
1 0 0) facets of W C and WC, respectively. In addition, the Fast
2
Fourier transform (FFT) images of given regions in blue and yellow
rectangle (Fig. 3b and c) were also analyzed, and the results man-
ifest the (h k l) planes ascribed to the diffractions of W C (blue
2
rectangle) and WC (yellow rectangle) facets, which indicates the
coexistence of different crystallographic planes. For more evidence,
a set of seven interplanar spacings, marked by double-headed
arrow in Fig. 3a, were carefully measured; and the results are
shown in Fig. 3d and e. The distance of regions i and ii are 1.589
W C@CS
W C@CS 189.7 kJ/mol
WC@CS
W@CS
2
x
1
45.5 kJ/mol
8
1
.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00
000/T / K
and 1.762 nm, respectively, owing to the W
(
(
2
C (1 0 1) and WC
1 0 0) facets. More interestingly, clear dark spots were observed
marked by white arrows) in these carbide nanoparticles (Fig. 3f–
1
Fig. 2. Arrhenius plots of reaction rate (ln(r)) versus 1/T for GUA conversion over
W-based catalysts.
h). These spots are ascribed to the C vacancies on the surface of

2
88
H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
Table 1
Catalytic performance (reaction rate, apparent Ea and TOF) of guaiacol conversion over W-based catalysts with varying numbers of active sites measured by CO chemisorption.
a
ꢀ1
b
ꢀ1 ꢀ1
ꢀ1
c
ꢀ1
Catalyst
C
carbides/CMatrix molar ratio /%
CO uptake/
l
mol gcat.
Rate /mol g
s
Ea/kJ mol
TOF /s
5
ꢀ2
ꢀ2
ꢀ2
ꢀ3
W
W
WC@CS
W@CS
2
C@CS
1.25C@CS
3.1
5.4
7.7
ꢂ0
36.3
29.8
23.6
40.7
2.1 ꢁ 10ꢀ
111.1
145.5
151.9
189.7
6.1 ꢁ 10
4.8 ꢁ 10
1.7 ꢁ 10
8.4 ꢁ 10
ꢀ
5
6
6
1.1 ꢁ 10
7.0 ꢁ 10
ꢀ
2.3 ꢁ 10ꢀ
a
b
c
Calculated based on the W and C contents.
Obtained by steady state values at 573 K; the conversion is below 10% by changing the WLHSV.
TOF values are based on the CO chemisorption uptake over different samples with <10% conversion; reaction conditions: T = 300 °C, P (H
2
2
) = 3.0 MPa, H /guaiacol molar
ratio = 50.
Table 2
2
Summary of adsorption energies (Eads) and average distance between the molecules and the surfaces (dmolecule-surface) on the W C, W1.25C and WC surfaces.
Molecule
E
ads (eV)
dmolecule-surface (Å)
2
W C
W
1.25
C
WC
2
W C
W
1.25
C
WC
Guaiacol
Catechol
Anisole
Phenol
ꢀ3.04
ꢀ2.68
ꢀ2.70
ꢀ3.33
ꢀ2.23
ꢀ2.11
ꢀ1.76
ꢀ1.80
ꢀ1.64
ꢀ1.05
ꢀ2.43
ꢀ2.37
–
2.60
2.44
2.58
2.23
2.33
2.72
2.60
2.84
2.70
2.82
2.75
2.72
–
2.56
2.70
ꢀ2.10
Benzene
ꢀ2.07
nanoparticles, in agreement with previous reports [17]. It is well-
known that, vacancies can be regarded as volume defects, and they
are capable of trapping and activating reactants. These results
reveal a transformation of diverse crystal defects with C-defect
sites by using the carburization of organic–inorganic hybrid
precursors.
XPS analyses were carried out, and the results are shown in
Fig. 4 and Table S4. As indicated from the deconvoluted W 4f spec-
trum in Fig. 4a, W@CS was featured with two peaks at 29.1 and
CO, and they can be restored through removal of CO, indicating the
CO reversible poisoning of the reactive sites activating substrates.
In our case, CO-TPD was performed to provide further evidence
of the W–C electron relocation as a function of the carburization
degree, as indicated in Fig. 5a. On the W@CS sample, CO desorbs
as two large broad peaks at 83 °C and 149 °C. Theoretical studies
have shown that CO bonding with C end down was favored on
the metallic sites, providing greater electron donation and leading
to a higher desorption energy [10]. Unlike W@CS, W
2
C@CS,
3
1.2 eV ascribed to the metal W. This finding implies that metallic
W1.25C@CS, and WC@CS display weaker absorption. With deeper
W is the majority species in the W@CS catalyst. By contrast, two
characteristic carbidic peaks appear in tungsten carbide samples.
The W C@CS displays two pronounced peaks centered at 31.4
2
carburization, the CO desorption peaks shifted to lower tempera-
tures from 83 °C to 70 °C and 149 °C to 132 °C, respectively. The
two broad desorption peaks were narrowed and divided in two
more obvious peaks in W1.25C@CS, indicating the existence of
two active sites. Interestingly, the WC@CS displays only a lower
desorption peak probably due to the complete formation of WC.
