Selectivity switch in transformation of CO2 from ethanol to methanol on Cu embedded in the defect carbon

Article abstract of DOI:10.1007/s11426-019-9730-5

The copper-based catalysts have been generally regarded as high-performance catalysts for CO₂ hydrogenation toward methanol, while the production of ethanol via C-C coupling on the copper-based catalysts is still challenging. Herein, we report a new catalyst where Cu nanoparticles are embedded in the carbon support with abundant defect sites, achieving a high selectivity for ethanol in the CO₂ hydrogenation. The experiments coupled with the theoretical studies show a clear map where carbon defects serve as anchor sites that can stabilize interfacial copper species, and interfacial Cu sites with low coordination numbers can adsorb two C₁ species and later convert them to a C₂ species via a hydrogenation-induced coupling reaction. Further adjacent Cu atoms of interfacial Cu sites can facilitate OH reduction reactions via the Cu-Cu bridge adsorption to assist the formation of ethanol. Especially, those specific active sites easily disappear in the reducing conditions and during the reaction, the major product can transform from ethanol to methanol.

Full text of DOI:10.1007/s11426-019-9730-5

SCIENCE CHINA
Chemistry
•
ARTICLES•
https://doi.org/10.1007/s11426-019-9730-5
SPECIAL ISSUE: Celebrating the 100th Anniversary of Chemical Sciences in Nanjing University
Selectivity switch in transformation of CO from ethanol to
2
methanol on Cu embedded in the defect carbon
1
†
2†
2,3,4*
1*
Shuohao Li , Govindarajan Saranya , Mingyang Chen
& Yan Zhu
1
School of Chemistry and Chemical Engineering, Nanjing University, Nanjing 210093, China;
2
Beijing Computational Science Research Center, Beijing 100193, China;
3
Center for Green Innovation, School of Materials Science and Engineering, University of Science and Technology Beijing,
Beijing 100083, China;
4
Shunde Graduate School of University of Science and Technology Beijing, Foshan 528000, China
Received December 13, 2019; accepted March 24, 2020; published online April 13, 2020
The copper-based catalysts have been generally regarded as high-performance catalysts for CO hydrogenation toward methanol,
2
while the production of ethanol via C–C coupling on the copper-based catalysts is still challenging. Herein, we report a new
catalyst where Cu nanoparticles are embedded in the carbon support with abundant defect sites, achieving a high selectivity for
ethanol in the CO hydrogenation. The experiments coupled with the theoretical studies show a clear map where carbon defects
2
serve as anchor sites that can stabilize interfacial copper species, and interfacial Cu sites with low coordination numbers can
adsorb two C species and later convert them to a C species via a hydrogenation-induced coupling reaction. Further adjacent Cu
1
2
atoms of interfacial Cu sites can facilitate OH reduction reactions via the Cu–Cu bridge adsorption to assist the formation of
ethanol. Especially, those specific active sites easily disappear in the reducing conditions and during the reaction, the major
product can transform from ethanol to methanol.
CO , ethanol, methanol, copper, defect carbon
2
Citation:
Li S, Saranya G, Chen M, Zhu Y. Selectivity switch in transformation of CO
China Chem, 2020, 63, https://doi.org/10.1007/s11426-019-9730-5
2
from ethanol to methanol on Cu embedded in the defect carbon. Sci
1
Introduction
catalysts has been reported for CO hydrogenation to ethanol
2
[11]. Meanwhile, the Pd-based catalysts have also attracted
Carbon dioxide has been largely used to make valuable
chemicals and low-carbon fuels, such as hydrocarbons, car-
much attention to produce ethanol owing to their capability
of promoting C–C couplings [11,12]. However, those noble
metal catalysts greatly limit extensive developments due to
their scarcities and expensive price [13,14]. Therefore, it is
highly desirable to obtain the inexpensive and abundant
bon monoxide, alcohols and C (n>2) oxygenates [1–7].
n
Among them, the hydrogenation of CO towards alcohols is
2
one of the most significant processes in the CO conversion
2
[2,8,9], as alcohols like ethanol can be served as a kind of
transition metals to drive CO conversion to ethanol.
2
renewable fuel additives and eminent industrial inter-
mediates [2,10]. However, it is challenging to generate
ethanol compared to common methanol products, due to the
hardship of C–C bond coupling. The Pt as well as its alloy
The low toxic, cheap and readily available copper-based
catalysts have been pursued as highly efficient catalysts for
CO hydrogenation toward methanol [15–17], e.g., typical
2
Cu/ZnO series catalysts [18–20]. In addition, the Cu nano-
particles imbedded in N-doped carbon nanospikes (CNS) can
provide an intimate contact between copper surfaces and
reactive sites in the carbon support, which can induce CO2
†
*
These authors contributed equally to this work.
Corresponding authors (email: mychen@ustb.edu.cn; zhuyan@nju.edu.cn)
©
Science China Press and Springer-Verlag GmbH Germany, part of Springer Nature 2020
chem.scichina.com link.springer.com

2
Li et al. Sci China Chem
electroreduction toward ethanol with a high selectivity
84%) [21]. It implies that the copper catalysts with specific
were grinded in a mortar about 20 min again, named for
Grind-2nd, and so on.
(
active sites are able to facilitate C–C coupling to obtain
ethanol in the CO conversion.
2
2
.2 Catalytic hydrogenation of CO2
Herein, we report a new copper-based catalyst, that is, Cu
nanoparticles embedded in a defect carbon support (denoted
as Cu@C), which is prepared in a fixed-bed reactor for CO2
hydrogenation. Interestingly, the product distribution can
change with reactions: the major product is gradually chan-
ged from ethanol to methanol. The experiments coupled with
the theoretical studies explain that the defects on the carbon
support serve as anchor sites to stabilize interfacial copper
species with low coordinated Cu nanoparticles, which can
assist C–C bond coupling with two CO intermediates ad-
sorbed on these copper species to initially form ethanol as a
major product. When these copper species are spent, the C–H
bond formation competes against the C–C coupling and
hence the methanol becomes the major product.
CO hydrogenation was performed in a fixed-bed reactor.
