nature catalysis

Article https://doi.org/10.1038/s41929-023-01002-6

Experimental evidence of distinct sites for

CO2-to-CO and CO conversion on Cu in the
electrochemical CO2 reduction reaction

Received: 25 October 2022 Wenqiang Gao1,2,3, Yifei Xu1,2,3, Linke Fu , Xiaoxia Chang1,2 & Bingjun Xu
1,2 1,2

Accepted: 12 July 2023

Published online: 3 August 2023 The electrochemical CO2 reduction reaction on Cu is widely believed to
occur via two consecutive and orthogonal reaction steps, that is, CO2-to-CO
Check for updates
and CO-to-C2+, on the same sites. Here we provide compelling experimental
evidence that challenges these long-held assumptions. We show that the
presence of CO2 promotes the electrochemical CO reduction reaction, and
there are at least two distinct types of Cu sites, with one (CuCO2) more active
in the CO2-to-CO conversion and the other (CuCO) favouring the further
reduction of CO to C2+ products. CO adsorbed on CuCO is at least six times
more active towards the formation of C2+ products than that on CuCO2.
Isotopic labelling experiments on Cu(111) and Cu(100) surfaces indicate
that CuCO2 and CuCO probably correspond to Cu(111)-like and defect sites,
respectively. These insights highlight the possibility of selectivity control in
the CO2 reduction reaction.

Electrochemical CO2 reduction reaction (CO2RR) represents a dissolved CO in the electrolyte. These two aspects are related. If the two
promising strategy to close the carbon circle by converting CO2 reactions take place on the same sites, then the source of CO would be
into valuable C2+ hydrocarbons and oxygenates powered by renew- irrelevant, and site competition between the two reactions is expected.
able electricity1–4. Despite extensive research, Cu-based materials In this mechanistic scenario, it is advantageous to spatially separate the
remain the only category of catalysts capable of selectively reduc- two reactions so that they can be individually optimized. This strategy
ing CO2 to C2+ products5–7. It is well established that CO2 is first con- could be implemented on the macroscopic level, that is, conducting
verted to CO on Cu before its further reduction, as adsorbed CO is a the two reactions in separate reactors in a tandem process13,14, or on
commonly reported reaction intermediate and a minor product8–10. the microscopic scale, that is, developing Cu-based tandem catalysts,
The CO2-to-CO and the CO-to-C2+ conversions are generally viewed as in which a CO-producing metal, for example, Au or Ag, is combined
two sequential reaction steps with distinct requirements for catalytic with Cu via a variety of synthetic strategies15–18. Alternatively, if the two
sites. As a result, the electrochemical CO reduction reaction (CORR) reactions occur on distinct sites on Cu, or certain sites favour a specific
has frequently been employed as a model reaction to understand reaction, then adsorbed CO produced from CO2 needs to migrate to the
reaction pathways in the CO2RR leading to C2+ products11,12. However, CORR sites either by surface diffusion or by desorption/re-adsorption.
experimental evidence supporting the orthogonality of the two cas- In this case, the facility with which adsorbed CO travels among differ-
cading reactions is scarce, and thus this commonly held assumption ent sites could limit the formation rate of C2+ products. Meanwhile,
deserves scrutiny. dissolved CO in the electrolyte could adsorb directly onto sites favour-
Understanding the connection between CO2-to-CO and CO-to-C2+ ing the CORR. Wang et al. reported enhanced ethylene formation rate
steps on Cu could provide key insights into the design of CO2RR cata- with co-electrolysis of a mixture of CO2 and CO compared with the
lysts and processes. Two aspects need to be clarified, that is, whether pure CO2RR or CORR, and proposed that CO2-to-CO and CO-to-C2+
the two reactions occur on the same sites on Cu, and whether the conversions occurred on distinct Cu sites19. While density functional
adsorbed CO formed via CO2 reduction is equivalent to that from theory calculations suggested that coupling between two CO adsorbed

College of Chemistry and Molecular Engineering, Peking University, Beijing, China. 2Beijing National Laboratory for Molecular Sciences, Beijing, China.
1

These authors contributed equally: Wenqiang Gao, Yifei Xu.

3
e-mail: [email protected]

Nature Catalysis | Volume 6 | October 2023 | 885–894 885

Article https://doi.org/10.1038/s41929-023-01002-6

a b
Ethylene Ethanol Den-Cu/Au film in –1.5 VSHE
60

Production rate (mol s–1 cm–2) × 10–10

Acetate n-Propanol 0.1 M KHCO3
C2+ total
CO
45

Absorbance (a.u.)
2,053

30

15
CO2

0
CO CO/CO2 CO/CO2 CO2 2,120 2,080 2,040 2,000 1,960
1 atm (4:1) (1:1) 1 atm
Wavenumber (cm–1)

c d
Ethylene Ethanol Den-Cu/Au film in –1.5 VSHE
Production rate (mol s–1 cm–2) × 10–10

60
Acetate n-Propanol 0.1 M KHCO3
C2+ total 0.8 CO + 0.2 CO2

45

Absorbance (a.u.)
2,051

30 0.8 CO + 0.2 Ar

15

0
CO/Ar CO/Ar/CO2 CO/CO2 2,120 2,080 2,040 2,000 1,960
(4:1) (8:1:1) (4:1)
Wavenumber (cm–1)

Fig. 1 | Co-electrolysis of CO/CO2 and in situ spectroscopic characterization. formation rates for ethylene, ethanol, acetate, n-propanol and total C2+ in the
a, Product formation rates of ethylene, ethanol, acetate, n-propanol and total co-electrolysis of CO/CO2 mixtures with varying compositions in 0.1 M KHCO3
C2+ in the co-electrolysis of CO/CO2 mixtures with varying compositions in on Den-Cu at −1.5 VSHE. d, In situ SEIRA spectra of adsorbed CO on Den-Cu at
0.1 M KHCO3 on Den-Cu at −1.5 VSHE in an H-type cell. b, In situ SEIRA spectra −1.5 VSHE in 0.1 M KHCO3 with the atmosphere gradually switching from CO/Ar
of adsorbed CO on Den-Cu at −1.5 VSHE in 0.1 M KHCO3 with the atmosphere (0.8 atm/0.2 atm) to CO/CO2 (0.8 atm/0.2 atm). Error bars represent standard
gradually switching from CO2 (blue trace) to CO (green trace). c, Product deviation (s.d.) calculated from three independent measurements.

on distinct types of sites was responsible for the enhanced ethylene Results
production, experimental evidence was needed to confirm this pre- Co-electrolysis of CO2 and CO
diction. This mechanistic scenario favours atomically mixed sites for Co-electrolysis of mixtures of CO2 and CO with varying compositions
CO2-to-CO and CO-to-C2+ conversions on the catalyst to facilitate rapid were conducted on dendritic Cu (Den-Cu) in 0.1 M KHCO3 at −1.5 V
diffusion of adsorbed CO produced from CO2RR to sites responsible to determine the effect of feed composition on the rate and product
for its further reduction. distribution (all potentials in this work are referenced to the standard
In this Article, we demonstrated the promotional effect of CO2 hydrogen electrode unless noted otherwise). Production rate of the
on the CORR and the presence of at least two distinct types of sites on overall C2+ products increased by ~90% when the CO2 fraction in the feed
Cu via a combination of co-electrolysis, spectroscopic and isotopic was raised from 0% to 20%, before declining with further increases in
labelling investigations. 13CO/12CO2 mixtures were employed as feed the CO2 fraction (Fig. 1a, red trace). Individual C2+ products, including
in the reactivity study to determine the carbon source of products, ethylene, ethanol, acetate and n-propanol, follow similar trends, sug-
while relative surface coverages of the adsorbed 12CO and 13CO were gesting a shared rate-determining step20,21. Interestingly, the rates are
probed by in situ surface-enhanced infra-red absorption spectroscopy comparable in the CORR and the CO2RR, that is, electrolysis with CO and
(SEIRAS). Isotopic distributions of products from the co-electrolysis CO2 as the sole feed, respectively. Plots of Faradaic efficiencies (FEs) of
of 13CO/12CO2 mixtures on dendritic Cu deviated substantially from products in the CORR and the CO2RR are included in Supplementary
the binomial distribution. By comparing the isotopic distribution of Information (Supplementary Fig. 1), while FEs of the co-electrolysis
adsorbed CO and products, we demonstrated that there were at least experiments cannot be determined because the fractions of prod-
two types of sites with distinct catalytic properties on dendritic Cu, ucts originating from the reduction of CO2 and CO, respectively, are
that is, one is more active for the CO2-to-CO conversion (CuCO2) whereas unknown. In addition, the amount of CO produced from the reduction
the other favours the further reduction of CO to C2+ products (CuCO). of CO2 in the co-electrolysis experiments is also uncertain because CO
A two-site model was proposed to quantitatively describe the reactivity is part of the feed. The electrolyte pH in CO and CO2 saturated 0.1 M
of the sites, which shows that CO adsorbed on CuCO is at least six times KHCO3 were determined to be 8.3 and 6.8, respectively. To rule out
more active towards the formation of C2+ products than that on CuCO2. the possibility that the electrolyte alkalinity was the main cause of
Finally, co-electrolysis results on Cu(111) and Cu(100) surfaces indicate the observed activity trend, similar experiments were conducted in
that CuCO2 sites are enriched on Cu(111) and CuCO sites probably involve a 0.1 M potassium phosphate electrolyte, in which the electrolyte pH
defect sites on Cu surfaces. (6.9) remained unchanged regardless of the CO/CO2 ratio in the feed.