The weaker CO absorption on carbides indicates that metallic sites
with lower electron density existed on the carbide surface. This
finding was also proven by XPS analysis.
and 33.7 eV, corresponding to the tungsten carbide species. A tiny
peak assigned to metallic W still existed, presumably due to slight
lack of carbon atoms on the surface. Moreover, the binding energy
of W 4f is shifted from 31.4 eV to 31.9 eV with deeper carburiza-
tion. As a consequence of surface passivation during the catalyst
preparation, the peaks assigned to the oxides (35.3 and 37.5 eV)
were inevitable on the surface of all samples. Indeed, the d-band
of the electronic density of states derived from our computer mod-
els show a shift of the occupied d-band center of 0.30 eV as a func-
tion of the carbonization level (Fig. S4). The higher binding energy
found in the XPS implies that W provides electrons to C species,
resulting in a greater accumulation of electron density around C.
This was also probed as evidence in Fig. 4b. The deconvolution of
C 1s energy level signals revealed two peaks, which were ascribed
to graphitic and carbidic species. The peak located at 284.5 eV
belongs to the graphitic C and is unchanged during the carburiza-
tion. However, the carbidic peak first centered at 282.1 eV in
2
Investigation of the absorption behavior of hydrogen in H -
related reactions, such as hydrogenolysis, is also important. As
shown in Fig. 5b, two broad hydrogen desorption peaks can be rec-
ognized in these samples. W@CS shows smaller and weaker hydro-
gen desorption peaks than those of carbides, and the major peak
2
was located at 76 °C. While W C@CS and W1.25C@CS display larger
peaks centered at 89 °C with a more obvious shoulder peak at a
higher temperature of 120 °C. With absolute carburization, WC@CS
demonstrated a slight decrease in the amount of hydrogen desorp-
tion. These results reveal that carbides bonded hydrogen more
strongly than that of metal W, affording more disassociated and
absorbed hydrogen on the carbide surface. DFT calculations and
experiments have shown that hydrogen can be activated, adsorbed,
and bonded to the hollow sites on the surface of metal carbides;
and the increase of the d-band density of states caused by the
insertion of carbon atoms is also beneficial for hydrogen activation
[10,53,54]. The difference on the chemisorption behaviors imply
that the C-defect sites can be tuned by carburizing level, in agree-
ment with the TEM analyses mentioned above. As a matter of fact,
2
W C@CS and then shifted to lower binding energy of 281.2 eV in
WC@CS with increasing carburization. The change in binding ener-
gies suggested that the carburization control is particularly impor-
tant and results in charge transfer between W and C.
A set of typical TPD measurements were carried out to gain
insight into the chemisorption behaviors. Carbon monoxide (CO)
is a typical probe molecule to investigate the relative electron
donating and accepting capability of the reactants, as well as sur-
face reactive sites of heterogeneous catalysts [8,52]. Lee et al. [8]
studied CO chemical titration to determine the reactive sites in
2
the CO-TPD and H -TPD results correlated well with the finding
that weaker CO absorbed sites and more H-occupied hollow sites
existed on the surface of carbides. Therefore, a rational carburiza-
tion of the catalyst offers an appropriated activation of reactants.
2
Mo C-catalyzed anisole hydrodeoxygenation. They have found that
the benzene synthesis rate remarkably declined in the presence of

H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
289
Fig. 3. HRTEM image of samples: (a and g) W1.25C@CS; (f) W
FFT) images of given regions in blue and yellow rectangle, respectively. (d and e) The measured d spacing of nanostructured units marked by double-headed arrow; and i, ii
reveal 7 interplanar spacing in different regions taken from (a).
2
C@CS; (h) WC@CS; the white arrows show C vacancies in carbide crystallites. (b and c) Fast Fourier transform
(
3
.4. Activity improvement by surface reconstruction of an inert bulk
Fig. S6. Specifically, the precursor with a careful carburization dis-
plays two typical peaks of W1+xC species by insufficient insertion of
carbon atoms, revealing successful modification with more C-
defect sites on the surface, namely W1+xC-Com-WC.
