Selectivity Map for the Late Stages of CO and CO2 Reduction to C2 Species on Copper Electrodes

Article abstract of DOI:10.1002/anie.202014060

The electrochemical CO and CO₂ reduction reactions (CORR and CO₂RR) using copper catalysts and renewable electricity hold promise as a carbon-neutral route to produce commodity chemicals and fuels. However, the exact mechanisms and structure sensitivity of Cu electrodes toward C₂ products are still under debate. Herein, we investigate ethylene oxide reduction (EOR) as a proxy to the late stages of CORR to ethylene, and the results are compared to those of acetaldehyde reduction to ethanol. Density functional theory (DFT) calculations show that ethylene oxide undergoes ring opening before exclusively reducing to ethylene via *OH formation. Based on generalized coordination numbers (CN), a selectivity map for the late stages of CORR and CO₂RR shows that sites with moderate coordination (5.9 < CN < 7.5) are efficient for ethylene production, with pristine Cu(100) being more active than defective surfaces such as Cu(311). In contrast, kinks and edges are more active for ethanol production, while (111) terraces are relatively inert.

Full text of DOI:10.1002/anie.202014060

Angewandte
Research Articles
Chemie
How to cite: Angew. Chem. Int. Ed. 2021, 60, 10784–10790
International Edition:
German Edition:
doi.org/10.1002/anie.202014060
doi.org/10.1002/ange.202014060
CO2 Reduction Reaction
Hot Paper
Selectivity Map for the Late Stages of CO and CO2 Reduction to C2
Species on Copper Electrodes
Oriol PiquØ+, Qi Hang Low+, Albertus D. Handoko, Boon Siang Yeo,* and Federico Calle-
Vallejo*
Abstract: The electrochemical CO and CO2 reduction reac-
tions (CORR and CO2RR) using copper catalysts and renew-
able electricity hold promise as a carbon-neutral route to
produce commodity chemicals and fuels. However, the exact
mechanisms and structure sensitivity of Cu electrodes toward
C2 products are still under debate. Herein, we investigate
ethylene oxide reduction (EOR) as a proxy to the late stages of
CORR to ethylene, and the results are compared to those of
acetaldehyde reduction to ethanol. Density functional theory
(DFT) calculations show that ethylene oxide undergoes ring
opening before exclusively reducing to ethylene via *OH
formation. Based on generalized coordination numbers (CN),
a selectivity map for the late stages of CORR and CO2RR
shows that sites with moderate coordination (5.9 < CN < 7.5)
are efficient for ethylene production, with pristine Cu(100)
being more active than defective surfaces such as Cu(311). In
contrast, kinks and edges are more active for ethanol
production, while (111) terraces are relatively inert.
a unique catalyst capable of catalyzing the production of
a variety of hydrocarbons and alcohols from CORR[4] and
CO2RR.[5] Interestingly, the relative abundance of certain
products depends strongly on the catalystꢀs morphology and
the applied potential. Specifically, methane tends to abound
when Cu electrodes have hexagonal symmetry (e.g. pristine
and stepped Cu(111)) (for the side view of different facets, see
Figure 1a). On the other hand, electrodes with square
symmetry, for example Cu(100), favor C2 products such as
ethylene and ethanol. Additionally, methane evolution com-
mences at considerably more negative potentials than those
required to form C2 compounds.[6–9] Such dissimilar behavior
indicates that the C1 and C2 reaction pathways are likely
mediated by different species.
Introduction
The electrochemical reduction of carbon monoxide and
carbon dioxide (CORR and CO2RR) using solar electricity
has the potential to supply useful chemicals and liquid fuels
sustainably while maintaining carbon neutrality.[1–3] Pioneer-
ing works from Hori and co-workers showed that copper is
[*] O. PiquØ,[+] Dr. F. Calle-Vallejo
Departament de Ci›ncia de Materials i Química Física & Institut de
Química Teòrica i Computacional (IQTCUB)
Universitat de Barcelona
Martí i Franqu›s 1, 08028 Barcelona (Spain)
E-mail: f.calle.vallejo@ub.edu
Q. H. Low,[+] Prof. Dr. B. S. Yeo
Department of Chemistry, Faculty of Science, National University of
Singapore, 3 Science Drive 3, Singapore 117543 (Singapore)
and
Solar Energy Research Institute of Singapore (SERIS)
National University of Singapore (NUS)
7 Engineering Drive 1, Building E3A, #06-01, Singapore 117574
(Singapore)
E-mail: chmyeos@nus.edu.sg
Dr. A. D. Handoko
Institute of Materials Research and Engineering
Agency for Science, Technology and Research (A*STAR)
2 Fusionopolis Way, Innovis, Singapore 138634 (Singapore)
Figure 1. a) Side views of the surface models used in this work.
b) Schematics of the different reaction pathways for ethylene oxide
reduction to ethylene and ethanol on Cu electrodes. The first step is
the chemisorption of ethylene oxide, whereas the other steps are
electrochemical. The green arrows mark the lowest-energy pathway.