Nature Catalysis | Volume 6 | October 2023 | 885–894 886

Article https://doi.org/10.1038/s41929-023-01002-6

Consistent with results obtained in KHCO3, C2+ production rate in the a b

phosphate buffer was higher with a CO/CO2 mixture (0.5 atm/0.5 atm) 30 amu 29 amu 28 amu Den-Cu/Au film in 0.8 13CO + 0.2 CO2
12 13 –1.5 VSHE 0.1 M KHCO3
Corrected 29 amu for C C
feed than both the CORR and the CO2RR (Supplementary Figs. 2 and 3). 12 12 2,003
Corrected 28 amu for C C
A similar trend, that is, co-electrolysis of CO2/CO mixtures exhibiting

Relative abundance
13

Absorbance (a.u.)
CO/CO2 = 4:1 m
I 30 = I30 13
higher rates than the electrolysis of either molecule, was reported by Ethylene CO

Strasser and co-workers19. Complementary SEIRAS was conducted at m

Im = I29 – 0.68 × I 30
the reaction potential (−1.5 V) in 0.1 M KHCO3. Potential mass transport 29

2,050
limitation of CO2 and CO to the catalyst surface at this condition was
I 28 = I28– 0.68 × (I 29 + I 30 )
m m m

investigated by monitoring the band of the linearly adsorbed CO (COL 12

CO

at ~2,060 cm−1) (ref. 22) at various stirring rates in a custom-designed

spectro-electrochemical cell. Integrated area of the COL band did not
change at stirring rates >600 rpm (Supplementary Fig. 4), suggesting 2,120 2,080 2,040 2,000 1,960 1,920
Wavenumber (cm–1)
negligible mass transport limitation of CO2 and CO in spectroscopic
experiments performed with a stirring rate of 750 rpm. Thus, spec- 13 12
Fig. 2 | Co-electrolysis of CO/ CO2 (4:1) on Den-Cu. a, Distribution of
troscopic results could be used to analyse the reactivity data col- isotopologues of ethylene produced in the co-electrolysis of 13CO/12CO2
lected at the same conditions. When the atmosphere was switched (0.8 atm/0.2 atm) on Den-Cu at −1.5 V in 0.1 M KHCO3. Procedure of the
from CO2 to CO, the COL band located at 2,053 cm−1 increased mono- mass fragmentation analysis is included in Methods. b, In situ SEIRA spectra
of adsorbed CO on Den-Cu at −1.5 V in 13CO/12CO2 (0.8 atm/0.2 atm) and
tonically by only 7%, probably because the CO coverage produced via
0.1 M KHCO3.
CO2 reduction at −1.5 V was slightly lower than that by CO supplied
externally (Fig. 1b). The relatively minor difference between the COL
band in CO, CO/CO2 and CO2 atmospheres suggests that the enhanced
formation of C2+ products in CO/CO2 could not be accounted for by liquid phase products were analysed with proton nuclear magnetic
variations in the COL coverage in different feed compositions. COL resonance (1H-NMR) based on the characteristic peak splitting due to
has been proved to be the only active adsorbed CO species in the reac- the presence of 13C directly attached to H or as the adjacent carbon.
tion on Cu (refs. 23,24). The insensitivity of rates to the COL coverage Vast majority of carbons (>95%) in the produced ethanol, acetate and
is consistent with the close zeroth order of CO in the CORR on Cu at n-propanol are 13C (Supplementary Fig. 9). A singlet was observed for
CO partial pressures close to 1 atm (refs. 20,21,25). It should be noted the formate (Supplementary Fig. 10), which was expected as it could
that the relative insensitivity of the COL coverage with respect to the only be formed via the reduction of 12CO2. Further, the MS signals for
13
composition of the CO/CO2 mixture could be attributed to the similar CO (29 amu) and 12CO2 (44 amu) were consistent with the feed com-
rates of COL production from the reduction of CO2 and the adsorption position (Supplementary Fig. 11), confirming the intended isotopic
of CO in the feed at the potential −1.5 V. When a similar experiment composition of the feed. Collectively, these results suggest that the
to that in Fig. 1b, that is, the atmosphere was switched from CO2 to vast majority of carbon in the products from the co-electrolysis came
CO, was conducted at −0.8 V, the COL band was absent in 1 atm CO2 from 13CO, and the contribution of CO2 to the enhanced production rate
and appeared upon introduction of CO in the feed (Supplementary observed in Fig. 1c was not via contributing an additional source of CO
Fig. 5). This is because −0.8 V is too positive a potential for the CO2-to-CO (in the feed of 0.8 atm 13CO/0.2 atm 12CO2). While the mechanism for
conversion to occur on Cu (ref. 6), so no COL could be produced in the CO2’s promotional effect deserves further investigations, it could be
CO2 atmosphere. inferred that presence of CO2 in the electrochemical interface changes
Co-electrolysis in feeds with a fixed CO partial pressure of 0.8 atm the microenvironment for the CORR.
and varying CO2 partial pressures (with balancing Ar) suggests that In situ SEIRAS results suggest that not all adsorbed CO L pos-
CO2 has a promotional effect on the formation rates of C2+ products sess the same reactivity. SEIRA spectrum collected in 13CO/12CO2
(Fig. 1c). The overall C2+ production rate increased by 50% in the (0.8 atm/0.2 atm) and 0.1 M KHCO3 at −1.5 V showed a COL band centred
CO/CO2/Ar feed (0.8 atm/0.1 atm/0.1 atm) compared with that with a at 2,003 cm−1 (Fig. 2b), which could be attributed to linearly adsorbed
13
pure CO feed, which was further increased by ~90% when the partial COL(refs. 21,26). There was a clear shoulder at the high wavenumber
pressure of CO2 in the feed increased to 0.2 atm (0.8 atm CO/ 0.2 atm side (~2,050 cm−1) attributable to 12COL. Peak deconvolution showed
CO2). The integrated area of the COL band increased only marginally that the 12COL band accounted for ~14% of the entire COL band. The
(by ~5%) when the feed was switched from CO/Ar (0.8 atm/0.2 atm) mismatch between the fraction of adsorbed 12CO (~14%) and the frac-
to CO/CO2 (0.8 atm/0.2 atm), confirming that the variations in the tion of 12C in the detected co-electrolysis products (<5%) indicates that
COL coverage were unlikely to be the main cause of the enhanced adsorbed 12CO is less active in the CORR than adsorbed 13CO. Since 12CO
C2+ production rates (Fig. 1d). and 13CO dissolved in the electrolyte are probably chemically equivalent
under reaction conditions in this work (Supplementary Note 1), it is
Promotional effect of CO2 on the CORR hypothesized that they occupy distinct surface sites, probably due to
Co-electrolysis of 13CO/12CO2 (0.8 atm/0.2 atm) further supports the the distinct pathways through which they are introduced to the surface.
hypothesis that CO2 has a promotional effect on the 13CORR. The dis- 12
CO was introduced to the surface via the reduction of 12CO2, and thus
tribution of isotopologues of ethylene produced in the co-electrolysis probably initially adsorbed on surface sites favourable towards the
experiment was plotted by normalizing it to the intensity of 30 amu CO2RR before further reaction or diffusion to other sites. In contrast,
13
signal with mass spectrometry (MS), corresponding to the parent CO can explore most of the surface sites, and thus statistically is more
ion of 13CH213CH2. Procedure to quantify isotopologues based on MS likely to encounter surface sites that are more active for the CORR.
signals via mass fragmentation analysis is included in Supplemen- These reasonings lead to the hypothesis that there are at least two
tary Information (Supplementary Fig. 6) and Methods. Strikingly, all types of sites on Cu. One type is more favourable for the CO2RR to CO,
produced ethylene was 13CH213CH2, and the corrected MS signals for and the other is better at the further conversion of CO. This hypothesis
12
CH213CH2 and 12CH212CH2 were comparable to the noise level (Fig. 2a). will be examined further below.
Distributions of isotopologues of other products were also consistent
with that of ethylene. Approximately 96% of detected methane was Co-electrolysis of 13CO/12CO2 mixtures
13
CH4 (Supplementary Figs. 7, 8 and 15), and the MS signal for 12CH4 Co-electrolysis of 13CO/12CO2 mixtures with 13CO molar fractions of 50%
was barely above the noise level. Distributions of isotopologues of and 20% further confirm that the 13C is preferentially incorporated