WC phase
To exclude the effect of supported carbon and gain insight into
the catalytic origin on the surface configuration, reconstruction
methods on the Com-WC were carried out to provide further
experimental evidence. No residual carbon was detected in Com-
WC sample from the Raman spectra (Fig. S5). To create C-defect
surfaces, tungsten precursors were deposited onto the surface of
Typical morphologies of Com-WC and W1+xC-Com-WC were
identified by SEM, TEM, HAADF, and HR-TEM measurements. As
indicated in Fig. S7, large Com-WC bulks were observed clearly
with smooth surfaces but became rougher with rich stacking faults
after modification, revealing the successful deposition of a noncon-
tinuous carbide phase. Structurally, Com-WC possesses bare lattice
fringe with an interplanar distance of 0.188 nm, fitting with the
typical (1 0 1) facets of WC (Fig. 6a). By contrast, as sketched from
Fig. 6b, W1+xC-Com-WC provides grain lattice facets, stacking
faults and C-defect sites. Clear lattice fringes of 0.227 and
6
Com-WC by hydrolysis of WCl , and then a careful carburization
control was further used for formation of W1+xC phase on the sur-
face. For comparison, a set of catalysts derived from Com-WC by
modification with other tungsten species were also prepared (Sec-
tion 2 for details). The XRD patterns of samples were used to inves-
tigate the crystallinity and phase composition, as illustrated in
2
0.249 nm, corresponding to the (1 0 1) facet of W C and the

2
90
H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
Oxides
WC@CS
Carbides
1.9
Metal
a
a
72
3
WC@CS
133
W1.25C@CS
70
W1.25C@CS
31.4
1
32
49
W C@CS
2
W C@CS
2
1
8
3
W@CS
W@CS
5
0
100 150 200 250 300 350
3
8
36
34
32
30
28
o
Binding energy / eV
Temperature / C
Graphitic
Carbidic
281.2
b
b
120
WC@CS
WC@CS
W1.25C@CS
W1.25C@CS
8
9
W C@CS
2
282.1
W C@CS
2
76
W@CS
250
W@CS
5
0
100
150
200
300
2
90 288 286 284 282 280 278
Binding energy / eV
o
Temperature / C
2
Fig. 5. (a) CO-TPD and (b) H -TPD profiles of W-based catalysts.
Fig. 4. XPS profiles of W-based catalysts: (a) W 4f and (b) C 1s.
duction of acid sites. However, W1+xC-Com-WC exhibited a signif-
icant enhancement both in guaiacol conversion and phenol
selectivity, reaching 65.3% and 85.3%, respectively. Moreover, with
further carburization, WC-Com-WC displayed a decreased conver-
sion although the phenol selectivity is maintained at a high level,
which is in agreement with the previous result of WC@CS. To bet-
ter understand the effect of surface reconstruction on reaction
pathways, we evaluated guaiacol over these samples at similar
conversion by changing the WLHSV. W1+xC-Com-WC displayed
17.1% guaiacol conversion with high phenol selectivity of 91.2 at
a WLHSV of 10 h . WC-Com-WC showed 10.3% guaiacol conver-
sion with slight increase of phenol selectivity from 70.9% to
77.4%. The different product distributions over these samples
under similar low conversions indicated the importance of surface
formulation of C and W atoms. C1 product selectivity was further
provided to confirm catalytic nature of metal tungsten and carbidic
tungsten species; tungsten carbides produce methanol as major C1
product while metal tungsten and tungsten oxides obtain methane
as major product. These finding are corresponding to the Table S5.
Therefore, the regioselective hydrogenolysis of guaiacol to phenol
depends on the surface reconfiguration between W and C atoms,
i.e., the incorporation of rational C-defect sites.
(
1 0 0) facet of WC, have newly emerged and were located at the
same region (yellow rectangle), revealing a coexistence of different
crystallographic planes. These diverse facets expose different crys-
tal defects and cause the formation of stacking faults and bound-
aries, leading to the atom dislocation and distinct configuration
between W and C atoms. A lattice fringe of 0.156 nm in red rectan-
gle is visible, and it is attributed to the (2 0 0) facet of W presum-
ably due to the uncompleted carburization. The FFT images of the
given regions in red, blue, and yellow rectangles also confirm the
diffractions of the (h k l) planes over these species. Most interest-
ingly, W1+xC-Com-WC was found to contain a large amount of C-
defect sites marked by white arrows, indicating the formation of
hollow active sites, which favor the activation of hydrogen and
substrates. That is to say, the reconstruction of the carbide surface
with abundant dislocations and crystal defects was successfully
achieved.
The catalytic performance for guaiacol hydrogenolysis was
evaluated to investigate the effect of these distinct modifications
on the Com-WC, as demonstrated in Table 3. Com-WC without
any modification displayed negligible activity. Upon comparison,
the activities changed after different modifications. W-Com-WC
exhibited a slightly increased conversion with high selectivity of
catechol. This reveals that the incorporation of metal W overlap
the hollow carbidic sites for the activation of hydrogen and is likely
to strengthen the absorption of guaiacol, resulting in catechol pro-
ꢀ1
The effect of surface reconfiguration was further investigated by
XPS measurements and chemisorption behavior. As shown in
Fig. S8a, two typical peaks centered at 31.8 and 33.9 eV are visible
on the W 4f profile of the Com-WC, corresponding to the carbidic
species of tungsten. Importantly, a slight negative shift of carbidic
W species occurred in W1+xC-Com-WC; such a shift might have
resulted from the formation of C-defects, affording an electron-
3 x
duction through cleavage of an aliphatic CH AO bond. WO -Com-
WC displayed similar activity with more trans-alkylated products,
such as cresols and methyl catechol, presumably due to the intro-

H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
291
Fig. 6. HR-TEM images: (a) Com-WC and (b) W1+xC-Com-WC.
Table 3
Catalytic performance of GUA conversion over Com-WC catalysts modified with different W species.
