Hydrodeoxygenation of anisole over different Rh surfaces

Article abstract of DOI:10.1016/S1872-2067(19)63345-0

The cleavage of the alkoxy (Ar–O–R) ether bond present in anisole is an interesting hydrodeoxygenation (HDO) reaction, since this asymmetric group contains two different C–O bonds, C_aryl–O or C_alkyl–O, which could potentially cleave. Recent work on the HDO of anisole over Pt, Ru, and Fe catalysts has shown that a common phenoxy surface intermediate is formed on all three metals. The subsequent reaction path of this intermediate varies from metal to metal, depending on the metal oxophilicity. Over the less oxophilic Pt, phenol is the only primary product. By contrast, on the more oxophilic Fe catalyst, the sole primary product is benzene instead of phenol. On Ru, with intermediate oxophilicity, both benzene and phenol are primary products. In this contribution, we have investigated Rh catalysts of varying surface nanostructures. A combination of experimental measurements and computational calculations was used to explore the effects of varying metal coordination number, an additional parameter that can be used to control the oxophilicity of a metal. The results confirm that metal oxophilicity is a good descriptor for HDO performance of metal catalysts and it can be controlled via selection of metal type and/or metal extent of coordination. Small Rh metal clusters with low coordination metal sites are more active for the deoxygenation pathway but also quickly deactivated while large clusters with high coordination sites are more active toward hydrogenation and more stable.

Full text of DOI:10.1016/S1872-2067(19)63345-0

Chinese Journal of Catalysis 40 (2019) 1721–1730
a v a i l a b l e a t w w w . s c i e n c e d i r e c t . c o m
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c h n j c
Article (Special Issue on Celebrating the 70th Anniversary of Dalian Institute of Chemical Physics, CAS)
Hydrodeoxygenation of anisole over different Rh surfaces
Nhung N. Duong a,b, Darius Aruho a, Bin Wang a,*, Daniel E. Resasco a,#
a School of Chemical, Biological and Materials Engineering (CBME) Center for Interfacial Reaction Engineering (CIRE) University of Oklahoma, Norman
OK 73019, USA
b School of Biotechnology, International University Vietnam National University, Ho Chi Minh City 700000, Vietnam
A R T I C L E I N F O
A B S T R A C T
Article history:
The cleavage of the alkoxy (Ar‒O‒R) ether bond present in anisole is an interesting hydrodeoxy-
genation (HDO) reaction, since this asymmetric group contains two different C–O bonds, Caryl–O or
Calkyl–O, which could potentially cleave. Recent work on the HDO of anisole over Pt, Ru, and Fe cata-
lysts has shown that a common phenoxy surface intermediate is formed on all three metals. The
subsequent reaction path of this intermediate varies from metal to metal, depending on the metal
oxophilicity. Over the less oxophilic Pt, phenol is the only primary product. By contrast, on the more
oxophilic Fe catalyst, the sole primary product is benzene instead of phenol. On Ru, with intermedi-
ate oxophilicity, both benzene and phenol are primary products. In this contribution, we have inves-
tigated Rh catalysts of varying surface nanostructures. A combination of experimental measure-
ments and computational calculations was used to explore the effects of varying metal coordination
number, an additional parameter that can be used to control the oxophilicity of a metal. The results
confirm that metal oxophilicity is a good descriptor for HDO performance of metal catalysts and it
can be controlled via selection of metal type and/or metal extent of coordination. Small Rh metal
clusters with low coordination metal sites are more active for the deoxygenation pathway but also
quickly deactivated while large clusters with high coordination sites are more active toward hydro-
genation and more stable.
Received 10 January 2019
Accepted 7 March 2019
Published 5 November 2019
Keywords:
Hydrodeoxygenation
Anisole
Alkoxy
Oxophilicity
Phenolic
Surface defects
© 2019, Dalian Institute of Chemical Physics, Chinese Academy of Sciences.
Published by Elsevier B.V. All rights reserved.
1. Introduction
ularly active for the cleavage of C–O bonds, and a large number
of studies have investigated catalysts containing different
Hydrodeoxygenation (HDO) is an effective reaction for up-
grading biomass to fuels and chemicals [1]. For example, bio-
mass pyrolysis oil contains significant fractions of phenolics
and other oxygenated aromatic compounds derived from lig-
nin, which require deoxygenation. Therefore, the HDO reaction
of these biomass-derived compounds has attracted attention of
researchers in recent years [2–5]. Transition metals are partic-
characteristics, such as varying types of metal [6,7], particle
size and coordination number [8,9], alloying elements [10–15],
supports [16–19], and bifunctionality in cooperation with the
acidic function of zeolites [20–27].
The two most abundant functionalities in these oxygenated
aromatic compounds are hydroxy (–OH) and alkoxy (–OR)
groups. In previous studies on the removal of the –OH func-
* Corresponding author. E-mail: wang_cbme@ou.edu
# Corresponding author. E-mail: resasco@ou.edu
This work was supported by the U.S. Department of Energy, DOE/EPSCOR (Grant DESC0004600). The computational research used the supercom-
puter resources of the National Energy Research Scientific Computing Centre (NERSC) and the OU Supercomputing Centre for Education & Research
(OSCER) at the University of Oklahoma.
DOI: S1872-2067(19)63345-0 | http://www.sciencedirect.com/science/journal/18722067 | Chin. J. Catal., Vol. 40, No. 11, November 2019

1722
Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
tionality in m-cresol, we have proposed that the metal’s oxophi-
licity (described as the M–O bond strength) is a major de-
scriptor of the HDO activity as well as the –OH removal mecha-
nism [6]. That is, over less oxophilic metals, such as Pt or Pd,
HDO proceeds via an indirect pathway involving tautomeriza-
tion-hydrogenation-dehydration, while over more oxophilic
metals (Ru, Fe, Rh), it happens via direct deoxygenation [6].