Atom key: Cu (gray), C (brown), O (red), H (pink).
[+] These authors contributed equally to this work.
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under:
https://doi.org/10.1002/anie.202014060.
10784
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790

Angewandte
Research Articles
Chemie
The C2 pathways seem to proceed via several common
drives the functionality of Cu catalysts toward specific C2
molecules.
Herein, we provide a computational-experimental study
intermediates, as evidenced by the experimental observation
that the Faradaic efficiencies of ethylene and ethanol shift
similarly when different alkaline cations are used in the
electrolyte.[10] Ethylene and ethanol are also known to have
identical or similar onset potentials,[11–13] which suggests that
they share a common potential-limiting step. Further insight
can be extracted from the pH independence in the standard
hydrogen electrode (SHE) scale of the CORR to ethylene
of EOR to ethylene on copper electrodes. The onset
potentials for ethylene formation were determined for
Cu(111), Cu(100), Cu(211) and Cu(311) single-crystal elec-
trodes. We find a quantitative correlation between the
experimental and calculated onset potentials and the geo-
metric structure of the active sites. Furthermore, we elucidate
the most likely reaction pathway for EOR to ethylene, where
*OH protonation is the potential-limiting step for all the
studied surface sites. More importantly, we build a selectivity
map for the late stages of CO2RR and CORR to C2 products
at Cu surfaces. The map shows how the selectivity toward
either ethanol or ethylene is modulated by the geometric
structure of the active sites.
and ethanol.[4] This observation suggests that C C bonds are
À
made via *CO coupling through an electron transfer followed
by a proton transfer.
Indeed, computational studies have indicated that during
CORR on Cu(100) to C2 species, the coupling of two *CO
mediated by an electron transfer to form *C2O2À is potential-
limiting.[14] Besides, ethylene and ethanol were found to share
the same intermediates up to the fifth proton-electron
transfer. The pathways to ethanol and ethylene then bifurcate,
with the latter being the dominant product on that facet. To Results and Discussion
improve the selectivity toward acetaldehyde and ethanol,
modified Cu catalysts such as oxide-derived Cu (OD-Cu)[15–18]
or CuAg composites have been used.[19–22]
Active sites and reaction pathways. The pathways through
which ethylene oxide may be reduced to either ethylene
(C2H4O(g) + 2H+ + 2 eÀ ! C2H4(g) + H2O(l)) or ethanol
(C2H4O(g) + 2H+ + 2eÀ ! C2H5OH(l)) via 2-electron catalytic
processes are shown in Figure 1b. We computed the possible
pathways on Cu(100), Cu(111) and four defective Cu surfaces,
namely, Cu(211), Cu(311), 4AD@Cu(100) (four Cu adatoms
on Cu(100)) and 4AD@Cu(111) (four Cu adatoms on
Cu(111)). A depiction of the different active sites under
study is shown in Figure 1a and the calculated adsorption
energies are presented in the Supporting Information,
Table S10. We concluded that the most favorable pathway
in all cases is the one that leads to ethylene via *OH, as shown
in Equations (1) to (3) (green pathway in Figure 1b). *OH is
at least ꢀ 0.2 eV more stable than the other two intermediates
(*OCH2CH3 and *CH2CH2OH). Free-energy diagrams com-
paring these three pathways for each surface model are found
in the Supporting Information, Figures S10–S15.
The intermediate formed upon the fifth proton-electron
transfer in the C2 pathway during CORR is *CH2CHO. If the
alpha carbon is hydrogenated, acetaldehyde (CH3CHO) is
produced. This molecule has been experimentally detected as
a CORR intermediate on OD-Cu,[15] and shown to reduce
selectively to ethanol on Cu electrodes.[23] If the carbonyl
carbon is hydrogenated, adsorbed oxygen (*O) and gaseous
ethylene are formed (*CH2CHO + H+ + eÀ ! *O +
C2H4(g)).[14] A closer inspection of the latter electrochemical
step shows that it convolutes the hydrogenation of the
À
carbonyl carbon and the scission of the C O bond. As the
kinetic barrier for the scission has been shown to be large,[24,25]
*OCH2CH2 might form at the surface of the electrodes. Such
an oxametallacycle species, known in gas-phase heteroge-
neous catalysis,[26] is a bidentate, adsorbed form of ethylene
oxide in which the epoxy ring has opened (Figure 1b). Thus,
the final stages of CO2RR and CORR to ethylene on Cu may
be regarded as the electrochemical ethylene oxide reduction
(EOR).[7]
*
*
C2H4O g
!