Catalyst
Conv./%
C6 product selec./%
C1 product selec./%
CH CH OH
PhOH
Anisole
Cresols
Benzene
Catechol
Others
4
3
Com-WC
7.1
51.0
19.1
85.3
91.2
70.9
77.4
22.5
0.6
0.0
0.0
0.1
0.0
0.0
0.0
9.3
2.0
9.8
5.3
10.6
7.9
6.9
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.5
70.3
0.0
0.0
0.0
36.3
8.6
4.9
21.5
80.2
15.7
11.2
18.5
17.7
73.3
64.6
13.4
75.4
83.5
62.3
69.8
14.9
a
W-Com-WC
12.6
65.3
17.1
24.0
10.3
15.5
W
W
1+xC-Com-WCb
1+xC-Com-WC
b*
3.7
c
WC-Com-WC
WC-Com-WC
18.5
14.7
14.8
c*
0.0
55.8
d
3
WO -Com-WC
Reaction conditions: weight liquid hourly space velocity (WLHSV) = 3.0 hꢀ1, P(H
2 2
) = 3.0 MPa, H /GUA molar ratio = 50, T = 300 °C.
a
2
Pretreated at 450 °C for 4 h at a rate of 5 °C/min under H .
b
b*
c
First pretreated from 20 to 450 °C at a rate of 5 °C/min under H
1+xC-Com-WC with the WLHSV of 10.0.
2
; then 450 to 750 °C at a rate of 1 °C/min for 1 h under 15%CH
4
/H
2
.
W
First pretreated from 20 to 450 °C at a rate of 5 °C/min under H
WC-Com-WC with the WLHSV of 6.
Pretreated at 450 °C for 4 h at a rate of 5 °C/min under Ar; See Section 2 for details.
2
; then 450 to 750 °C at a rate of 1 °C/min for 6 h under 15% CH
4
/H
2
.
c*
d
rich W species. The binding energy of W-Com-WC was lower than
that of W1+xC-Com-WC, which can be deconvoluted to carbidic and
metallic species. A greater divergence can be observed among
these samples from the C 1s profiles in Fig. S8b. Com-WC contains
two obvious deconvoluted peaks corresponding to the graphitic
and carbidic C species. Interestingly, the amount of carbidic C spe-
cies in the W1+xC-Com-WC decreased significantly; whereas the
graphitic C species are almost unchanged. The carbidic C species
of W-Com-WC are almost vanished. These results suggested that
C-defects are formed in the W1+xC-Com-WC, whereas W-Com-
WC was coated by metal W species on the surface. Further inves-
and hydrogen, resulting in a remarkable enhancement in catalytic
hydrogenolysis of guaiacol.
3.5. Correlation between phenol STY and surface C/W atomic ratio of
the catalysts
There is an agreement that photoelectrons escaping depth can
be penetrated up to a few nanometers; thus, XPS measurements
might not truly reflect the composition on the topmost layer. We
further explored the outer surface information combined with
HS-LEIS (Fig. 7a). Only three C, O, and W peaks exist over these
samples. The C peak of W-Com-WC is nearly disappeared, suggest-
ing an almost W-terminations on the topmost surface. With
increasing carburizing level, the peaks of carbon became larger,
but the W and O peaks are almost unchanged. These samples with
reconstruction treatment display more obscure than that of Com-
WC, indicating successful formation of C-defect sites. For exact
quantitative analysis, the C/W atomic ratio on the surface was cal-
culated by integrating the relative intensity of W and C from both
HS-LEIS and XPS measurements, as indicated in Fig. 7b. The C/W
atomic ratio of Com-WC by XPS is 1.17:1, which is close to the bulk
composition of pure WC. It showed a decline of 0.44/1 with recon-
struction of C-defect species on the surface. The results convey that
C-defect surfaces were formed by our treatments. However, W-
Com-WC displayed a quite low C/W atomic ratio of 0.15:1, suggest-
ing a surface enrichment of metal W. By contrast, the HS-LEIS
2
tigations on CO and H -TPD have been conducted to provide more
evidence on the surface defects dependent activity. In Fig. S9a,
Com-WC demonstrated weak desorption peak at approximately
8
7 °C. Meanwhile, the W1+xC-Com-WC displayed bigger and
broader peaks. Notably, a shoulder peak at higher temperature is
more obvious. These observations indicate more C vacancies (or
W-terminations) on the surface of W1+xC-Com-WC. It is worth
2
stressing that the H –TPD profiles display quite different desorp-
tion behaviors (Fig. S9b). On Com-WC, hydrogen desorbed as a tiny
broad peak, which is in agreement with the previous study [30].
However, when the Com-WC was modified by W1+xC on the sur-
face, hydrogen desorbed as two big and broad signals, indicating
the facilitation of hydrogen activation. These results reveal that
the reconstruction of surface enriched the formation of hollow car-
bidic sites and promote the capability for activation of substrate

2
92
H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
results illustrate quite different C/W atomic ratios compared with
the XPS analysis. The Com-WC possesses higher C/W atomic ratio
on the topmost surface, reaching approximately 16.0:1 and reveal-
ing C-terminations on the topmost atomic layer of Com-WC. The
ratios in W1+xC-Com-WC rapidly declined to 7.2:1, which is attrib-
uted to the insufficient carburization, leading to the formation C-
defects. Comparatively speaking, W-Com-WC displayed the lowest
C/W atomic ratio, obtaining a surface with rich W-terminations.
