APPLIED PHISICAL
SCIENCES
Cu metal embedded in oxidized matrix catalyst to
promote CO
2
activation and CO dimerization for
electrochemical reduction of CO
2
Hai Xiao
a
, William A. Goddard III
a,1
, Tao Cheng
a
, and Yuanyue Liu
a,b
a
Materials and Process Simulation Center and Joint Center for Artificial Photosynthesis, California Institute of Technology, Pasadena, CA 91125;
and
b
The Resnick Sustainability Institute, California Institute of Technology, Pasadena, CA 91125
Contributed by William A. Goddard III, May 9, 2017 (sent for review February 13, 2017; reviewed by Timo Jacob and Bruce Parkinson)
We propose and validate with quantum mechanics methods a
unique catalyst for electrochemical reduction of CO
2
(CO
2
RR) in
which selectivity and activity of CO and C
2
products are both
enhanced at the borders of oxidized and metallic surface regions.
This Cu metal embedded in oxidized matrix (MEOM) catalyst is
consistent with observations that Cu
2
O-based electrodes improve
performance. However, we show that a fully oxidized matrix
(FOM) model would not explain the experimentally observed per-
formance boost, and we show that the FOM is not stable under
CO
2
reduction conditions. This electrostatic tension between the
Cu
+
and Cu
0
surface sites responsible for the MEOM mechanism
suggests a unique strategy for designing more efficient and selec-
tive electrocatalysts for CO
2
RR to valuable chemicals (HCO
x
), a
critical need for practical environmental and energy applications.
electrochemical reduction of CO
2
|
Cu metal embedded in oxidized
matrix
|
density functional theory
|
CO
2
activation
|
CO dimerization
E
lectrochemical reduction of CO
2
(CO
2
RR) to valuable
chemicals is an essential strategy to achieve industrial-scale
reduction of the carbon footprint under mild conditions and to
provide a means of storing electrical power from intermittent
renewable sources into stable chemical forms (1). Cu is the pro-
totype electrocatalyst for CO
2
RR, because it is the only pure
metal that delivers appreciable amounts of methane and ethy-
lene plus minor alcohol products (2–7), but it suffers from high
overpotentials and very significant hydrogen evolution reactions
(HERs). Consequently, tremendous efforts are being made to
develop more efficient and selective electrocatalysts, for exam-
ple by surface modification (8) and by nanoparticle (9, 10) and
nanowire (11) engineering.
We examine here the mechanism by which Cu
2
O-based elec-
trodes are observed to improve both efficiency and selectivity for
C
2
products (12–15), which also suppresses HERs by several-
fold. Because Cu
2
O is subject to reduction (back to Cu metal)
under CO
2
RR conditions, the improved performance was ini-
tially attributed to Cu metal surface morphology (8, 16). But a
more recent experiment (15) showed that Cu
+
sites can survive
on the Cu surface for the course of CO
2
RR. Importantly, a Cu
sample that is first oxidized and then reduced using an H2 plasma
leads to performance substantially worse than that of the oxi-
dized sample, despite both having similarly roughened surfaces.
This provides solid evidence that surface Cu
+
plays an essential
role in promoting the efficiency and selectivity of CO
2
RR. How-
ever, experiments have provided no clue about how surface Cu
+
affects the mechanisms of CO
2
RR. Moreover, no previous theo-
retical efforts have elucidated its role.
To understand the promising results achieved with Cu
2
O-
based electrodes, we investigated three distinct models aimed at
unraveling the role of surface Cu
+
in shaping the free energy
profiles of two key steps for CO
2
RR. Here we carry out quan-
tum mechanics (QM) calculations at constant potential by using
our grand canonical methodology (17, 18) that uses the charge-
asymmetric nonlocally determined local-electric (CANDLE)
implicit solvation model (19) to achieve constant electrochemical
potential (not constant number of electrons) within the frame-
work of joint density functional theory (JDFT) (20, 21) (details
in
Computational Details
). The three key steps we focus on are
(
i
) CO
2
activation, which we previously showed to be the rate-
determining step (RDS) for CO production on pure Cu (22);
(
ii
) CO dimerization, which we previously showed to be the RDS
for forming C
2
products from CO on pure Cu (17, 18); and
(
iii
) C
1
product formation, which we find to compete with C
2
products for pure Cu.
