Experimental and computational investigation of Au25 clusters and CO2: A unique interaction and enhanced electrocatalytic activity

Article abstract of DOI:10.1021/ja303259q

Atomically precise, inherently charged Au₂₅ clusters are an exciting prospect for promoting catalytically challenging reactions, and we have studied the interaction between CO₂ and Au₂₅. Experimental results indicate a reversible Au₂₅-CO₂ interaction that produced spectroscopic and electrochemical changes similar to those seen with cluster oxidation. Density functional theory (DFT) modeling indicates these changes stem from a CO₂-induced redistribution of charge within the cluster. Identification of this spontaneous coupling led to the application of Au₂₅ as a catalyst for the electrochemical reduction of CO₂ in aqueous media. Au₂₅ promoted the CO₂ → CO reaction within 90 mV of the formal potential (thermodynamic limit), representing an approximate 200-300 mV improvement over larger Au nanoparticles and bulk Au. Peak CO₂ conversion occurred at -1 V (vs RHE) with approximately 100% efficiency and a rate 7-700 times higher than that for larger Au catalysts and 10-100 times higher than those for current state-of-the-art processes.

Full text of DOI:10.1021/ja303259q

Article
pubs.acs.org/JACS
Experimental and Computational Investigation of Au25 Clusters and
CO2: A Unique Interaction and Enhanced Electrocatalytic Activity
Douglas R. Kauffman,*,†,‡ Dominic Alfonso,† Christopher Matranga,† Huifeng Qian,§ and Rongchao Jin§
†National Energy Technology Laboratory (NETL), United States Department of Energy, Pittsburgh, Pennsylvania 15236, United
States
‡URS, P.O. Box 618, South Park, Pennsylvania 15129, United States
§Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, United States
S
* Supporting Information
ABSTRACT: Atomically precise, inherently charged Au25 clusters are
an exciting prospect for promoting catalytically challenging reactions,
and we have studied the interaction between CO2 and Au25.
Experimental results indicate a reversible Au25−CO2 interaction that
produced spectroscopic and electrochemical changes similar to those
seen with cluster oxidation. Density functional theory (DFT)
modeling indicates these changes stem from a CO2-induced
redistribution of charge within the cluster. Identification of this
spontaneous coupling led to the application of Au25 as a catalyst for
the electrochemical reduction of CO2 in aqueous media. Au25
promoted the CO2 → CO reaction within 90 mV of the formal potential (thermodynamic limit), representing an approximate
200−300 mV improvement over larger Au nanoparticles and bulk Au. Peak CO2 conversion occurred at −1 V (vs RHE) with
approximately 100% efficiency and a rate 7−700 times higher than that for larger Au catalysts and 10−100 times higher than
those for current state-of-the-art processes.
nonaqueous spectroelectrochemistry was used to manipulate
the charge state of Au25 and establish a benchmark for
adsorbate-induced spectroscopic changes. These benchmarking
studies were required to gain insight into the subtle electronic
structure changes noted during the coupling of Au25 and CO2
in dimethylformamide (DMF). Specifically, the introduction of
CO2 reversibly induced spectroscopic and electrochemical
changes that were similar to those seen during Au25 oxidation.
We also used density functional theory (DFT) to model the
interaction between Au25 and CO2. This type of study is a
valuable complement to experimental efforts because it allows
direct, atomic-scale determination of favorable binding sites and
adsorption structures. Previous computational efforts have
considered the binding of adsorbates such as O2 and CO on
ligand-free or partially ligand-protected clusters.20,21 In our
calculations, several physisorbed linear CO2 states were
identified on fully S−CH3 protected Au25 clusters. From a
theoretical standpoint, first-principles investigations focusing on
the adsorption properties of fully ligand-protected clusters are
scarce, and relatively little is known about the effects of
molecular physisorption on the cluster’s electronic structure.
Our results correlate experimental and computational data to
demonstrate that molecular physisorption can reversibly perturb
the Au25 electronic structure and impact optical and electro-
chemical properties.
