Electrocatalytic conversion of carbon dioxide to methane and methanol on transition metal surfaces

Article abstract of DOI:10.1021/ja505791r

Fuels and industrial chemicals that are conventionally derived from fossil resources could potentially be produced in a renewable, sustainable manner by an electrochemical process that operates at room temperature and atmospheric pressure, using only wate

Full text of DOI:10.1021/ja505791r

Article
pubs.acs.org/JACS
Electrocatalytic Conversion of Carbon Dioxide to Methane and
Methanol on Transition Metal Surfaces
Kendra P. Kuhl, Toru Hatsukade, Etosha R. Cave, David N. Abram, Jakob Kibsgaard,
and Thomas F. Jaramillo*
Department of Chemical Engineering, Stanford University, Stanford, California 94305, United States
S
* Supporting Information
ABSTRACT: Fuels and industrial chemicals that are conven-
tionally derived from fossil resources could potentially be
produced in a renewable, sustainable manner by an electro-
chemical process that operates at room temperature and
atmospheric pressure, using only water, CO2, and electricity as
inputs. To enable this technology, improved catalysts must be
developed. Herein, we report trends in the electrocatalytic
conversion of CO2 on a broad group of seven transition metal
surfaces: Au, Ag, Zn, Cu, Ni, Pt, and Fe. Contrary to
conventional knowledge in the field, all metals studied are capable of producing methane or methanol. We quantify reaction rates
for these two products and describe catalyst activity and selectivity in the framework of CO binding energies for the different
metals. While selectivity toward methane or methanol is low for most of these metals, the fact that they are all capable of
producing these products, even at a low rate, is important new knowledge. This study reveals a richer surface chemistry for
transition metals than previously known and provides new insights to guide the development of improved CO2 conversion
catalysts.
materials and promoters have previously been examined for
CO2RR catalysis, including ionic liquids,5,6 proteins,7 organo-
metallic complexes,8,9 organic compounds,10 and semiconduc-
tors.11 The majority of studies have focused on transition
metals,11−14 sparked by a 1985 report from Hori and Suzuki
that showed methane (CH4) and ethylene (C2H4) as the main
products of the CO2RR on a copper electrode.15 Several reports
since have compared the activity of different transition metals
and have found that the product yield and composition of the
CO2RR depend on the transition metal’s binding energy of CO,
believed to be an important intermediate in the reduction of
CO2.16 Metals that bind CO strongly produce few CO2RR
products because they are poisoned by CO or other
intermediates that form during CO2RR, and consequently,
hydrogen (H2), evolved from the competing water reduction, is
the main product observed. On the other hand, metals that
bind CO weakly produce mostly CO, as the CO is released
from the surface before it can be further reduced to products
such as alcohols and hydrocarbons. Cu possesses an
intermediate binding energy for CO, which is believed to be
the reason for its ability to catalyze the formation of more
reduced products that require more than a two-electron
reduction.17−19
INTRODUCTION
■
The modern global energy economy and chemical industry are
heavily dependent on fossil resources, but sustainable
alternatives need to be developed to secure long-term
economic growth while mitigating socio-environmental prob-
lems potentially associated with increasing anthropogenic
emissions of CO2.1,2 Incorporation of renewable electricity
from wind and solar into the global energy supply is one
promising avenue; however, electricity storage is costly for
these intermittent resources.3 One potential solution to
alleviate this concern while simultaneously addressing rising
concentrations of atmospheric CO2 is the electrochemical
reduction of carbon dioxide to carbon-based products.1,4 With
renewable electricity as an input, carbon dioxide and water
could be converted in a sustainable fashion into fuels and
industrial chemicals, for example, hydrocarbons and alcohols
(Figure 1a). Such a process allows energy from an intermittent
renewable source to be consumed immediately and in a manner
that provides the same types of chemical compounds for which
there already exists global-scale demand and infrastructure.
Figure 1b shows the half reactions and reduction potentials of
several possible CO2 reduction reactions (CO2RRs) along with
that of the hydrogen evolution reaction (HER).