We find that the surface Cu
+
by itself actually deteriorates
the performance of CO
2
RR. Instead we show that it is synergy
between surface Cu
+
and surface Cu
0
that improves significantly
the kinetics and thermodynamics of both CO
2
activation and
CO dimerization, while making C
1
unfavorable, thereby boosting
the efficiency and selectivity of CO
2
RR. These results provide a
unique concept for designing improved electrocatalysts. To illus-
trate this synergy we consider the case with an applied potential
U =
−
0.9 V [referenced to standard hydrogen electrode (SHE)],
which is where CO production reaches the peak and C
2
produc-
tion begins on the oxide electrode (15). The free energies at any
other U can be calculated using
Table S1
.
Fig. 1 shows the three surface models we used to probe the
role of surface Cu
+
in CO
2
RR. Fig. 1
A
shows the metallic matrix
(MM), where the pristine Cu(111) surface serves as a reference
model for pure MM that has only Cu
0
on the surface Fig. 1
B
shows the fully oxidized matrix (FOM), where the stoichiometric
nonpolar Cu
2
O(111) surface serves as a model for a FOM with
only Cu
+
on the surface. Here we find two types of Cu
+
surface
Significance
A most promising approach to boosting both efficiency and
selectivity for electrochemical reduction of CO
2
(CO
2
RR) is
using Cu
2
O-based electrodes, and the surface Cu
+
is believed
to play an essential role that is totally unclear from both
experiment and theory. We find that the surface Cu
+
by itself
actually deteriorates the performance of CO
2
RR. Instead we
propose a Cu metal embedded in oxidized matrix (MEOM)
model and show that it is synergy between surface Cu
+
and
surface Cu
0
present in the MEOM model that improves sig-
nificantly the kinetics and thermodynamics of both CO
2
acti-
vation and CO dimerization, thereby boosting the efficiency
and selectivity of CO
2
RR. The MEOM model serves as a unique
platform for design of better electrocatalysts for CO
2
RR.
Author contributions: H.X. and W.A.G. designed research; H.X. and T.C. performed
research; H.X., W.A.G., T.C., and Y.L. analyzed data; and H.X. and W.A.G. wrote the paper.
Reviewers: T.J., Universit
̈
at Ulm; and B.P., University of Wyoming.
The authors declare no conflict of interest.
Freely available online through the PNAS open access option.
1
To whom correspondence should be addressed. Email: wag@wag.caltech.edu.
This article contains supporting information online at
www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1702405114/-/DCSupplemental
.
www.pnas.org/cgi/doi/10.1073/pnas.1702405114
PNAS
|
June 27, 2017
|
vol. 114
|
no. 26
|
6685–6688
A
BC
MM
FOM
MEOM
(Cu
+
)
CUS
(Cu
+
)
CUS
(Cu
+
)
CUS
(Cu
+
)
CUS
metallic Cu
0
region
Fig. 1.
Top and side views of the three surface models. (
A
) A 4
×
4 Cu(111) surface, the model for a metallic matrix (MM). (
B
) A 2
×
2 Cu
2
O(111) surface,
the model for fully oxidized matrix (FOM). (
C
) Metal embedded in oxidized matrix (MEOM) derived by reducing one-quarter of a 2
×
2 Cu
2
O(111) surface.
Here Cu is dark blue, with the active (Cu
+
)
CUS
marked light blue, and red is O. Purple dashes mark the border between Cu
0
and Cu
+
regions. (Please refer
to
Fig. S7
for Bader charge analysis of surface Cu sites on FOM and MEOM models.)
sites: (Cu
+
)
CSS
, a coordinatively saturated Cu
+
site that is
bonded to two O atoms, and (Cu
+
)
CUS
, a coordinatively unsat-
urated site that is bonded to only the one O atom directly below
it (examples are marked in Fig. 1
B
). The (Cu
+
)
CUS
is believed
to be the active site (ref. 23 and references therein). However,
it has been suggested by both theory (24) and experiment (25)
that such (Cu
+
)
CUS
sites are likely missing under oxygen-rich
conditions (e.g., oxygen plasma treatment). More recent theoret-
ical work (26) has shown that (Cu
+
)
CUS
sites are favored under
CO
2
RR conditions at neutral pH. Here we focus on the role of
active surface Cu
+
in shaping the energetics and mechanisms of
CO
2
RR, rather than stability. Fig. 1
C
shows metal embedded in
oxidized matrix (MEOM), a partially reduced Cu
2
O(111) sur-
face in which one-quarter of the surface is reduced. This serves
as a conceptual model for our MEOM in which both Cu
0
and
Cu
+
are present on the surface. We find that this leads to active
(Cu
+
)
CUS
sites at the edge of metallic Cu
0
regions that play an
essential role in the enhanced activity. Our MEOM catalyst site
mimics the case for CO
2
RR operation, where the majority of
surface stays oxidized but some reduced regions are created.
We focus here on the (111) surface orientation, because it is
the most stable among Cu
2
O surfaces (27) and has the fastest
kinetics for Cu surface oxidation (28) (thus the most likely oxide
surface orientation from oxidation of Cu foil). In the experiment
that directly compared the metal with oxide surfaces (15), the
measured onset potentials for C
2
H4 production on Cu metal
surfaces are
−
1.2 V to
−
1.1 V (this is the value for both elec-
tropolished and roughened surfaces obtained by first oxidizing
and then reducing the Cu foil with hydrogen plasma), which are
the same as the onset potentials measured on a Cu(111) single
crystal electrode (
−
1.2 V to
−
1.1 V), but very different from the
value of
−
0.8 V to
−
0.6 V measured on Cu(100) (29, 30).
CO
2
Activation
MM Model.
We find that physisorption of CO
2
(CO
2
,phys
) on
the MM model leads to a noncovalent bond distance of 3.84
̊
A
between the C atom of linear CO
2
and the Cu surface (C-Cu
s
;
Fig. S1
A
), which is similar to our previous study (22). Forming
chemisorbed bent CO
2
(CO
2
,chem
) from this CO
2
,phys
involves a
transition state (TS) that bends CO
2
with C-Cu
s
= 2.36
̊
A, and
the resulting CO
2
,chem
is asymmetrically adsorbed, with a surface
CuO = 2.04
̊
A, whereas the second O atom pointing away from
the surface to form a hydrogen bond (1.52
̊
A) to a surface H
2
O
bonded to a nearby Cu
0
(
Fig. S1
C
and Fig. 2). On the MM model
the activation free energy barrier is
∆
G
6
=
= 0.49 eV at 298 K, sim-
ilar to the value (0.43 eV) for Cu(100) in our previous study (22).
FOM Model.
It was proposed (23) that in the FOM model CO
2
can adsorb at the (Cu
+
)
CUS
site with a 2.09-
̊
A bond of one O
to (Cu
+
)
CUS
. However, exposed to the electrolyte, the (Cu
+
)
CUS
sites are mostly occupied by H
2
O molecules [strong electronic
binding energy of
∆
E =
−
0.98 eV, much larger than that for
CO
2
(
∆
E =
−
0.31 eV)]. Thus, the initial structure for CO
2
acti-
vation on the FOM model is still physisorption of CO
2
with a
4.07-
̊
A distance between the C atom of linear CO
2
and O
s
, the
closest surface O atom (
Fig. S2
A
). We find a
∆
G
6
=
= 0.56 eV to
convert this CO
2
,phys
to a surface carbonate (Fig. 2), which can
subsequently be released into the electrolyte, thereby reducing
0.27
−0.26
0.56
0.12
0.49
0.20
physisorbed CO
2
TS
chemisorbed CO
2
U = −0.9 V (SHE)
C
O
O
O
H
H
+
Cu
+
+
Cu
0
O
C
O
O
Cu
+
C
O
O
O
H
H
+
Cu
0
1.47 Å
1.52 Å
Fig. 2.