100 mg catalyst (20–40 mesh) was mixed with quartz sands
placed in the middle of the reactor. First of all, the catalyst
2
was reduced in 99.99% H at 150 °C for 2 h (flow rate=
2
20 mL/min). And then, the flow of the reactant gas (H2/
CO =3/1) with a flow rate of 20 mL/min was started. The
2
reaction temperature was programmed by a temperature
controller and ramped at a rate of 5 °C/min from 150 to
325 °C (the reactants and products were analysed at various
temperatures, such as 150, 200, 225, 250, 275, 300 and
325 °C) at the typical pressure of 2 MPa. The products were
analysed with a gas chromatograph (GC-9860) instrument
equipped with a column FFAP (50 m×0.32 mm×0.25 μm)
and a thermal conductivity detector (TCD) with a PQ col-
umn. Nitrogen at a mole fraction of 4% in the reactant
mixture was used as an internal standard for the calculation
2
Experimental
of CO conversion.
2
2
.1 Preparation of the catalysts
A certain amount of nitrates and cellulose (Table S1, Sup-
porting Information online) were together dissolved into
acetone (20 mL) until full dissolution; then the mixture was
dried at 100 °C and carbonized at 500 °C for 2 h with a
temperature rate of 2 °C/min. Finally, the catalyst was de-
noted as n%M@C(T), where M meant metal elements
2
.3 Characterizations
The nitrogen adsorption-desorption isotherm at 77 K was
obtained on a micromeritics ASAP 3020 volumetric ad-
sorption analyser, for which 100 mg of sample was out-
gassed at 573 K for 3 h in the degas port prior to tests.
Specific surface area of the sample was calculated by the
BET (Brunauer-Emmett-Teller) method using adsorption
(M=Cu, Zn, Fe, Ni, Co and Zr), n meant the different ad-
ditive amounts (n=3, 5, 7, 10 and 15 wt%), T meant carbo-
nization temperature (T=300, 400, 600, 700 and 1100 °C).
Samples without additive metal salts were prepared follow-
data acquired in the relative pressure range (P/P ) from 0.04
to 0.2, while the total pore volume was determined from the
0
ing the same procedure denoted as the C support. Cu O
amounts adsorbed at a P/P of about 0.99. The morphology
x
0
catalyst was obtained from 7%Cu@C carbonized at 500 °C
for 2 h to remove the carbon supports.
of the sample was observed by a scanning electron micro-
scope (SEM) using a Hitachi S4800 FE-SEM system (Japan)
with 10 kV accelerating voltage and 10 mA of beam cur-
rents. Transmission electron microscopy (TEM) analysis was
carried out on a JEM-2100 electron microscope (Japan)
operating at 200 kV. The X-ray absorption fine structure
(XAFS) spectra at Cu K-edge were measured at BL14W1
beamline of Shanghai Synchrotron Radiation Facility
(SSRF). They were operated at 3.5 GeV under a “top-up”
mode with a constant current of 260 mA. The energy was
calibrated according to the absorption edge of a pure Cu foil.
Athena and Artemis codes were used to analyse the data and
fit the profiles. The X-ray photoelectron spectroscopy (XPS)
was performed on a PHI 5000 Versa Probe instrument using
Al-Kα as the excitation source. Infrared Fourier transform
spectroscopy (FTIR) spectra of CO adsorbed were recorded
using an FTIR NICOLET iS10 (Thermo Fisher Scientific
Inc., USA) provided with a DRIFT cell (Praying Mantis,
Harrick) and a liquid nitrogen cooled MCT detector. The
Cu(NO ) •3H O of 0.317 g and C supports of 0.125 g were
3
2
2
dissolved into acetone (20 mL) and then carbonized at
00 °C for 2 h with a temperature rate of 2 °C/min. The
5
obtained catalyst was denoted as the Cu/C catalyst. More-
over, this C support was also replaced with graphene (G) and
carbon nanotubes (CNT) as a reference, and the samples
were denoted as Cu/G and Cu/CNT. Besides, the fresh 7%
Cu@C catalyst was treated at 2 MPa for 12 h in H or N
2
2
atmosphere and the treated catalysts were named as 7%
Cu@C-H and 7%Cu@C-N , respectively. Meanwhile the
2
2
7
%Cu@C-LiAlH catalyst was obtained by treating 7%
4
Cu@C in a mixed solution of LiAlH and tetrahydrofuran
4
(THF).
For the regeneration test, the used 7%Cu@C catalysts
were grinded in a mortar for 20 min at room temperature,
denoted as Grind-1st. When the Grind-1st 7%Cu@C cata-
lysts can produce methanol as major products, the catalysts

Li et al. Sci China Chem
3
catalysts were loaded into the cell and then purged under Ar
flow for 2 h at 150 °C. After the catalysts cooled to the
analysis temperature, a background spectrum was first col-
lected in Ar atmosphere and then exposed to CO (10%) at
hydrogenation reactions were constructed and added to the
SA and IF sites and the reaction pathways towards ethanol
were predicted. Three major types of adsorption configura-
tions were considered and found for the CO hydrogenation
2
1
2
2
98 K for 30 min. After the saturation adsorption, the data
reactions over the SA and IF Cu sites, i.e., dual-, η -, and η -
was collected under Ar flow with 20 mL/min. Time-resolved
types. In the dual-adsorption configuration, two C (one-
1
in situ IR data were recorded using a Bruker FTIR TENSOR
carbon) ligands adsorb at the same Cu without direct inter-
actions between the two ligands. The coupling of the two C1
ligands of the dual-adsorption configuration leads to a C2
(Germany), provided with a DRIFT cell, both equipped with
a transmission accessory and a high-temperature/pressure
reaction chamber. The DRIFT cell was used as a reactor for
the catalytic experiments. Prior to catalytic testing, the
samples (ca. 12 mg) were placed in the reactor/DRIFT cell
and treated in flowing Ar at 150 °C for 1 h to remove the
moisture, followed by heating up 250 °C in flowing Ar. A
background spectrum was collected in Ar before the catalyst
exposure to the CO /H mixture gas. The samples were
1
(two-carbon) adsorbate that can be adsorbed via either η - or
2
1
η -configurations. In the η -adsorption configuration, only
one C of the C ligand is bonded to Cu, and hydrogenation is
2
only considered for α-C unless there is an uncommon in-
2
teraction between Cu and β-C. In the η -adsorption config-
uration, both C atoms of the C species may be activated by
2
Cu via direct Cu–C interactions, and hence the hydrogena-
tion of both C atoms is considered. For a considered C atom,
two types of hydrogenation reactions are considered, i.e.,
adding an H to C or removing an OH from C.