OCH2CH2
eÀ
1
2
3
*
*
OH
OCH2CH2
H
!
C2H4 g
A previous study by Schouten et al. showed this possibility
when they examined various possible precursors to ethylene,
such as glyoxal, glycolaldehyde, ethylene glycol and ethylene
oxide. They found that only ethylene oxide was reduced to
ethylene on polycrystalline Cu.[7] Additionally, they showed
that the onset potential for the EOR occurs earlier on
Cu(100) as compared to Cu(111), which indicates that EOR is
more efficiently catalyzed by Cu(100).[27] This observation
agrees with the results for CO2RR and CORR, where
ethylene forms on Cu(100) with lower onset potentials than
on Cu(111).[1,3,11,27] It has been hypothesized that the reason
behind the lower EOR onset potential on Cu(100) is the
enhanced stability of the oxametallacycle intermediate
formed during ethylene oxide reduction on this surface.[7]
Given that some intermediates are possibly shared, but
ethylene is usually more abundantly produced than ethanol
during CO2RR and CORR,[28] it is critical to understand what
eÀ ! H2O l
*
OH
H
Where * is a free surface site. The first step is the
chemisorption of ethylene oxide, which involves the opening
of the epoxy ring and a bidentate adsorption configuration
[Equation (1)]. Ethylene and *OH are formed after the first
proton-electron transfer [Equation (2)]. Finally, *OH is
reduced to H2O(l) upon the second proton-electron transfer,
such that the active site is free again to start a new catalytic
cycle [Equation (3)].
It is important to note that if a given site does not bind
ethylene oxide strong enough [Equation (1)], it would be
difficult for EOR to proceed. This could occur as a result of
the competitive adsorption of *H, which is involved in the
hydrogen evolution reaction. Hence, before we analyze the
electrochemical steps in Equations (2) and (3), a comparison
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790
ꢀ 2021 Wiley-VCH GmbH
www.angewandte.org 10785

Angewandte
Research Articles
Chemie
intermediates in Equations (1) to (3) for the six surface
models under study are shown in the Supporting Information,
Figure S9. In Figure 3a, we condense these data in the form of
a coordination-activity plot, in which the limiting potential for
each site is correlated with its respective generalized coordi-
nation number. Extrapolating the blue line in Figure 3a, we
find that the top of the coordination-activity plot for EOR on
Cu is at CN = 16.6. Since CNmax = 12 for a face-centered cubic
crystal, neither the top nor the weak-binding side of the EOR
coordination-activity plot are accessible on Cu. Consequently,
all Cu sites share the same potential-limiting step, namely
*OH hydrogenation to H2O [Equation (3)]. In Figure 3a, the
Figure 2. Adsorption free energies of *OCH2CH2 (blue; see [Equa-
tion (1)]), *H at 0 V vs. RHE (red; with respect to H+ + eÀ), and *H at
À0.35 V vs. RHE (orange), as a function of the generalized coordina-
tion number (CN) of the active sites. À0.35 V vs. RHE represents the
least negative experimental onset potential in Figure 3. A linear
regression for *OCH2CH2 along with the corresponding Pearson
correlation coefficient is shown, without including Cu(100) in view of
its known ensemble effects.[29,30] The red and orange lines are the
arithmetic mean of the respective *H adsorption energies. A gray band
of Æ2 MAE is provided around the lines. MAE: mean absolute error of
the points with respect to the linear fit.
between hydrogen and ethylene oxide adsorption is neces-
sary. These adsorption energies are plotted in Figure 2 as
a function of the generalized coordination number (CN) of
the active sites (for details on the assessment of CN, see the
Supporting Information, Section S7). *H and *OCH2CH2
behave differently, as the adsorption energies of the former
are potential dependent (Figure 2), and only the latter
displays strong structural sensitivity. Interestingly, at 0 V vs.
RHE, the adsorption energies of *H and *OCH2CH2 are
comparable on Cu(111) terrace sites. Consequently, Cu(111)
will probably have a higher *H coverage as compared to the
other Cu facets, since these latter surfaces adsorb *OCH2CH2
more strongly than *H at the two potentials analyzed in
Figure 2. As ethylene oxide adsorption requires binding to
several surface atoms, its adsorption is likely hindered to
a large extent by the high coverage of *H on Cu(111).
Furthermore, we note that the adsorption energy of
ethylene oxide on Cu(100) noticeably departs from the linear
trend in Figure 2. As *OCH2CH2 is a bidentate adsorbate, this
departure can be attributed to the ensemble effects originat-
ing from its double-bridge adsorption configuration (Sup-
porting Information, Figure S8), in line with previous reports
for similar bidentate adsorbates at square-symmetry sites.[29,30]
According to those works, ensemble effects should generally
be present on (100)-like sites but tend to be less pronounced
at undercoordinated sites. Indeed, we observe this for
4AD@Cu(100) in Figure 2, whereby the deviation from the
linear trend is within a confidence interval of 83% set at Æ 2
MAE (MAE: mean absolute error).