In this study, we focus on six transition metals in addition to
copper, presenting new results on Au, Ag, and Zn, which bind
CO weakly, and Ni, Pt, and Fe, which bind CO strongly (Table
A cost-effective process for the electrochemical conversion of
carbon dioxide to value-added products requires electro-
catalysts that are efficient, selective, and stable. To date, no
known catalyst meets these criteria; the development of new
catalysts is needed, and to do so, deeper insights into the
reaction chemistry are required. Toward this goal, a variety of
Received: June 9, 2014
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
results were averaged to give the data reported herein.
Deionized water from a Millipore system (18.2 MΩ·cm) was
used throughout.
Cu. The foil was purchased from Alfa Aesar (1.0 mm thick,
product #42975, 99.999% metals basis). Before each electrolysis
experiment, the foil was mechanically polished (400G sand-
paper, 3M) and then was electrochemically polished at 2.1 V
versus a graphite foil counter electrode in 85% H3PO4 followed
by rinsing with water.
Ni. The foil was purchased from Alfa Aesar (0.1 mm thick,
product #12046, lot #B11X051 and G29U009, 99.994% metals
basis). Before each electrolysis experiment, the foil was
mechanically polished (400G sandpaper, 3M) and was rinsed
with water.
Fe. The foil was purchased from Alfa Aesar (0.1 mm thick,
product #40493, lot #B27U015, 99.99% metals basis). Before
each electrolysis experiment, the foil was mechanically polished
(400G sandpaper, 3M) and was rinsed with water.
Pt. The foil was purchased from Alfa Aesar (0.05 mm thick,
product #42456, lot #J19R006, 99.99% metals basis). Before
each electrolysis experiment, the foil was electrochemically
cleaned by applying +4 V versus a Pt mesh in 30% nitric acid
and by rinsing with water, and the PtO surface was
potentiostatically reduced at −0.20 V versus Ag/AgCl for 3
min in the electrolysis cell immediately prior to electrolysis.
Zn. The foil was purchased from Alfa Aesar (1.0 mm thick,
product #11915, lot #J08S021, 99.9985% metals basis). Before
each electrolysis experiment, the foil was mechanically polished
(400G sandpaper, 3M) and was rinsed with water.
Figure 1. Electrochemical reduction of CO2 coupled to renewable
electricity sources, such as wind or solar, can potentially enable a CO2-
neutral energy cycle in which CO2 is converted to fuels and industrial
chemicals in a renewable and sustainable manner.
Ag. The foil was purchased from Alfa Aesar (0.1 mm thick,
product #12126, lot #D22W016, 99.998% metals basis). Before
each electrolysis experiment, the foil was mechanically polished
(400G sandpaper, 3M) and was rinsed with water.
Au. The foil was purchased from Alfa Aesar (0.1 mm thick,
product #11391, lot #C01Y014, 99.9975+% metals basis). Au
foils were left overnight in concentrated nitric acid and were
rinsed with water prior to each experiment. Au foils were also
held at 2.3 V versus RHE (reversible hydrogen electrode) for
30 min prior to the start of CO2 electroreduction; similar
results on Au were observed for samples that did not undergo
the anodic hold.
S1 of the Supporting Information), among the many transition
metals worthy of deeper exploration. Utilizing new exper-
imental methods that we have developed,20 we were able to
gather the most consistent data on each metal shown to date. A
critical discovery was made during the course of this study: we
observed that every one of the seven transition metals
examined has the ability to produce either methane or
methanol, or both. This is a new insight previously unreported.
Furthermore, for metals that can produce both methane and
methanol, we find a strong correlation in their rates of
production, suggesting commonalities in their mechanistic
pathways. In addition, we find that the overall CO2RR activities
of the metals studied follow a volcano relationship with respect
to their binding energies for CO, with the ideal value near that
of Au. Selectivity is another matter, as we find that in order to
facilitate the formation of products that require greater than
two electrons, for example, methane and methanol, a stronger
binding energy for CO is needed. A deeper understanding of
the mechanism of methane and methanol formation and
associated correlations between the two products, as well as
their relationship to CO binding affinity, will help guide the
design of improved CO2RR catalysts.