INTRODUCTION
■
The chemistry of Au surfaces and Au nanoparticles has been
the focus of intense study,1−3 but recent synthetic advances
have introduced a new class of “small” ligand-protected Au
clusters with unique chemical and electronic properties.2,4−6
Sub-2-nm clusters differ from larger nanoparticles because their
energy levels become quantized and they develop molecule-like
electronic structures.2,4 Crystallographic efforts have confirmed
that such small Au clusters form into atomically precise
structures and that ligand-protected Au25 clusters, possess an
inherent anionic (negative) charge.7,8 Ligand-protected Au25
clusters are a unique platform to study catalytic reactions
because they bridge the size gap between molecules and larger
nanoparticles, they possess an anionic charge, and their surface
structure is precisely known. Despite these features, the
catalytic activities of Au25 and similar atomically precise clusters
have only been investigated experimentally for a handful of
reactions, such as the oxidation of styrene and cyclohexane,9−12
the hydrogenation of aldehydes and ketones,13,14 and the
electrochemical reduction of O2.15 One particularly appealing
catalytic challenge to consider for the negatively charged Au25
cluster is the reduction of carbon dioxide. Not only is CO2 an
important greenhouse gas, but it also represents an abundant
starting material for the generation of fine chemicals and
fuels.16−19
Herein we report a spontaneous and reversible electronic
interaction between CO2 and ligand-protected
Au25(SC2H4Ph)18 clusters (abbreviated as Au25). In situ
Received: April 4, 2012
Published: May 22, 2012
−
© 2012 American Chemical Society
10237
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243

Journal of the American Chemical Society
Article
The observation of spontaneous coupling between the
negatively charged Au25 cluster and CO2 motivated us to
investigate Au25 as a catalyst for the electrochemical reduction
of CO2. Typical electrocatalysts require large overpotentials to
convert CO2 into useful products,16−19 ultimately creating a
challenge for large-scale deployment. In this application, Au25
catalyzed the two-electron conversion of CO2 into CO within
90 mV of the formal potential (thermodynamic limit) of
−0.103 V vs the reversible hydrogen electrode (RHE).19 The
low overpotential is significant because it represents an
approximate 200−300 mV reduction in potential compared
to the larger Au nanoparticles and bulk Au tested in this study
and those in previously published reports.18,19,22 Moreover,
Au25 showed peak CO2 → CO conversion at −1.0 V with
approximately 100% Faradaic efficiency and a rate 7−700 times
higher than those for the larger Au catalysts tested in this study
and 10−100 times higher than those for current state-of-the-art
processes.23 In practical terms, CO is a very useful chemical
that can be converted into a variety of valuable hydrocarbon
species, and a low-voltage, high-efficiency process for
converting CO2 into CO could be instrumental in developing
new carbon management technologies.16,17
RESULTS AND DISCUSSION
■
Spectroelectrochemical Properties of Au25. Figure 1
presents the structure, energy level diagram, and optical spectra
of Au25. The structure of Au25 contains a Au13 core surrounded
by a shell of six Au2S3 semiring structures (the organic C2H4Ph
ligands and tetraoctylammonium counterion are shown in
Figure S1 of the Supporting Information).7,8 The origins of the
Au25 optical absorption spectrum have been interpreted,8 albeit
some uncertainty remains because electronic and geometric
coupling between core and shell atoms contribute to the
spectral features.4,24,25 The onset of the absorption spectrum at
approximately 1.4 eV corresponds to the energy gap (Eg)
between the highest occupied and lowest unoccupied molecular
orbitals (HOMO and LUMO).8,26,27 Recent computational and
experimental studies have identified the a′ and a absorbance
features as HOMO → LUMO transitions originating from
closely spaced, nondegenerate orbitals.24,25,28 Absorbance
Figure 1. Structure, energy level diagram and spectroelectrochemical
properties of Au25. (a) Au25 structure: the C2H2Ph ligands and
tetraoctylammonium (TOA+) counterion are shown in Figure S1 in
the Supporting Information.7,8 (b) Au25 energy level diagram. (c) In
situ spectroelectrochemistry demonstrating oxidation-induced changes
to Au25 optical properties in N2 purged DMF + 0.1 M TBAP; the
labeled absorbance features correspond to the transitions identified in
panel b.
1
feature b is attributed to both HOMO → LUMO + /2 and
presented in Figure 1c. The application of an oxidizing
potential bleached the a′, a, and b absorbance features due to
electronic depletion of the HOMO.27 Furthermore, an
oxidation-induced geometric change increased both the
absorbance at 2.04 eV and the peak area of absorbance feature
c.29 The origins of the oxidation-induced PL increase are still
debated in the literature,27,30,33 but the blue-shifted PL
maximum suggests that oxidation may inhibit photoexcited
electrons from relaxing into midgap states. Both the absorbance
and PL changes were qualitatively reversible with the
application of a sufficiently reducing potential, but the PL
blue shift did not reverse regardless of the applied potential
(Figure S5).
HOMO − 2 → LUMO transitions, and absorbance feature c
represents a HOMO − 5 → LUMO transition. The Au25
photoluminescence (PL) profile remained unchanged if excited
at absorbance feature b (2.78 eV) or a (1.81 eV), although
excitation at absorbance feature a produced lower PL intensity
(Figure S2). Equivalent emission profiles at different excitation
energies indicate LUMO → HOMO emission,27 and the PL
shoulder at 1.38 eV corresponded to the expected Eg in a
variety of solvents. PL below 1.38 eV suggests relaxed emission
from midgap states,27 and the apparent spectral structure below
1.1 eV results from the optical transmittance of the solvent
(Figure S3).
The Au25 optical absorbance and PL spectra are very
sensitive to charge transfer, and characteristic spectral changes
have been observed after chemical oxidation with oxygen29,30 or
dissolved cations,31,32 ligand exchange,30,33−35 and the
application of electrochemical potentials.27 Accordingly, elec-
trochemical modification of the Au25 optical properties provides
an excellent benchmark to compare adsorbate-induced changes.