2
2
treated with the CO /H mixture gas (1:3) at 250 °C and the
2
2
data were collected every 10 min. The optical properties of
samples were investigated by a Raman spectrometer (Horiba,
+
LabRAMAramis, excited with an Ar line at 532 nm). The
All of the first-principles calculations were performed
using the Vienna Ab Initio Simulation Package (VASP) [23]
with the projector-augmented-wave (PAW) [24] potentials.
The structure optimization calculations were performed with
the Perdew-Burke-Ernzerhof (PBE) [25] exchange-correla-
tion functional. The energy cut-off was set to 500 eV and a
3×4×1 Monkhorst-Pack grid [26] was used to samples in the
Brillouin zones (BZ). A force convergence criterion of 0.01
eV/angstrom on each atom was used. The activation barriers
were determined from transition state search calculations
using the climbing-image nudged elastic band method with 5
intermediate images [27].
1
H NMR spectra were recorded on a Bruker ARX-400
spectrometer. Chemical shifts were reported in ppm relative
1
3
C
to the internal solvent peak (δ=4.79 ppm, for D O). The
2
NMR spectra were recorded on a Bruker ARX-600 spec-
trometer and all of the chemical shifts were reported down-
field in ppm relative to the carbon resonance of deuterated
solutions.
2
.4 Computational methods
The Cu–C interface of the Cu@C catalyst was modelled
using a Cu(111)@C nanobelt model (slab) with periodic
boundary conditions (PBC). The supercell of the Cu@C
model was composed of a Cu(111) nanobelt and an amor-
phous carbon nanobelt that were parallelly juxtaposed and
joined. The Cu(111) nanobelt was cut from the optimized Cu
crystal structures, and expanded into a 4×3 slab, containing 4
atom layers and a total number of 48 Cu atoms. The amor-
phous carbon nanobelt, consisting of 78 C atoms, was con-
structed based on the nanostructure reported by Margraf et
al. [22]. The initial supercell was first optimized with two
bottom layers of atoms that were frozen and fixed lattice
parameters, and the as-optimized supercell was further op-
timized without any constraints on atom positions. The
Cu(111)@C slab has a size of 15.3 Å×10.2 Å with a vacuum
of 15 Å to separate mirror images along the vertical direc-
tion. The widths of both the Cu and C nanobelts are ~5 Å. In
our Cu(111)@C model, two types of Cu sites with low Cu–C
3 Results and discussion
3
.1 Physical properties
SEM images of the C support showed that the defect carbon
was likely to be the amorphous bulk with a compact structure
and smooth surface (Figure S1, Supporting Information on-
line). Cu nanoparticles were embedded in the C support,
looking like fruits embedded in a cake (Figure 1(a, b)).
Meanwhile, TEM images provided a direct observation of
7%Cu@C (7 wt% Cu loading) catalysts, in which the Cu
nanoparticles with an approximate size of 5 nm were well
dispersed in the C support. The lattice spacing of 0.204 nm is
assigned to the {111} lattice fringes of the cubic Cu (Figure 1
(c, d)) [28]. Nitrogen sorption experiments gave a BET
2
surface area of 223.9 m /g for 7%Cu@C, which was smaller
2
coordination numbers were considered for the CO hydro-
than that of pure C supports (463.3 m /g) (Table S2).
2
genation reaction, a single-atom (SA) site resulting from the
diffusion of a single Cu atom into carbon and an interfacial
Meanwhile, the BET surface areas of a series of Cu@C
samples with the different Cu loading decreased with the
additive amount of copper salts increasing. Thus, the in-
troduction of Cu nanoparticles changed the internal structure
(IF) site with Cu at the interface between the Cu nanobelt and
the C nanobelt. Reactive intermediate species for the CO2

4
Li et al. Sci China Chem
of the C support. The X-ray absorption near edge structure
XANES) experiments showed the copper particles in the
(
7
0
%Cu@C catalysts were mostly assigned to Cu species due
to the reduction of concomitant C supports (Figure 1(e)). The
extended X-ray absorption fine structure (EXAFS) studies
indicated local coordination environments in the 7%Cu@C,
where the Cu species anchored in the carbon support had low
coordination numbers (Figure 1(f)). Besides, Raman spectra
3
indicated that this C support had amorphous sp forms of
carbon (Figure 1(g)), because the existence of D peak and
large ID/IG intensity ratio confirmed the lower degree of
graphitization with abundant defects [29]. Meanwhile, the
introduction of Cu nanoparticles did not change the defective
structure of the carbon support, which was consistent with C
1
s XPS of 7%Cu@C catalysts that contained numerous
oxygen functional groups accompanied by these defects
Figure 1(h)) [30].
(
3
.2 Catalytic properties
Now we discussed the catalytic properties of the Cu@C
catalysts for CO hydrogenation in a fixed-bed reactor. From
2
Figure 2, the Cu@C catalysts gave rise to the four products
(ethanol, methanol, methyl formate and methane). The major
liquid products were further confirmed by the nuclear mag-
netic resonance (NMR). The reaction mixture was collected
1
by a cold trap and detected by H NMR (Figure S2). The
peak at δ=3.34 ppm is assigned to *CH of methanol and the
3
peaks at δ=1.17 and 3.65 ppm are assigned to *CH and
Figure 1 (a, b) SEM and (c, d) TEM images of the 7%Cu@C catalyst.
The white circles in C highlight the Cu nanoparticles embedded in carbon
supports. (e) XANES and (f) EXAFS profiles of 7%Cu@C. (g) Raman
spectra of the 7%Cu@C and C. (h) C 1s XPS profile of 7%Cu@C (color
online).
3
13
13
*
CH groups of ethanol [31]. When using C-labeled CO /
2 2
H as the feed gas, the peaks at δ=16.92 and 57.49 ppm
2
correspond to *CH and *CH groups of ethanol, and the
3
2
peak at δ=48.95 ppm is assigned to methanol (Figure S3)
32], respectively, confirming that methanol and ethanol are
[
(CNT) obtained under higher carbonization temperatures,
the ethanol was almost undetectable over the Cu/G and Cu/
CNT catalysts while methanol became the major product
with a >99% selectivity (Figure 2(c)). Meanwhile, 7%Cu/C
prepared by the post-loading method also led to a significant
decrease in the ethanol selectivity, owing to the much low
contact between Cu nanoparticles and C supports.
produced from the CO gas, rather than C supports.