Nature Catalysis | Volume 6 | October 2023 | 885–894 887

Article https://doi.org/10.1038/s41929-023-01002-6

a 3
b further supporting the hypothesis that 13COL is the more active in the
Den-Cu/Au film in 0.5 13CO + 0.5 CO2
Observed distribution Ethylene
–1.5 VSHE 0.1 M KHCO3
further conversion of CO.
Binomial distribution (MS)
Binomial distribution (SEIRAS) 2,052
Analysis of the isotopic composition of acetate and ethanol
2,005
13
in the liquid phase with 1H-NMR in the co-electrolysis of 13CO/12CO2

Absorbance (a.u.)
CO:CO2 = 1:1
2
Product ratio

12
CO
13
CO (0.5 atm/0.5 atm) led to additional mechanistic information. For ace-
tate, there is only one type of proton active in the 1H-NMR (H in the
methyl group), so it is relatively straightforward to analyse (Fig. 4a). The
1
peak at 1.79 ppm is attributed to 12CH312COO− (purple singlet), while the
green doublet could be assigned to 12CH313COO−, with the peaking split-
ting caused by the adjacent 13C. The red doublet at 1.66 and 1.91 ppm
0
12
C12C 12
C13C 13
C13C 2,120 2,080 2,040 2,000 1,960 can be attributed to 13CH312COO− due to the peak splitting caused by
Wavenumber (cm–1) the 13C of the methyl group. The doublets around 1.65 and 1.90 ppm are
attributed to 13CH313COO− caused by the presence of two 13C atoms. The
c d overall 12C/13C ratio (0.36) of the acetate is similar to that of ethylene
6 Den-Cu/Au film in 0.2 13CO + 0.8 CO2
Observed distribution Ethylene
Binomial distribution (MS)
–1.5 VSHE 0.1 M KHCO3 (0.29). Importantly, the ratio between 12CH313COO− and 13CH312COO- is
5 2,055
Binomial distribution (SEIRAS) close to unity (Fig. 4d), suggesting that there is no difference between
13
Absorbance (a.u.)

CO:CO2 = 1:4 12
4
12
CO
CO or 13CO in becoming the methyl or the carboxylate carbon.
Product ratio

A similar observation was reported by Ager and co-workers27. Peak

3
deconvolution of ethanol is more involved because there are four
2
2,007 isotopologues with two types of protons bonded to C, that is,
12
13
CO CH312CH2OH, 13CH312CH2OH, 12CH313CH2OH and 13CH313CH2OH. Inte-
1 grated peak areas of isotopologues of ethanol show that the ratio
0
among 12C-12C/12C-13C/13C-13C isotopologues of ethanol (0.25/0.63/1.00)
12
C12C 12
C13C 13
C13C 2,120 2,080 2,040 2,000 1,960 is similar to that of both ethylene and acetate (Fig. 4c). Similar to the
Wavenumber (cm–1) case of acetate, comparable amounts of 13CH312CH2OH and 12CH313CH2OH
Fig. 3 | Distribution of ethylene isotopologues in the co-electrolysis of formed were detected, indicating that 12CO and 13CO were equally likely
13
CO/12CO2. a,c, Observed distribution (blue bars) and expected binomial to become the methyl and methylene carbons. In addition, the 12C/13C
distributions of isotopologues of ethylene produced based on the 12CH212CH2/13 ratio in methane is also consistent with that of C2+ products (Supple-
CH213CH2 ratio determined by MS (green bars), and 12COL/13COL ratio determined mentary Fig. 15), further confirming the consistency of the isotopic
by SEIRAS (grey bars) in the co-electrolysis of 13CO/12CO2 (0.5 atm/0.5 atm) (a) composition among products. It is noted that the doublet at 1.08 and
and 13CO/12CO2 (0.2 atm/0.8 atm) (c) on Den-Cu at −1.5 V in 0.1 M KHCO3. Intensity
1.09 ppm (labelled with stars in Fig. 4b) is due to the contamination of
of the 13CH213CH2 signal is set to unity in the plots to allow for direct comparison.
a small amount of isopropanol introduced in the catalyst preparation
b,d, In situ SEIRA spectra of adsorbed CO on Den-Cu at −1.5 V in 13CO/12CO2
procedure, which was confirmed by a control experiment (Supplemen-
(0.5 atm/0.5 atm) (b) and 3CO/12CO2 (0.2 atm/0.8 atm) (d) on Den-Cu at −1.5 V
tary Fig. 21). Collectively, co-electrolysis of 13CO/12CO2 mixtures indicate
in 0.1 M KHCO3. Error bars represent s.d. calculated from three independent
measurements, and the mean value (centre point) is presented together with that the electrochemical reduction of 13CO is consistently favoured
s.d. (bar height). compared with the conversion of 12CO (produced via the reduction
of 12CO2) in the formation of methane and C2+ products on Den-Cu,
compared with the relative coverages of 12COL and 13COL determined
spectroscopically. It could be inferred that 12COL and 13COL are probably
in C2+ products on Den-Cu. In a 13CO/12CO2 (0.5 atm/0.5 atm) feed adsorbed on surface sites with distinct CORR activities, probably due
(Supplementary Fig. 12), the ratio of 13CH213CH2/12CH213CH2/12CH212C to their different origins.
H2 is 1/0.48 (±0.08)/0.12 (±0.05) (Fig. 3a and Supplementary Fig. 13). Competition for the access to surface Cu sites between the
Clearly, there were substantially more 13C incorporated in ethylene than CO2-to-CO conversion and the CORR is unlikely to be a major factor
12
C. SEIRA spectrum collected at the same condition showed that the affecting the observed reactivity trend. Our recent ambient pressure
12
COL and 13COL bands were comparable in size (Fig. 3b), supporting the and high pressure SEIRAS investigations on Cu surfaces suggest that CO
hypothesis that 13COL adsorbed on sites that were more active in the coverage at conditions close to CO(2)RR is at or below 0.05 monolayer
CORR. 75% of methane detected was 13CH4, which was similar to that (ML) (ref. 28). This claim is further supported by the close-to-zero
of ethylene, with a 13C fraction of 79% (Supplementary Figs. 14 and 15). standard CO adsorption enthalpy on Den-Cu and OD Cu (ref. 29). Thus,
To further exclude the possibility that the electrolyte alkalinity was Cu surfaces are probably mostly devoid of specifically adsorbed reac-
the primary cause of the observed isotopic distribution of products, tion intermediates under CO(2)RR conditions (except for water), which
similar experiments were also conducted in a 0.1 M potassium phos- precludes site competition.
phate electrolyte (pH 6.9). A similar trend, that is, co-electrolysis of
the CO2/CO mixture (0.5 atm/0.5 atm) exhibiting a higher fraction Observed isotopic distribution versus binomial distribution
of 13C incorporated in ethylene and methane than 12C (Supplementary Isotopologue distributions of C2 products from the co-electrolysis of
13
Figs. 16–18), was observed. CO/12CO2 mixtures on Den-Cu deviate substantially from the binomial
Additional co-electrolysis of the 13CO/12CO2 (0.2 atm/0.8 atm) feed distribution. Since the isotopologue distributions of C2 products are
produced a ratio of 13CH213CH2/12CH213CH2/12CH212CH2 is 1/1.73 (±0.09) generally consistent (Fig. 4c), the following discussion uses ethylene
/1.24 (±0.12) (Fig. 3c and Supplementary Fig. 19), which was consistent as a representative example for further analysis. Binomial distribution
with the analysis above. 41% of methane detected was 13CH4, which was of isotopologues is expected when isotopologues of adsorbed CO
similar to that of ethylene, with a 13C fraction of 46% (Supplementary have the identical chemical properties, and all surface sites possess
Figs. 15 and 20). Peak deconvolution and integration of the SEIRA identical catalytic properties or adsorbed CO diffuses sufficiently fast
spectrum collected with the same feed and at the same condition sug- on the surface so that all sites possess the same probability of bond-
gested that 12COL and 13COL accounted for 77% and 23%, respectively, of ing to 12CO or 13CO. This is supported by the proposed two-site model
the entire COL band (Fig. 3d). Although the 12COL/13COL ratio is 3.3 times presented in the next section. There is extensive discussion in the
higher, the 12C/13C ratio is just above unity in the produced ethylene, literature on the possibility of sites with distinct catalytic properties