Nonaqueous electrochemistry was used to apply oxidizing or
reduction potentials to Au25 clusters in DMF (Figure S4),27,36
and the in situ spectroelectrochemical oxidation of Au25 is
Interaction between Au25 and CO2. Figure 2a and b
presents the optical spectra of an initially N2 purged Au25
solution that was subsequently saturated with CO2. The
introduction of CO2 bleached the a′, a, and b absorbance
features and increased the absorbance at both 2.04 eV and
feature c. Additionally, the normalized PL maximum showed an
increase and blue shift; non-normalized PL spectra are shown
in Figure S6 for reference. The spectroscopic changes were
reproducible from sample to sample (Table S1) and consistent
with those noted during spectroelectrochemical oxidation
10238
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243

Journal of the American Chemical Society
Article
Figure 2. Experimental results and density functional theory (DFT) modeling of the Au25−CO2 couple. (a) Optical absorbance and PL spectra, and
(b) difference spectra (Δabs and ΔPL) of Au25 in N2 purged and CO2 saturated DMF; PL spectra were normalized to the absorbance peak area at
λex = 2.78 eV (i.e., absorbance feature b). (c) Square wave voltammetry of Au25 in N2 purged and CO2 saturated DMF + 0.1 M TBAP; anodic
(oxidizing) currents are negative, and scans were collected at 0.05 V s−1. (d) DFT model of stable CO2 adsorption where an O atom of CO2 interacts
with three S atoms in the Au25 shell. (e) Bader charge analysis showing the change in Au25 valence electrons upon CO2 adsorption; negative values
indicate electron loss. Additional DFT results are presented in Figures S14−S17.
(Figure 1c). The apparent interaction between Au25 and CO2
was somewhat unexpected because CO2 shows little electronic
interaction with traditional gold surfaces.37 Control experi-
ments have shown that the optical changes did not result from
drift in the spectrometer signal, the purge gas, solvent
evaporation, solvent polarity, instability of Au25 optical
properties, or the inadvertent introduction of trace atmospheric
moisture (Figures S7−S12). Furthermore, pH-induced spectral
changes are unlikely in aprotic solvents,38 and we confidently
attribute the phenomenon to an interaction between Au25 and
CO2. We note that the CO2-induced spectral changes were
somewhat smaller than those in previous reports of Au25
oxidation.29−33,39 Nonetheless, the above cited examples of
Au25 oxidation were not easily reversed, and the restoration of
Au25 optical properties, when attempted, required strong
reducing agents or electrochemical potentials. In the current
study, the CO2-induced optical changes were reversed by
simply purging the solution with N2 (Figure S13), suggesting a
comparatively weak interaction between the Au25 cluster and
CO2.
species).34 The electrochemical and spectroscopic results are
consistent with the energy level diagram presented in Figure 1b,
since depletion of the HOMO is expected to induce
concomitant spectral bleaching and positive redox shifts. We
+1/0
could not identify CO2-induced shifts in the Au25
redox
wave because this feature was masked by a broad current
increase. We hypothesize this phenomenon was related to
potential-induced desorption of CO2 from the Au25 surface.
Lastly, the electrochemical changes were reversed by purging
the solution with N2 (Figure S13). In comparison to previous
studies, both Murray and co-workers34,35 and Devadas et al.41
have reported that electron-withdrawing ligand groups can
0/−1
irreversibly shift the Au25
wave by several hundred
millivolts. The observed CO2-induced changes are smaller
than those for the above cited examples, but again, the smaller
and more reversible shift in redox potential indicates a
comparatively weak interaction between Au25 and CO2.
Impressive strides have been made in the theoretical
modeling of Au clusters.6,8,20,21,28,35,42,43 However, investigation
of strongly bound adsorbates like O2 and CO required
activating the cluster by removing ligand groups.20,21 Some
molecules, such as CO2, are not expected to chemisorb to the
Au25 surface, and the effects of weaker physisorption-type
interactions have remained unaddressed. In our work, an
extensive search for stable CO2 adsorption configurations was
undertaken using DFT. Specifically, our model utilized Au25
clusters capped with 18 S−CH3 ligands to mimic the fully
ligand-protected cluster used in our experiments. CO2
adsorption was found to occur at a specific “pocket”, or site,
CO2 also affected the electrochemical properties of Au25. The
redox waves presented in Figure 2c represent quantified charge
injection into the Au25 HOMO.27,36 We found that the
introduction of CO2 induced a small but significant 21 2 mV
shift in the Au250/−1 redox wave (>99% confidence level, CL; n
= 3). Positive potentials are considered oxidizing,40 and a
0/−1
positive shift in the Au25
redox potential is consistent with
electronic depletion of the Au25 HOMO (e.g., a larger potential
will be required to withdraw electrons from an oxidized
10239
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243

Journal of the American Chemical Society
Article
Figure 3. Electrochemical reduction of CO2 in aqueous 0.1 M KHCO3; cathodic (reducing) currents are positive, all scans were collected at a rate of
0.01 V s−1, and error bars are from three separate runs with freshly prepared samples. (a) Linear sweep voltammograms (LSVs) of carbon black (CB)
supported Au25 in quiescent (unstirred) N2 purged (pH = 9) and CO2 saturated (pH = 7) 0.1 M KHCO3. (b) Potential-dependent H2 and CO
formation rates for Au25/CB; solutions were stirred at a constant rate during electrolysis runs to prevent product (bubble) buildup on the electrode
surface. (c) LSVs of various Au catalysts in quiescent CO2 saturated 0.1 M KHCO3 (pH = 7). (d) Potential-dependent CO formation rates for the
various Au catalysts; solutions were stirred at a constant rate during electrolysis runs to prevent product (bubble) buildup on the electrode surface.