Figure 3. a) Coordination-activity plot for ethylene oxide reduction to
ethylene. The calculated (blue) and measured (green) onset potentials
are plotted as a function of generalized coordination numbers (CN)
for various electrodes. The dashed line is the equilibrium potential
(0.81 V vs. RHE) and the red line marks the separating point, whereby
*OCH2CH2 adsorbs more strongly than *H for surfaces with CN on
the left side of the red line. Detailed free-energy diagrams are provided
in Figure S9. The large departure of experimental Cu(111) from the
trends suggests that its activity originates from defect sites. b) Ethyl-
ene signals during EOR on Cu(111), Cu(100), Cu(311), and Cu(211) at
different applied potentials in 0.1 M potassium phosphate buffer
solution (pH 7). X marks the potentials where ethylene evolved from
EOR was not generated in detectable amounts.
Experimental and computational onset potentials. The
free-energy diagrams with the relative stability of the reaction
10786 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790

Angewandte
Research Articles
Chemie
calculated onset potentials for Cu(111), Cu(100), Cu(311),
This conclusion goes along the same lines of previous studies
Cu(211), 4AD@Cu(100) and 4AD@Cu(111) are, respectively,
À0.04, À0.34, À0.46, À0.50, À0.48, and À0.52 V vs. RHE.
Next, we determined experimentally the onset potentials
for EOR on copper surfaces (Figure 3; Table S2). The
reaction was performed at pH 7 in 0.1 M potassium phosphate
buffer electrolyte. An electrolyte with neutral pH was chosen
to avoid the side reactions of ethylene oxide to give ethylene
glycol and polyethylene glycol under acidic and basic
conditions.[31] The surfaces of the Cu electrodes used were
characterized using cyclic voltammetry and X-ray diffraction
(XRD; Supporting Information, Figures S1–S4). During elec-
trolysis, ethylene oxide (2.5 mol% in N2 gas balance) was
constantly bubbled into the cathodic compartment. The EOR
products were analyzed by gas and liquid chromatography.
Ethylene was the only EOR product detected, while ethanol
was not observed, even at considerably more negative
potentials such as À1.00 V vs. RHE (Supporting Information,
Figure S6). This agrees with the DFT results, which show that
ethylene is the more favorable EOR product (Figure 1;
Table S10).
The experimental EOR onset potential was determined to
be the potential at which the ethylene signal begins to be
stronger than the baseline signal (Supporting Information,
Section S3). The onset potentials for Cu(100) and Cu(111)
were À0.35 and À0.50 V vs. RHE respectively (Figure 3b).
Compared to the calculated values of À0.34 and À0.04 V vs.
RHE on these two surfaces, it is striking that there is only
good agreement for Cu(100). Judging by the simulated onset
potentials for the four defective surfaces, which are all in the
narrow range of À0.46 to À0.52 V vs. RHE (Figure 3a), we
hypothesized that the discrepancy could stem from the fact
that defects on the Cu(111) crystal, rather than the Cu(111)
terraces themselves, catalyze the EOR.
We evaluated this hypothesis by performing EOR on
Cu(211) and Cu(311). These two surfaces have, respectively
3-atom- and 2-atom-wide (111) terraces, separated by mono-
atomic (100) steps. The experimental EOR onset potential on
Cu(211) was À0.50 V vs. RHE (Figure 3b), which is the same
as that of Cu(111). For Cu(311), the EOR onset potential,
which was at À0.45 V vs. RHE, is slightly earlier than that on
Cu(211) and Cu(111) (Figure 3b). This result is in good
agreement with the presence of a higher density of step edges
on Cu(311), which can increase the amount of ethylene
formed from EOR, thus resulting in a somewhat earlier EOR
onset. Overall, similar onset potentials were determined for
Cu(111), Cu(211) and Cu(311), which strongly indicates that
the “discrepancy” between the experimental and calculated
onset of EOR on Cu(111) was due to (i) the presence of
defective sites such as step edges, and (ii) the inability of
Cu(111) to sustain an appreciable coverage of adsorbed
ethylene oxide.
for similar organic reactions, for example, acetone reduction,
where (111) terraces were also observed to be rather inactive
in view of the competition of the adsorbates with *H for
adsorption sites at potentials close to 0 V vs. RHE.[23,32–34]
Implications for CO2RR and CORR. Numerous works on
roughened and oxide-derived Cu catalysts have invoked the
presence of step and defect sites as being responsible for
enhanced ethylene formation from CO2RR or CORR.[35–37]
Our experimental and computational onset potentials in
Figure 3 suggest that this general claim is likely inaccurate, as
Cu(311), which has abundant step edges, actually catalyzed
ethylene formation less readily than Cu(100). This observa-
tion also concurs with a previous work on CuCl-derived Cu
mesocrystals.[38] The surfaces of these catalysts, which consist
mainly of various atomic steps and (100) terraces, reduced
CO2 to ethylene more selectively than Cu nanoparticles.