Nature Catalysis | Volume 6 | October 2023 | 885–894 888

Article https://doi.org/10.1038/s41929-023-01002-6

a b
13 13
CH3 COOH 13 13
CH3 CH2OH

12 12
CH3 COOH 12 12
CH3 CH2OH

Intensity (a.u.)

Intensity (a.u.)
13 12
CH3 COOH
13
13
CH3 CH2OH
12 CH3
12 13
CH3 COOH 12 13 in PrOH
CH3 CH2OH

1.95 1.90 1.85 1.80 1.75 1.70 1.65 1.20 1.15 1.10 1.05 1.00 0.95 0.90 0.85

Chemical shift (ppm) Chemical shift (ppm)

c 1.5
d 2.0
12 12 12 13 13 13 12 13 12 13
C C C C C C CH3 COOH CH3 CH2OH
13 12 13 12
CH3 COOH CH3 CH2OH

1.5
1.0
Product ratio

Product ratio
1.0

0.5
0.5

0 0
Ethylene Acetate Ethanol Acetate Ethanol

Fig. 4 | Isotopologue distribution of C2 products in the co-electrolysis of by the dashed line, to allow for direct comparison) (c), and acetate and ethanol
13
CO/12CO2. a,b, Deconvolution of 1H-NMR spectra of acetate (a) and ethanol (b) produced by co-electrolysis of 13CO/12CO2 (0.5 atm/0.5 atm) on Den-Cu at −1.5 V
produced in the co-electrolysis (0.5 atm 13CO/0.5 atm 12CO2) on Den-Cu at −1.5 V in 0.1 M KHCO3 (d). Error bars represent s.d. calculated from three independent
in 0.1 M KHCO3. Peaks corresponding to different isotopologues are marked. measurements, and the mean value (centre point) is presented together with the
c,d, Distribution of isotopologues of ethylene, acetate and ethanol produced s.d. (bar height).
(product ratios are normalized to the 13C–13C coupling product, as indicated

on Cu surfaces, including the hypothesized grain boundary and (sub) environments, even on single crystal surfaces, at CO(2)RR conditions are
surface oxides27. Neither the binomial distribution of ethylene iso- well documented30–32, so the presence of Cu sites with different catalytic
topologues based on the 12COL/13COL ratio determined by SEIRAS, activities is unsurprising. To the best of our knowledge, the impact of
nor that derived from the observed ratio of 12CH212CH2 and 13CH213CH2 the surface Cu site heterogeneity on the CO2RR has yet been rational-
is able to predict the amount of 12CH213CH2 produced (Fig. 3a,c). The ized on the basis of their distinct activities in the CO2-to-CO conver-
MS signal (29 amu) corresponding to 12CH213CH2 is consistently lower sion and the CORR. The exclusive incorporation of 13C in the products
than the value predicted by the binomial distribution. The discrepan- formed in the co-electrolysis of 13CO/12CO2 (0.8 atm/0.2 atm, Fig. 2a and
cies in the expected binomial distributions based on SEIRAS and MS Supplementary Fig. 8), combined with the observed 13COL/12COL ratio
data (Fig. 3a,c) stem from the different assumptions made in the data (86/14) with SEIRAS (Fig. 2b), suggests that 13CO is more active in the C-C
analysis, that is, the binomial distributions based on SEIRAS results coupling reaction. Importantly, the expected binomial distribution of
assume identical activity of 12COL and 13COL, while those based on MS isotopologues of ethylene deviates drastically from observed results
results assume all surface sites have identical catalytic properties. in all co-electrolysis experiments with different 13CO/12CO2 ratios (Sup-
Both assumptions will be proved untenable in the discussion below. plementary Fig. 24). It can be inferred that there are at least two types
We note that this observation is consistent with the results reported of Cu sites with distinct catalytic properties (Supplementary Note 2),
by Wang et al. at similar potentials (Supplementary Fig. 22), though which gives rise to a two-site model: one type of site is active for the
this type of analysis was not performed in this reference19. The similar CORR (referred to as CuCO), and the other favours the CO2-to-CO con-
amounts of 13CH312COO− and 12CH313COO−, as well as 13CH312CH2OH version but is relatively inert for the further conversion of CO (referred
and 12CH313CH2OH (Fig. 4d), produced in the co-electrolysis suggest to as CuCO2). CO2 is primarily converted on CuCO2, and the produced CO
that statistically there is no difference between 12COL and 13COL when initially adsorbs on CuCO2 due to proximity (Fig. 5a), which could be
participating the C–C coupling reaction, that is, their isotopic iden- either further reduced at CuCO2 or leave the site by diffusion or desorp-
tity does not affect the activity. The kinetic isotope effect of 12CO tion. The deviation in the distribution of the isotopologues from the
and 13CO is expected to be minor due to their similar reduced masses binomial distribution is evidence that adsorbed CO tends to desorb or
(Supplementary Note 1). This is also supported by our control experi- be converted on the same type of site as the one it initially occupies,
ments showing that 12CO/12CO2 and 13CO/12CO2 (0.5 atm/0.5 atm) lead regardless whether the adsorbed CO is formed via CO2 reduction or
to similar amounts of ethylene formed within experimental errors adsorbed from the electrolyte.
(Supplementary Fig. 23). Considering the modest calculated CO binding energy on Cu
(ref. 33), as well as the slightly positive CO adsorption enthalpy deter-
The two-site model mined at electrochemical conditions29, it appears that the CO diffu-
We propose a two-site model to rationalize the observed reactivity and sion would be fast on the molecular scale, and thus slow CO diffusion
spectroscopic results. Surface Cu sites with different coordination is unlikely the primary cause for the observed deviations from the

Nature Catalysis | Volume 6 | October 2023 | 885–894 889

Article https://doi.org/10.1038/s41929-023-01002-6

a
CO(aq)
Adsorption

CO2(aq) Reduction *CO

*CO
*CO *CO

Reduction

Den-Cu Slow mixing

C2+ products
CuCO2 CuCO

b log a c
100 6.0 10
13
CO:12CO2 = 1:1
13
CO:12CO2 = 1:4
80 5.0 8

60 4.0 6

log γ
y (%)