Product values are listed in Tables S2 and S3 in the Supporting Information section.
on the Au25 surface with binding energies that ranged between
−0.07 and −0.14 eV (Figure S14). Our calculations predict
similar CO2 binding at finite temperatures, and comparable
binding energies were also found at 300 K. Figure 2d presents
one stable configuration where an oxygen atom of CO2
interacts with three sulfur atoms in the Au25 shell. This specific
adsorption configuration had a calculated binding energy of
0.13 eV and a O−S separation between 3.31 and 3.45 Å. We
found the bound CO2 to be essentially linear, and the
calculated O−C−O angle differs by only 1−2° with respect
to the free CO2. The small perturbation in the internal
structure of the molecule is consistent with CO2 physisorption.
Analysis of the electronic properties indicated that CO2
adsorption only induced minor perturbations to the Au25
density of states (Figure S15) and withdrew less than 0.1
electrons from the Au25 cluster. However, complementary
Bader charge analysis44,45 revealed the major consequence of
CO2 adsorption was charge redistribution within the Au25
cluster. Figure 2e presents the change in Au25 valence electrons
after the adsorption of a CO2 molecule, clearly showing the
depletion of S atom electrons. The three predicted CO2
adsorption configurations consistently depleted electron
density from the S atoms, although charge redistribution
between the core and shell Au atoms varied between the
different Au25−CO2 configurations (Figure S16).
adsorption on partially ligand-protected Au25 clusters, and O2
binding energies of −0.62 and −0.72 eV were found after Au25
was activated by removing one or two ligand groups.21 In the
present case, the Au25 cluster was not artificially activated by
ligand removal. Under these conditions, our work provides
theoretical evidence of weakly bound CO2, and this result is in
line with the observation of reversible CO2-induced spectro-
scopic and electrochemical changes. Finally, the computational
results indicate the major consequence of CO2 adsorption was
the redistribution of charge within the Au25 cluster. This finding
is interesting because Murray and co-workers have suggested
that polar ligand groups can induce oxidation-like effects
through charge redistribution and Au−S bond polariza-
tion.33−35 This hypothesis is reasonable because most Au25
orbitals contain a significant S atom contribution (Figure S17),8
meaning charge redistribution and/or Au−S bond polarization
could disrupt geometric and electronic coupling within the
cluster and produce the experimentally observed optical and
electrochemical changes noted above.4,24
CO2 is not a polar molecule, but it does have a rather strong
quadrupole moment and it can couple with anionic species.46
We suspect that CO2 adsorption was promoted, in part, by an
electrostatic attraction to the negatively charged Au25 cluster.
The electrostatic potential that developed between the
adsorbed CO2 quadrupole and Au25 redistributed charge within
the cluster to produce reversible oxidation-like optical and
electrochemical phenomena. Finally, the Au25 electronic
structure was restored by simply purging the solution with
N2 to desorb the weakly bound CO2.
Direct comparison to literature values is not straightforward
because previous DFT efforts utilized models that were based
on bare or partially ligand-protected clusters. For example,
intermediately bound states with energies of approximately
−1.0 eV were found for CO adsorption on bare Aun (n = 16−
35) clusters.20 A similar bound state was also predicted for O2
Electrochemical Reduction of CO2. The unique inter-
action between Au25 and CO2 provided an exciting opportunity
10240
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243

Journal of the American Chemical Society
Article
to test CO2 reduction with a negatively charged nanocatalyst.
From a practical standpoint, the electrochemical reduction of
aqueous CO2 is attractive because the mechanisms are well
understood and water represents an inexpensive and environ-
mentally benign solvent for scaled-up processes.18,19 Figure 3
compares the electrocatalytic activity of Au25 (∼1 nm)7,8 and
larger Au catalysts for the electrochemical reduction of CO2 in
aqueous 0.1 M KHCO3. Potential-dependent product analysis
(Figure 3b) identified significant and reproducible CO
formation at an onset potential of −0.193 V vs RHE (>95%
CL, n = 3). Electrolysis in N2 purged KHCO3 ruled out
spurious CO evolution from Au25’s organic ligands or the
carbon black support. Remarkably, the onset of CO formation
was within 90 mV of the CO2 → CO formal potential (−0.103
V vs RHE).19 This low overpotential constitutes an
approximate 200−300 mV reduction compared to the case of
the larger Au catalysts in this study (Figure 3d) and those in
previous literature reports.18,19,22
Peak CO production from the Au25 catalyst was found at
−1.0 V vs RHE with approximately 100% Faradaic efficiency
(FE) and a rate 7−700 times higher than 2−5 nm Au
nanoparticles and bulk Au (Figure 3d and Table S2). Our
previous experiments with CB-supported catalysts showed CO
production from the CB support was negligible compared to
the case of Au25 (at −0.9 V CB: 5 × 10−9 mol CO h−1 vs Au25:
6.2 × 10−6 mol CO h−1);47 please note these rates are not
normalized to the electrochemical surface area of Au. FE relates
the amount of reaction product to the total number of electrons
passed through the electrode (Figure S18). For Au25, CO
formation at −1.0 V occurred with approximately 100% FE,
meaning almost every electron injected into the catalyst layer
was utilized for CO2 reduction. Au is known to selectivity
reduce CO2 into CO,18,19 and CO selectivities ranged between
80.8 and 99.6% for Au25, 71.0−96.9% for the larger Au
nanoparticles and 26.9−92.9% for bulk Au, depending on the
applied voltage (Table S3). However, the higher FE of the Au25
cluster enhanced its CO production rate compared to the those
for the other Au catalysts (Figure 3d and Table S2).