For quantitative analysis, the phenol STY, as a function of the
surface C/W atomic ratio, is summarized in Fig. 8. The phenol
STY changed in a volcano-type trend with the increase of C/W
atomic ratio. At low C/W ratio, the surface exposed abundant W-
terminations. These surface W atoms displayed an acceptable gua-
surface control with rich hollow C-deficient sites benefits the phe-
nol production owing to the optimal textural structure and elec-
tronic behavior.
At present, there might be some shortcomings in our study
because our samples were not synthesized and characterized
under operando conditions due to the unavailability of facilities
under extremely high temperature (over 850 °C) and inevitability
of surface oxidation of carbides. Table S6 shows that surface oxida-
tion of carbides decrease the guaiacol conversion with nearly
unchanged phenol selectivity, which indicates that these oxide
species mainly suppress the number of active sites, in agreement
with precious reports [8,55,56]. To eliminate the interference of
surface oxidation, we treated all as-prepared materials with 1%
iacol absorption and quite low H
2
absorption, resulting in a low
O
2
2
/99%N for 30 min prior to exposure to air to ensure the consis-
phenol production. As the ratio increased, the guaiacol conversion
can be significantly promoted, which is attributed to the formation
of hollow sites, as well as the electronic interaction between W and
C atom from the molecules, enhancing the catalytic absorption
behavior. The phenol STY reached the maximum value of
tency of oxycarbides (confirmed by Fig. 4).
3.6. Understanding of the origin of catalytic performance
ꢀ1
ꢀ1
In the present work, the W C@CS catalysts demonstrated
x
1
3.6 mmol gcat
h
at the C/W ratio of 7.2. Then it would decline
appropriate regioselectivity upon the hydrogenolysis of guaiacol.
Our kinetic study also demonstrated that appropriate Ea derives
a high selectivity of phenol by hydrogenolysis of aryl CAO bond,
particularly on the optimal W1.25C@CS catalyst. Apparently, the
carburizing level and configuration of W and C atoms are particu-
larly crucial in this reaction. In fact, we have shown that both
further when the surface was fully occupied by C atoms. These
results are in full agreement with the interaction energies of gua-
iacol and phenol with the slab models (Fig. 9 and Table 2). The gua-
iacol adsorption energies changed from ꢀ3.04 eV to ꢀ2.11 eV and
to ꢀ2.43 eV on W
2
C{1 0 1}, W1.25C{1 1 0}, and WC{1 1 0}, respec-
tively. Similar volcano seems to rise from the interaction between
phenol and the model surfaces, which exhibited interaction
energies of ꢀ3.33, ꢀ1.64, and ꢀ2.10 eV. That is to say, appropriate
W1.25C@CS and W1+xC-Com-WC possessed abundant crystallo-
graphic planes, boundaries, C-defects, and stacking faults, as con-
firmed by the XRD, TEM, and HR-TEM measurements. The
carburization control and the reconstruction on the surface of
Com-WC facilitated tuning the electronic property, which can be
further confirmed by XPS. An obvious electron transfer occurred
from W to C during the carburization process or reconstruction
treatment, whereas C trapped the electrons from W, resulting in
electronic richness on C. Indeed, the Bader analysis of atomic
charge distribution on the computational models showed an accu-
a
C
mulation on the carbons of 1.4, 1.2, and 1.0 electrons for in W
2
C,
O
W-Com-WC
W1.25C, and WC, respectively. Thus, although the number of elec-
W
trons transferred is lower in absolute value, the total charge per
carbon increases considerably with the level of carburization. The
electronic property can be changed subtly when these surface C
are partially etching. Our TPD measurements also proved the
appropriate activation of both hydrogen and substrate plays a vital
role in transforming guaiacol to phenol.
W1+xC-Com-WC
Com-WC
Consequently, the textural transformation of carbides involving
the insertion of carbon atoms provides a significant redistribution
5
00
1000 1500 2000 2500 3000
Energy / eV
2
1
1
0
2.0
1.6
1.2
0.8
b
1
1
1
4
2
0
8
6
4
2
0
O
OH
6
2
8
4
0
OH
W1+xC-Com-WC
0.4
0.0
Com-WC
W-Com-WC
4
6
8
10
12
14
16
18
Increase of surface C
Surface C/W atomic ratio
x
Fig. 7. (a) HS-LEIS profile of C-defect W C-Com-WC catalysts with increase of
surface C; (b) surface W/C atomic ratio of C-defect
measured by HS-LEIS and XPS.
W
x
C-Com-WC catalysts
Fig. 8. Phenol STY for guaiacol hydrogenolysis as a function of surface C/W atomic
ratio in W1+xC-Com-WC catalysts with controlled carburization.