2
As the loading amount of Cu species increased from 3 to
1
5 wt%, as shown in Figure 2(a), the CO conversion in-
2
creased (from 1.0 to 2.6%), whereas ethanol selectivity im-
proved initially and then decreased, with a maximum
selectivity of ethanol to be about 74.4% on the 7%Cu@C
catalyst. Furthermore, it is noting that the defect degree of C
supports played a key role in the catalytic performances.
Figure 2(b) showed that low carbonization temperatures for
Furthermore, catalytic properties for the comparative cat-
alysts that had other metal species (e.g., Zn, Fe, Ni, Co and
Zr) doping in C supports were also evaluated for CO reac-
2
7
%Cu@C was much favorable for CO hydrogenation to
tions with H . As shown in Figure 2(d), ethanol cannot be
2
2
ethanol, possibly as a carbon precursor under low carboni-
zation temperatures can generate numerous defects and
oxygen functional groups [33–35].
obtained on these catalysts under identical reaction condi-
tions. The 7%Zn@C and 7%Zr@C catalysts generated me-
thanol with the selectivity of 91.5% and 50.9%, respectively.
Both 7%Ni@C and 7%Co@C catalysts gave the >95% se-
lectivity for methane. The 7%Fe@C exhibited the product
selectivity for 25.1%, 21.9 %, 15.2% and 37.8% for metha-
nol, methyl formate, methane and carbon monoxide, re-
spectively.
Notably, pure C supports were little active (Figure 2(c)).
The Cu O catalyst from calcined 7%Cu@C gave methanol as
x
the major product with a selectivity of 97.5%, which was
similar to typical Cu/ZnO/Al O catalysts utilized for CO
2
3
2
hydrogenation (Figure 2(c)) [18]. When this kind of C sup-
ports was replaced with graphene (G) or carbon nanotubes
More interestingly, the product distribution on the 7%

Li et al. Sci China Chem
5
Grind-2nd 7%Cu@C catalyst lost the selectivity for ethanol,
the Grind-3nd 7%Cu@C catalyst exhibited 37.2% initial
selectivity for ethanol. The results show that the grinding
method is feasible for the Cu@C catalyst to recover the se-
lectivity of ethanol to some extent. By the solid phase
grinding, the active species of the Cu@C catalyst, that is,
0
Cu (CO) , can be re-formed on the interface of C and Cu,
2
which is indicated by FTIR spectra of CO adsorption onto
the Grind-1st 7%Cu@C catalyst (Figure S4(b)).
3
.3 Insights into the catalytic mechanism
We next turned our attention to the potential mechanism to
unveil the mystery that the 7%Cu@C catalyst can initially
give rise to ethanol as major products and subsequently give
methanol as major products. Firstly, the Cu nanoparticles
embedded in C supports after reactions (spent 7%Cu@C)
remained well-dispersed and were similar to the fresh cata-
lysts (Figure S5), suggeasting that Cu nanoparticles did not
significantly gather and grow into large particles. Then,
FTIR spectra of adsorbed CO onto the catalysts was carried
out to determine the active sites. As seen from Figure 3(a),
two kinds of Cu species can be observed by CO adsorbed on
−
1
the 7%Cu@C catalyst: the band at 2,170 cm is assigned to
+
−1
Cu –CO species [36] and the bands at 2,040 and 2,117 cm
Figure 2 Catalytic performances of CO hydrogenation over a series of
0
2
are attributed to Cu –CO species [37]. Especially, the ad-
jacent band at 2,040 cm is assigned to Cu (CO) geminal
catalysts. (a) Cu@C with different Cu loadings; (b) 7%Cu@C prepared at
different carbonization temperatures. (c) Comparison of catalytic properties
−
1
0
2
of different catalysts: C support, Cu O, and Cu nanoparticles supported on
x
complexes that were formed on the low coordinated copper
sites [38], which indicated that CO can be strongly absorbed
onto the unsaturated copper sites where the π-back-donation
was sterically facilitated. As a result, another CO molecule
was coordinated to the highly unsaturated sites, and then the
graphene (G) and carbon nanotubes (CNT). (d) Catalytic properties of other
metal species like Zn, Fe, Ni, Co, and Zr doped in C supports with fixed
doping amount of 7 wt%. (e) The long-term performance of 7%Cu@C
catalysts for CO2 hydrogenation. Reaction conditions: 100 mg catalysts;
2
MPa feed gas pressure with a molar composition of CO /H =1/3; a feed
2 2
flow rate=20 mL/min; a reaction temperature=250 °C. Error bars corre-
spond to deviations from three independent area integrals (color online).
C–C coupling occurred on the catalyst to form the ethanol.
0
However, it was found that the intensity of Cu (CO) species
2
simultaneously reduced after reactions, meaning that these
copper active sites were disappeared gradually with reactions
(Figure 3(b)). It might account for the transformation of the
Cu@C catalyst changed with the reaction time, although the
CO conversion held steady at 2.5% during the time-on-
2
stream experiment (Figure 2(e)). The selectivity for ethanol
changed from 74.4% to 2.5%, implying that the reaction
major products from ethanol to methanol with CO hydro-
2
genation on 7%Cu@C catalysts. Besides, there was no
0
pathway of CO was altered during the reaction. Notably, the
characteristic peaks assigned to Cu (CO) species for the
2
2
+
high selectivity for ethanol can be re-obtained using the spent
catalysts by the solid phase grinding. The 7%Cu@C catalyst
that first lost the high selectivity for ethanol was grinded in
the mortar for 20 min at room temperature, denoted as Grind-
Cu O catalyst (Figure 3(c)), beside Cu –CO species
x
−
1
0
−1
(2,179 cm ) and Cu –CO species (2,123 cm ) [36]. In other
words, the thermal-treated 7%Cu@C catalyst had no Cu (CO)2
0
species, and finally no ethanol was produced on the treated
one. The pure C support had no active sites for the CO
adsorption (Figure 3(d)), that is, no dicarbonyl was absorbed
onto the C support, and hence no ethanol was produced.