Potential-dependent CO2 reduction rates have also been
noted by Chen and Kanan for Sn-oxide catalysts,48 but the
decreased rates beyond −1.0 V likely stem from gaseous
products blocking the Au25 surface (Figure 3b,d). On the basis
of the peak CO production rate of 1.26 mmol cm−2 h−1, we can
estimate a maximum turnover frequency (TOF) of 87 CO
molecules site−1 s−1 for the Au25 catalyst; sites are defined as
accessible Au atoms and determined from electrochemical
surface area measurements.49 This TOF value is approximately
10−100 times higher than those of current state-of-the-art
electrochemical processes23 and is comparable to previous
reports of CO oxidation on ligand-free Aun clusters (n = 4−
19).50
The retention of characteristic optical spectra after CO2
reduction at −1.0 V ruled out Au25 decomposition under
electrocatalytic conditions (Figure S19 of the Supporting
Information).14,51−53 Specifically, the optical spectra contain
contributions from states derived from both core and shell
atoms,4,24,25 and structural deterioration, i.e. ligand desorption
or destruction of the −S−Au−S−Au−S− bonding motif in the
cluster’s shell, should severely alter the absorbance and/or PL
spectra.14,51−53 In contrast to Au25’s ligands, alkanethiol
monolayers can desorb from traditional Au surfaces during
the application of even modest electrochemical potentials (c.a.
−0.4 V vs RHE).54 However, Au25 is not a traditional Au
surface, and the stability of its −SR ligands has been noted
before.51,52 Au25’s stability stems from the unique −S−Au−S−
Au−S− bonding motif in the cluster shell (Figure 1a). This
shell stabilizes the organic ligands and protects the cluster from
deterioration during chemical oxidation or reduction,51,52
catalytic reactions,9−14 and the application of electrochemical
potentials (Figure S19).15
CO and H2 were the only reaction products detected, and
the potential-dependent product distribution in Figure 3b
provides insight into the electrocatalytic mechanism. CO is the
major CO2 reduction product for Au electrodes, and the
reaction proceeds along a two-electron, two-proton pathway
through an adsorbed •CO2 intermediate;18,19 please see
−
equations S1−S6 in the Supporting Information for further
details. In the low potential regime (below −0.5 V), sequential
•
−
proton capture and electron transfer converts adsorbed CO2
into •COOH(ads) before forming CO and water. A sharp
increase in CO production occurred with the onset of H2
evolution at approximately −0.5 V. In this potential range, the
formation of Hads occurs simultaneously with H2 evolution, and
a CO2 → CO pathway based on the direct reduction of •CO2
−
with Hads is likely. CO evolution onset potentials for the larger
Au catalysts were comparable to previous results,18,19,22 and
their equivalent values suggest the presence of similar active
sites. Alternatively, the smaller CO evolution potential of Au25
points to a unique catalytic site capable of promoting the CO2
→ CO reaction closer to the thermodynamic limit. Jin and co-
workers have suggested that Au25 contains a reactive site
capable of promoting both CO bond activation and Hads
formation.13,14 Our computationally identified CO2 adsorption
site is consistent with the site proposed by Jin and co-workers
(Figure S20 of the Supporting Information). Accordingly, we
hypothesize that enhanced electrocatalytic activity stems from
Au25’s anionic charge promoting CO2 adsorption and its unique
reactive site facilitating CO bond activation and Hads
formation.
CONCLUSIONS
■
We have shown spectroscopic, electrochemical, and computa-
tional evidence to support a reversible electronic interaction
between CO2 and atomically precise Au25 clusters. Specifically,
CO2 adsorption redistributed charge within the Au25 cluster to
produce optical and electrochemical changes similar to those
observed during cluster oxidation. The successful correlation of
experimental and computational data is exciting because it
provides insight into the electronic interactions between Au25
clusters and weakly bound adsorbates. Lastly, we have shown
that the Au25 clusters can perform as superior catalysts for the
electrochemical conversion of CO2 into CO. Specifically, Au25
promoted the CO2 → CO reaction within 90 mV of the formal
potential (thermodynamic limit) and showed peak CO
production at −1.0 V vs RHE that was 7−700 times higher
than that for larger Au catalysts and 10−100 times higher than
those for current state-of-the-art processes.
EXPERIMENTAL SECTION
■
Au25 was synthesized as previously reported,55 and all organic solvents
were dried over 3 Å molecular sieves. Absorption spectroscopy was
performed with Perkin-Elmer Lambda 1050 and Agilent 8453
spectrometers. PL spectroscopy was performed with a Jobin Yvon
Horiba Fluorolog 332 spectrometer equipped with a liquid N2 cooled
InGaAs detector. All presented PL spectra are the average of 10 scans
with an excitation energy of 2.78 eV (447 nm) unless otherwise noted,
10241
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243

Journal of the American Chemical Society
Article
and a 2.26 eV (550 nm) long-pass optical filter was placed between the
sample and the detector to prevent the appearance of higher order
excitation peaks in the PL spectrum. Au25 solutions in DMF were
placed in sealable cuvettes and initially purged overnight with N2,
bubbled with ultrahigh purity CO2 for 1.5 h, and finally purged with
N2 again for 1.5 h; experiments with p-xylene as a solvent used an
initial 1.5 h N2 purge to reduce solvent evaporation. During gas
exposure experiments, the PL spectra were normalized to the
absorbance peak area at λex = 2.78 eV (i.e., absorbance feature b);
please see Figure S6 in the Supporting Information for further details.