H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
293
W
2
C
W
1.25
C
WC
Fig. 9. Side and top representations of the most stable phenol configuration adsorbed on W
2
C, W1.25C and WC from left to right respectively. Colour scheme: clear grey is W,
dark grey is C, red is O and white is H.
of electron density, as well as surface configuration between W and
C atoms, altering their catalytic performance. Gong et al. [6]
0
W C
2
reported that W C possessed a larger number of states at the Fermi
2
-
1
2
W1.25C
energy than WC by theoretic calculation, which is in agreement
with our calculations. Wang et al. [54] reported a carbon modifica-
tion method on the metal Ni surface for tuning electronic property
of metallic sites, which is highly selective for CAO bond
hydrogenolysis reactions. Posada–Pérez and co-workers [57] stud-
ied the effect of metal/C ratio in supported molybdenum carbides
WC
-
A
A
F
B
G
A
C
C
B
C
B
-3
E D
for CO
ratio raised the number of metal centers exposed on the surface,
affecting the bonding modes of CO , and resulting in the formation
of CH as a side product.
2
conversion. They found that the increase of the metal/C
A: guanine(ads)
-
-
-
4
5
6
E
B: anisole(ads) + water(g)
C: phenol(ads) + mathanol(g)
D: phenol(ads) + mathane(g) + water(g)
E: benzene(ads) + mathanol(g) + water(g)
F: phenol(g) + mathane(g) + water(g)
G: benzene(g) + mathane(g) + 2*water(g)
D
2
4
D
The C-defect sites modified the catalytic properties of the sur-
face, affecting the relative stability of intermediates along the gua-
iacol hydrogenolysis pathway. The clean surface on the W
possesses rich C-defects and exposed more W-terminations. These
neighboring atoms facilitate the absorption of oxygen-
containing groups (CarylAOH and CarylAOCH ) in the guaiacol
molecular and intermediates due to the different electronegativi-
ties of W and C (2.36 versus 2.55). Indeed, the attraction of oxygen
2
C
Fig. 10. Thermodynamic energy profile for guaiacol hydrogenolysis on W
and WC surfaces clarified by dark, red, and blue lines, respectively.
2
C, W1.25C
W
3
to W
version of guaiacol and regioselective towards phenol. Contrary to
C (and WC), these catalysts stabilize the phenol intermediate
X
C@CS catalysts, which have presented a good catalytic con-
W
2
groups by the W
2
C surface is clearly represented in Fig. 9; the oxy-
over the anisole. Both pathways follow a smooth downhill energy
profile towards phenol on the surface (D), which explains the good
selectivity (Fig. 11b). The complete carburization reduces the over-
all catalytic activity (Fig. 11c). The hydrogenolysis of anisole leads
to benzene (E), which is a less favorable intermediate than phenol
gen from the phenol molecule directly interacts with the surface,
while on W1.25C and on WC, the HOA group remains parallel to
the surfaces. The strong interactions result in the cleavage of Caryl
AOH and CarylAOCH bonds, rendering the production of anisole
and phenol (B and C, respectively, in Fig. 9).
-
3
3
(D) produced from the cleavage of the CarylAOCH bond. The small
According to the computed energy profile in Fig. 10, the anisole
pathway is thermodynamically favored compared with the hydro-
genation of CarylAOCH group, which can explain the presence of
3
anisole in the products. Further hydrogenation is driven by the for-
mation of phenol (intermediate D), which desorbs (F) or reduces to
benzene in an intermediate state between D and G. The phenol gen-
erated through intermediate C may further reduce to benzene or
energy difference between these intermediates may limit the selec-
tivity of the hydrogenation process as shown by the experiments.
The surface configuration with careful carburization control is
of particular importance to designing catalysts for efficient
hydrogenolysis. The results of this study deepen our understanding
on the structure of tungsten carbides, as well as the structure-
activity relationship, appealing to wider catalytic applications in
the field of hydrogenolysis, surface science, and device design.
Based on the superiority of identified tungsten carbide catalyst
and deep understanding of structure-activity relationship, the
upgrading of complex lignin-derived compounds and lignin bio-
oils is under way of progress.
desorb. Both pathways, through cleavage of CarylAOH and Caryl
AOCH bonds, lead to downhill profiles, which, together with the
favorable stability of benzene and reduced products in gas phase
G), explains the presence of benzene. This process is represented
by the diagram in Fig. 11a. Further carburization of the surface leads
-
3
(

2
94
H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
(
a) W
2
C
<
(b) W C (1 << x 2)
x
(c) WC
Fig. 11. Plausible scheme for the guaiacol hydrogenolysis on tungsten carbides.
[
[
11] J.L. Calais, Adv. Phys. 26 (1977) 847–885.
12] J.R. Kitchin, J.K. Nørskov, M.A. Barteau, J.G. Chen, Catal. Today 105 (2005) 66–
4
. Conclusions
7
3.
x
A series of W C@CS catalysts was fabricated by carburization of
[13] M.G. Quesne, A. Roldan, N.H. de Leeuw, C.R.A. Catlow, Phys. Chem. Chem. Phys.