Free energy profiles (at U =
−
0.9 V) for CO
2
activation on the MM
(blue), FOM (red), and MEOM (green) models, including the resulting chemi-
sorbed CO
2
structures. Note that FOM leads to a surface carbonate product.
6686
|
www.pnas.org/cgi/doi/10.1073/pnas.1702405114
Xiao et al.
APPLIED PHISICAL
SCIENCES
the FOM surface. Therefore, CO
2
activation in the FOM model
has a barrier 0.07 eV higher than the MM model and involves a
different mechanism that does not lead to the key intermediate
(chemisorbed CO
2
) for CO production. This indicates that the
experimentally observed promotion of CO production using oxi-
dized electrodes (12, 13, 15) cannot be explained with the pres-
ence only of surface Cu
+
.
MEOM Catalyst.
In contrast, the MEOM surface has a metallic
Cu
0
region bordered by the Cu
+
oxide matrix. Here physisorbed
CO
2
is favored on top of the Cu
0
region (
Fig. S3
A
), and the
activation of CO
2
proceeds through a TS that bends CO
2
just
as in the MM case (
Fig. S3
B
), leading to the asymmetrically
chemisorbed CO
2
on the Cu
0
region. But now the free energy bar-
rier is
∆
G
6
=
= 0.27 eV, which is 0.22 eV lower than for the MM
model. Moreover, the chemisorbed CO
2
is
∆
G =
−
0.26 eV more
stable than physisorbed CO
2
on the MEOM catalyst (Fig. 2).
This drastic improvement in both kinetics and thermodynam-
ics for the MEOM catalyst is due to the presence of (Cu
+
)
CUS
sites that bind H
2
O molecules at the edge of the Cu
0
region. This
H
2
O molecule on the (Cu
+
)
CUS
site forms strong hydrogen bonds
to the CO
2
, stabilizing both the TS and the final state (FS) (Fig.
2). This opens a channel in which the negative charge accumu-
lated on the O atom of the CO
2
during activation is distributed
to the Cu
+
region, thus stabilizing both the TS and the FS.
Summarizing, only this MEOM catalyst with both surface Cu
0
regions (binds to activated CO
2
) and Cu
+
(dilutes negative
charge) sites has the ability to enable promotion of CO
2
activa-
tion, with favorable kinetics and thermodynamics. This explains
the experimental observation that both the onset potential and
the peak Faradaic efficiency for CO production are improved
for CO
2
RR on oxide-based electrodes (12, 13, 15). We propose
that our MEOM catalyst might also provide the mechanism by
which partially oxidized atomic cobalt layers improve formate
production (31).
CO Dimerization
MM Model.
CO dimerization on the MM model has been thor-
oughly studied by us and others (17, 32). The initial structure
of two well-separated adsorbed CO molecules (
Fig. S4
A
) goes
through a TS that tilts and draws the two COs close (
Fig. S4
B
),
with
∆
G
6
=
= 1.10 eV, to form an OCCO surface species with a
1.52-
̊
A C-C bond (
Fig. S4
C
).
FOM Model.
In contrast, CO dimerization in the FOM model
takes a distinctly different path. A CO molecule introduced
near the FOM surface (either direct or from CO
2
RR) binds
to the (Cu
+
)
CUS
site by
∆
E =
−
1.62 eV, displacing the H
2
O
(
∆
E =
−
0.98 eV). Thus, CO dimerization starts with two
strongly adsorbed CO molecules on neighboring (Cu
+
)
CUS
sites
(
Fig. S5
A
), which proceeds through an asymmetric TS (
Fig. S5
B
)
that rotates both CO molecules to have O atoms bonded to the
(Cu
+
)
CUS
sites (initially C atoms were bonded) with the C atom
of one CO molecule bonded to the (Cu
+
)
CSS
site in between. The
resulting OCCO surface species (
Fig. S5
C
and Fig. 3) is formed
with a C=C double bond (1.30
̊
A) with each C atom bonded to
the middle (Cu
+
)
CSS
. This is the only stable CO dimer on the
FOM surface, but the formation barrier is
∆
G
6
=
= 3.15 eV, and
the product has a free energy unstable by
∆
G = 2.25 eV. Thus,
the presence of only Cu
+
at the active surface cannot explain the
experimental observation that C
2
products are promoted with
oxidation-treated electrodes (13–15).
MEOM Catalyst.
On the MEOM surface, CO also adsorbs on the
(Cu
+
)
CUS
site (CO@Cu
+
), more stable by 0.48 eV than on the
metallic Cu
0
region (CO@Cu
0
).With MEOM, CO dimerization
from two neighboring CO@Cu
+
is the same in nature as that on
0.80
CO + H
1.13
0.71
0.12
3.15
2.25
1.10
0.86
CO + CO
TS
OCCO
U = −0.9 V (SHE)
C
O
C
O
+
Cu
0
CHO
C
O
C
O
+
Cu
+
+
Cu
0
C
C
O
O
+
Cu
+
C
O
C
O
+
Cu
+
+
Cu
0
δ+
δ−
Fig. 3.
Free energy profiles (at U =
−
0.9 V) of CO dimerization in the MM
(blue), FOM (red), and MEOM (green) models and for CO hydrogenation
to form surface CHO species in the MEOM model (gray green) at pH 7.
Right
shows resulting surface OCCO structures, whereas
Left
shows the ini-
tial structure on the MEOM model, which shows that the C atoms of the
two COs on the Cu
+
and Cu
0
regions are positively and negatively charged,
respectively, which assists the C-C coupling.
the FOM surface, so it would lead to the same
∆
G
6
=
(3.15 eV) as
for the FOM. However, CO dimerization starting with CO@Cu
0
and a neighboring CO@Cu
+
(
Fig. S6
A
) has a modest barrier
of
∆
G
6
=
= 0.71 eV to form the OCCO surface species, in which
the two C atoms are still bonded to the Cu
+
and Cu
0
regions
(
Fig. S6
C
), leading to
∆
G = 0.12 eV (Fig. 3). The favorable ener-
getics of this C-C coupling can be understood by noting that the
C atom of CO@Cu
+
is positively charged (Mulliken charge of
+0.11) whereas the C atom of CO@Cu
0
is negatively charged
(Mulliken charge of
−
0.31) due to back donation. Thus, the
attractive electrostatics between the two Cs assists C-C bond for-
mation. It is this favorable dimerization process on the MEOM
model that improves both kinetics and thermodynamics of the
RDS for C
2
products, compared with the traditional MM model
(Fig. 3). Thus, we propose that promotion of C
2
products for
oxidation-treated electrodes arises from the MEOM surface via
the mechanism described above (13–15).
The major C
2
products from CO
2
RR have been reported to
be either ethylene (13, 15) or ethanol (14), where the major dif-
ference in these experiments is the pH (neutral pH for ethylene
and basic pH for ethanol). We have shown recently (18) that the
energetics of surface water determine the selectivity of alcohol
vs. hydrocarbon products. We found that at neutral pH it is favor-
able for surface water to donate a proton for dehydroxylation to
form hydrocarbon products. Whereas in basic pH the ability of
surface water to dehydroxylate the surface species is suppressed
(because the product OH
−
is less favorable), favoring instead
the alcohol product (ethanol).
C
1
Pathways
Next we consider the possible pathways for forming C
1
with the
MEOM surface. Here we expect CO@Cu
+
and H@Cu
0
. Inter-
estingly, the COH pathway previously proposed by us (17, 33)
is an unreasonable option, because COH@Cu
+
is higher than
CHO@Cu
+
by
∆
G = 1.86 eV. This also eliminates the CO-
COH pathway for C
2
products we previously proposed (17).
On the other hand, the CHO pathway has a reasonable free
energy barrier of
∆
G
6
=
= 1.13 eV (Fig. 3) at neutral pH, which
is still significantly higher than the
∆
G
6
=
= 0.71 eV for C-C cou-
pling with CO@Cu
+
and CO@Cu
0
. Consequently the stability
of CO@Cu
+
(which is more resistant to hydrogenation) blocks
the C
1
products. This selectivity for C
2
over C
1
is intrinsic and
not due to the external local high-pH effect as speculated previ-
ously (13, 15).
Xiao et al.
PNAS
|
June 27, 2017
|
vol. 114
|
no. 26
|
6687
Notes on the MEOM Catalyst
The MEOM concept is in fact a synergistic metal and oxidized
matrix cocatalyst, with both ingredients directly participating in
catalysis. Thus, for the MEOM model to be effective, it is nec-
essary for the metal surface to be level with the oxidized matrix
surface so that the surface species can interact via proper geome-
tries. Therefore, a suitable scheme to generate the MEOM cata-
lyst is deriving the metal directly from the oxidized matrix surface
as in our construction of the MEOM model, which is also consis-
tent with the current experimental strategy. This scheme can be
naturally extended to starting with a mixed oxidized matrix (e.g.,
Cu
2
O/Ag
2
S) for alternative oxidized matrices.
Summary
We present the MEOM model for a partially oxidized Cu sur-
face and show that this model leads to plausible mechanisms
to explain the experimental findings that CO
2
RR can be made
more efficient and selective, using oxidized electrodes. How-
ever, MEOM requires that we only partially oxidize the sur-
face. This MEOM model presents a unique guideline for design
of improved CO
2
RR electrocatalysts. In contrast to previous
speculations, we find that the active surface Cu
+
sites alone do
not improve the efficiencies of CO
2
RR and indeed deteriorate
the efficiency. Instead the synergy between active surface Cu
+
and Cu
0
regions present in the MEOM model is responsible for
improving significantly the kinetics and thermodynamics of both
CO
2
activation and CO dimerization while impeding C
1
path-
ways, the key steps for efficiency and selectivity of CO
2
RR.
Based on our MEOM model, we conclude that the oxidized
matrix (Cu
2
O) is unstable under CO
2
RR working conditions.
We find that the role of the Cu
2
O is mainly electrostatic in dilut-
ing the negative charge built up on the CO
2
as it transitions from
physisorbed to chemisorbed structures, which in turn makes the
C atom of CO positively charged. This MEOM model suggests
alternative oxidized matrices (like Ag
2
S) could also deliver sim-
ilar electrostatic contributions, leading to much improved elec-
trochemical stabilities.
ACKNOWLEDGMENTS.
This research was supported by the Joint Center for
Artificial Photosynthesis, a Department of Energy (DOE) Energy Innovation
Hub, supported through the Office of Science of the US DOE under Award
DE-SC0004993. This work used the computational resources of Zwicky (at
California Institute of Technology).
1.
Zhu DD, Liu JL, Qiao SZ (2016) Recent advances in inorganic heterogeneous electro-
catalysts for reduction of carbon dioxide.
Adv Mater
28:3423–3452.
2.
Hori Y, Kikuchi K, Suzuki S (1985) Production of CO and CH
4
in electrochemical reduc-
tion of CO
2
at metal electrodes in aqueous hydrogencarbonate solution.
Chem Lett
14:1695–1698.
3.
Hori Y, Wakebe H, Tsukamoto T, Koga O (1994) Electrocatalytic process of CO selectiv-
ity in electrochemical reduction of CO
2
at metal electrodes in aqueous media.
Elec-
trochim Acta
39:1833–1839.
4.
Hori Y (2008) Electrochemical CO
2
reduction on metal electrodes.
Modern Aspects
of Electrochemistry
, eds Vayenas C, White R, Gamboa-Aldeco M (Springer, New York)
Vol 42, pp 89–189.
5.
Gattrell M, Gupta N, Co A (2006) A review of the aqueous electrochemical reduction
of CO
2
to hydrocarbons at copper.
J Electroanal Chem
594:1–19.
6.
Kuhl KP, Cave ER, Abram DN, Jaramillo TF (2012) New insights into the electrochemical
reduction of carbon dioxide on metallic copper surfaces.
Energy Environ Sci
5:7050–
7059.
7.
Kuhl KP, et al. (2014) Electrocatalytic conversion of carbon dioxide to methane and
methanol on transition metal surfaces.
J Am Chem Soc
136:14107–14113.
8.
Kim YG, Javier A, Baricuatro JH, Soriaga MP (2016) Regulating the product distribu-
tion of CO reduction by the atomic-level structural modification of the cu electrode
surface.
Electrocatalysis
7:391–399.
9.
Roberts FS, Kuhl KP, Nilsson A (2015) High selectivity for ethylene from carbon dioxide
reduction over copper nanocube electrocatalysts.
Angew Chem Int Ed
54:5179–5182.
10.
Loiudice A, et al. (2016) Tailoring copper nanocrystals towards C
2
products in electro-
chemical CO
2
reduction.
Angew Chem Int Ed
55:5789–5792.
11.
Ma M, Djanashvili K, Smith WA (2016) Controllable hydrocarbon formation from
the electrochemical reduction of CO
2
over cu nanowire arrays.
Angew Chem Int Ed
55:6680–6684.
12.
Li CW, Kanan MW (2012) CO
2
reduction at low overpotential on cu electrodes result-
ing from the reduction of thick Cu
2
O films.
J Am Chem Soc
134:7231–7234.
13.
Kas R, et al. (2014) Electrochemical CO
2
reduction on Cu
2
O-derived copper nanopar-
ticles: Controlling the catalytic selectivity of hydrocarbons.
Phys Chem Chem Phys
16:12194–12201.
14.
Li CW, Ciston J, Kanan MW (2014) Electroreduction of carbon monoxide to liquid fuel
on oxide-derived nanocrystalline copper.
Nature
508:504–507.
15.
Mistry H, et al. (2016) Highly selective plasma-activated copper catalysts for carbon
dioxide reduction to ethylene.
Nat Commun
7:12123.
16.
Dutta A, Rahaman M, Luedi NC, Mohos M, Broekmann P (2016) Morphology matters:
Tuning the product distribution of CO
2
electroreduction on oxide-derived cu foam
catalysts.
ACS Catal
6:3804–3814.
17.
Xiao H, Cheng T, Goddard WA, Sundararaman R (2016) Mechanistic explanation of
the pH dependence and onset potentials for hydrocarbon products from electro-
chemical reduction of CO on Cu(111).
J Am Chem Soc
138:483–486.
18.
Xiao H, Cheng T, Goddard WA (2017) Atomistic mechanisms underlying selectivities
in C
1
and C
2
products from electrochemical reduction of CO on Cu(111).
J Am Chem
Soc
139:130–136.
19.
Sundararaman R, Goddard WA (2015) The charge-asymmetric nonlocally determined
local-electric (CANDLE) solvation model.
J Chem Phys
142:064107.
20.
Petrosyan SA, Rigos AA, Arias TA (2005) Joint density-functional theory:
Ab initio
study of Cr
2
O
3
surface chemistry in solution.
J Phys Chem B
109:15436–15444.
21.
Letchworth-Weaver K, Arias TA (2012) Joint density functional theory of the
electrode-electrolyte interface: Application to fixed electrode potentials, interfacial
capacitances, and potentials of zero charge.
Phys Rev B
86:075140.
22.
Cheng T, Xiao H, Goddard WA (2016) Reaction mechanisms for the electrochemical
reduction of CO
2
to CO and formate on the Cu(100) surface at 298 K from quantum
mechanics free energy calculations with explicit water.
J Am Chem Soc
138:13802–
13805.
23.
Bendavid LI, Carter EA (2013) CO
2
adsorption on Cu
2
O(111): A DFT+U and DFT-D
study.
J Phys Chem C
117:26048–26059.
24.
Soon A, Todorova M, Delley B, Stampfl C (2007) Thermodynamic stability and
structure of copper oxide surfaces: A first-principles investigation.
Phys Rev B
75:
125420.
25.
̈
Onsten A, G
̈
othelid M, Karlsson UO (2009) Atomic structure of Cu
2
O(111).
Surf Sci
603:257–264.
26.
Nie X, Griffin GL, Janik MJ, Asthagiri A (2014) Surface phases of Cu
2
O(111) under CO
2
electrochemical reduction conditions.
Catal Commun
52:88–91.
27.
Bendavid LI, Carter EA (2013) First-principles predictions of the structure, stability,
and photocatalytic potential of Cu
2
O surfaces.
J Phys Chem B
117:15750–15760.
28.
Lian X, Xiao P, Yang SC, Liu R, Henkelman G (2016) Calculations of oxide formation
on low-index Cu surfaces.
J Chem Phys
145:044711.
29.
Schouten KJP, Qin Z, Gallent EP, Koper MTM (2012) Two pathways for the forma-
tion of ethylene in CO reduction on single-crystal copper electrodes.
J Am Chem Soc
134:9864–9867.
30.
Schouten KJP, P
́
erez Gallent E, Koper MTM (2014) The influence of pH on the reduc-
tion of CO and to hydrocarbons on copper electrodes.
J Electroanal Chem
716:
53–57.
31.
Gao S, et al. (2016) Partially oxidized atomic cobalt layers for carbon dioxide elec-
troreduction to liquid fuel.
Nature
529:68–71.
32.
Montoya JH, Shi C, Chan K, Nørskov JK (2015) Theoretical insights into a CO dimer-
ization mechanism in CO
2
electroreduction.
J Phys Chem Lett
6:2032–2037.
33.
Nie X, Esopi MR, Janik MJ, Asthagiri A (2013) Selectivity of CO
2
reduction on cop-
per electrodes: The role of the kinetics of elementary steps.
Angew Chem Int Ed
52:
2459–2462.
34.
Kresse G, Hafner J (1993) Ab initio molecular dynamics for liquid metals.
Phys Rev B
47:558–561.
35.
Kresse G, Furthmuller J (1996) Efficiency of ab-initio total energy calculations for met-
als and semiconductors using a plane-wave basis set.
Comput Mater Sci
6:15–50.
36.
Kresse G, Furthmuller J (1996) Efficient iterative schemes for ab initio total-energy
calculations using a plane-wave basis set.
Phys Rev B
54:11169–11186.
37.
Perdew JP, Burke K, Ernzerhof M (1996) Generalized gradient approximation made
simple.
Phys Rev Lett
77:3865–3868.
38.
Kresse G, Joubert D (1999) From ultrasoft pseudopotentials to the projector
augmented-wave method.
Phys Rev B
59:1758–1775.
39.
Henkelman G, Uberuaga BP, J
́
onsson H (2000) A climbing image nudged elastic band
method for finding saddle points and minimum energy paths.
J Chem Phys
113:
9901–9904.
40.
Henkelman G, J
́
onsson H (1999) A dimer method for finding saddle points on
high dimensional potential surfaces using only first derivatives.
J Chem Phys
111:
7010–7022.
41.
Bertie JE, Whalley E (1967) Optical spectra of orientationally disordered crystals.
II. Infrared spectrum of ice Ih and ice Ic from 360 to 50 cm
−
1
.
J Chem Phys
46:
1271–1284.
42.
Liu H, Wang Y, Bowman JM (2013) Vibrational analysis of an ice Ih model from 0
to 4000 cm
−
1
using the ab initio WHBB potential energy surface.
J Phys Chem B
117:10046–10052.
43.
Sundararaman R, Gunceler D, Letchworth-Weaver K, Schwarz KA, Arias TA (2012)
JDFTx. Available at jdftx.org/. Accessed November 2, 2016.
44.
Garrity KF, Bennett JW, Rabe KM, Vanderbilt D (2014) Pseudopotentials for high-
throughput DFT calculations.
Comput Mater Sci
81:446–452.
45.
Arias TA, Payne MC, Joannopoulos JD (1992) Ab initio molecular dynamics: Analyti-
cally continued energy functionals and insights into iterative solutions.
Phys Rev Lett
69:1077–1080.
46.
Freysoldt C, Boeck S, Neugebauer J (2009) Direct minimization technique for metals
in density functional theory.
Phys Rev B
79:241103.
6688
|
www.pnas.org/cgi/doi/10.1073/pnas.1702405114
Xiao et al.