Unlike the Cu mesocrystals, the Cu nanoparticle surfaces
were found to have high-index planes composed of abundant
steps, but with no terraces. The present results further stress
the importance of Cu(100) terraces for the selective forma-
tion of ethylene during CO2 electrolysis.
The standard potential for the reduction of ethylene oxide
to ethylene is 0.81 V vs. RHE. Judging by the calculated and
measured onset potentials, we conclude that, despite its
apparent simplicity, the EOR on Cu requires large over-
potentials of no less than ꢀ 1.15 V. Knowing that Equation (3)
(*OH + H+ + eÀ ! H2O(l)) is the potential-limiting step for
all the studied models (Figure 3a), destabilizing *OH would
likely lead to lower overpotentials. Different strategies may
be used to this end, for example, changing Cu-Cu distances via
strain,[39–41] making use of different electrolytes so that cation
and/or anion effects modify *OH adsorption energies,[42,43]
using nonaqueous solvents such as acetonitrile to avoid the
stabilization granted by *OH-H2O hydrogen bonds,[44,45] and
alloying a noble metal such as Pt with Cu, as the former binds
*OH more weakly than the latter.
These strategies are also important for the CORR to
ethylene on Cu(100), as the onset potential is around À0.40 V
vs. RHE and is usually attributed to the formation of a *CO
dimer.[11,14] Once the dimer formation is experimentally
optimized so as to require a significantly less negative
potential, the next target will probably be *OH, as its
conversion to H2O requires À0.35 V vs. RHE according to
Figure 3. However, we note that the adsorption energies of
*OH and *OCH2CH2 are roughly correlated in a linear
fashion (Supporting Information, Figure S7), such that if *OH
adsorption is too weak, ethylene oxide might not adsorb on
the surface of the catalyst.
Finally, we compare the EOR results to those of the
acetaldehyde reduction reaction (ARR) on Cu.[23] Experi-
mentally, the ARR leads exclusively to ethanol (CH3CHO(l)
+ 2H+ + 2eÀ ! C2H5OH(l)) and its equilibrium potential is
0.24 V vs. RHE. On Cu electrodes, the ARR is limited by the
formation of a monodentate, O-bound ethoxy intermediate
(* + CH3CHO(l) + H+ + eÀ ! *OCH2CH3), the hydro-
genation of which is usually downhill in energy to produce
ethanol (*OCH2CH3 + H+ + eÀ ! * + C2H5OH(l)).
In this context, the ethylene signals obtained on the
different facets at À0.50 V vs. RHE are noteworthy (Fig-
ure 3b): Cu(100) produces the highest amount of ethylene
from EOR amongst all the facets studied, followed by
Cu(311), Cu(211) and lastly Cu(111). This finding further
highlights the importance of (100) facets for the EOR to
ethylene and the fact that (111) terraces are likely inactive.
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790
ꢀ 2021 Wiley-VCH GmbH
www.angewandte.org 10787

Angewandte
Research Articles
Chemie
nated defects onto facets within this category makes the
limiting potential more negative. This is what we observed
experimentally in Figure 3b. Finally, adsorbed hydrogen
prevails over adsorbed ethylene oxide on Cu(111) terraces
and more coordinated sites (CN > 7.5), which renders them
inactive for the EOR. This is also in line with the exper-
imental results shown in Figure 3b.
We note in Figure 4a that the opposite slopes for the right
leg of ARR (in blue, 3.7 < CN < 5.9) and the EOR region (in
green, 5.9 < CN < 7.5) suggest that undercoordination has
opposite effects on the catalytic activities. Interestingly, these
observations coincide with many previous findings for
CO2RR and CORR: (I) Pristine Cu(100) is usually selective
toward ethylene.[6,27,28] (II) A rougher Cu surface, which
intuitively is a surface with more defective sites such as steps
and kinks, will enhance the relative production of ethanol
over ethylene as compared to smoother Cu surfaces (Fig-
ure 4b; Supporting Information, Section S4).[13,16,47–50] (III)
Cu(111) produces small amounts of C2 species.[6,27,28] In this
order of ideas, electrodes where a multiplicity of sites coexist,
will likely lead to a mix of C2 products from CO2RR and
CORR.