1
st. When the Grind-1st 7%Cu@C catalyst gave methanol as
major products, the catalyst was again grinded in the mortar
about 20 min for again, named for Grind-2nd, and so on. As
shown in Figure S4(a), when the 7%Cu@C catalyst gave
only <10% selectivity for ethanol, the Grind-1st 7%Cu@C
catalyst gave rise to 66.9% initial selectivity for ethanol.
When the Grind-1st 7%Cu@C catalyst lost the selectivity for
ethanol with the reaction time, the Grind-2nd 7%Cu@C
catalyst gave 53.8% initial selectivity for ethanol. When the
To further determine whether a close correlation was be-
0
tween the product selectivity and Cu (CO) species, we in-
2
vestigated the catalytic performances and the CO-IR spectra
of the 7%Cu@C catalysts pretreated by different conditions.
As shown in Figure S6(a), a distinct difference was found in
hydrogenation products between 7%Cu@C-N (pretreated in
2

6
Li et al. Sci China Chem
−
1
−1
2
,127 cm ) were found and the IR band at 2,109 cm as-
signed to the adsorbed CO showed negligible CO products
−
1
[
40]. Importantly, the band at 2,038 cm species assigned to
0
0
Cu (CO) geminal complexes was undetected beside Cu –
CO species (2,109 cm ), which might be due to the low
stability of Cu (CO) intermediates. Thus, one possible
2
−
1
0
2
pathway of CO hydrogenation on the 7%Cu@C catalyst is
2
0
that one of CO intermediates in the Cu (CO) species is re-
duced to the surface-bound CHO* species (1,150 cm )
12,41], and ethanol is likely formed via the coupling of the
2
−
1
[
two CHO* species with the assistance of the unique space
superiority. By contrast, the Cu O catalyst showed the IR
x
bands corresponding to various oxygenated C intermediates
1
while C species were completely undetectable (Figure 4(c,
2
d)), which explained methanol as the major product formed
on the Cu O catalyst.
x
Figure 3 FTIR spectra of CO adsorbed onto fresh 7%Cu@C (a), spent
7
%Cu@C (b), fresh Cu O (c), and fresh C support (d), respectively (color
x
online).
3.4 Proposed reaction pathways for the ethanol for-
mation on Cu@C
N ) and 7%Cu@C-H (pretreated in H ) catalysts: the former
2
2
2
maintained the high selectivity towards ethanol of 65.4%,
whereas the latter lost the ability of producing ethanol
The reaction pathways for the CO hydrogenation reaction
2
over Cu@C catalysts were modelled using density functional
theory (DFT) computations with a Cu(111)@C nanobelt
model that contains single-atoms (SA) and interfacial (IF)
low-coordinate Cu sites (Figure S7). Similar to the reaction
(
1.1%). Moreover, the 7%Cu@C-LiAlH catalyst reduced by
4
a strong reductant (LiAlH ) was unable to produce ethanol
4
(
only 1.5% selectivity) but methanol (92.3% selectivity).
0
Meanwhile, the IR intensity of the crucial Cu (CO) species
on the Cu surface, CO is the first hydrogenated and con-
2
2
in the 7%Cu@C-H catalyst also decreased simultaneously
verted into COOH. DFT calculations show that on the
Cu(111) surface, a COOH is shared by two Cu atoms, while a
SA or IF Cu site can adsorb two COOH ligands to form dual-
*(COOH)COOH (* denotes the surface Cu); this is con-
sistent with the experimental CO adsorption results where
2
0
(Figure S6(b)). The results indicated that not only the Cu (CO)
2
species are crucial to the ethanol formation on the 7%Cu@C
0
catalyst, but also the Cu (CO) species are not stable under
2
0
the reductive conditions. With the amount of Cu (CO)
2
species decreasing, the selectivity of ethanol decreased and
the selectivity of methanol increased. This accounted for the
catalytic result of the long-term test in Figure 2(e), the high-
geminal Cu(CO) is exclusively formed over Cu@C. At the
SA site (Figure 5(a) and Figure S7), subsequent hydro-
genation of a COOH adsorbate of dual-*(COOH)COOH
2
pressure H reduction atmosphere resulted in the gradual
leads to dual-*(CHO)COOH and a leaving H O (∆E=
2
2
0
disappearing of Cu (CO) species, and thus the ethanol
−0.90 eV). Next, the remaining COOH is hydrogenated,
which leads not only the formation of the second CHO but
also the C–C coupling of the two CHO adsorbates, and result
2
dropped and the methanol went up in the products.
To investigate the reaction pathway of the CO hydro-
2
2
genation on the 7%Cu@C catalyst, the reaction intermediate
species were detected by the time-resolved in situ FTIR
spectroscopy. As shown in Figure 4(a, b), the obvious band at
in η -*CHO–CHO (∆E=−0.65 eV). Facile formations of the
2
two-carbon (C ) intermediates such as η -*CHO–CHO at the
2
low-coordinate Cu (for which the formation is exothermic)
−
1
δ−
1
,228 cm corresponding to carboxylates (CO2 ) appeared
might lead to the formation of C products. The hydro-
2
2
1
at the initial reaction due to the primary adsorption and the
genation of η -*CHO–CHO (∆E=−0.30 eV) yields η -
*CHOH–CHO with the two Cu–C distances being 2.08 and
2.58 Å, where there may be a weak interaction between Cu
activation of CO [38]. The trace peaks of unidentate car-
2
−
1
−1
bonates (1,396 cm for υ (OCO) and 1,457 cm for
υas(OCO)) and carbonate species (1,510 cm for CO3 ) were
s
−1
2−
1
and β-C. The hydrogenation at the β-C of η -*CHOH–CHO
2
2
subsequently observed with reactions [39]. Meanwhile, new
yields the η -*CHOH–CHOH analogous to the typical η
−
1
IR peaks were observed: the IR bands at 1,016 and 1,074 cm
corresponded to the adsorbed ethoxy and methoxy species,
ethylene adsorption on a transition metal atom, which results
in a stabilization effect (∆E=−1.15 eV). Hydrogenation of
−
1
2
respectively; a broad peak at 1,309–1,396 cm was assigned
η -*CHOH–CHOH with ½ H may either add a H to C atom
2
−
1
to formate (HCOO ) species [8,39]; the peaks at 2,854, 2,917
or remove a OH from C to form H O, and form η -*CHOH–
2
−
1
1
and 2,958 cm were assigned for CH stretch modes (ν(CH)).