RESULTS AND DISCUSSION
■
For each metal studied, potentiostatic electrolysis experiments
were conducted over a range of potentials; the average current
density is shown in Figure 2a. The products of CO2RR in the
gas phase and in the liquid phase and the amount of hydrogen
formed by competing water reduction were measured at each
potential. Figure 2b shows the percentage of the total current at
each potential that goes toward CO2RR instead of making
hydrogen. This data confirms what has been reported
previously: Au, Ag, and Zn, which bind CO weakly, exhibit a
high current efficiency for CO2 reduction whereas Ni, Pt, and
Fe, which bind CO strongly, are poor CO2 reduction catalysts
as they mostly produce hydrogen. Cu, with a moderately strong
CO binding energy, shows a high current efficiency for CO2RR
but less than that of Au, Ag, or Zn. Our measurements as a
function of potential reveal that the current efficiency for
CO2RR on Au, Ag, Zn, and Cu generally increases as
overpotential increases, until reaching a peak after which the
current efficiency decreases. The decrease in CO2RR current
efficiency at high overpotentials could result from CO2 mass
transport limitations or could indicate that the CO2RR is
EXPERIMENTAL SECTION
■
Experimental methods used in this study have been outlined
previously.20 Briefly, using a custom electrochemical cell,
electrolysis experiments were carried out in 0.1 M KHCO3
electrolyte at constant voltage with IR-correction performed by
the potentiostat. Gas chromatography was used to quantify the
amount of gas-phase products, and NMR was used to quantify
the concentration of liquid-phase products. From the
concentrations, it was possible to calculate current efficiencies
and partial current densities for each product. Multiple
electrolysis experiments were run at each potential, and the
B
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 2. (a) Current densities during 1 h long potentiostatic CO2
electrolysis experiments. (b) Current efficiencies (%) for the CO2
reduction reaction (CO2RR) at each potential.
Figure 3. Partial current densities of methane and methanol.
partial current densities for methane and methanol as a
function of applied potential for each of the seven metals;
partial current density is a measure of reaction rate that scales
linearly with turnover frequency (TOF). Among the seven
metals, there is a wide variation in onset potentials for methane
and methanol, defined for the purposes of this discussion as the
earliest potential at which the reaction product is observed by
our analytical methods. For instance, Fe exhibits the earliest
onset for any product that requires greater than two electrons,
producing methane at −0.53 V versus RHE, whereas Ag
requires a voltage of −1.22 V versus RHE to produce methane
and a voltage of −1.35 V versus RHE to produce methanol.
Interestingly, for each of the five metals that can produce
both methane and methanol, only a small difference in onset
potential is observed between the two products for that
particular surface. Generally, methane emerges at lower
potentials by a small margin, which could be explained in
part by our experimental methods’ somewhat greater sensitivity
to methane than to methanol.20 Figure 3 also shows that the
partial current densities of methane and methanol generally
correlate with one another on each metal surface. These
observations suggest commonalities in their mechanistic
pathways. For instance, each product could have a rate-
determining step (rds) in the reaction sequence that is similarly
difficult to surmount; after the rds, both products are
thermodynamically downhill. It is also plausible that methane
and methanol may share several common intermediates. To
produce methanol, a catalyst needs to break just one C−O
bond in CO2, while to produce methane, both C−O bonds
must be broken. If a catalyst breaks both C−O bonds early in
the reaction, producing an oxygenated product such as
methanol would not be expected at these reducing potentials;
methane would be the only plausible C1 product. However, as
most metals in this study were found to catalyze the formation
of both methane and methanol, with the product formation
outcompeted by hydrogen evolution at such negative potentials
(see Supporting Information for further discussion).
While Figure 2 shows important activity trends for the
transition metals studied, a deeper understanding may be
gained by examining the product distribution, that is, selectivity,
in greater detail over the potential range. Using our methods,
sensitive to the identification and quantification of reaction
products, here we report several new products for these metal
surfaces. In particular, we found that Cu is not unique in the
ability to catalyze hydrocarbon and alcohol formation. All of the
metals tested in this study produced methane, methanol, or
both, along with several other minor products. To rule out the
possibility that liquid-phase products observed in this study
come from sources other than CO2 reduction, for example,
carbonaceous contamination,21,22 isotope-labeled 13CO2 was
used to study the 13CO2RR on each metal at a potential where
novel minor products were observed. The incorporation of the
13C label into the products was confirmed by the peak splitting
observed in 1H NMR; see Figure S2 of the Supporting
Information.