Conclusion
Despite extensive research on CO2RR and CORR, there
is not yet a consensus about the complete reaction mecha-
nism. Herein, we studied the ethylene oxide reduction (EOR)
as a proxy for the late stages of the CO2RR and CORR
pathway to ethylene on Cu. *OH hydrogenation is the
potential-limiting step in all studied cases, and the most
efficient surface for EOR is Cu(100), with an onset potential
of À0.35 V vs. RHE. The EOR onsets for Cu(111) and
Cu(211) were identical at À0.50 V vs. RHE, but the measured
current densities were smaller on the former, suggesting that
only defect sites on Cu(111) are active.
Comparing the acetaldehyde reduction reaction (ARR)
with EOR on Cu, we conclude that the former is favored at
highly undercoordinated sites (CN < 5.9, typical of step edges
and kinks), the latter is preferred at moderately undercoordi-
nated sites (5.9 < CN < 7.5, for example, Cu(100)), and
hydrogen adsorption precludes ethylene oxide reduction on
Cu(111) and more coordinated sites (CN > 7.5).
In broader terms, our findings indicate that the most
active sites for CO2RR and CORR are different for each C2
product. Cu(100) terraces favor ethylene formation, while
steps, kinks and akin undercoordinated defects favor ethanol
evolution. As the EOR and the ARR lead exclusively to
ethylene and ethanol, respectively, we conclude that the lack
of CO2RR product selectivity on Cu stems from surface
inhomogeneities and is dictated by the site-specific formation
of either *OCH2CH2 or CH3CHO.
Figure 4. a) Selectivity map for the late stages of CO and CO2 electro-
reduction on Cu electrodes. The calculated limiting potentials (UL) are
plotted as a function of the generalized coordination number (CN) of
the active sites. The upper x-axis shows the studied active sites.
Acetaldehyde reduction (blue) to ethanol is most favorable on low-
coordination sites (CN < 5.9), whereas ethylene oxide reduction
(green) to ethylene is most favorable on intermediate coordination
sites (5.9 < CN < 7.5). For CN > 7.5, only hydrogen evolution
(orange) proceeds. To the left of the red line (CN=7.5), the catalytic
sites adsorb ethylene oxide more strongly than *H. The dashed lines
mark the equilibrium potentials; the potential-limiting steps are
provided in each case. b) Ratio of jethanol/jethylene of various Cu catalysts
plotted as a function of catalysts’ roughness factor. All data used are
obtained from previous CO2RR (i–v) or CORR (vi) studies (Supporting
Information, Section S4): (i) reference [13], (ii) reference [47], (iii)
reference [48], (iv) reference [49], (v) reference [50], (vi) reference [16].
In Figure 4a, we provide a selectivity map of Cu sites for
the ARR, EOR and hydrogen evolution reaction (HER). The
map connects the generalized coordination of the active sites
to the preferred reaction products and limiting potentials
(UL). ARR is thermodynamically more favorable than EOR
at highly undercoordinated Cu sites (CN < 5.9), which are
characteristic of step edges and kinks. Note that this result
helps to explain why in a recent work, Cu3 and Cu4 clusters
could reduce CO2 to ethanol with Faradaic efficiencies of up
to 91%.[46] On the other hand, ethylene formation is favorable
at moderate-coordination sites (5.9 < CN < 7.5). Cu(100)
terraces belong to this category and introducing undercoordi-
Acknowledgements
F.C.-V. acknowledges funding from Spanish MICIUN
RTI2018-095460-B-I00, RYC-2015-18996 and María de
10788 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790

Angewandte
Research Articles
Chemie
[18] O. PiquØ, F. ViÇes, F. Illas, F. Calle-Vallejo, ACS Catal. 2020, 10,
Maeztu MDM-2017-0767 grants, and partly by Generalitat de
Catalunya 2017SGR13. O.P. thanks the Spanish MICIUN for
a PhD grant (PRE2018-083811). We thank Red EspaÇola de
Supercomputación (RES) for supercomputing time at
SCAYLE (projects QS-2019-3-0018, QS-2019-2-0023, and
QCM-2019-1-0034), MareNostrum (project QS-2020-1-
0012), and CENITS (project QS-2020-2-0021). The use of
supercomputing facilities at SURFsara was sponsored by
NWO Physical Sciences, with financial support by NWO.
Q.H.L. and B.S.Y. acknowledge funding from the National
University of Singapore (R-143-000-B52-114 and R-143-000-
A64-114), National University of Singapore Flagship Green
Energy Program (R-143-000-A55-733 and R-143-000-A55-
646), and the Solar Energy Research Institute of Singapore
(SERIS).
10488 –10494.
[19] L. R. L. Ting, O. PiquØ, S. Y. Lim, M. Tanhaei, F. Calle-Vallejo,
B. S. Yeo, ACS Catal. 2020, 10, 4059 –4069.
[20] E. L. Clark, C. Hahn, T. F. Jaramillo, A. T. Bell, J. Am. Chem.