40 3.0 4

13 12
CO: CO2 = 1:1
20 2.0 2
13
CO:12CO2 = 1:4

0 1.0 0
5 10 15 20 25 30 1 2 3 4 5 6

x (%) log a

Fig. 5 | The two-site model and its predictions. a, Schematic the proposed and supplementary equation (26)). The dotted line means that x and y can only
two-site model on Cu. b, Relation between the fractional contribution of CuCO site approach but not reach 0. c, Log plot of the activity ratio between CO adsorbed
for 12COL ( x ) and 13COL ( y ) bands with the corresponding ratio ( a) between the on CuCO and CuCO2 ( γ ) versus a (calculated with equation (3) and supplementary
amount of ethylene produced on CuCO and CuCO2 (calculated with equation (2) equations (28) and (29)).

binomial distribution34. The observed non-binomial distributions of C2 12

C1 and 13C1 on the surface, a binomial distribution would entail the
isotopologues are probably caused by the substantial spatial segrega- ratio of 12C–12C/12C–13C/13C–13C to be x2/2xy/y2.
tions of CuCO2 and CuCO sites into different domains (Fig. 5a). Isotopic Under the simplified assumption that the C–C coupling only
labelling results suggest that the size of domains with enriched CuCO2 occurs between C1 species adsorbed on the same type of sites, each
and CuCO sites are substantially larger than the distance CO is able to type of site would lead to an independent binomial distribution, which
diffuse on the surface before either reaction or desorption (Supple- is denoted as D1 and D2 for isotopologues of ethylene produced on
mentary Note 2). It could be inferred that the extent of deviation of the CuCO2 and CuCO, respectively. The experimentally observed distribution
observed isotopologue distribution from the binomial distribution is of ethylene isotopologues (denoted as Do) is the sum of the two bino-
a qualitative measure of the degree of spatial segregation of CuCO2 or mial contributions of ethylene isotopologues from the two types of
CuCO sites. Unfortunately, neither SEIRAS nor Raman spectroscopy is sites. Assuming the ratio between the amount of ethylene produced
able to differentiate CuCO and CuCO2. The broad lineshapes of the vibra- on CuCO and CuCO2 is set to be a (to account for the relative density of
tional bands for COL limit the resolving power of these techniques. the two types of sites), Do could be expressed in equation (1):
In addition, the correlation between the CO adsorption enthalpy or
Do = a⋅D1 + D2 (1)
the CO(2)RR activity remains unverified, so that it is not unreason-
able for sites with similar CO adsorption energy, and in turn similar To derive an analytical expression of equation (1), we set the frac-
COL bands, to have distinct reactivities (Supplementary Note 3). In tions of 12COL and 13COL bands attributed to CuCO to x % and y %, respec-
this regard, co-electrolysis experiments with 12CO2/13CO as feed is a tively (Supplementary Table 1), and those on CuCO2 to (100 − x) %
more sensitive and reliable probe for different sites’ ability to mediate and (100 − y) % (detailed derivations included in Supplementary
specific reactions. Note 5). In the co-electrolysis with the feed composition of 13CO/12CO2
Additional mechanistic insights could be gained when the two-site (0.5 atm/0.5 atm), combining the measured 12COL/13COL ratio of 1:1 by
model is considered more quantitatively. Although the identity of the SEIRAS, equation (2) could be obtained on the basis of the binomial
rate-determining step in the formation of C2 products in the CORR distributions of D1 and D2:
remains a topic of discussion21,35, it must involve the coupling of two C1
2
species. Importantly, the distribution of isotopologues of C2 products [ax2 + (100 − x) ] ∶ [2axy + 2 (100 − x) (100 − y)]
does not depend on the kinetic models on which kinetic derivations (2)
2
are based (Supplementary Note 4). We first assume that C2 products ∶ [ay2 + (100 − y) ] = 0.12 ∶ 0.48 ∶ 1.00#
can only be formed by coupling of two C1 species adsorbed on the same
type of site, that is, no C–C coupling between one adsorbed C1 inter- Equation (2) shows that a, x and y are subject to the constraint of
mediate, for example, CO or COH, on CuCO and the other on CuCO2. two independent equations (Supplementary Note 5), so there is only
Coupling between one 12C1 and one 13C1 species adsorbed on the same one independent variable among the three. With the boundary condi-
type of site would lead to a binomial distribution of isotopologues of tions of 0 ≤ x ≤ 100 and 0 ≤ y ≤ 100, the correlation among a, x and
the C2 products. For example, there are x % and y %, respectively, of y in the co-electrolysis of 13CO/12CO2 (0.5 atm/0.5 atm) is plotted in

Nature Catalysis | Volume 6 | October 2023 | 885–894 890

Article https://doi.org/10.1038/s41929-023-01002-6

a b
15 3
Observed distribution Ethylene Observed distribution Ethylene
Binomial distribution (MS) Cu (111) Binomial distribution (MS) Cu (100)
13 13
11.84 CO:CO2 = 1:1 CO:CO2 = 1:1
12

2.05
2

Product ratio

Product ratio
9
7.32 6.87

6 1.05 1.09
1.00
1

1.00
1
0 0
12
C12C 12
C13C 13
C13C 12
C12C 12
C13C 13
C13C

Fig. 6 | Distribution of ethylene isotopologues on single crystal Cu surface. a,b, Distribution of ethylene isotopologues produced in the co-electrolysis of
13
CO/12CO2 (0.5 atm/0.5 atm) on Cu(111) (a) and Cu(100) (b) at −1.5 V in 0.1 M KHCO3.