20 (2018) 6905–6916.
organic–inorganic hybrid precursors and evaluated for the guaiacol
hydrogenolysis. The optimal W1.25C@CS exhibited high activity and
selectivity for the phenol production by direct cleavage of aryl CAO
bond. The impressive activity was attributed the appropriate con-
figuration of surface W and C atoms and electronic properties,
which enriched with rich C-defect sites and boundaries. These
were further characterized by detailed textural analysis using
XRD, TEM, XPS, and HS-LEIS techniques. Controlled experiments
by surface reconstruction of pure Com-WC were used to prove
the effect of the surface configuration. The phenol production
was directly affected by the change of the C/W atomic ratio on
the surface and demonstrated a volcano-type curve. Specifically,
[
14] H.H. Fang, J.M. Du, C.C. Tian, J.W. Zheng, X.P. Duan, L.M. Ye, Y.Z. Yuan, Chem.
Commun. 53 (2017) 10295–10298.
15] C.Z. Li, M.Y. Zheng, A.Q. Wang, T. Zhang, Energy Environ. Sci. 5 (2012) 6383–
6390.
16] D.R. Stellwagen, J.H. Bitter, Green Chem. 17 (2015) 582–593.
17] R. Liu, M. Pang, X.Z. Chen, C. Li, C.J. Xu, C.H. Liang, Catal. Sci. Technol. 7 (2017)
[
[
[
1333–1341.
[18] P.K. Shen, S. Yin, Z. Li, C. Chen, Electrochim. Acta 55 (2010) 7969–7974.
[
[
19] A.V. Nikiforov, I.M. Petrushina, E. Christensen, N.V. Alexeev, A.V. Samokhin, N.J.
Bjerrum, Int. J. Hydrogen Energy 37 (2012) 18591–18597.
20] D.V. Esposito, J.G. Chen, Energy Environ. Sci. 4 (2011) 3900–3912.
[21] C.M. Friend, J.G. Serafin, E.K. Baldwin, P.A. Stevens, R.J. Madix, J. Chem. Phys. 87
1987) 1847–1850.
22] S.T. Hunt, T. Nimmanwudipong, Y. Roman-Leshkov, Angew. Chem. Int. Ed. 53
2014) 5131–5136.
[23] J. Lemaître, B. Vidick, B. Delmon, J. Catal. 99 (1986) 415–427.
(
[
(
the
W1+xC-Com-WC possessed the highest phenol STY of
ꢀ1
ꢀ1
[
24] R.L. Miller, P.T. Wolczanski, A.L. Rheingold, J. Am. Chem. Soc. 115 (1993)
0422–10423.
1
3.6 mmol gcat
h . DFT calculations were performed, and a plausi-
1
ble scheme for guaiacol hydrogenolysis was further proposed to
explain the catalytic pathway. This work provides protocols to
realize the surface control on tungsten carbides and understand
the relationship between structure and catalytic performance, ben-
efiting fundamental studies and rational design of carbide catalysts
on hydrogenolysis, especially in transformation of lignin-derived
compounds.
[
25] S.T. Hunt, M. Milina, A.C. Alba-Rubio, C.H. Hendon, J.A. Dumesic, Y. Roman-
Leshkov, Science 352 (2016) 974–978.
[26] Q.G. Yan, Y.W. Lu, F. To, Y.B. Li, F. Yu, Catal. Sci. Technol.
270–3280.
5 (2015)
3
[
[
27] H.B. Wu, B.Y. Xia, L. Yu, X.Y. Yu, X.W. Lou, Nat. Commun. 6 (2015) 6512.
28] Y.T. Xu, X.F. Xiao, Z.M. Ye, S.L. Zhao, R.A. Shen, C.T. He, J.P. Zhang, Y.D. Li, X.M.
Chen, J. Am. Chem. Soc. 139 (2017) 5285–5288.
[
29] H.H. Fang, J.W. Zheng, X.L. Luo, J.M. Du, A. Roldan, S. Leoni, Y.Z. Yuan, Appl.
Catal. A: Gen. 529 (2017) 20–31.
[
30] A.G. Sergeev, J.F. Hartwig, Science 332 (2011) 439–443.
[
31] A.J. Ragauskas, G.T. Beckham, M.J. Biddy, R. Chandra, F. Chen, M.F. Davis, B.H.
Davison, R.A. Dixon, P. Gilna, M. Keller, P. Langan, A.K. Naskar, J.N. Saddler, T.J.
Tschaplinski, G.A. Tuskan, C.E. Wyman, Science 344 (2014) 709.
Acknowledgments
We acknowledge the financial support from the National Key
R&D Program of China (2017YFA0206801), the National Natural
Science Foundation of China (91545115 and 21473145), and the
Program for Innovative Research Team in Chinese Universities
[32] W.S. Lee, Z.S. Wang, W.Q. Zheng, D.G. Vlachos, A. Bhan, Catal. Sci. Technol. 4
2014) 2340–2352.
33] Q. Lu, C.J. Chen, W. Luc, J.G. Chen, A. Bhan, F. Jiao, ACS Catal. 6 (2016) 3506–
514.
[34] S.T. Oyama, X.M. Zhang, J.Q. Lu, Y.F. Gu, T. Fujitani, J. Catal. 257 (2008) 1–4.
(
[
3
[
35] J.R. Kitchin, J.K. Nørskov, M.A. Barteau, J.G. Chen, J. Chem. Phys. 120 (2004)
0240–10246.