Soc. 2017, 139, 15848 –15857.
[21] L. Wang, D. C. Higgins, Y. Ji, C. G. Morales-Guio, K. Chan, C.
Hahn, T. F. Jaramillo, Proc. Natl. Acad. Sci. USA 2020, 117,
12572 –12575.
[22] D. Higgins, A. T. Landers, Y. Ji, S. Nitopi, C. G. Morales-Guio, L.
Wang, K. Chan, C. Hahn, T. F. Jaramillo, ACS Energy Lett. 2018,
3, 2947 –2955.
[23] I. Ledezma-Yanez, E. P. Gallent, M. T. M. Koper, F. Calle-
Vallejo, Catal. Today 2016, 262, 90 –94.
[24] W. Luo, X. Nie, M. J. Janik, A. Asthagiri, ACS Catal. 2016, 6,
219 –229.
[25] D. Torres, N. Lopez, F. Illas, R. M. Lambert, J. Am. Chem. Soc.
2005, 127, 10774 –10775.
[26] S. Linic, M. A. Barteau, J. Am. Chem. Soc. 2002, 124, 310 –317.
[27] K. J. P. Schouten, E. PØrez Gallent, M. T. M. Koper, ACS Catal.
2013, 3, 1292 –1295.
Conflict of interest
[28] Y. Hori, I. Takahashi, O. Koga, N. Hoshi, J. Mol. Catal. Chem.
2003, 199, 39 –47.
The authors declare no conflict of interest.
[29] H. Li, F. Calle-Vallejo, M. J. Kolb, Y. Kwon, Y. Li, M. T. M.
Koper, J. Am. Chem. Soc. 2013, 135, 14329 –14338.
[30] H. Li, Y. Li, M. T. M. Koper, F. Calle-Vallejo, J. Am. Chem. Soc.
2014, 136, 15694 –15701.
Keywords: coordination-activity plot ·copper ·
density functional calculations ·electrocatalysis ·
ethylene oxide reduction
[31] G. A. Melhem, A. Gianetto, M. E. Levin, H. G. Fisher, S.
Chippett, S. K. Singh, P. I. Chipman, Process Saf. Prog. 2001,
20, 231 –246.
[1] S. Nitopi, E. Bertheussen, S. B. Scott, X. Liu, A. K. Engstfeld, S.
Horch, B. Seger, I. E. L. Stephens, K. Chan, C. Hahn, J. K.
Nørskov, T. F. Jaramillo, I. Chorkendorff, Chem. Rev. 2019, 119,
7610 –7672.
[32] C. J. Bondue, F. Calle-Vallejo, M. C. Figueiredo, M. T. M. Koper,
Nat. Catal. 2019, 2, 243 –250.
[33] Y.-J. Zhang, V. Sethuraman, R. Michalsky, A. A. Peterson, ACS
Catal. 2014, 4, 3742 –3748.
[2] O. S. Bushuyev, P. D. Luna, C. T. Dinh, L. Tao, G. Saur, J.
van de Lagemaat, S. O. Kelley, E. H. Sargent, Joule 2018, 2, 825 –
832.
[34] Z. P. Jovanov, H. A. Hansen, A. S. Varela, P. Malacrida, A. A.
Peterson, J. K. Nørskov, I. E. L. Stephens, I. Chorkendorff, J.
Catal. 2016, 343, 215 –231.
[3] Y. Y. Birdja, E. PØrez-Gallent, M. C. Figueiredo, A. J. Gçttle, F.
Calle-Vallejo, M. T. M. Koper, Nat. Energy 2019, 4, 732 –745.
[4] Y. Hori, R. Takahashi, Y. Yoshinami, A. Murata, J. Phys. Chem.
B 1997, 101, 7075 –7081.
[5] Y. Hori, H. Wakebe, T. Tsukamoto, O. Koga, Electrochim. Acta
1994, 39, 1833 –1839.
[6] K. J. P. Schouten, Z. Qin, E. PØrez Gallent, M. T. M. Koper, J.
Am. Chem. Soc. 2012, 134, 9864 –9867.
[7] K. J. P. Schouten, Y. Kwon, C. J. M. van der Ham, Z. Qin,
M. T. M. Koper, Chem. Sci. 2011, 2, 1902 –1909.
[8] G. L. De Gregorio, T. Burdyny, A. Loiudice, P. Iyengar, W. A.
Smith, R. Buonsanti, ACS Catal. 2020, 10, 4854 –4862.
[9] Y. Huang, A. D. Handoko, P. Hirunsit, B. S. Yeo, ACS Catal.
2017, 7, 1749 –1756.
[10] A. Murata, Y. Hori, Bull. Chem. Soc. Jpn. 1991, 64, 123 –127.