Table S2 of the Supporting Information compares previous
literature reports to the data collected in this study. The main
difference between the data sets is that we observe alcohols or
hydrocarbons with all the metal catalysts tested. This can be
partially explained by the sensitivity of our methods for product
detection. In many cases, however, the potential where we first
detected the formation of more reduced products was more
negative than the potential explored in previous reports. This
also highlights the importance of studying catalysts over a wide
potential range in order to gain important insights into activity.
While all the metals tested could produce methane or
methanol, most produced both. Au and Fe are the two
exceptions, producing only methanol and methane, respec-
tively, within the potential range tested. Figure 3 shows the
C
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
rates generally tracking one another across the potential range,
it follows that the second C−O bond cleavage step that leads to
methane likely occurs later in the reaction pathway.
against the density functional theory (DFT) calculated binding
energy of CO onto step sites for each metal surface (Table S1
of the Supporting Information).26,27 A vertical line labeled
CO*|CO(g) is included in Figure 4 for reference to indicate the
thermodynamics of chemical CO adsorption/desorption;19
metal surfaces that favor CO in an adsorbed state, CO*, are
located to the left of the line while those that favor CO
desorption, CO(g), are located to the right. Figure 4 is
presented as a framework for understanding activity and
selectivity for different CO2RR products on the basis of CO
binding energy as a descriptor for the reaction.
A useful figure of merit to compare CO2RR activity among
catalysts is by means of their partial current density for the
CO2RR (to any product) at a common potential. Given the
vastly different current−voltage profiles of Figure 2, there was
no common potential for which CO2RR activity could be
measured on all seven metals. Measurements on six of the
seven metals (Pt, Ni, Cu, Au, Ag, Zn), however, could be
obtained at a common potential, −0.80 0.05 V versus RHE.
Figure 4a shows this data, plotting the partial current densities
for the CO2RR at this potential versus CO binding energy,
where a volcano-shaped trend emerges. As seen in Figure 4a,
Au has the highest partial current density for the CO2RR and
thus represents the peak of the plot with activity decreasing for
metals on either side. The volcano-shaped activity trend for
CO2RR has been explained previously by DFT calculations19
that suggest that metals with a lower CO binding energy than
Au, such as Ag and Zn, exhibit lower activity than Au for the
CO2RR because of slower activation of CO2, the first step in
the CO2RR reaction sequence. Conversely, metals that bind
CO more tightly, namely Cu, Ni, Pt, and Fe, while effective at
activating CO2 are limited by slow CO desorption or further
reaction of CO to other products because of the strong
surface−CO bond.28−31 Thus, overall CO2RR activity is
observed when CO can desorb from the surface or when
overpotentials are reached that can continue to reduce
adsorbed CO to other products that can emerge from the
surface, for example, methane or methanol.
The favorability of cleaving of the final C−O bond could
explain the product selectivity observed for Fe and Au, with Fe
favoring methane with no methanol detected and vice versa for
Au. Interestingly, Fe and Au lie at the strong and weak extremes
of oxygen-binding energies, respectively (Table S1 of the
Supporting Information). With this in mind, the Fe surface
could favor breaking the second C−O bond early in the
reaction as Fe exhibits a strong binding affinity for oxygen
(Table S1 of the Supporting Information). In fact, ultrahigh
vacuum (UHV) experiments of CO on Fe surfaces have shown
that CO dissociates to adsorbed Cads and Oads, even at low
temperatures, because of its high oxophilicity.23,24 If Fe breaks
both C−O bonds early in the reaction sequence, it may explain
why methane is observed as a product but not methanol. Au, on
the other hand, exhibits a very weak affinity for oxygen (Table
S1 of the Supporting Information) which may favor leaving the
C−O bond intact, leading to methanol instead of methane. The
selectivity of Fe for hydrocarbons instead of alcohols at a low
overpotential suggests that hydrocarbon production could be
improved by engineering a surface that binds Oads and Cads
strongly, which would facilitate C−O bond dissociation, but
that has a weak binding energy for Hads, so that HER is
disfavored. The oxygen-binding energies of the other five metal
surfaces are all in-between those of Au and Fe, which could
explain, in part, why those metals produce measurable amounts
of both methane and methanol rather than just one or the
other.