Electrochemical measurements were performed with a Biologic SP-
150 potentiostat equipped with a low-current option. Nonaqueous
electrochemistry was conducted in DMF + 0.1 M tetrabutylammo-
nium perchlorate (TBAP) with Pt wire working and counter
electrodes and a nonaqueous Ag/Ag+ reference electrode (0.01 M
AgNO3 + 0.1 M TBAP in CH3CN); potentials were calibrated into the
standard hydrogen electrode (SHE) scale using ferrocene (Fc/Fc+ =
0.7112 V vs SHE in DMF + 0.1 M TBAP).56 In situ
spectroelectrochemical measurements were conducted with a
commercially available quartz cell from BASi.
Aqueous electrochemical experiments were conducted in 0.1 M
KHCO3 solutions prepared with ultrapure water (≥18.0 MΩ·cm).
Catalyst inks were prepared by sonicating acetone-solvated Au25 into a
mixture of methanol, Vulcan XC-72R carbon black, and Nafion;
catalyst inks were also prepared in an identical manner using
commercially available 2 and 5 nm Au nanoparticles from BBInterna-
tional. The catalyst ink was deposited onto a glassy carbon (GC)
working electrode and allowed to dry in air. Electrochemical
experiments with bulk Au were conducted with a 99.99% Au wire
that was cleaned in 1.0 M H2SO4, rinsed with ultrapure water, and
used as a working electrode. A Pt wire counter electrode and a
conventional Ag/AgCl reference electrode (3.0 M NaCl) were used to
complete the aqueous electrochemical setup. The Ag/AgCl reference
electrode was regularly calibrated against a commercially available
Hydroflex reversible hydrogen electrode; E° = 0.000 V − 0.059pH (ref
40). The solution pH was measured after each experiment to convert
the Ag/AgCl reference electrode potentials into the RHE scale; typical
pH values for N2 and CO2 saturated 0.1 M KHCO3 were 7.0 and 9.0.
The linear sweep voltammograms presented in Figure 3a and c were
conducted in quiescent (unstirred) solutions.
Electrochemical CO2 reduction reactions were performed in a
gastight, two-compartment H-cell. The working and counter electro-
des were housed in different compartments that were separated by a
0.1778 mm (0.007 in.) thick Nafion 117 cation exchange membrane;
this experimental configuration prevented CO2 reduction products
from being oxidized at the counter electrode. During CO2 electrolysis
runs, a particular voltage was applied for 1 h and the solution was
stirred at a constant rate to prevent product (bubble) buildup on the
electrode surface. After 1 h of electrolysis, products were analyzed with
a Perkin-Elmer Clarus 600 gas chromatograph equipped with TCD
and FID detectors. Multiple blank injections accompanied each
electrolysis run to establish an instrumental baseline for the CO peak
area, and we have defined the CO formation onset potential as the
potential where CO formation was significant above the baseline at a
95% confidence level. Electrochemical currents and product formation
rates were normalized to the catalyst electrochemical surface area
(ECSA), as measured by integrating the so-called Au oxide stripping
peak in N2 purged 0.1 M KHCO3 (ref 49); this approach has
previously been used to measure the ECSA of Aun cluster
electrocatalysts (n = 11−140) in aqueous media.15 We found the
Au25 clusters retained characteristic optical absorbance and PL spectral
features after oxide formation/stripping (Figure S19 of the Supporting
Information).
energy.59 The electron−ion interaction was described by the projector-
augmented wave (PAW) method.60
−1
The model for the Au25(SCH3)18 cluster was constructed using
−1
the DFT relaxed Au25(SH)18 structure determined by Aikens as a
starting point.42 The terminating H atoms were then replaced by CH3
to give rise to a Au25(SCH3)18−1 structure. The cluster was placed in a
cubic box with dimensions of 24 Å sides, and a uniform compensating
background charge was assumed. Various binding configurations of
CO2 adsorbate on the cluster were explored, and Γ-point calculations
were carried out with an energy cutoff of 600 eV. Geometries were
relaxed using a quasi-Newton variable metric algorithm until the
atomic forces were less than 0.03 eV/Å. The CO2 binding energy was
calculated using the expression Ebind = ECO +cluster − (Ecluster + ECO ).
2
2
Here ECO +cluster is the total energy of the relaxed CO2−Au25 cluster
2
system, while Ecluster and ECO are the total energies of the relaxed bare
2
cluster and the gas phase CO2, respectively. On the basis of this
convention, a negative Ebind corresponds to an exothermic reaction
(stable adsorption configuration).
ASSOCIATED CONTENT
■
S
* Supporting Information
Additional experimental and computational data. This material
is available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
Douglas.Kauffman@CONTR.NETL.DOE.GOV
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors thank Professor J. P. Lewis for fruitful discussions.
This technical effort was performed in support of the NETL’s
ongoing research in CO2 utilization under RES Contract DE-
FE0004000. R.J. acknowledges financial support from the Air
Force Office of Scientific Research under AFOSR Award No.
FA9550-11-1-9999 (FA9550-11-1-0147). Reference in this
work to any specific commercial product is to facilitate
understanding and does not necessarily imply endorsement
by the United States Department of Energy. This report was
prepared as an account of work sponsored by an agency of the
United States Government. Neither the United States Govern-
ment nor any agency thereof, nor any of their employees,
makes any warranty, express or implied, or assumes any legal
liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process
disclosed, or represents that its use would not infringe privately
owned rights. Reference herein to any specific commercial
product, process, or service by trade name, trademark,
manufacturer, or otherwise does not necessarily constitute or
imply its endorsement, recommendation, or favoring by the
United States Government or any agency thereof. The views
and opinions of authors expressed herein do not necessarily
state or reflect those of the United States Government or any
agency thereof.
REFERENCES
■
Computational Methodolgy. The adsorption of CO2 on the
(1) Daniel, M.-C.; Astruc, D. Chem. Rev. 2004, 104, 293.
(2) Sardar, R.; Funston, A. M.; Mulvaney, P.; Murray, R. W.
Langmuir 2009, 25, 13840.
−1
model Au25(SCH3)18 cluster was examined using the Vienna ab
initio simulation package (VASP) which employs density functional
theory with a basis set constructed from plane waves.57,58 The
Perdew−Burke−Enzerhoff (PBE) generalized gradient approximation
(GGA) functional was employed to calculate the exchange-correlation
(3) Rodriguez, P.; Kwon, Y.; Koper, M. T. M. Nature Chem. 2012, 4,
177.
(4) Jin, R. Nanoscale 2010, 2, 343.
10242
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243

Journal of the American Chemical Society
Article
(5) Parker, J. F.; Fields-Zinna, C. A.; Murray, R. W. Acc. Chem. Res.
2010, 43, 1289.
(6) Jupally, V. R.; Kota, R.; Dornshuld, E. V.; Mattern, D. L.;
Tschumper, G. S.; Jiang, D.-e.; Dass, A. J. Am. Chem. Soc. 2011, 133,
20258.
(7) Heaven, M. W.; Dass, A.; White, P. S.; Holt, K. M.; Murray, R. W.
J. Am. Chem. Soc. 2008, 130, 3754.
(8) Zhu, M.; Aikens, C. M.; Hollander, F. J.; Schatz, G. C.; Jin, R. J.
Am. Chem. Soc. 2008, 130, 5883.
(9) Zhu, Y.; Qian, H.; Jin, R. J. Mater. Chem. 2011, 21, 6793.
(10) Liu, Y.; Tsunoyama, H.; Akita, T.; Tsukuda, T. Chem. Commun.
2010, 46, 550.
(39) Venzo, A.; Antonello, S.; Gascon
R. D.; Perera, N. V.; Sousa, A.; Zamuner, M.; Zanella, A.; Maran, F.
Anal. Chem. 2011, 83, 6355.
(40) Bard, A. J.; Faulkner, L. R. Electrochemical methods: fundamentals
and applications, 2nd ed.; John Wiley & Sons: New York, 2001.
(41) Devadas, M. S.; Kwak, K.; Park, J.-W.; Choi, J.-H.; Jun, C.-H.;
Sinn, E.; Ramakrishna, G.; Lee, D. J. Phys. Chem. Lett. 2010, 1, 1497.
(42) Aikens, C. M. J. Phys. Chem. C 2008, 112, 19797.
(43) Pei, Y.; Pal, R.; Liu, C.; Gao, Y.; Zhang, Z.; Zeng, X. C. J. Am.
Chem. Soc. 2012, 134, 3015.
́
, J. A.; Guryanov, I.; Leapman,
(44) Henkelman, G.; Arnaldsson, A.; Jon
2006, 36, 354.
́
sson, H. Comput. Mater. Sci.
(45) Sanville, E.; Kenny, S. D.; Smith, R.; Henkelman, G. J. Comput.
Chem. 2007, 28, 899.
(46) Arnold, D. W.; Bradforth, S. E.; Kim, E. H.; Neumark, D. M. J.
Chem. Phys. 2008, 102, 3510.
(47) Kauffman, D. R.; Ohodnicki, P. R.; Kail, B. W.; Matranga, C. J.
Phys. Chem. Lett 2011, 2, 2038.
(48) Chen, Y.; Kanan, M. W. J. Am. Chem. Soc. 2012, 134, 1986.
(49) Trasatti, S.; Petrii, O. A. Pure Appl. Chem. 1991, 63, 711.
(50) Wallace, W. T.; Wetten, R. L. J. Am. Chem. Soc. 2002, 124, 7499.
(51) Negishi, Y.; Chaki, N. K.; Shichibu, Y.; Whetten, R. L.; Tsukuda,
T. J. Am. Chem. Soc. 2007, 129, 11322.
(52) Wu, Z.; Jin, R. ACS Nano 2009, 3, 2036.
(53) Shichibu, Y.; Negishi, Y.; Tsunoyama, H.; Kanehara, M. Small
2007, 3, 835.
(11) Turner, M.; Golovko, V. B.; Vaughan, O. P. H.; Abdulkin, P.;
Berenuer-Murcia, A.; Tikhov, M. S.; Johnson, B. F. G.; Lambert, R. M.
Nature 2008, 454, 981.
(12) Liu, Y.; Tsunoyama, H.; Akita, T.; Xie, S.; Tsukuda, T. ACS
Catal. 2011, 1, 2.
(13) Zhu, Y.; Qian, H.; Drake, B. A.; Jin, R. Angew. Chem., Int. Ed.
2010, 49, 1295.
(14) Zhu, Y.; Wu, Z.; Gayathri, C.; Qian, H.; Gil, R. R.; Jin, R. J.
Catal. 2010, 271, 155.
(15) Chen, W.; Chen, S. Angew. Chem., Int. Ed. 2009, 48, 4386.
(16) Arakawa, H.; Aresta, M.; Armor, J. N.; Barteau, M. A.; Beckman,
E. J.; Bell, A. T.; Bercaw, J. E.; Creutz, C.; Dinjus, E.; Dixon, D. A.;
Domen, K.; DuBois, D. L.; Eckert, J.; Fujita, E.; Gibson, D. H.;
Goddard, W. A.; Goodman, D. W.; Keller, J.; Kubas, G. J.; Kung, H.
H.; Lyons, J. E.; Manzer, L. E.; Marks, T. J.; Morokuma, K.; Nicholas,
K. M.; Periana, R.; Que, L.; Rostrup-Nielson, J.; Sachtler, W. M. H.;
Schmidt, L. D.; Sen, A.; Somorjai, G. A.; Stair, P. C.; Stults, B. R.;
Tumas, W. Chem. Rev. 2001, 101, 953.
(17) Peterson, A. A.; Nørskov, J. K. J. Phys. Chem. Lett. 2012, 3, 251.
(18) Hori, Y. Electrochemical CO2 reduction on metal electrodes;
Springer: New York, 2008; Vol. 42.
(19) Gattrell, M.; Gupta, N.; Co, A. J. Electroanal. Chem. 2006, 594,
1.
(54) Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.;
Whitesides, G. M. Chem. Rev. 2005, 105, 1103.
(55) Zhu, M.; Lanni, E.; Garg, N.; Bier, M. E.; Jin, R. J. Am. Chem.
Soc. 2008, 130, 1138.
(56) Connelly, N. G.; Geiger, W. E. Chem. Rev. 1996, 96, 877.
(57) Kresse, G.; Furthmuller, J. Phys. Rev. B 1996, 54, 11169.
̈
(58) Hafner, J. J. Comput. Chem. 2008, 29, 2044.
(59) Perdew, J. P.; Burke, K.; Wang, Y. Phys. Rev. B 1996, 54, 16533.
(60) Kresse, G.; Joubert, D. Phys. Rev. B 1999, 59, 1758.
(20) Gao, Y.; Shao, N.; Pei, Y.; Chen, Z.; Zeng, X. C. ACS Nano
2011, 5, 7818.
(21) Lopez-Acevedo, O.; Kacprzak, K. A.; Akola, J.; Hakkinen, H.
̈
Nature Chem. 2010, 2, 329.
(22) Hori, Y.; Murata, A.; Katsuhei, K.; Suzuki, S. J. Chem. Soc., Chem.
Commun. 1987, 728.
(23) 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.
(24) Devadas, M. S.; Bairu, S.; Qian, H.; Sinn, E.; Jin, R.;
Ramakrishna, G. J. Phys. Chem. Lett. 2011, 2, 2752.
(25) Qian, H.; Sfeir, M. Y.; Jin, R. J. Phys. Chem. C 2010, 114, 19935.
(26) Please note that some early reports misidentified anionic Au25 as
neutral Au38.
(27) Lee, D.; Donkers, R. L.; Wang, G.; Harper, A. S.; Murray, R. W.
J. Am. Chem. Soc. 2004, 126, 6193.
(28) Aikens, C. M. J. Phys. Chem. Lett. 2010, 1, 2594.
(29) Zhu, M.; Eckenhoff, W. T.; Pintauer, T.; Jin, R. J. Phys. Chem. C
2008, 112, 14221.
(30) Wu, Z.; Jin, R. Nano Lett. 2010, 10, 2568.
(31) Choi, J.-P.; Fields-Zinna, C. A.; Stiles, R. L.; Balasubramanian,
R.; Dougals, A. D.; Crowe, M. C.; Murray, R. W. J. Phys. Chem. C
2010, 114, 15890.
(32) Liu, Z.; Zhu, M.; Meng, X.; Xu, G.; Jin, R. J. Phys. Chem. Lett.
2011, 2, 2104.
(33) Wang, G.; Guo, R.; Kalyuzhny, G.; Choi, J.-P.; Murray, R. W. J.
Phys. Chem. B 2006, 110, 20282.
(34) Guo, R.; Murray, R. W. J. Am. Chem. Soc. 2005, 127, 12140.
(35) Parker, J. F.; Kacprzak, K. A.; Lopez-Acevedo, O.; Hakkinen, H.;
̈
Murray, R. W. J. Phys. Chem. C 2010, 114, 8276.
́
(36) García-Raya, D.; Madueno, R.; Blazquez, M.; Pineda, T. J. Phys.
̃
Chem. C 2009, 113, 8756.
(37) Freund, H.-J.; Roberts, M. W. Surf. Sci. Rep. 1996, 25, 225.
(38) Gennaro, A.; Isse, A. A.; Vianello, E. J. Electroanal. Chem. 1990,
289, 203.
10243
dx.doi.org/10.1021/ja303259q | J. Am. Chem. Soc. 2012, 134, 10237−10243