Fig. 5b. x and y cannot both be 0, as that would entail that all CO This result confirms that CO adsorbed on CuCO is at least a factor of 6
adsorbs on CuCO2. As discussed above, if all CO adsorbs on one type of more active towards C2+ products than CuCO2. The two-site model prob-
sites (CuCO2), a binomial distribution of ethylene isotopologues ( D2 in ably represents a (over)simplification of the well-recognized site het-
this case) would be expected, which is inconsistent with the erogeneity on Cu surfaces under CO(2)RR conditions36,37; however,
non-binomial distribution observed experimentally. Same analysis analysis of this simple model yields the insight that categorizing surface
conducted for the distribution of ethylene isotopologues in the sites into two groups with distinct activities is able to capture key fea-
co-electrolysis of 13CO/12CO2 (0.2 atm/0.8 atm) yields a similar trace in tures of the reactivity and spectroscopic results.
Fig. 5b (Supplementary Note 5). The consistent results from two dif- The observed activity trend with the feed composition in the
ferent feed compositions indicate that the calculated trends reflect co-electrolysis experiments (Fig. 1a) could be rationalized by the two
the intrinsic properties of Den-Cu. The results show that the contribu- roles played by CO2 in the reaction. At low partial pressures of CO2
tion of the CuCO to the 12COL band is less than 25% ( x < 25 in Fig. 5b), (≤0.2 atm), the presence of CO2 exhibits a pronounced promotional
indicating that the majority of 12CO (>75%) remains adsorbed on the effect on the CORR of CO adsorbed on CuCO (Fig. 1c). At high partial pres-
site where it is initially formed, that is, CuCO2, on the reaction time scale. sures, CO2 suppresses the formation rates of C2+ products by directing
Since the CO diffusion is facile on the molecular scale29, substantial an increasing fraction of CO reduction via the slower CuCO2 mediated
spatial segregation of domains with enriched CuCO and CuCO2 is a likely pathway. The overall impact of the two countervailing effects of CO2 is
cause, that is, the majority of CO produced in the CO2RR cannot diffuse manifested in the volcano-shaped dependence of C2+ production rates
from CuCO2 to CuCO before they are converted or desorbed due to the on the CO/CO2 ratio in the feed.
large size of domains with enriched CuCO2.
Relative activity of CO adsorbed CuCO and CuCO2 could be estimated Co-electrolysis of 13CO/12CO2 on single crystal Cu surfaces
from the value of a. Our recent electrokinetic analysis showed that the Co-electrolysis of 13CO/12CO2 (0.5 atm/0.5 atm) was conducted on
formation rate of C2+ products exhibited a close to first-order depend- Cu(111) and Cu(100) to gain insights into the identity of CuCO2 and CuCO
ence of CO coverage in near-neutral electrolyte20,21. Let the CO adsorbed sites proposed in the previous section. X-ray diffraction (XRD) patterns
on CuCO be γ times more likely to react in the CO-to-C2 conversion than and cyclic voltammograms (CVs) of single crystals show expected
on CuCO2, equation (3) can be obtained: features (Supplementary Figs. 25 and 26)27,31, confirming that Cu(111)
and Cu(100) are indeed the exposed facets. On Cu(111), the majority
EthyleneCuCO /COCuCO a of the produced ethylene is 12CH212CH2, followed by 12CH213CH2 and
γ= = (3)
EthyleneCuCO2 /COCuCO2 COCuCO /COCuCO2 13
CH213CH2 (Fig. 6a), which is the opposite to the trend on Den-Cu. Impor-
tantly, the isotopologue distribution of ethylene produced on Cu(111)
where EthyleneCuCO and EthyleneCuCO2 represent the amount of ethyl­ follows the binomial distribution. This is consistent with the expecta-
ene produced on CuCO and CuCO2, respectively, and COCuCO and COCuCO2 tion that Cu surfaces with primarily one type of sites would lead to a
stand for the amount of CO adsorbed on CuCO and CuCO2, respectively. statistical mixture of different isotopologues of products. CV collected
EthyleneCu /COCuCO(2) in equation (3) represents the amount of eth- on Cu(111) after reaction does not show detectable difference from that
CO(2)
ylene produced per unit amount of adsorbed CO on COCuCO(2) . Since collected before the reaction (Supplementary Fig. 27), suggesting the
the ratio between EthyleneCuCO and EthyleneCuCO2 is the defini­tion of absence of surface reconstruction. Given the high proportion of 12CO
a, an analytical expression of γ ’s dependence on a could be obtained incorporation in ethylene, it could be inferred that Cu(111) contains
when the 12COL/13COL ratio is determined by SEIRAS, for example, primarily CuCO2 sites. This claim is supported by the low C2+ selectivity
Fig. 3b,d (Supplementary Note 5). The log γ versus log a plots for two on Cu(111) reported in multiple publications32,38,39, as CuCO2 is proposed
feed compositions are shown in Fig. 5c. With the constraint of y ≤ 100, to be active for CO2-to-CO conversion and relatively inactive for the
the minimum γ value is ~6, suggesting that CO adsorbed on CuCO is at further conversion of CO. 12C and 13C-containing ethylene is evenly
least six times more likely to be reduced further than those on CuCO2. distributed in the co-electrolysis on Cu(100), and the isotopologue
A key assumption in the analysis above is that C2 products only form distribution deviates from the binomial distribution (Fig. 6b). CV on
via coupling of two C1 intermediates on the same type of sites Cu(100) after the co-electrolysis indicates appreciable surface recon-
(homo-coupling), analysis of the two-site model taking into the pos- struction during the reaction (Supplementary Fig. 27), which probably
sibility of C–C coupling between one C1 intermediate adsorbed on CuCO creates sites with properties distinct from those on pristine Cu(100).
and the other on CuCO2 (cross-coupling) shows that the incorporation The lower fraction of 12C incorporation in ethylene on Cu(100) than
of cross-coupling leads to a higher value of γ (Supplementary Note 6). Cu(111) in the co-electrolysis indicates that a lower fraction of surface

Nature Catalysis | Volume 6 | October 2023 | 885–894 891

Article https://doi.org/10.1038/s41929-023-01002-6

sites on Cu(100) than Cu(111) are CuCO2. This is in line with the higher (SP-150e) electrochemical station. Before the CO2RR or CORR, all elec-
ethylene production rate on Cu(100) than on Cu(111) by a factor of ~6.8, trodes were pre-treated for 5 min by applying −1.5 V in the Ar-saturated
as a larger fraction of surface sites on Cu(100) are CuCO. Further, the electrolyte. The feed was continuously introduced into the H-type
fraction of ethylene produced on Den-Cu containing 12C is lower than cell (15 min) to saturate the electrolyte and fill the headspace before
that of Cu(100) indicating that Den-Cu contains an even lower fraction sealing the cell. The working electrode compartment had 18 ml of elec-
of CuCO2 sites than Cu(100). The highly curved surface on particulated trolyte and a headspace of 12.5 ml. Electrochemical measurements
Cu samples could suppress the formation of Cu(111)-like sites (CuCO2), were performed at a constant potential, and the iR (current-resistance)
suggesting that CuCO could be related to defect sites, for example, steps, correction was conducted automatically by the potentiostat (85% com-
adatoms, vacancy and kinks, on Cu surfaces. This hypothesis is sup- pensation of the uncompensated resistance, Ru). Gaseous products were
ported by multiple computational studies40–42. The wide accessible quantified by injecting 0.5 ml of gas from the headspace into a gas chro-
range of 12C/13C ratio in the products of co-electrolysis on different Cu matograph (Agilent Technologies 8890B) equipped with an FID (flame
surfaces highlights the potential of varying the distribution of CuCO2 ionization detector) and a TCD (thermal conductivity detector). Liquid
and CuCO sites to tune the selectivity in the CO2RR. products were analysed by 1H-NMR spectroscopy (Bruker AVIII500).
A total of 100 μl dimethyl sulfoxide (0.05 mM)/D2O was introduced as
Conclusions the internal standard in 500 μl of electrolyte. The CVs were determined
In summary, we demonstrated that the presence of CO2 had a pro- in Ar-saturated 0.1 M KOH solution in the potential range of −1.3 V to
nounced beneficial effect on the CORR without increasing the surface −0.45 V versus Ag/ AgCl at a rate of 50 mV s−1.
CO coverage through combined reactivity, in situ spectroscopic and
isotopic labelling investigations. Analysis of isotopologue distributions Isotopic labelling experiments
of products in the co-electrolysis of mixtures of 13CO/12CO2 showed All electrodes were pre-treated for 5 min by applying −1.5 V in Ar satu-
substantial deviations from those expected of the binomial distribu- rated under the electrolyte. The different proportions of 13CO/CO2
tion. Combined with the relative surface coverages of 12CO and 13CO mixture were continuously introduced into the H-type cell (15 min)
determined with SEIRAS, we showed there were at least two distinct to saturate the electrolyte and fill the headspace before sealing the
types of sites on Den-Cu, and the migration of adsorbed CO among cell. The electrolysis was performed at −1.5 V, and the IR correction
different types of sites was slow on the reaction time scale. On the basis was conducted automatically by the potentiostat (85% compensation
of these observations, we proposed a two-site model, in which one of Ru). The gas-phase products were sampled using a gas-tight syringe
type of site was more efficient for the CO2-to-CO conversion (CuCO2) (Agilent) and analysed using a home-built gas chromatograph (Agilent
and the other favoured the further reduction of CO to C2+ products 8890B)–mass spectroscopy (Pfeiffer Omnistar GSD-350) system. The
(CuCO), which was able to rationalize the observed reactivity trend. liquid-phase products were analysed with 1H-NMR as described above.
The two-site model suggests that CO adsorbed on CuCO is more active
towards the formation of C2+ products than that on CuCO2 by a factor of Preparation of working electrodes for SEIRAS
at least 6. Co-electrolysis experiments on Cu(111) and Cu(100) surfaces Au film underlayers were pre-deposited onto Si ATR crystal by chemi-
indicate that CuCO2 and CuCO probably correspond to Cu(111)-like and cal deposition following established procedure43. The prism was first
defect sites, respectively. polished with 0.05 μm Al2O3 slurry and sonicated in acetone and water,
respectively. Then, the reflecting plane of the prism was immersed
Methods in a 40% NH4F solution for 5 min to create a hydrogen-terminated
Materials surface. The prism was then immersed into a mixture of 2% HF and Au
Dendritic Cu powder (<45 μm, 99.7% trace metals basis), potassium plating solution consisting of 5.75 mM NaAuCl4·2H2O, 0.025 M NH4Cl,
carbonate (K2CO3, 99.9% trace metals basis), potassium phosphate 0.025 M Na2S2O3·5H2O, 0.075 M Na2SO3 and 0.026 M NaOH at 60 °C for
(KH2PO4, 99.7%), dipotassium hydrogen phosphate (K2HPO4, 99.7%), 10 min. The obtained Au film on Si prism was rinsed with de-ionized
Chelex 100, isopropanol (99.999% trace metal basis) and Nafion solu- water and dried at ambient conditions. The Au film was activated with
tion (5 wt. %) were purchased from Sigma-Aldrich. Single crystals of potential scans from −0.2 to 1.5 VRHE in 0.1 M NaHCO3 at a scanning
Cu exposing the Cu(111) and Cu(100) facets were purchased from rate of 100 mV s−1 to improve the surface enhancement effect. The
Princeton Scientific Corporation. carbon dioxide (CO2, 99.999%), Den-Cu electrodes were prepared by dropping the Den-Cu ink onto the
Carbon monoxide (CO, 99.999%) and argon (Ar, 99.999%) were pur- activated Au film. The Den-Cu ink was prepared by dispersing 100 mg
chased from Air Liquide. The carbon fibre paper support (Sigracet of commercial Den-Cu powder in 2.5 ml of isopropanol, followed by
29 BC) was purchased from the Fuel Cell Store. The electrolyte solutions adding 40 μl of 5% Nafion solution into the mixture. After sonicating
were prepared using Milli-Q water (18.2 MΩ cm). for 30 min, 75 μl of the ink was uniformly dropped onto 1 cm2 of the Au
film, followed by drying at room temperature overnight.
Preparation of Cu powder electrodes
Den-Cu (8 mg) was mixed with isopropanol (2.5 ml) and sonicated for In situ SEIRAS experiments
20 min to form an ink solution, which was dropped onto a carbon fibre A home-designed spectro-electrochemical cell with a three-electrode
paper (29BC, loading of 1 mg cm−2). Then, 200 μl of a 2.5 wt. % Nafion configuration was employed for the in situ SEIRAS test (Supplemen-
solution was uniformly dropped on the catalyst layer and dried under tary Fig. 28). Au supported Den-Cu electrode was used as the work-
vacuum for 6 h. A nickel wire was attached to the electrode with silver ing electrode, a graphite rod as the counter electrode and a 3.5 M
epoxy as the current collector. Neither the nickel wire nor the epoxy KCl Ag/AgCl electrode as the reference electrode. The cell was inte-
was exposed to the electrolyte during reaction. grated into the Bruker INVENIO FTIR Spectrometer equipped with
a liquid-nitrogen-cooled mercury cadmium telluride detector. The
Electrochemical measurements potential on the cell was controlled by a Biologic SP150 potentiostat.
Electrochemical CO2RR, CORR and co-electrolysis measurements were All spectra were collected at a 4 cm−1 spectral resolution and cor-
performed in a gas-tight two-compartment, three-electrode H-type cell. responded to 64 co-added scans. During the test, Ar, CO, CO2 or the
The working and counter electrode compartments were separated by a gas mixture was kept bubbling into the electrolyte and the system
piece of anion-conducting membrane. An Ag/AgCl (3.5 M KCl) electrode was mechanically stirred. Gas flow rates were controlled by mass
and a graphite rod were used as reference and counter electrodes, flow controllers (SevenStar) and calibrated by an Agilent ADM2000
respectively. Potential of the system was controlled using a Biologic universal flow meter.