Given the importance of CO as a reactive intermediate in the
CO2RR,19,25 to better understand the behavior of these
catalysts, we construct Figure 4 in which a plot of the
CO2RR activity and selectivity data from Figures 2 and 3
In considering overall CO2RR activity (converting CO2 to
any product), the data in Figure 4a suggest that Au has a CO
binding energy closer to the ideal value than any of the other
metals investigated in this study. However, selectivity is also an
important concern. The major product of CO2RR on Au is CO,
whereas Cu, located on the strong-binding side of Au, exhibits
greater selectivity toward more reduced products that can
readily serve as fuels or industrial chemicals, for example,
hydrocarbons and alcohols.20 This suggests that to favor such
CO2RR products with greater than two electrons transferred,
catalyst surfaces with CO binding energies stronger than that of
Au are desirable.
To explore selectivity further, we examined onset potential as
another important figure of merit that can describe catalytic
activity and selectivity. For each of the seven metals
investigated, Figure 4b plots two different CO2RR onset
potentials with respect to CO binding energy: one for the
overall CO2RR (i.e., for the first CO2RR product to be
detected) and one for the specific CO2RR products of methane
or methanol (denoted as methane/methanol). Many trends in
onset potential are apparent. First, one can observe an overall
CO2RR activity volcano that is similar in shape to that
presented in Figure 4a above it, again with Au near the peak.
The fact that the same trend in overall CO2RR activity is
observed in Figure 4b as in Figure 4a, with the two plots using
Figure 4. (a) Volcano plot of partial current density for CO2RR at
−0.8 V vs CO binding strength. (b) Three distinct onset potentials
plotted vs CO binding energy: the HER, the overall CO2RR, and
methane or methanol. Dashed lines are to guide the eye.
D
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
completely distinct experimentally derived figures of merit for
CO2RR catalysis, further emphasizes the relevance of this type
of relationship between overall CO2RR activity and CO binding
energy.
expected on the polycrystalline metals investigated in this study.
For the sake of comparison, Table S1 of the Supporting
Information also shows DFT-calculated CO binding energies
for other possible surface sites in addition to experimentally
derived values for CO binding energy. There is general
agreement in the trends for CO-binding, whether using values
from theory or experiment, or among different types of surface
sites, which facilitates analysis. Figure S4 of the Supporting
Information plots the CO2RR activity and selectivity data
against these additional CO binding values and shows identical
trends as those observed in Figure 4; thus, the same
conclusions can be drawn irrespective of the particular surface
site, further supporting the relevance of CO binding energy as a
descriptor for the CO2RR.
No discussion of CO2RR kinetics is complete without
addressing the HER, as H2 production is undesirable when
CO2RR products are the target. HER is widely regarded as a
more kinetically facile reaction that can compete against
CO2RR, decreasing CO2RR selectivity.19 For all seven metals,
H2 is observed as a product either simultaneously with, or even
before, any CO2RR products. Interestingly, among the seven
metals examined, we find that Pt, generally considered the most
active HER catalyst,35 is not the most active for the HER under
CO2RR conditions. Fe clearly exhibits the highest HER current
density at any given applied potential (Figure 2a), despite the
fact that Fe is not known to be particularly active for the HER
under more conventional conditions.35 This major change in
the ranking of activities among HER catalysts when examined
under CO2RR conditions suggests the surface chemistry of
adsorbed species is indeed complex, with coverage effects and
adsorbate−adsorbate interactions potentially playing significant
roles.36
The methane/methanol onset potentials plotted in Figure
4b, however, reveal very different behavior with respect to CO
binding energy compared to that for the overall CO2RR. The
broad range of metals can be classified in two distinct regimes:
metals that bind CO weakly (Au, Ag, and Zn) located to the
right of the CO*|CO(g) line and the metals on the left that
bind CO more tightly (Cu, Ni, Pt, and Fe). The first
observation to note is that metals on the right exhibit much
later onset potentials for methane/methanol than the metals on
the left of the CO*|CO(g) line. A second observation is that
the metals on the left, the strong CO-binding side, all exhibit
fairly similar onset potentials for methane/methanol irrespec-
tive of their particular value of CO binding energy, as indicated
by the effectively flat trend line drawn to guide the eye. The
distinct behavior observed across the two regimes of CO
binding can be explained by two crucial phenomena: (1) The
ability for each metal to reduce adsorbed CO and (2) the
coverage of CO on each surface under CO2RR conditions.