CH OH (∆E=0.21 eV) and η -*CH–CHOH (∆E=1.05 eV)
2
1
1
Moreover, no adsorption peaks of gaseous CO (2,173 and
respectively. Both of η -*CHOH–CH OH and η -*CH–
2

Li et al. Sci China Chem
7
found that the pathway involves highly endothermic reac-
2
1
tions steps, such as η -*CH –CHOH→η -*CH–CH (∆E=
2
2
1
.32 eV), and therefore the formation of alkane C products
2
is inhibited, which is in good agreement with the experi-
mental observations (ethane was not detected in the pro-
ducts).
The predicted reaction pathway at the IF site is shown in
Figure 5(b). The reactivity of CO hydrogenation species
2
over the IF sites is different from that over the SA sites. For
1
example, the C intermediate, *CHO–CHO, has the η -ad-
2
sorption configuration with a bridge O atom on Cu–C, in
2
contrast to the η adsorption of CHO–CHO at the SA site.
This indicates a strong interaction between Cu and α-C
where the Cu not only activates C but also direct affects
certain atoms on α-C at IF Cu. We have shown that the high
2
endothermicity for removing OH from α-C (during the η -
1
*
CHOH–CHOH → η -*CH –CH OH steps) hinders the CO
2
2
2
Figure 4 Time-resolved in situ FTIR spectra of CO hydrogenation on
hydrogenation towards ethanol at the SA sites; conversely,
we found that the removal of OH from η -*CHOH–CHOH
2
2
7
%Cu@C (a, b) and Cu O (c, d) catalysts at 250 °C in 0.2 MPa CO /H at a
x 2 2
molar ratio of 1:3 (color online).
over the IF sites is exothermic (∆E=−0.46 eV). This is be-
1
cause, in the resultant η -bridge-*CH–CHOH species over
the IF Cu sites, the α-C is bonded to two Cu atoms at the
bridge position, and therefore the α-C at IF Cu sites has a
more favorable coordination (coordination number=4) than
1
the corresponding η -*CH–CHOH does (coordination num-
ber=3) over the SA Cu. Consequently, the reaction pathway
towards ethanol is essentially spontaneous without a highly
endothermic step. We propose that the IF Cu sites contribute
most to the ethanol formation in the CO hydrogenation re-
2
actions with the Cu@C catalysts. The reaction pathway to-
wards the alternative alkane C H product is also predicted
2
6
for the IF Cu sites. Similar to the SA reaction, the IF reaction
2
path towards C H involves a highly endothermic η -*CH –
2
6
2
1
CHOH→η -CH–CH step with ∆E=1.05 eV. Our DFT re-
sults thus provide reasonable explanations for the high
2
ethanol conversion and selectivity for the CO hydrogenation
2
over the Cu@C catalyst. The Cu@C catalyst contains a large
amount of IF Cu sites which provides several advantages: (1)
IF Cu with the low coordination number can adsorb two C1
Figure 5 Reaction energy profiles at the DFT level for the hydrogenation
of CO (g) to CH CH OH at single-atom (SA) sites (a) and interfacial (IF)
2
3
2
sites (b) of the Cu(111) @ C. For each reaction step ½ H is added unless
2
species and later convert them to a C species via hydro-
2
otherwise indicated. H O molecules formed during the reaction steps are
2
genation-induced coupling reactions; (2) adjacent Cu atoms
of IF Cu can facilitate OH reduction reactions via the Cu–Cu
omitted. The reaction pathway towards CH CH OH is illustrated in red,
3
2
whereas the pathway towards C H is illustrated in black. Insets are opti-
2
6
mized geometries at the DFT level with the PBE functional for the inter-
bridge adsorption; (3) IF sites also inhibit C alkane pro-
2
mediate species for CO hydrogenation reactions. Color codes: O, red; H,
2
ducts, as the key intermediate (*CH–CH ) for the alkane
white; C, gray; Cu, orange (color online).
2
1
formation has η -adsorption configurations with OH on the
1
CHOH can be further hydrogenated to form η -*CH –CH -
inactive β-C.
2
2
OH and subsequently the final CH CH OH product, via the
3
2
1
1
η -*CH–CH OH (∆E=0.76 eV) and η -*CH –CHOH (∆E=
2.16 eV) intermediates respectively. Both paths have total
2
2
3
.5 Atomic-level understanding of the reactivity of
−
Cu@C
endothermicities of over 1 eV, which suggests that the SA-
Cu sites may not significantly contribute to the production of
ethanol over Cu@C. In addition, we also explore the reaction
As demonstrated by many prior experimental and compu-
tational studies, the hydrogenation of CO using a Cu solid or
2
pathway to the ethane C product at the SA-Cu site. It is
nanocrystal can facilely lead to the formation of CH [42–
2
4

8
Li et al. Sci China Chem
4
4], whereas the formation of CH CH OH is rather rare.
3 2
There are several possible pathways for CO hydrogenation
2
reactions towards CH on Cu, including the formate and
4
reverse water gas shift (RWGS) pathways, both via an
*
H CO intermediate (from *H COOH→*H CO and
2 2 2
*
CHO→*H CO, respectively) [45]. The former is much
2
more accepted as the formation of CH via CO hydro-
4
2
genations is found experimentally to be much faster than the
conversion via CO hydrogenations over the Cu/ZnO cata-
lysts [46]. The latter pathway is suggested to be inhibited for
the catalytic reactions over pure Cu, as the *CHO, precursor
intermediate for *(H CO), tends to dissociate into *CO and
2
*
H [47]. The IF and SA sites are significantly different from
the surface sites of clean Cu crystals. For the low-coordinate
IF and SA sites, the RWGS pathways may present with the
formation of a comparatively stable *CHO intermediate,
while the formate pathway can be suppressed as the formate
pathway involves more than one Cu sites and bidentate ad-
sorption. The SA sites have no neighboring Cu, whereas SA
Cu sites of Cu@C have approximately half of maximally
allowed the number of surface Cu neighbors, which may
suppress the formate pathway, especially for the second C1
intermediate adsorbate. The question is whether the terminal
product *CO for pure Cu can further react to form more
reactive intermediates on Cu@C and hence extend the
RWGS pathway. To find this out, the activation barriers for
Figure 6 (a) Predicted reaction energy and activation barriers for the
dissociation reactions of CHO at the IF site of Cu@C and at the Cu(111)
site at the DFT level. (b) The optimized geometry for the *CHO, *CO+*H,
and the transition state (TS) for CHO dissociations. (c) Predicted reaction
energy and activation barriers for the coupling reaction of two CHO ad-
sorbates at different Cu sites of Cu@C and Cu crystals at the DFT level. (d)
The optimized geometry for the *(CHO) , *(OHC–CHO), and the transition
2
state (TS) for CHO-CHO coupling. The Cu site at the Cu(111) surface, two
interfacial sites (IF and IF′) of Cu@C nanobelts, and a single-atom (SA)
site of Cu@C nanobelts are considered. Color codes: O, red; H, white; C,
gray; Cu, orange (color online).
riers in the range of 0.3–0.4 eV, which are much lower than
the rating limiting barrier heights along the formate or
RWGS pathways towards the formation of CH (~1.5 eV)
4
*
CHO dissociation into *CO and *H were predicted for
[19,48]. Thus, the critical CHO-CHO coupling elementary
step in the proposed reaction mechanism for the formation of
Cu@C is likely viable.
Cu(111) and the IF site of Cu@C. The results readily show
that the CHO dissociation barrier is considerably higher for
the IF reaction than that for the Cu(111) reaction (0.39 eV vs.
Last, we need to address whether the CO hydrogenation at
2
0
.22 eV). In addition, the dissociation at the IF sites of
Cu@C exhibits a lower exothermicity (−0.33 eV vs.
0.66 eV). These predictions suggest that, in contrast to the
the low-coordinate Cu sites produces CH OH. The reaction
3
energy profile of the CO hydrogenation towards CH OH at
2
3
−
the IF site of Cu@C is given in Figure S8. Despite that the
reaction pathway is exothermic for all elementary steps, the
Cu(111) reaction, *CHO could be a reaction intermediate at
the IF sites of Cu@C which may lead to the key *(OHC-
CHO) species in our proposed mechanism, as well as the
further hydrogenation, in contrast to the pure Cu cases. The
comparatively high stability of *CHO on Cu@C is also
suggested by the predicted CHO binding energies, as given
in Table S3. The binding of CHO at the low-coordinate sites
of Cu@C is found to be much stronger than the binding of
CHO on Cu(111). The CHO binding is strongest at tshe IF
site (among the investigated sites), at which the binding
energies are evaluated to be 2.34 and 1.60 eV for *CHO and
reaction is most likely inhibited, as the *H CO intermediate,
2
which is required by both formate and RWGS pathways, is
not bound to Cu. We suggest that H CO can only be adsorbed
2
at multiple-Cu sites. Thus, even if H CO is produced at low-
2
coordinate Cu sites of Cu@C, they could migrate to the Cu
nanoparticle surface and become an intermediate for the
CH OH formation on the Cu nanoparticle surface, but such
3
contributions are expected to be small. Experimental results
show that CH OH is also significantly produced from the
3
catalytic reaction over Cu@C. From TEM images (Figure 1
(d)), we clearly see defined Cu(111) crystal facets of the
*
(CHO) , versus the values of 1.15 and 1.10 eV for the
2
corresponding species for Cu(111).
embedded Cu nanoparticles, which produces CH OH as the
3
To further show that the stabilized *CHO on Cu@C can
main product. The CH OH could result from the formate
3
lead to the formation of C products, we calculated the ac-
pathways at these Cu(111) sites.
2
tivation barriers for the CHO-CHO coupling reactions at IF
sites, IF' (a second interfacial site with different coordina-
tions from IF), and SA of the nanobelt structures (Figure 6).
It is shown that the CHO-CHO coupling reactions at the IF
and SA sites only require surmounting low activation bar-
4 Conclusions
We report a novel Cu-based catalyst, that is, Cu nanoparticles

Li et al. Sci China Chem
9
Z, Luo Z, Chen J, Cui Y, Ozin GA, Gong J. Angew Chem Int Ed,
018, 57: 15415–15419
Watari R, Kayaki Y. Asian J Org Chem, 2018, 7: 2005–2014
Behrens M, Studt F, Kasatkin I, Kühl S, Hävecker M, Abild-Pedersen
F, Zander S, Girgsdies F, Kurr P, Kniep BL, Tovar M, Fischer RW,
Nørskov JK, Schlögl R. Science, 2012, 336: 893–897
embedded in the carbon support with abundant defect sites,
2
for CO hydrogenation. We demonstrate that the carbon
2
1
1
7
8
defects can act as anchor sites to stabilize copper species
with low coordination numbers, which assists C–C bond
coupling to form ethanol. With the decrease of the interfacial
Cu species under the reductive conditions, the major product
can change ethanol to methanol. This article provides deep
insights into the contribution of interfacial site Cu on based-
1
2
2
9
0
1
Kattel S, Ramírez PJ, Chen JG, Rodriguez JA, Liu P. Science, 2017,
355: 1296–1299
Arena F, Italiano G, Barbera K, Bordiga S, Bonura G, Spadaro L,
Frusteri F. Appl Catal A-Gen, 2008, 350: 16–23
Song Y, Peng R, Hensley DK, Bonnesen PV, Liang L, Wu Z, Meyer
Iii HM, Chi M, Ma C, Sumpter BG, Rondinone AJ. Chemistry Select,
Cu catalysts to the reaction pathways of CO hydrogenation
2
and is expected to develop a feasible strategy to enable
highly efficient CO conversion toward C or C products.
2
016, 1: 6055–6061
22 Margraf JT, Strauss V, Guldi DM, Clark T. J Phys Chem B, 2015, 119:
258–7265
2
1
2
7
2
2
2
3
4
5
Kresse G, Furthmüller J. Phys Rev B, 1996, 54: 11169–11186
Kresse G, Joubert D. Phys Rev B, 1999, 59: 1758–1775
Perdew JP, Burke K, Ernzerhof M. Phys Rev Lett, 1996, 77: 3865–
Acknowledgements This work was supported by the National Natural
Science Foundation of China (21773109, 91845104, U1930402), the Sci-
entific and Technological Innovation Foundation of Shunde Graduate
School, University of Science and Technology Beijing (BK19BE024).
3868
2
2
6
7
Monkhorst HJ, Pack JD. Phys Rev B, 1976, 13: 5188–5192
Henkelman G, Uberuaga BP, Jónsson H. J Chem Phys, 2000, 113:
Conflict of interest The authors declare that they have no conflict of
interest.
9901–9904
28
Hansen PL, Wagner JB, Helveg S, Rostrup-Nielsen JR, Clausen BS,
Topsøe H. Science, 2002, 295: 2053–2055
Gong X, Guo Z, Wang Z. Energy Fuels, 2009, 23: 4547–4552
Li R, Mansukhani ND, Guiney LM, Ji Z, Zhao Y, Chang CH, French
CT, Miller JF, Hersam MC, Nel AE, Xia T. ACS Nano, 2016, 10:
Supporting information The supporting information is available online at
http://chem.scichina.com and http://link.springer.com/journal/11426. The
supporting materials are published as submitted, without typesetting or
editing. The responsibility for scientific accuracy and content remains en-
tirely with the authors.
2
3
9
0
10966–10980
3
1
Fulmer GR, Miller AJM, Sherden NH, Gottlieb HE, Nudelman A,
Stoltz BM, Bercaw JE, Goldberg KI. Organometallics, 2010, 29:
2176–2179
1
2
3
4
5
6
7
8
Wang D, Bi Q, Yin G, Zhao W, Huang F, Xie X, Jiang M. Chem
Commun, 2016, 52: 14226–14229
Goeppert A, Czaun M, Jones JP, Surya Prakash GK, Olah GA. Chem
Soc Rev, 2014, 43: 7995–8048
Bi QY, Lin JD, Liu YM, He HY, Huang FQ, Cao Y. Angew Chem Int
Ed, 2016, 55: 11849–11853
Wang W, Wang S, Ma X, Gong J. Chem Soc Rev, 2011, 40: 3703–
32 Angamuthu R, Byers P, Lutz M, Spek AL, Bouwman E. Science,
2010, 327: 313–315
33 Lemus-Yegres LJ, Such-Basáñez I, Román-Martínez MC, de Lecea
CSM. Appl Catal A-Gen, 2007, 331: 26–33
34 Montes-Morán MA, Suárez D, Menéndez JA, Fuente E. Carbon,
2004, 42: 1219–1225
35 Lim DH, Jo JH, Shin DY, Wilcox J, Ham HC, Nam SW. Nanoscale,
2014, 6: 5087–5092
36 Hadjiivanov K, Venkov T, Knözinger H. Catal Lett, 2001, 75: 55–59
37 Hadjiivanov K, Tsoncheva T, Dimitrov M, Minchev C, Knözinger H.
Appl Catal A-Gen, 2003, 241: 331–340
38 Wang F, He S, Chen H, Wang B, Zheng L, Wei M, Evans DG, Duan
X. J Am Chem Soc, 2016, 138: 6298–6305
39 Wang L, Wang L, Zhang J, Liu X, Wang H, Zhang W, Yang Q, Ma J,
Dong X, Yoo SJ, Kim JG, Meng X, Xiao FS. Angew Chem Int Ed,
2018, 57: 6104–6108
3
727
Pang SH, Schoenbaum CA, Schwartz DK, Medlin JW. Nat Commun,
013, 4: 2448
Sun X, Wang P, Shao Z, Cao X, Hu P. Sci China Chem, 2019, 62:
686–1697
2
1
Hu C, Zhang L, Li L, Zhu W, Deng W, Dong H, Zhao ZJ, Gong J. Sci
China Chem, 2019, 62: 1030–1036
Graciani J, Mudiyanselage K, Xu F, Baber AE, Evans J, Senanayake
SD, Stacchiola DJ, Liu P, Hrbek J, Fernández Sanz J, Rodriguez JA.
Science, 2014, 345: 546–550
9
0
Sun X, Zhu Q, Kang X, Liu H, Qian Q, Zhang Z, Han B. Angew Chem
Int Ed, 2016, 55: 6771–6775
Wang G, Luo R, Yang C, Song J, Xiong C, Tian H, Zhao ZJ, Mu R,
Gong J. Sci China Chem, 2019, 62: 1710–1719
40 Chen M, Kumar D, Yi CW, Goodman DW. Science, 2005, 310: 291–
293
41 Weigel J, Koeppel RA, Baiker A, Wokaun A. Langmuir, 1996, 12:
5319–5329
1
1
1
He Z, Qian Q, Ma J, Meng Q, Zhou H, Song J, Liu Z, Han B. Angew
Chem Int Ed, 2016, 55: 737–741
Bai S, Shao Q, Wang P, Dai Q, Wang X, Huang X. J Am Chem Soc,
42 Liu W, Wang D, Ren J. Appl Surf Sci, 2020, 505: 144528
43 Cheng D, Negreiros FR, Aprà E, Fortunelli A. ChemSusChem, 2013,
6: 944–965
1
2
2
017, 139: 6827–6830
44 Sakong S, Gross A. J Phys Chem A, 2007, 111: 8814–8822
45 Li MMJ, Tsang SCE. Catal Sci Technol, 2018, 8: 3450–3464
46 Chinchen GC, Denny PJ, Parker DG, Spencer MS, Whan DA. Appl
Catal, 1987, 30: 333–338
1
1
3
4
Choi YM, Liu P. J Am Chem Soc, 2009, 131: 13054–13061
Gong K, Du F, Xia Z, Durstock M, Dai L. Science, 2009, 323: 760–
764
1
5
6
Chen A, Yu X, Zhou Y, Miao S, Li Y, Kuld S, Sehested J, Liu J, Aoki
T, Hong S, Camellone MF, Fabris S, Ning J, Jin C, Yang C, Nefedov
A, Wöll C, Wang Y, Shen W. Nat Catal, 2019, 2: 334–341
47 Yang Y, Evans J, Rodriguez JA, White MG, Liu P. Phys Chem Chem
Phys, 2010, 12: 9909–9917
48 Tameh MS, Dearden AK, Huang C. J Phys Chem C, 2018, 122:
17942–17953
1
Chang X, Wang T, Zhao ZJ, Yang P, Greeley J, Mu R, Zhang G, Gong