Nature Catalysis | Volume 6 | October 2023 | 885–894 892

Article https://doi.org/10.1038/s41929-023-01002-6

Mass fragmentation analysis 8. Wang, L. et al. Electrochemical carbon monoxide reduction on

To evaluate the ratio of three isotopologues in the ethylene produced polycrystalline copper: effects of potential, pressure, and pH
in co-electrolysis of 13CO/12CO2 mixtures, we focused on the MS intensi- on aelectivity toward multicarbon and oxygenated products.
ties of 30, 29 and 28 amu, corresponding to the parent ions of ACS Catal. 8, 7445–7454 (2018).
13
CH213CH2, 12CH213CH2 and 12CH212CH2, respectively. The 29 amu MS 9. Hori, Y., Murata, A., Takahashi, R. & Suzuki, S. Electroreduction of
signal contains the contribution from the parent ion of 12CH213CH2 and CO to CH4 and C2H4 at a copper electrode in aqueous-solutions
MS fragment of 13CH213CH2 (13C2H3). Similarly, the 28 amu MS signal at ambienttemperature and pressure. J. Am. Chem. Soc. 109,
contains contributions from the parent ion of 12CH212CH2 and MS frag- 5022–5023 (1987).
ments of 13CH213CH2 and 12CH213CH2. MS fragmentation patterns of neat 10. Li, H., Jiang, K., Zou, S. Z. & Cai, W. B. Fundamental aspects in CO2
12
CH212CH2 and 13CH213CH2 were determined (Supplementary Fig. 6), electroreduction reaction and solutions from in situ vibrational
both of which showed the ratio of C2H2/C2H3/C2H4 was 0.68/0.68/1. spectroscopies. Chin. J. Catal. 43, 2772–2791 (2022).
Thus, the intensities of MS signal attributable to the parent ion of 11. Wuttig, A., Yoon, Y., Ryu, J. & Surendranath, Y. Bicarbonate is not a
13
CH213CH2, 12CH213CH2 and 12CH212CH2, referred to as Im , Im and Im
30 29 28
, general acid in Au-catalyzed CO2 electroreduction. J. Am. Chem.
respectively, were deconvoluted by subtracting the contributions from Soc. 139, 17109–17113 (2017).
MS fragments of other isotopologues from the measured MS signals 12. Xiao, H., Cheng, T., Goddard, W. A. 3rd & Sundararaman, R.
of 30, 29 and 28 amu (referred to as I30, I29 and I28, respectively): Mechanistic explanation of the pH dependence and onset
potentials for hydrocarbon products from electrochemical
Im = I30 (4)
30 reduction of CO on Cu (111). J. Am. Chem. Soc. 138, 483–486 (2016).
13. Wu, G. L. et al. Selective electroreduction of CO2 to n-propanol
Im = I29 − 0.68 × Im (5) in two-step tandem catalytic system. Adv. Energy Mater. 12,
29 30
2202054 (2022).
14. Overa, S. et al. Tandem and hybrid processes for carbon dioxide
Im
28
= I28 − 0.68 × (Im
29
+ Im
30
) (6) utilization. Joule 5, 8–13 (2021).
15. Huang, J. F., Mensi, M., Oveisi, E., Mantella, V. & Buonsanti, R.
Then the ratio of 13CH213CH2/12CH213CH2/12CH212CH2 is equal to Structural sensitivities in bimetallic catalysts for electrochemical
m m m
I30/ I29/ I28. CO2 reduction revealed by Ag–Cu nanodimers. J. Am. Chem. Soc.
Similar MS fragmentation analysis was conducted to determine 141, 2490–2499 (2019).
the ratio of 12CH4 and 13CH4. MS fragmentation pattern of 12CH4 and 13CH4 16. Ting, L. R. L. et al. Enhancing CO2 electroreduction to ethanol
were obtained, with CH3/CH4 = 0.83/1 (Supplementary Fig. 6). The on copper–silver composites by opening an alternative catalytic
intensities of parent ions of 13CH4 ( Im
17
) and 12CH4 ( Im
16
) were determined pathway. ACS Catal. 10, 4059–4069 (2020).
as follows: 17. Lv, X. M. et al. Electron-deficient Cu sites on Cu3Ag1 catalyst
promoting CO2 electroreduction to alcohols. Adv. Energy Mater.
Im = I17 (7)
17 10, 2001987 (2020).
18. Jia, H. L. et al. Symmetry-broken Au–Cu heterostructures and
Im = I16 − 0.83 × Im (8) their tandem catalysis process in electrochemical CO2 reduction.
16 17
Adv. Funct. Mater. 31, 2101255 (2021).
19. Wang, X. et al. Mechanistic reaction pathways of enhanced
Data availability ethylene yields during electroreduction of CO2–CO co-feeds
The data that support the findings of this study are available from on Cu and Cu-tandem electrocatalysts. Nat. Nanotechnol. 14,
the corresponding author upon reasonable request. Source data are 1063–1070 (2019).
provided with this paper. 20. Li, J. et al. Electrokinetic and in situ spectroscopic investigations
of CO electrochemical reduction on copper. Nat. Commun. 12,
References 3264 (2021).
1. Centi, G., Quadrelli, E. A. & Perathoner, S. Catalysis for CO2 21. Chang, X. et al. C–C coupling is unlikely to be the rate-determining
conversion: a key technology for rapid introduction of renewable step in the formation of C2+ Products in the copper-catalyzed
energy in the value chain of chemical industries. Energy Environ. electrochemical reduction of CO. Angew. Chem. Int. Ed. 134,
Sci. 6, 1711 (2013). e202111167 (2022).
2. Weber, R. S. Effective use of renewable electricity for making 22. Chang, X. X. et al. Determining intrinsic stark tuning rates
renewable fuels and chemicals. ACS Catal. 9, 946–950 (2019). of adsorbed CO on copper surfaces. Catal. Sci. Technol. 11,
3. Luna, P. D. et al. What would it take for renewably powered 6825 (2021).
electrosynthesis to displace petrochemical processes? Science 23. Chang, X. et al. Understanding the complementarities of surface-
364, 350 (2019). enhanced infrared and Raman spectroscopies in CO adsorption
4. Chu, S., Cui, Y. & Liu, N. The path towards sustainable energy. and electrochemical reduction. Nat. Commun. 13, 2656 (2022).
Nat. Mater. 16, 16–22 (2017). 24. Gunathunge, C. M., Ovalle, V. J., Li, Y., Janik, M. J. & Waegele, M.
5. Hori, Y. et al. Production of methane and ethylene in electro­ M. Existence of an electrochemically inert CO population on Cu
chemical reduction of carbon dioxide at copper electrode electrodes in alkaline pH. ACS Catal. 8, 7507–7516 (2018).
in aqueous hydrogencarbonate solution. Chem. Lett. 15, 25. Lu, X. F., Shinagawa, T. & Takanabe, K. Product distribution control
897–898 (1986). guided by a microkinetic analysis for CO reduction at high-flux
6. Kuhl, K. P., Cave, E. R., Abram, D. N. & Jaramillo, T. F. New electrocatalysis using gas-diffusion Cu electrodes. ACS Catal. 13,
insights into the electrochemical reduction of carbon 1791–1803 (2023).
dioxide on metallic copper surfaces. Energy Environ. Sci. 5, 26. Hollins, P. & Pritchard, J. Interactions of CO molecules adsorbed
7050–7059 (2012). on Cu (111). Surf. Sci. 89, 486–495 (1979).
7. Raciti, D. & Wang, C. Recent advances in CO2 reduction 27. Lum, Y. & Ager, J. W. Evidence for product-specific active sites on
electrocatalysis on copper. ACS Energy Lett. 3, 1545–1556 oxide-derived Cu catalysts for electrochemical CO2 reduction.
(2018). Nat. Catal. 2, 86–93 (2019).

Nature Catalysis | Volume 6 | October 2023 | 885–894 893

Article https://doi.org/10.1038/s41929-023-01002-6

28. Hou, J. J., Chang, X. X., Li, J., Xu, B. J. & Lu, Q. Correlating CO 43. Dunwell, M. et al. The central role of bicarbonate in the
coverage and CO electroreduction on Cu via high-pressure in situ electrochemical reduction of carbon dioxide on Gold. J. Am.
spectroscopic and reactivity investigations. J. Am. Chem. Soc. Chem. Soc. 139, 3774–3783 (2017).
144, 22202–22211 (2022).
29. Xiong, H. C. et al. Correlating the experimentally determined Acknowledgements
CO adsorption enthalpy with the electrochemical CO reduction This work is supported by the Beijing National Laboratory for
performance on Cu surfaces. Angew. Chem. Int. Ed. 62, Molecular Sciences. W.G. acknowledges the support of the Project
e202218447 (2023). funded by China Postdoctoral Science Foundation (2022M710183).
30. Hori, Y., Takahashi, I., Koga, O. & Hoshi, N. Electrochemical The authors thank Y. Zhang and C. Fan for the assistance with
reduction of carbon dioxide at various series of copper single numerical analysis.
crystal electrodes. J. Mol. Catal. A 199, 39–47 (2003).
31. Zhong, D. Z. et al. Coupling of Cu (100) and (110) facets promotes Author contributions
carbon dioxide conversion to hydrocarbonsand alcohols. Angew. W.G., Y.X. and B.X. conceived the idea and designed experiments
Chem. Int. Ed. 60, 4879–4885 (2021). in this study. W.G. synthesized and characterized the catalysts,
32. Hori, Y., Wakebe, H., Tsukamoto, T. & Koga, O. Adsorption of CO and conducted the electrochemical tests. Y.X. performed SEIRAS
accompanied with simultaneous charge transfer on copper single experiments. L.F. performed part of the electrochemical tests. W.G.,
crystal electrodes related with electrochemical reduction of CO2 Y.X., X.C. and B.X. analysed data, and co-wrote the manuscript, with
to hydrocarbons. Surf. Sci. 335, 258–263 (1995). input from all other co-authors.
33. Bagger, A., Ju, W., Varela, A. S., Strasser, P. & Rossmeisl, J.
Electrochemical CO2 reduction: a classification problem. Competing interests
ChemPhysChem 18, 3266–3273 (2017). The authors declare no competing interests.
34. Li, J. H., Maresi, I., Lum, Y. W. & Ager, J. W. Effects of surface
diffusion in electrocatalytic CO2 reduction on Cu revealed by Additional information
kinetic Monte Carlo simulations. J. Chem. Phys. 155, 164701 (2021). Supplementary information The online version
35. Schreier, M., Yoon, Y., Jackson, M. N. & Surendranath, Y. contains supplementary material available at
Competition between H and CO for active sites governs https://doi.org/10.1038/s41929-023-01002-6.
copper-mediated electrosynthesis of hydrocarbon fuels.
Angew. Chem. Int. Ed. 57, 10221–10225 (2018). Correspondence and requests for materials should be addressed to
36. Lee, S. H. et al. Oxidation state and surface reconstruction Bingjun Xu.
of Cu under CO2 reduction conditions from in situ X-ray
characterization. J. Am. Chem. Soc. 143, 588–592 (2021). Peer review information Nature Catalysis thanks Joel Ager III,
37. Simon, G. H., Kley, C. S. & Cuenya, B. R. Potential-dependent Wen-Bin Cai and the other, anonymous, reviewer for their
morphology of copper catalysts during CO2 electroreduction contribution to the peer review of this work.
revealed by in situ atomic force microscopy. Angew. Chem. Int.
Ed. 60, 2561–2568 (2021). Reprints and permissions information is available at
38. Huang, Y., Handoko, A. D., Hirunsit, P. & Yeo, B. S. Electrochemical www.nature.com/reprints.
reduction of CO2 using copper single-crystal surfaces: effects of
CO* coverage on the selective formation of ethylene. ACS Catal. Publisher’s note Springer Nature remains neutral with regard to
7, 1749–1756 (2017). jurisdictional claims in published maps and institutional affiliations.
39. Nitopi, S. et al. Progress and perspectives of electrochemical
CO2 reduction on copper in aqueous electrolyte. Chem. Rev. 119, Springer Nature or its licensor (e.g. a society or other partner)
7610–7672 (2019). holds exclusive rights to this article under a publishing
40. Liu, X. et al. Understanding trends in electrochemical carbon agreement with the author(s) or other rightsholder(s); author
dioxide reduction rates. Nat. Commun. 8, 15438 (2017). self-archiving of the accepted manuscript version of this article is
41. Tang, W. et al. The importance of surface morphology in solely governed by the terms of such publishing agreement and
controlling the selectivity of polycrystalline copper for CO2 applicable law.
electroreduction. Phys. Chem. Chem. Phys. 14, 76–81 (2021).
42. Jiang, K. et al. Metal ion cycling of Cu foil for selective C–C coupling © The Author(s), under exclusive licence to Springer Nature Limited
in electrochemical CO2 reduction. Nat. Catal. 1, 111–119 (2018). 2023

Nature Catalysis | Volume 6 | October 2023 | 885–894 894