There is spectroscopic evidence to suggest that during CO2
reduction there is a high surface coverage of adsorbed CO on
Cu,25,32 Ni,28,29 Pt,33 and Fe.29 High CO coverage on these
metals could imply that the reduction of adsorbed CO is the
rate-determining step to produce desirable reduced products
such as methane/methanol, a view that has been supported by
recent theory.19,34 DFT calculations have also suggested that
for metal surfaces, the thermochemistry of the reaction step in
which adsorbed CO is reduced is rather insensitive with respect
to CO binding energy.19 Hence, a flat trend line in methane/
methanol onset potentials would be expected for metals
operating at fairly high CO coverage across a wide range of
CO binding energies, which is exactly what is observed
experimentally for all metals to the left of the CO*|CO(g) line
in Figure 4b, with onset potentials of approximately −0.65 V
0.1 V versus RHE. These are fairly negative potentials for all
metals.
As mentioned earlier and shown in Figure 4b, the ability to
catalyze the production of methane/methanol is quite different
for metals that bind CO weakly, namely, Au, Ag, and Zn. The
arguments above, where the thermochemistry of reducing
adsorbed CO is fairly insensitive to CO binding energy, would
imply that the weak CO binding metals would also be expected
to be as proficient in executing this step as metals that bind CO
strongly. However, another important factor plays a role: For
metals that bind CO weakly, CO reduction competes against
the very kinetically fast process of CO desorption. Once CO2 is
converted to adsorbed CO on these metals, the CO readily
desorbs before it could be further reduced to alcohol or
hydrocarbon products. This explains the large difference
observed, approximately 0.3−0.6 V, in the measured onset
potential for the overall CO2RR (where CO is the first
product) versus that for methane/methanol; only at extremely
high overpotentials can the rate of CO reduction begin to
compete with the fast rate of CO desorption, consistent with
overpotentials in the range of 1.0−1.2 V needed to produce
methane/methanol at a measurable rate for the weak CO
binding metals.
CONCLUSIONS
■
In this report, we investigate the activity and selectivity for the
CO2 reduction reaction (CO2RR) among seven transition
metal surfaces: Au, Ag, Zn, Cu, Ni, Pt, and Fe. Bringing to bear
new experimental methods with unprecedented sensitivity to
identify and quantify reaction products, we have provided new
insights into CO2 reaction chemistry. Adding important new
knowledge to the field, we have found that all of these
transition metals are capable of producing methane or
methanol, with at least five of the metals (Zn, Ag, Cu, Ni,
and Pt) able to produce both. Quantitative measurements of
the partial current density of these two products as a function
of applied potential revealed a close correlation in their reaction
rates. This allowed us to conclude that methane and methanol
share commonalities in their mechanisms. More generally, it
was found that many of the trends observed in activity and
selectivity for CO2RR products, and particularly for methane
and methanol, can be explained at least in part by each surface’s
binding energy for CO. Our results confirm the importance of
this reaction intermediate and provide greater insight as to how
the CO binding energy impacts catalyst behavior. Ongoing
work exploring the activity of transition metals for the
reduction of CO, formate, and other small organic molecules
will further increase our understanding of the roles of important
intermediates and activity trends in the CO2RR.
We also find that the oxophilicity of the surface, as
determined by binding energy of Oads, could play an important
role in the reaction, in particular for determining selectivity
between methane and methanol. We detect no methane but
only methanol on Au, the least oxophilic metal of the group of
While Figure 4 plots activity and selectivity data versus CO
binding energies calculated for stepped sites, we note that the
analysis is robust with respect to other types of surface sites
E
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
seven, whereas for Fe, the most oxophilic of the group, we
detect methane but not methanol. This provides key insights as
to how one might engineer a surface to improve selectivity
toward one product or the other by modifying the binding
energy for Cads and Oads to favor or disfavor C−O bond
breakage. Controlling the Hads binding energy will also be key
for disfavoring competing HER. These adsorbate binding
characteristics are particularly important to investigate under
the reaction conditions relevant to the CO2RR.