[11] E. PØrez-Gallent, G. Marcandalli, M. C. Figueiredo, F. Calle-
Vallejo, M. T. M. Koper, J. Am. Chem. Soc. 2017, 139, 16412 –
16419.
[12] K. P. Kuhl, E. R. Cave, D. N. Abram, T. F. Jaramillo, Energy
Environ. Sci. 2012, 5, 7050 –7059.
[13] D. Ren, Y. Deng, A. D. Handoko, C. S. Chen, S. Malkhandi, B. S.
Yeo, ACS Catal. 2015, 5, 2814 –2821.
[35] Y. Hori, I. Takahashi, O. Koga, N. Hoshi, J. Phys. Chem. B 2002,
106, 15 –17.
[36] W. Tang, A. A. Peterson, A. S. Varela, Z. P. Jovanov, L. Bech,
W. J. Durand, S. Dahl, J. K. Nørskov, I. Chorkendorff, Phys.
Chem. Chem. Phys. 2012, 14, 76 –81.
[37] R. Kas, R. Kortlever, A. Milbrat, M. T. M. Koper, G. Mul, J.
Baltrusaitis, Phys. Chem. Chem. Phys. 2014, 16, 12194 –12201.
[38] C. S. Chen, A. D. Handoko, J. H. Wan, L. Ma, D. Ren, B. S. Yeo,
Catal. Sci. Technol. 2015, 5, 161 –168.
[39] F. Calle-Vallejo, A. S. Bandarenka, ChemSusChem 2018, 11,
1824 –1828.
[40] A. Khorshidi, J. Violet, J. Hashemi, A. A. Peterson, Nat. Catal.
2018, 1, 263 –268.
[41] M. Escudero-Escribano, P. Malacrida, M. H. Hansen, U. G. Vej-
Hansen, A. Velµzquez-Palenzuela, V. Tripkovic, J. Schiøtz, J.
Rossmeisl, I. E. L. Stephens, I. Chorkendorff, Science 2016, 352,
73 –76.
[42] B. Garlyyev, S. Xue, M. D. Pohl, D. Reinisch, A. S. Bandarenka,
ACS Omega 2018, 3, 15325 –15331.
[43] K. Holst-Olesen, M. Reda, H. A. Hansen, T. Vegge, M. Arenz,
ACS Catal. 2018, 8, 7104 –7112.
[44] A. Fortunelli, W. A. Goddard, Y. Sha, T. H. Yu, L. Sementa, G.
Barcaro, O. Andreussi, Angew. Chem. Int. Ed. 2014, 53, 6669 –
6672; Angew. Chem. 2014, 126, 6787 –6790.
[45] F. Calle-Vallejo, A. Krabbe, J. M. García-Lastra, Chem. Sci.
2017, 8, 124 –130.
[46] H. Xu, D. Rebollar, H. He, L. Chong, Y. Liu, C. Liu, C.-J. Sun, T.
Li, J. V. Muntean, R. E. Winans, D.-J. Liu, T. Xu, Nat. Energy
2020, 5, 623 –632.
[14] F. Calle-Vallejo, M. T. M. Koper, Angew. Chem. Int. Ed. 2013, 52,
7282 –7285; Angew. Chem. 2013, 125, 7423 –7426.
[15] E. Bertheussen, A. Verdaguer-Casadevall, D. Ravasio, J. H.
Montoya, D. B. Trimarco, C. Roy, S. Meier, J. Wendland, J. K.
Nørskov, I. E. L. Stephens, I. Chorkendorff, Angew. Chem. Int.
Ed. 2016, 55, 1450 –1454; Angew. Chem. 2016, 128, 1472 –1476.
[16] C. W. Li, J. Ciston, M. W. Kanan, Nature 2014, 508, 504 –507.
[17] Y. Lum, J. W. Ager, Nat. Catal. 2019, 2, 86 –93.
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790
ꢀ 2021 Wiley-VCH GmbH
www.angewandte.org 10789

Angewandte
Research Articles
Chemie
[47] D. Ren, N. T. Wong, A. D. Handoko, Y. Huang, B. S. Yeo, J. Phys.
[50] D. Ren, J. Fong, B. S. Yeo, Nat. Commun. 2018, 9, 925.
Chem. Lett. 2016, 7, 20 –24.
[48] Y. Lum, B. Yue, P. Lobaccaro, A. T. Bell, J. W. Ager, J. Phys.
Chem. C 2017, 121, 14191 –14203.
[49] Y. Kwon, Y. Lum, E. L. Clark, J. W. Ager, A. T. Bell, Chem-
ElectroChem 2016, 3, 1012 –1019.
Manuscript received: October 20, 2020
Revised manuscript received: January 16, 2021
Accepted manuscript online: February 1, 2021
Version of record online: April 1, 2021
10790 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10784 –10790