Oxophilicity is a direct function of the position of the d-band
center with respect to the Fermi level. Therefore, it can be ad-
justed, not only by changing the type of metal used, but also by
varying the coordination number of metal atoms [28–32]. For
example, we have shown that small Rh clusters with higher
density of low-coordination atoms are more active for –OH
removal but also more prone to deactivation while larger Rh
clusters with higher coordination numbers are less active but,
at the same time, more stable [8].
the phenoxy intermediate C6H5O* to phenol, which is the dom-
inant primary product. As conversion increases phenol con-
verts to benzene in a secondary path. This trend is in good
agreement with our DFT calculations that showed an exceed-
ingly high barrier for Caryl–O cleavage of the phenoxy interme-
diate on Pt (270 kJ/mol). By contrast, the hydrogenation of the
phenoxy intermediate has a much lower energy barrier and it
produces phenol as the primary product (17 kJ/mol).
An opposite behavior is observed on the more oxophilic Fe
catalyst, over which the phenoxy intermediate prefers to un-
dergo direct Caryl–O cleavage, producing benzene. On this metal,
the phenol-to-benzene ratio is found to be zero at all anisole
conversions, indicating that benzene rather than phenol is the
only primary product. This is also in good agreement with the
DFT calculations which show that the energy barrier for direct
Caryl–O cleavage (101 kJ/mol) is lower than that for hydrogena-
The mechanism for the cleavage of the alkoxy (Ar‒O‒R)
bond is more complicated than the corresponding to –OH due
to the asymmetric nature of this ether group, which contains
two different C‒O bonds, Caryl‒O and Calkyl‒O. Cleavage of the
tion (147 kJ/mol) of the phenoxy intermediate over Fe.
Finally, on a metal with intermediate oxophilicity like Ru,
this ratio is in between those found on Pt and Fe, dropping with
conversion. This trend shows that both benzene and phenol are
primary products and more benzene is formed in a secondary
path. This experimental trend was also in perfect agreement
with DFT calculations, which showed that the energy barriers
for C–O bond cleavage and hydrogenation of the phenoxy in-
termediate have comparable values, which are 165 and 134
kJ/mol, respectively.
In addition to the type of metal, the structure of metal sur-
face also has significant impact on the C–O bond cleavage activ-
ity since coordination of surface metal atoms controls the metal
oxophilicity. Metal catalysts with different metal cluster sizes
such as Ni/SiO2, Rh/SiO2, Ru/NbOPO4, etc. have been investi-
gated for HDO activity of many different phenolic molecules
[8,9,33,34]. A general trend was recently observed in which the
smaller metal particles showed higher activity for C–O bond
cleavage. In this contribution, we have further explored the
influence of metal nanostructures on the HDO mechanism of
anisole and its physical origin. By combining experimental
measurements and DFT calculations, we investigate the effects
of varying surface coordination number, a parameter that can
be used to control the oxophilicity of a metal catalyst. Rhodium
metal catalysts with different particle sizes have been chosen
since Rh possesses intermediate oxophilicity between Pt and
Fe; therefore the particle size effects would be more clearly
discerned, as observed in our previous study with m-cresol [8].
Ru also possesses intermediate oxophilicity between Pt and Fe,
but Ru is also very active for hydrogenolysis, giving high yields
of methane [6,7]. Since the purpose of this work is to focus on
the C–O bond cleavage activity at two positions Caryl–O and
Caryl‒O bond eliminates the oxygen functionality from the aro-
matic ring while cleavage of the Calkyl‒O leads to formation of a
phenolic compound. In a recent work [7], we compared the
HDO reaction of anisole over Pt, Ru, and Fe catalysts. No meth-
anol was observed as a product on any of the three catalysts.
Therefore, neither experimental nor density functional theory
(DFT) calculations gave evidence that Caryl‒O or Calkyl‒O was
cleaved initially, but rather the C–H bond in the methyl group
was activated first. Only when the ‒CH3 group was dehydro-
genated enough, it was possible to break the Calkyl‒O bond. As
a result, a common phenoxy surface intermediate, C6H5O* is
expected to be formed on all three metals. However, the sub-
sequent reaction path of this intermediate varies from metal to
metal (Scheme 1). It was hypothesized that the mechanisms
responsible for the different outcomes on the different metals
are directly related to the metal oxophilicity [7]. Indeed, when
the product ratio of phenol-to-benzene was compared as a
function of anisole conversion over the three metals, a very
diverse behavior is observed, as shown in Fig. S1. That is, over
the less oxophilic Pt, a very high phenol-to-benzene ratio is
observed at low conversions, but it becomes lower as conver-
sion increases.
This trend indicates that Pt greatly favors hydrogenation of
Deoxygenation
Calkyl-O
cleavage
C
alkyl–O, the best candidate for this study is Rh metal because it
is active for C–O bond cleavage while its cracking activity is
limited.
phenoxy
2. Experimental
Hydrogenation
2.1. Catalyst preparation
Three Rh/SiO2 samples with different metal particle sizes
Scheme 1. Reaction pathway for anisole conversion.

Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
1723
were prepared by incipient wetness impregnation of SiO2 sup-
2.4. DFT calculations
port (Hi-sil 210, BET area 135 m2/g) using Rhodium (III)
nitrate aqueous solution (10 wt% Rh in > 5 wt% HNO3, pur-
chased from Sigma-Aldrich). The metal particle sizes were con-
trolled using different metal loading and treatment conditions.
That is, we made small Rh metal clusters with low metal load-
ings and mild treatment conditions, while we made large Rh
clusters with high metal loading and more severe treatment
conditions. After impregnation, the low-loading catalyst (0.6
Plane-wave DFT calculations were performed using the Vi-
enna ab initio Simulation Package (VASP) [35,36]. The
Perdew-Burke-Ernzerhof (PBE) functional in the generalized
gradient approximation (GGA) was used to calculate the ex-
change-correlation energy [37]. The vdW corrections were
included using the DFT-D3 method [38]. The Rh(111) and
Rh(533) surfaces were modeled with a lattice constant of 3.81
Å of the Rh bulk, close to a reported value of 3.803 Å [39]. A 4 ×
4 unit cell of Rh(111) was chosen to minimize lateral interac-
tion between neighboring anisole molecules. Both repeated
slabs of Rh (533) and Rh (111) consisted of 80 Rh atoms with a
vacuum space of 15 Å in the z-direction to separate successive
layers. All of reactants, intermediates and products structures
were optimized until the forces on each atom were less than
0.01 eV/Å. A kinetic cutoff energy of 400 eV was used in the
calculations. The top two layers were allowed to fully relax in
all calculations, whereas the bottom layers were fixed to the
bulk distance of the metal. The adsorption energy of anisole on
a Rh surface was defined as:
o
and 2.5 wt% Rh/SiO2) was heated under H2 at 200 C for 2 h,
while the high-loading catalyst (7 wt% Rh/SiO2) was heated
o
under N2 at 550 C for 2 h, and then for an additional 12 h pe-
riod at 550 oC under H2 flow.
2.2. Catalyst characterization
The catalysts were characterized by high resolution trans-
mission electron microscopy (HR-TEM) and diffuse reflectance
infrared Fourier transform spectroscopy (DRIFTS) of adsorbed
CO. The TEM samples were first dispersed in ethanol with a
Cole-Parmer horn sonicator operating at 25% amplitude for 10
min and then deposited onto lacey carbon coated copper grids
(300 mesh) before analysis on a TEM JEOL 2000 field emission
system operated at 200 kV. DRIFTS of adsorbed CO was per-
formed using a Perkin-Elmer Spectrum 100 FTIR spectrometer,
equipped with a MCT detector. The diffuse reflectance cell is
from Harrick Scientific, type HVC-DR2 equipped with CaF2
windows for in-situ reduction. The IR spectra were taken at a
resolution of 4 cm−1, collecting 256 scans for each spectrum. In
a typical DRIFTS experiment, the catalyst was reduced under
E
ads = ETOT(anisole + surface) – ETOT(anisole) – ETOT(surface) (1)
where ETOT(anisole + surface) is the DFT-calculated total energy
of anisole adsorbed on the metal surface. ETOT(anisole) defines
the total energy of an isolated anisole molecule, and
ETOT(surface) is the total energy of the bare slab. The adsorption
was analyzed at highly symmetric adsorption sites of the metal
surface. Spin polarized calculations were performed for the
atomic O and H atoms in the gas phase. Transition states were
calculated using the Nudge Elastic Band (NEB) method [40–42],
and verified by calculating the vibrational frequencies.
o
H2 for 1 h at 300 C, then cooled down to room temperature
under He flow and purged in He for 30 min. The background
was recorded at this time. Prior to obtaining the scans, the cat-
alyst was exposed to a flow of 5% CO in He for 30 min at room
temperature and purged in He for 30 min.
3. Results and discussions
3.1. Catalyst characterization
2.3. Catalytic activity measurements
The three Rh/SiO2 catalysts of different metal loading (0.6,
2.5 and 7 wt% Rh) and thermal pre-treatments were found to
have varying particle sizes, as expected. As demonstrated in
Fig. 1, HR-TEM analysis of the catalysts clearly reflects the dif-
ferent ranges of metal cluster sizes and shapes. The particle
size decreases in the expected order (7 > 2.5 > 0.6 wt% Rh).
The corresponding histograms are summarized in Fig. 2.
The 7 wt% Rh catalyst has particles that range in size up to
about 7 nm with an average of 4 nm; by contrast, the 2.5 wt%
Rh catalyst exhibits a much smaller average size (2 nm) and a
range in which no particles are larger than 4 nm; finally, the
lowest loading catalyst (0.6 wt% Rh) has a much narrower size
distribution centered at 1 nm.
In correspondence with these differences in particle sizes,
the DRIFT spectra of adsorbed CO exhibited a vastly different
distribution of intensity of the characteristic IR absorption
bands (Fig. S5). Table 1 summarized the relative intensities
observed on two different samples with varying particle sizes
for the three different modes, gem (G) dicarbonyl, linear (L),
and bridge (B). The gem dicarbonyl band is typically used as a
fingerprint for the presence of small Rh cluster, while the
All catalytic tests were conducted using a micro-pulse reac-
tor, which allows us to precisely monitor deactivation and
study the catalytic activity, starting from a pristine surface and
following with gradual deactivation after each consecutive
pulse, as described elsewhere [8]. In a typical experiment, the
catalyst (pellet size 250‒350 μm) was packed in a vertical
quartz tube reactor (20 cm L × 0.6 cm D) and pre-reduced
in-situ under continuous flow of 100 sccm H2 at 400 oC for 1 h.
Subsequently, it was cooled down to the selected reaction
o
temperature (270 C), keeping the catalyst under H2 flow dur-
ing the entire experiment. Only a small amount of feed is in-
troduced periodically through the catalytic bed. The products
were analyzed online by GC (Agilent HP 5890) equipped with a
FID detector and a HP-Innowax capillary column. The benefit of
this type of reactor has been demonstrated in previous work, in
which the device enabled us to directly observe struc-
ture/activity relationship on clean metal surfaces and how the
product distribution changes as the surface gradually becomes
deactivated [8].

1724
Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
10nm
7% Rh
2.5% Rh
0.6% Rh
Fig. 1. HR-TEM images of 7, 2.5 and 0.6 wt% Rh/SiO2 catalysts.
bridge mode is a fingerprint for large Rh clusters. The gem twin
band arises when two CO molecules adsorb on one Rh site;
thus, the doublet represents the symmetric and asymmetric
vibrations of the CO molecules. A linear band corresponds to
one CO molecule adsorbed atop one Rh atom, while the bridge
band appears when one CO molecule adsorbs on a multiplet of
Rh atoms [43–47].
As reported in Table 1, over the small metal particles in the
2.5 wt% Rh catalyst, the gem bands have significantly more
than a third of the intensity of the linear band, taken as a refer-
ence, indicating a very high density of low-coordination Rh
sites. By contrast, they totally disappear on the 7 wt% Rh sam-
ple. On the other hand, the bridge band is less than 10% of the
linear band on the 2.5 wt% Rh/SiO2 catalyst, but it rises to 25%
on the 7 wt% Rh sample, showing a good agreement with TEM
results mentioned above. Due to the low loading, it was not
possible to get a DRIFT spectrum with acceptable signal/noise
for the 0.6 wt% Rh.
3.2.1. C1 products from anisole
As mentioned above, along with the aromatic products
(phenol and benzene), the HDO reaction of anisole also pro-
duces C1. If, as reported in the previous study, methane is the
only C1 product obtained, the ratio of C1/C6-ring should be
one, provided that no significant amounts of C1 products re-
main on the catalyst surface [7]. On the other hand, if methanol
is also produced, one can expect that some of it may decompose
to CO or CO2, not detectable by FID.
As shown in Fig. 3, contrary to the previous observations,
we do observe some methanol in the products of anisole con-
version. While the observed amounts of methanol are much
lower than those of methane, there is still a detectable fraction.
At the same time, Fig. 4 shows the C1/C6-ring ratio for the
three Rh catalysts at varying anisole conversion. In this case, C1
is calculated as the sum of the amounts of methane and meth-
anol observed.
While the C1/C6-ring ratio is close to 1, there is a consistent
difference among the three catalysts that is worth discussing.
The highest and closest to 1 ratio is observed over the catalyst
with large Rh cluster (7 wt% Rh). At the same time, this catalyst
is the one that produces less methanol. By contrast, the cata-
lysts with smaller Rh clusters (and rougher surfaces) not only
produce more methanol, but also C1 products (CO, CO2) that we
cannot detect via FID. The formation of methanol suggests that
anisole could undergo direct C-O bond cleavage on the metal
3.2. Catalytic conversion of anisole
0.7
0.6% Rh
2.5% Rh
7% Rh
0.6
0.5
0.4
0.3
0.2
0.1
0
0-1
1-2
2-3
3-4
4-5
5-6
6-7
Particle size (nm)
Fig. 2. Particle size distribution of 7, 2.5 and 0.6 wt% Rh/SiO2 catalysts.
Table 1
Relative Intensity of the DRIFTS bands of different CO adsorption
modes on two Rh/SiO2 catalysts of varying metal particle sizes.
CO adsorption
Catalysts
mode
Gem
Linear
Bridge
(2030 and
2100 cm‒1
35 and 42
(2050‒2070 cm‒1)(1800‒1900 cm‒1
)
)
2.5 wt% Rh/SiO2
100
100
6.4
25
Fig. 3. Ratio of methanol/methane product at varying anisole conver-
sion over the 0.6, 2.5, and 7 wt% Rh/SiO2 catalysts at 270 oC.
7 wt% Rh/SiO2
—

Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
1725
flects the similar rates of the phenoxy intermediate to either
hydrogenation to phenol or deoxygenation to benzene.
Since Rh has an intermediate oxophilicity, similar to Ru
metal, it is reasonable that both reaction paths are comparable.
Interestingly, for this catalyst, the phenol-to-benzene ratio only
starts decreasing at high conversion (larger mass). The third
observed (minor) product, cyclohexane methyl ether, arises
from the hydrogenation of the aromatic ring of anisole, without
C‒O or C‒C cleavage. Clearly, this is a much less preferred reac-
tion pathway.
Fig. 6 compares products selectivity at the same anisole
conversion (10%) over the three different Rh catalysts. The
data is taken from the first pulse of anisole through the catalyst
bed. Thus, it represents the activity of pristine surfaces. We
can see that as the Rh metal cluster size decreases from 4 > 2 >
1 nm for 7 > 2.5 > 0.6 wt% Rh, respectively, the selectivity to
phenol and its derivatives (cyclohexanol and cyclohexanone)
decreases while the selectivity to benzene and its derivatives
(cyclohexane) increases. As mentioned above, at such a low
anisole conversion benzene is only formed by primary HDO of
anisole. That is, this selectivity represents the preference of the
phenoxy intermediate to undergo either hydrogenation to
phenol or deoxygenation to benzene on each Rh surface. Clear-
ly, smaller Rh clusters with a higher density of surface defects
(steps, kinks) exhibits a higher selectivity to deoxygenation
while larger Rh clusters with smooth surfaces (terraces) favor
hydrogenation of the phenoxy intermediate to phenol.
Fig. 4. Ratio of C1/C6 ring product at varying anisole conversion over
the 0.6, 2.5, and 7 wt% Rh/SiO2 catalysts at 270 °C.
surface to produce benzene and methanol. Based on the de-
scending order of C1/C6 ratio of 0.6 > 2.5 > 7 wt% Rh/SiO2 as
seen in Fig. 3, it seems that small Rh particles favor this path-
way more than larger Rh particles. This reaction pathway will
also be investigated in the DFT theoretical calculation section
below.
To further confirm the above observation, we compared the
number of anisole turnovers per exposed Rh site as a function
of the number of pulses over the three different catalysts. The
number of turnovers was calculated as the moles of anisole
converted in each pulse to a specific product relative to the
moles of surface Rh atoms present in the catalyst. These
numbers are compared in Fig. 7 for the three Rh catalysts of
different particle sizes. It can be seen that, regardless of the
small changes in activity exhibited after each consecutive pulse,
the specific phenoxy hydrogenation activity increases with
particle size, while, by contrast, the specific phenoxy deoxy-
genation activity decreases with particle sizes. Fig. 7 also illus-
3.2.2. Phenol and benzene production from anisole
Fig. 5 shows the yield evolution of the ring products ob-
tained over the pristine 2.5 wt% Rh/SiO2 catalyst from the first
anisole pulse in the micro-pulse reactor as a function of catalyst
mass. Since the independent variable in a pulse reactor is space
time (W/F), one can vary space time by keeping mass constant
and varying the flow rate, or vice versa. The three products are
benzene, phenol and cyclohexane methyl ether.
The positive slope for all three products that can be ob-
served near zero conversion, that is approaching zero mass of
catalyst, indicates that they are all primary products. Also, the
almost identical initial production of benzene and phenol re-
100%
20%
Phenol and derivatives
Benzene and derivatives
80%
Benzene
Anisole ring hydrogenation
15%
60%
Phenol
10%
40%
20%
0%
5%
Cyclohexane methyl ether
0%
0
3
6
9
12
Mass of catalyst (mg)
7%Rh
2.5%Rh
0.6% Rh
Fig. 5. Yield of benzene, phenol and cyclohexane methyl ether from
anisole conversion in the first pulse as a function of mass of catalyst of
2.5 wt% Rh/SiO2 at Trxn = 270 °C.
Fig. 6. Products selectivity of 10% anisole conversion over different Rh
catalyst including 7, 2.5 and 0.6 wt% Rh/SiO2 at Trxn = 270 °C.

1726
Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
sorb with the O‒CH3 away from the rhodium atoms on the step
(Fig. S2), probably caused by steric hinderance. However, the
oxygen atom forms a covalent bond of 2.6 Å with the step Rh
atom, a configuration that favors deoxygenation, as discussed
below. The adsorption energy of Rh(533) is ‒278 kJ/mol. Pre-
vious studies have shown that though “atop” sites are unfavor-
able, they can be favored in defected surfaces [48]. Notice that
anisole adsorption on Rh(533) is weaker than on Rh(111). This
difference may be due to two possible reasons: first, the pres-
ence of the step results in a lower number of carbons attached
to the surface than on the flat terrace; and second, the oxygen
proximity to the step does not make up for the loss of about
three C–Rh bonds, as seen in Fig. S2. Previous studies on phenol
support this explanation, showing that it adsorbs more favora-
bly on Rh(111) terraces than on Rh(211) stepped surfaces [49].
Fig. 7. Number of turnovers of anisole per exposed Rh site to different
products; that is, phenol, benzene and their derivatives at increasing
number of pulses over the three catalysts 0.6, 2.5, and 7 wt% Rh/SiO2 at
270 °C.
3.3.2. Reaction path of anisole HDO on Rh(111)
The energy barriers and reaction energies for different re-
action pathways of anisole on Rh(111) are shown in Fig. 8. We
first considered direct C‒O bond cleavage on the surface. One
proposed pathway is to directly cleave the methyl (–CH3) over
Rh(111). This pathway turns out to be thermodynamically fa-
vorable with reaction energy of ‒19 kJ/mol but it remains ki-
netically unfavorable with a high activation barrier of 258
kJ/mol. The Caryl‒O bond may also be directly cleaved to form a
phenyl intermediate, which could then be hydrogenated to
benzene. In order to cleave ‒OCH3 from anisole on Rh(111), the
trates the extent of catalyst deactivation as a function of num-
ber of anisole pulses sent over the catalyst. While no significant
deactivation is observed for the phenoxy hydrogenation path-
way, the drop in activity for deoxygenation, particularly on the
smallest Rh cluster 0.6 wt% Rh/SiO2 is significant during the
initial pulses.
Apparently, the 0.6 wt% Rh/SiO2 catalyst contains very ac-
tive sites for deoxygenation of the phenoxy intermediate, but
they are prone to deactivation. On the other hand, the 2.5 and 7
wt% Rh catalysts contain less active but more stable sites. Fi-
nally, it is observed that hydrogenation is not significantly de-
activated in any of the catalysts. As previously proposed, deox-
ygenation of the phenoxy intermediate on the more oxophilic
sites also produces O* species, which need to be removed to
regenerate the active site. Our DFT calculations have shown
that the energy barriers necessary to remove the thus gener-
ated surface O* via hydrogenation are higher on
low-coordination sites on the stepped surface Rh(533) than on
flat high-coordination terraces Rh(111), which agrees with the
experiments [8].
Caryl‒O bond increases from 1.37 Å in the reactant to 1.93 Å in
the transition state. The activation barrier is also very high at
226 kJ/mol for this endothermic (89 kJ/mol) reaction.
We have discussed above the possibility of sequentially de-
hydrogenating the ‒CH3 moiety before the C‒O cleavage. In-
deed, as shown in Fig. 8, the first ‒H removal is more favorable
with an activation barrier of 175 kJ/mol, which is much lower
than those for demethylation or demethoxylation. Fig. S3 com-
pares energy barriers and energy of reactions for three differ-
ent possible pathways of anisole on Rh(111), which clearly
3.3. DFT calculations on Rh(111) and Rh(533)
3.3.1. Adsorption of anisole on Rh(111) and Rh(533)
In the following we apply DFT calculations to investigate the
adsorption and reaction of anisole on Rh(111) and Rh(533)
surfaces. The former is an approximate model of the flat sur-
faces in larger particles, while the latter represents the
low-coordination stepped surfaces in small nanoparticles. Fig.
S2 shows the side view of anisole adsorption on the terrace
Rh(111) and the step of the Rh(533) surface. Adsorption of
anisole on Rh(111) favors a bridge configuration with an ad-
sorption energy of ‒283 kJ/mol, followed by the hcp hollow
site. In the lowest energy configuration, anisole lays flat on the
surface with the six-carbon atom ring shared amongst four Rh
surface atoms and the O–CH3 group tilted away from the sur-
face. The distance between the oxygen atom and the Rh(111)
surface is 2.98 Å. On the Rh(533) surface, anisole tends to ad-
Fig. 8. Layout of DFT calculated reactant, transition state and product
structures for Caromatic–O bond breaking, Calky–O and dehydrogenation
reactions of different Ph‒O‒CHx (1 ≤ x ≤ 3) on Rh(111) surface. (Activa-
tion energy/energy of reaction).

Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
1727
demonstrates that dehydrogenation is the most favorable
pathway. After the first H removal, we analyze the possible
paths for Ph‒O–CH2*, which could be cleaved at the Calkyl‒O or
the Caryl‒O, or could lose a second H. We find that the second
dehydrogenation has an energy barrier of 91 kJ/mol, which is
lower than 137 and 196 kJ/mol for the other pathways. This
successive dehydrogenation is both kinetically and thermody-
namically favorable (indicated by the green arrows in the fig-
ure).
As seen in the reaction scheme of Fig. 8, removing the 3rd
H-atom would create a carbon that needs to bind three rhodi-
um atoms. It would in turn have to shift the whole molecule to
accommodate for this highly reactive product. This third dehy-
drogenation is significantly more difficult with an activation
barrier of 166 kJ/mol on Rh(111). Instead, the Ph‒O–CH* can
now be cleaved at the Calkyl‒O bond, which forms a thermody-
namically stable product with a reaction energy of ‒55 kJ/mol
and a lower activation barrier of 143 kJ/mol. Successive dehy-
drogenation of the –CH3 on Rh(111) has significantly reduced
the energy requirement to form a phenoxy intermediate, simi-
lar to the reaction mechanisms described before on Pt(111)
and Ru(0001) [7]. Also, demethoxylation is almost always ki-
netically unfavorable, which is in line with our experimental
results showing that only a small amount of methanol is
formed.
Fig. 9. Layout of DFT calculated reactant, transition state and product
structures for Caromatic—O bond breaking, Calky—O and dehydrogenation
reactions of different Ph—O—CHx (1 ≤ x ≤ 3) on Rh(533) surface. (Acti-
vation energy/energy of reaction).
the most favorable path is dehydrogenation followed by
C
alkyl‒O cleavage to form the phenoxy, we find relatively low
barriers for the direct cleavage of the Caryl–O bond of adsorbed
anisole (141 kJ/mol), which leads to the formation of benzene
and methanol. This is in line with the experimental observa-
tions of higher methanol to methane ratio on the smaller na-
noparticles (Fig. 3). This relatively lower barrier for Caryl–O
bond cleavage may be explained in terms of oxophilicity of the
metal.
Specifically, we have found that the Rh(533) surface has
higher affinity for oxygen atoms than Rh(111), Pt(111) and
slightly lower than Ru(0001). It is well known that oxophilicity
of metals is determined by their individual d-bands. In this case,
the differences between the flat terraces and stepped surfaces
are very clear. Fig. 10 shows that the density of states of the
Rh(533) step atoms becomes narrower and sharper and shifts
towards the Fermi level; this upshifted d-band can explain the
enhanced adsorption of oxygen at the step atoms, which helps
cleaving the C–O bond as the Rh–O interaction becomes
3.3.3. Reaction paths for HDO of anisole over Rh(533)
In the following we discuss reactions on the stepped surface
Rh(533), which exhibits similar pathways as the more oxophil-
lic metals previously studied, i.e. Ru(0001) and Fe(110) with
some subtle differences. As we did for the flat surface, we first
consider direct cleavage of the anisole C‒O bonds on the
stepped surface. Similar to the case on Rh(111), the Caryl–O
bond on Rh(533) stretches from 1.42 Å to 1.70 Å in the transi-
tion state with an energy barrier of 141 kJ/mol. This value is a
significantly lower compared to 226 kJ/mol observed for the
terrace surface. It is also lower than the previously reported
values for Pt(111) (258 kJ/mol) but higher than those for
Ru(0001) (99 kJ/mol) and Fe(110) (79 kJ/mol) [7]. Upon
cleavage, the ‒OCH3 binds atop of one Rh atom to form a Rh‒O
bond with length of 2.15 Å. The barriers for demethoxylation
and demethylation are very similar, both of them lower than
the corresponding values calculated for Rh(111).
As shown in Fig. 9, removal of one H-atom from ‒CH3 re-
mains a favorable pathway with a much lower activation barri-
er than the two previous pathways. With an energy barrier of
79 kJ/mol, Rh(533) can catalyze the successive dehydrogena-
tion of anisole to remove the first H-atom. However, unlike the
reactions on Rh(111), due to the high reactivity of Rh(533),
C‒O bond cleavage can occur after the first dehydrogenation,
rather than after the second one as was the case for Rh(111), to
form the crucial surface phenoxy intermediate. The preferred
steps are shown with green arrows, in Figs. 9 and S4.
To compare these calculations with the experimental ob-
servations, we note that on catalysts with small Rh nanoparti-
cles we observed higher yields of deoxygenation products. The
DFT calculations show that, although on the stepped surface
Fig. 10. Density of states for a step atom on Rh(533) (blue curve) and
from a terrace atom on Rh(111) (orange curve). d-band centers calcu-
lated as: ‒1.8 eV for Rh(111) and ‒1.5 eV for Rh(533).

1728
Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
300
250
200
150
100
Deoxygenation
Fig. 11. Side view and top views of DFT calculated reactant, and prod-
uct structures for deoxygenation/hydrogenation of phenoxy interme-
diate on Rh(111) terrace surface. (Activation energy/energy of reac-
tion).
50
0
Hydrogenation
-750
-700
-650
-600
-550
-500
-450
stronger. Experimentally, this is the case observed with the
catalysts containing smaller metal particles, which are shown
E
(O) (kJ/mol)
Eads(O) (kJ/mol)
ads
Fig. 13. Correlation between activation energies for deoxygenation of
surface phenoxy (Ph–O*) on different metal surfaces and the oxophilic-
ity (O-affinity).
to contain
a higher density of steps, edges and other
low-coordination sites.
3.3.4. Reactions of the phenoxy intermediate
By combing previous experimental and theoretical results
on Pt, Ru and Fe catalysts [7] with the current observations we
can conclude that metal oxophilicity serves as a good de-
scriptor for the HDO activity and selectivity of metals. That is,
hydrogenation follows the trend Pt(111) > Rh(111) > Rh(533)
> Ru(0001) > Fe(110), while the opposite trend is found for the
deoxygenation reaction. Remarkably, both trends correlate
with the oxophilicity of the metal. Fig. 13 shows that the values
obtained in the current calculations for the different surfaces of
Rh are in excellent match with the correlations previously
found for Pt, Ru and Fe metal catalysts.
Once anisole has been sequentially dehydrogenated, a sur-
face phenoxy remains as a crucial intermediate, on both the
Rh(111) and Rh(533) surfaces. As demonstrated in Scheme 1,
this intermediate can undergo either hydrogenation to form
phenol or direct Caryl‒O cleavage to form surface phenyl, which
subsequently hydrogenates to become benzene. In the follow-
ing we analyze the surface reactions of the phenoxy species.
Experiments carried out by Pasco et al. [50] showed that ac-
tivation barriers for adsorption of hydrogen on Fe (111), (100),
(110) surfaces were relatively low, averaging 48 kJ/mol.
Therefore, the rate limiting step for hydrogenation should be
the actual surface reaction of the Ph‒O* intermediate and H* to
obtain phenol. As shown in Fig. 11, over Rh(111) surface, hy-
drogenation of phenoxy is slightly exothermic while deoxygen-
ation is endothermic. The energy barrier to hydrogenate the
surface phenoxy on Rh(111) is 136 kJ/mol, much lower than
that for deoxygenation (190 kJ/mol). This explains why we
observed phenol as the main product resulting from the surface
phenoxy on catalysts with larger Rh particles. At the same time,
DFT calculations show that hydrogenation on Rh(111) is less
favorable than on Pt(111) which was found to be 17 kJ/mol. On
the other hand, this value is about the same as that found for
Ru(0001), 134 kJ/mol, less than that for Fe(110), 147 kJ/mol.
By contrast, as shown in Fig. 12, the energy barrier to hy-
drogenate the surface phenoxy on Rh(533) is 166 kJ/mol,
which is higher than that for deoxygenation on this surface
(155 kJ/mol). This explains the experimental trend shown in
Fig. 6 with the benzene-to-phenol ratio increasing with de-
creasing size of Rh nanoparticles.
4. Conclusions
It is clearly shown here that metal oxophilicity is a good de-
scriptor for HDO performance of metal catalysts and, more
importantly, it can be manipulated by changing the d-band
center, which in turn can be tailored via changing either the
metal type or the coordination of metal atoms. The experi-
mental results obtained with metal nanoparticles of varying
sites are in good agreement with DFT calculations, which use
Rh(111) to demonstrate terraces in large metal clusters and
Rh(533) to represent step sites in small metal clusters. Over Rh
metal catalyst, the most energetically favorable pathway of
anisole is to form a phenoxy intermediate, which can later un-
dergo either hydrogenation to phenol or deoxygenation to
benzene. Small Rh metal clusters with low coordination metal
sites are more active for the deoxygenation pathway but also
quickly deactivated while large clusters with high coordination
sites are more active toward hydrogenation and more stable.
The formation of small amounts of CH3OH suggests that a small
fraction of anisole can undergo direct Caryl–O cleavage, which is
not energetically favorable but still possible, especially on small
Rh clusters.
References
[1] D. E. Resasco, B. Wang, D. Sabatini, Nat. Catal., 2018, 1, 731–735.
[2] J. A. Herron, T. Vann, N. Duong, D. E. Resasco, S. Crossley, L. L.
Lobban, C. T. Maravelias, Energy Technol., 2017, 5, 130–150.
Fig. 12. Side view and top views of DFT calculated reactant, and prod-
uct structures for deoxygenation/hydrogenation of phenoxy interme-
diate over Rh(533). (Activation energy/energy of reaction).

Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
1729
Graphical Abstract
Chin. J. Catal., 2019, 40: 1721–1730 doi: S1872-2067(19)63345-0
Hydrodeoxygenation of anisole over different Rh surfaces
Nhung N. Duong, Darius Aruho, Bin Wang *, Daniel E. Resasco *
University of Oklahoma, USA; Vietnam National University, Vietnam
Rh(533)
Rh(111)
O
O
H
Hydrogenation
Deoxygenation
Manipulating the surface coordination of rhodium atoms can greatly affect the metal oxophilicity and in turn control its activity for C‒O
bond cleavage. In the hydrodeoxygenation of anisole, a step surface favors deoxygenation to benzene while a smooth terrace prefers hy-
drogenation to phenol.
[3] D. E. Resasco, S. P. Crossley, Catal. Today, 2015, 257, 185–199.
[4] Q. Bu, H. Lei, A. H. Zacher, L. Wang, S. Ren, J. Liang, Y. Wei, Y. Liu, J.
Tang, Q. Zhang, R. Ruan, Bioresource Technol., 2012, 124,
470–477.
Technol., 2018, 8, 4083–4096.
[20] E. Furimsky, Appl. Catal. A, 2000, 199, 147–190.
[21] A. J. R. Hensley, Y. C. Hong, R. Q. Zhang, H. Zhang, J. M. Sun, Y. Wang,
J. S. McEwen, ACS Catal., 2014, 4, 3381–3392.
[5] T. Prasomsri, T. Nimmanwudipong, Y. Roman-Leshkov, Energy
Environ. Sci., 2013, 6, 1732–1738.
[22] C. Zhao, Y. Kou, A. A. Lemonidou, X. B. Li, J. A. Lercher, Angew.
Chem. Int. Edit., 2009, 48, 3987–3990.
[6] Q. H. Tan, G. H. Wang, L. Nie, A. Dinse, C. Buda, J. Shabaker, D. E.
Resasco, ACS Catal., 2015, 5, 6271–6283.
[23] C. Zhao, Y. Z. Yu, A. Jentys, J. A. Lercher, Appl. Catal. B, 2013, 132,
282–292.
[7] Q. Tan, G. Wang, A. Long, A. Dinse, C. Buda, J. Shabaker, D. E. Re-
sasco, J. Catal, 2017, 347, 102–115.
[24] X. L. Zhu, L. Nie, L. L. Lobban, R. G. Mallinson, D. E. Resasco, Energy
Fuels, 2014, 28, 4104–4111.
[8] N. Duong, Q. H. Tan, D. E. Resasco, C. R. Chim., 2018, 21,
155–163.
[25] C. Zhao, J. He, A. A. Lemonidou, X. Li, J. A. Lercher, J. Catal., 2011,
280, 8–16.
[9] F. Yang, D. Liu, Y. Zhao, H. Wang, J. Han, Q. Ge, X. Zhu, ACS Catal.,
2018, 8, 1672–1682.
[26] C. Zhao, Y. Kou, A. A. Lemonidou, X. Li, J. A. Lercher, Chem. Com-
mun., 2010, 46, 412–414.
[10] D. J. Rensel, S. Rouvimov, M. E. Gin, J. C. Hicks, J. Catal., 2013, 305,
256–263.
[11] P. T. M. Do, A. J. Foster, J. Chen, R. F. Lobo, Green Chem., 2012, 14,
1388–1397.
[27] C. Zhao, J. A. Lercher, ChemCatChem, 2012, 4, 64–68.
[28] J. Greeley, J. K. Nørskov, Surf. Sci., 2005, 592, 104–111.
[29] R. M. Watwe, R. D. Cortright, J. K. Nørskov, J. A. Dumesic, J. Phys.
Chem. B, 2000, 104, 2299–2310.
[12] K. Lee, G. H. Gu, C. A. Mullen, A. A. Boateng, D. G. Vlachos,
ChemSusChem, 2015, 8, 315–322.
[13] A. Robinson, G. A. Ferguson, J. R. Gallagher, S. Cheah, G. T. Beck-
ham, J. A. Schaidle, J. E. Hensley, J. W. Medlin, ACS Catal., 2016, 6,
4356–4368.