The results of this study offer the most consistent and
complete picture available to date regarding the CO2RR activity
and selectivity for these elemental transition metal catalysts.
The insights gained demonstrate the importance of measuring
CO2RR activity over a wide range of potentials with methods
capable of identifying and quantifying even minor products.
Our findings show that the surface chemistry of elemental
transition metals is richer than previously thought and that
electrocatalysts that can produce hydrocarbons or alcohols are
not as elusive as previously believed. The deeper understanding
of transition-metal catalysts presented here provides important
guidance in searching for catalysts with higher CO2RR activity
and selectivity by means of tuning adsorbate binding energies
appropriately, perhaps in the form of transition-metal alloys.
Improved CO2RR catalysts could enable new technologies with
the ability to address the intermittent nature of renewable
electricity while recycling CO2 to produce fuels and industrial
chemicals that carry global importance in a sustainable manner,
free of fossil fuels.
Anders Nilsson and Prof. Jens K. Nørskov for helpful
discussions.
REFERENCES
■
(1) Lewis, N. S.; Nocera, D. G. Proc. Natl. Acad. Sci. U.S.A. 2006, 103,
15729.
(2) Olah, G. A.; Prakash, G. K. S.; Goeppert, A. J. Am. Chem. Soc.
2011, 133, 12881.
(3) Barnhart, C. J.; Benson, S. M. Energy Environ. Sci. 2013, 6, 1083.
(4) Gattrell, M.; Gupta, N.; Co, A. Energy Convers. Manage 2007, 48,
1255.
(5) Lin, J.; Ding, Z.; Hou, Y.; Wang, X. Sci. Rep. 2013, 3, DOI:
10.1038/srep01056.
(6) Rosen, B. A.; Salehi-Khojin, A.; Thorson, M. R.; Zhu, W.;
Whipple, D. T.; Kenis, P. J. A.; Masel, R. I. Science 2011, 334, 643.
(7) Woolerton, T. W.; Sheard, S.; Reisner, E.; Pierce, E.; Ragsdale, S.
W.; Armstrong, F. A. J. Am. Chem. Soc. 2010, 132, 2132.
(8) Hull, J. F.; Himeda, Y.; Wang, W.-H.; Hashiguchi, B.; Periana, R.;
Szalda, D. J.; Muckerman, J. T.; Fujita, E. Nat. Chem. 2012, 4, 383.
(9) Angamuthu, R.; Byers, P.; Lutz, M.; Spek, A. L.; Bouwman, E.
Science 2010, 327, 313.
(10) Cole, E. B.; Lakkaraju, P. S.; Rampulla, D. M.; Morris, A. J.;
Abelev, E.; Bocarsly, A. B. J. Am. Chem. Soc. 2010, 132, 11539.
(11) Dhakshinamoorthy, A.; Navalon, S.; Corma, A.; Garcia, H.
Energy Environ. Sci. 2012, 5, 9217.
(12) Salehi-Khojin, A.; Jhong, H.-R. M.; Rosen, B. A.; Zhu, W.; Ma,
S.; Kenis, P. J. A.; Masel, R. I. J. Phys. Chem. C 2012, 117, 1627.
(13) Schouten, K. J. P.; Kwon, Y.; van der Ham, C. J. M.; Qin, Z.;
Koper, M. T. M. Chem. Sci. 2011, 2, 1902.
(14) Chen, Y.; Li, C. W.; Kanan, M. W. J. Am. Chem. Soc. 2012, 134,
19969.
(15) Hori, Y.; Kikuchi, K.; Suzuki, S. Chem. Lett. 1985, 1695.
(16) Gattrell, M.; Gupta, N.; Co, A. J. Electroanal. Chem. 2006, 594,
1.
ASSOCIATED CONTENT
■
S
* Supporting Information
(17) Handbook of Fuel Cells: Fundamentals, Technology and
Application; Hori, Y., Ed.; VHC-Wiley: Chichester, U.K., 2003; Vol. 2.
(18) Peterson, A. A.; Abild-Pedersen, F.; Studt, F.; Rossmeisl, J.;
Norskov, J. K. Energy Environ. Sci. 2010, 3, 1311.