(
IRT_14R31).
1
[
[
36] M.P. Humbert, C.A. Menning, J.G. Chen, J. Catal. 271 (2010) 132–139.
37] J.P. Perdew, A. Ruzsinszky, G.I. Csonka, O.A. Vydrov, G.E. Scuseria, L.A.
Constantin, X.L. Zhou, K. Burke, Phys. Rev. Lett. 100 (2008) 136406.
38] N.D. Mermin, Phys. Rev. 137 (1965) A1441.
Appendix A. Supplementary material
[
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.jcat.2018.11.020.
[39] G. Kresse, D. Joubert, Phys. Rev. 59 (1999) 1758.
[
40] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 132 (2010) 154104.
41] D. Santos-Carballal, A. Roldan, R. Grau-Crespo, N.H. de Leeuw, Phys. Chem.
Chem. Phys. 16 (2014) 21082–21097.
[
[
[
42] N.Y. Dzade, A. Roldan, N.H. de Leeuw, Phys. Chem. Chem. Phys. 16 (2014)
References
1
5444–15456.
43] D. Santos-Carballal, A. Roldan, R. Grau-Crespo, N.H. de Leeuw, Phys. Rev. 91
2015) 195106.
[
[
[
1] G.S. Ranhotra, A.T. Bell, J.A. Reimer, J. Catal. 108 (1987) 40–49.
2] E. Furimsky, Appl. Catal. A: Gen. 240 (2003) 1–28.
3] R.W. Gosselink, D.R. Stellwagen, J.H. Bitter, Angew. Chem. Int. Ed. 125 (2013)
(
[
[
[
44] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.
45] P.E. Blöchl, O. Jepsen, O.K. Andersen, Phys. Rev. B 49 (1994) 16223.
46] T. Epicier, J. Dubois, C. Esnouf, G. Fantozzi, P. Convert, Acta Metall. 8 (1988)
5
193–5196.
4] S.A.W. Hollak, R.W. Gosselink, D.S. van Es, J.H. Bitter, ACS Catal. 3 (2013) 2837–
844.
5] H. Ren, W.T. Yu, M. Salciccioli, Y. Chen, Y.L. Huang, K. Xiong, D.G. Valchos, J.G.
Chen, ChemSusChem 6 (2013) 798–801.
6] Q.F. Gong, Y. Wang, Q. Hu, J.G. Zhou, R.F. Feng, P.N. Duchesne, P. Zhang, F.J.
Chen, N. Han, Y.F. Li, C.H. Jin, Y.G. Li, S.T. Lee, Nat. Commun. 7 (2016) 13216.
7] A.L. Jongerius, R.W. Gosselink, J. Dijkstra, J.H. Bitter, P.C.A. Bruijnincx, B.M.
Weckhuysen, ChemCatChem 5 (2013) 2964–2972.
[
[
[
[
1903–1921.
2
[
47] K.D. Litasov, A. Shatskiy, Y.W. Fei, A. Suzuki, E. Ohtani, K. Funakoshi, J. Appl.
Phys. 108 (2010) 053513.
[
[
48] P.W. Tasker, J. Phys. C: Solid State Phys 12 (1979) 4977.
49] G.W. Watson, E.T. Kelsey, N.H. de Leeuw, D.J. Harris, S.C. Parker, J. Chem. Soc.,
Faraday Trans. 92 (1996) 433–438.
[
50] T. Nimmanwudipong, C. Aydin, J. Lu, R.C. Runnebaum, K.C. Brodwater, N.D.
Browning, D.E. Block, B.C. Gates, Catal. Lett. 142 (2012) 1190–1196.
51] Y. Nakagawa, M. Ishikawa, M. Tamura, K. Tomishige, Green Chem. 16 (2014)
[
[
8] W.S. Lee, A. Kumar, Z.S. Wang, A. Bhan, ACS Catal. 5 (2015) 4104–4114.
9] J.G. Chen, Chem. Rev. 96 (1996) 1477–1498.
[
2197–2203.
[
10] H.H. Hwu, J.G. Chen, Chem. Rev. 105 (2005) 185–212.

H. Fang et al. / Journal of Catalysis 369 (2019) 283–295
295
[
[
52] W.S. Lee, Z. Wang, R.J. Wu, A. Bhan, J. Catal. 319 (2014) 44–53.
53] J.R. Kitchin, J.K. Nørskov, M.A. Barteau, J.G. Chen, Phys. Rev. Lett. 93 (2004)
56801.
54] M. Wang, X.C. Zhang, H.J. Li, J.M. Lu, M.J. Liu, F. Wang, ACS Catal. 8 (2018)
614–1620.
[55] M.M. Sullivan, A. Bhan, ACS Catal. 6 (2016) 1145–1152.
[56] M.M. Sullivan, C.J. Chen, A. Bhan, Catal. Sci. Technol 6 (2016) 602–616.
[57] S. Posada-Pérez, P.J. Ramírez, J. Evans, F. Viñes, P. Liu, F. Illas, J.A. Rodriguez, J.
Am. Chem. Soc. 138 (2016) 8269–8278.
1
[
1