[30] J. Kleis, J. Greeley, N. A. Romero, V. A. Morozov, H. Falsig, A. H.
Larsen, J. Lu, J. J. Mortensen, M. Dułak, K. S. Thygesen, J. K. Nør-
skov, K. W. Jacobsen, Catal. Lett., 2011, 141, 1067–1071.
[31] M. Mavrikakis, B. Hammer, J. K. Nørskov, Phys. Rev. Lett., 1998, 81,
2819–2822.
[14] A. M. Robinson, L. Mark, M. J. Rasmussen, J. E. Hensley, J. W. Med-
lin, J. Phys. Chem. C, 2016, 120, 26824–26833.
[32] J. R. Kitchin, J. K. Nørskov, M. A. Barteau, J. G. Chen, Phys. Rev. Lett.,
2004, 93, 156801.
[15] X. Liu, W. An, Y. Wang, C. H. Turner, D. E. Resasco, Catal. Sci. Tech-
nol., 2018, 8, 2146–2158.
[33] H. Wang, S. Chen, L. Cao, W. X. Li, J. Lu, X. K. Gu, X. Zheng, H. Pan, J.
Zhu, Sci. Adv., 2019, 5, eaat6413.
[16] P. M. de Souza, R. C. Rabelo-Neto, L. E. P. Borges, G. Jacobs, B. H.
Davis, D. E. Resasco, F. B. Noronha, ACS Catal., 2017, 7,
2058–2073.
[17] P. M. de Souza, L. Nie, L. E. Borges, F. B. Noronha, D. E. Resasco,
Catal. Lett., 2014, 144, 2005–2011.
[18] M. B. Griffin, G. A. Ferguson, D. A. Ruddy, M. J. Biddy, G. T. Beckham,
J. A. Schaidle, ACS Catal., 2016, 6, 2715–2727.
[19] V. Jain, Y. Bonita, A. Brown, A. Taconi, J. C. Hicks, N. Rai, Catal. Sci.
[34] L. Dong, L. L. Yin, Q. Xia, X. Liu, X. Q. Gong, Y. Wang, Catal. Sci.
Technol., 2018, 8, 735–745.
[35] G. Kresse, J. Hafner, Phys. Rev. B, 1993, 47, 558.
[36] G. Kresse, J. Furthmüller, Phys. Rev. B, 1996, 54, 11169.
[37] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett., 1996, 77,
3865–3868.
[38] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys., 2010, 132.
[39] J. W. Arblaster, Platinum Metals Rev., 1996, 40, 62.

1730
Nhung N. Duong et al. / Chinese Journal of Catalysis 40 (2019) 1721–1730
[40] G. Henkelman, H. Jónsson, J. Chem. Phys., 2000, 113, 9978–9985.
[41] G. Henkelman, B. P. Uberuaga, H. Jonsson, J. Chem. Phys., 2000,
113, 9901–9904.
[42] D. Sheppard, G. Henkelman, J. Comput. Chem., 2011, 32,
1769–1771.
[43] H. F. J. van't Blik, J. B. A. D. van Zon, T. Huizinga, J. C. Vis, D. C. Kon-
ingsberger, R. Prins, J. Phys. Chem., 1983, 87, 2264–2267.
[44] D. J. C. Yates, L. L. Murrell, E. B. Prestridge, J. Catal., 1979, 57,
41–63.
[45] C. A. Rice, S. D. Worley, C. W. Curtis, J. A. Guin, A. R. Tarrer, J. Chem.
Phys., 1981, 74, 6487–6497.
[46] G. L. Haller, D. E. Resasco, Adv. Catal., 1989, 36, 173–235.
[47] S. Trautmann, M. Baerns, J. Catal., 1994, 150, 335–344.
[48] M. Saeys, M. F. Reyniers, G. B. Marin, M. Neurock, J. Phys. Chem. B,
2002, 106, 7489–7498.
[49] M. L. Honkela, J. Bjork, M. Persson, Phys. Chem. Chem. Phys., 2012,
14, 5849–5854.
[50] R. W. Pasco, P. J. Ficalora, Surf. Sci., 1983, 134, 476–498.
不同Rh表面上苯甲醚加氢脱氧
Nhung N. Duong a,b, Darius Aruho a, Bin Wang a,*, Daniel E. Resasco a,#
a俄克拉荷马大学化学生物和材料工程学院(CBME), 界面反应工程中心(CIRE), 诺尔曼OK 73019, 美国
b越南国立大学, 国际大学生物技术学院, 胡志明市700000, 越南
摘要: 苯甲醚中烷氧基(Ar-O-R)醚键的断裂是一个有趣的加氢脱氧反应, 因为这个不对称的基团包含两个不同的C–O键,
Caryl–O或Calkyl O, 它们可能会断裂. 最近有关在铂、钌和铁催化剂上苯甲醚HDO的研究表明, 在所有三种金属上都形成了
一种共有的苯氧基表面中间体. 该中间体的后续反应路径因金属而不同, 这取决于金属的亲氧性. 在亲氧性较低的铂上反
应, 苯酚是唯一的主要产物. 相反, 在亲氧性较强的铁催化剂上, 唯一的初级产物是苯而不是苯酚. 在亲氧性适中的Ru上,
苯和苯酚都是主要产物. 本文研究了不同表面纳米结构的铑催化剂. 采用实验测量和理论计算相结合的方法, 研究了可用
来控制金属亲氧性的参数, 即金属配位数的影响. 结果证实了金属亲氧性是描述金属催化剂催化HDO性能的很好的指标,
它可以通过选择金属类型和/或金属配位程度来加以控制. 含低配位数的金属位的小Rh金属簇合物对脱氧路径活性更高,
但很快失活, 而含高配位数的大的簇合物对加氢活性更高, 且更加稳定.
关键词: 加氢脱氧; 苯甲醚; 烷氧基; 亲氧性; 酚类; 表面缺陷
收稿日期: 2019-01-10. 接受日期: 2019-03-07. 出版日期: 2019-11-05.
*通讯联系人. 电子信箱:wang_cbme@ou.edu
#通讯联系人. 电子信箱: resasco@ou.edu
本文的电子版全文由Elsevier出版社在ScienceDirect上出版(http://www.sciencedirect.com/science/journal/18722067).