(19) Peterson, A. A.; Norskov, J. K. J. Phys. Chem. Lett. 2012, 3, 251.
(20) Kuhl, K. P.; Cave, E. R.; Abram, D. N.; Jaramillo, T. F. Energy
Environ. Sci. 2012, 5, 7050.
(21) Yang, C.-C.; Yu, Y.-H.; van der Linden, B.; Wu, J. C. S.; Mul, G.
J. Am. Chem. Soc. 2010, 132, 8398.
(22) Yui, T.; Kan, A.; Saitoh, C.; Koike, K.; Ibusuki, T.; Ishitani, O.
ACS Appl. Mater. Interfaces 2011, 3, 2594.
Binding strength of CO and O, mass transport limitations,
minor products of CO2RR, isotope-labeled 13CO2 reduction,
trace metal impurities, additional volcano plots. This material is
available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
jaramillo@stanford.edu
Author Contributions
(23) Gonzalez, L.; Miranda, R.; Ferrer, S. Surf. Sci. 1982, 119, 61.
(24) Seip, U.; Tsai, M. C.; Christmann, K.; Kuppers, J.; Ertl, G. Surf.
Sci. 1984, 139, 29.
(25) Hori, Y.; Murata, A.; Yoshinami, Y. J. Chem. Soc., Faraday Trans.
1991, 87, 125.
The manuscript was written through contributions of all
authors.
Notes
The authors declare no competing financial interest.
(26) Jiang, T.; Mowbray, D. J.; Dobrin, S.; Falsig, H.; Hvolbæk, B.;
Bligaard, T.; Nørskov, J. K. J. Phys. Chem. C 2009, 113, 10548.
(27) Hummelshøj, J. S.; Abild-Pedersen, F.; Studt, F.; Bligaard, T.;
Nørskov, J. K. Angew. Chem., Int. Ed. 2012, 51, 272.
(28) Hori, Y.; Koga, O.; Aramata, A.; Enyo, M. Bull. Chem. Soc. Jpn.
1992, 65, 3008.
(29) Koga, O.; Matsuo, T.; Yamazaki, H.; Hori, Y. Bull. Chem. Soc.
Jpn. 1998, 71, 315.
(30) Hori, Y.; Takahashi, R.; Yoshinami, Y.; Murata, A. J. Phys. Chem.
B 1997, 101, 7075.
ACKNOWLEDGMENTS
■
This material is based upon work supported by the National
Science Foundation under Grant Number 1066515 and by the
Global Climate & Energy Project (GCEP) at Stanford
University. T.F.J. gratefully acknowledges partial support
provided by the MDV Innovator Award program at Mohr
Davidow Ventures. K.P.K. and E.R.C. acknowledge support by
the National Science Foundation Graduate Research Fellow-
ship under Grant No. (DGE-1147470). E.R.C. acknowledges
support from the Ford Foundation. D.N.A. acknowledges
support from a Stanford Graduate Fellowship. The authors
would like to thank Dr. Corey Liu of the Stanford Magnetic
Resonance Laboratory and Dr. Stephen R. Lynch of the
Stanford Chemistry for technical assistance with NMR
experiments. The authors would also like to thank Dr. Zhebo
Chen, Dr. Arnold Forman, Prof. Andrew A. Peterson, Prof.
(31) Hansen, H. A.; Varley, J. B.; Peterson, A. A.; Nørskov, J. K. J.
Phys. Chem. Lett. 2013, 4, 388.
(32) Hori, Y.; Wakebe, H.; Tsukamoto, T.; Koga, O. Surf. Sci. 1995,
335, 258.
(33) Taguchi, S.; Aramata, A.; Enyo, M. J. Electroanal. Chem. 1994,
372, 161.
(34) Nie, X.; Esopi, M. R.; Janik, M. J.; Asthagiri, A. Angew. Chem.,
Int. Ed. 2013, 52, 2459.
(35) Greeley, J.; Markovic, N. M. Energy Environ. Sci. 2012, 5, 9246.
F
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(36) Shi, C.; Hansen, H. A.; Lausche, A. C.; Norskov, J. K. Phys.
Chem. Chem. Phys. 2014, 16, 4720.
G
dx.doi.org/10.1021/ja505791r | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX