Selective photoelectrochemical reduction of aqueous CO2 to CO by solvated electrons

Article abstract of DOI:10.1002/anie.201404328

Reduction of CO₂ by direct one-electron activation is extraordinarily difficult because of the -1.9 V reduction potential of CO ₂. Demonstrated herein is reduction of aqueous CO₂ to CO with greater than 90 % product selectivity by direct one-electron reduction to CO₂^.- by solvated electrons. Illumination of inexpensive diamond substrates with UV light leads to the emission of electrons directly into water, where they form solvated electrons and induce reduction of CO ₂ to CO₂^.-. Studies using diamond were supported by studies using aqueous iodide ion (I^-), a chemical source of solvated electrons. Both sources produced CO with high selectivity and minimal formation of H₂. The ability to initiate reduction reactions by emitting electrons directly into solution without surface adsorption enables new pathways which are not accessible using conventional electrochemical or photochemical processes.

Full text of DOI:10.1002/anie.201404328

.
Angewandte
Communications
DOI: 10.1002/anie.201404328
Photochemistry
Hot Paper
Selective Photoelectrochemical Reduction of Aqueous
CO2 to CO by Solvated Electrons**
Linghong Zhang, Di Zhu, Gilbert M. Nathanson, and Robert J. Hamers*
Angewandte
Chemie
9746
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 9746 –9750

Angewandte
Chemie
Abstract: Reduction of CO2 by direct one-electron activation is
extraordinarily difficult because of the ꢀ1.9 V reduction
potential of CO2. Demonstrated herein is reduction of aqueous
CO2 to CO with greater than 90% product selectivity by direct
one-electron reduction to CO2Cꢀ by solvated electrons. Illumi-
nation of inexpensive diamond substrates with UV light leads
to the emission of electrons directly into water, where they form
solvated electrons and induce reduction of CO2 to CO2Cꢀ.
Studies using diamond were supported by studies using
aqueous iodide ion (Iꢀ), a chemical source of solvated
electrons. Both sources produced CO with high selectivity
and minimal formation of H2. The ability to initiate reduction
reactions by emitting electrons directly into solution without
surface adsorption enables new pathways which are not
accessible using conventional electrochemical or photochem-
ical processes.
reduction on diamond has been inferred previously,[5] but
until our recent work on N2 reduction[6] the direct role of
solvated electrons in important reduction reactions had not
been established. Unlike conventional electrochemical or
photochemical approaches where reactants must adsorb to
a surface, this approach does not require surface adsorption
and is instead based on direct emission of electrons into the
aqueous reactants.
Figure 1 shows an energy level diagram for diamond[6,7]
and some redox reactions of importance.[3,7,8] When the
surface of diamond is terminated with H atoms its conduction
The development of methods to efficiently convert atmos-
pheric CO2 into chemical feedstocks or other environmentally
benign products is one of the most pressing problems today.
Attempts to reduce CO2 under aqueous conditions by photo-
chemical and electrochemical methods typically yield poor
selectivity and poor efficiency because of competing reduc-
tion of H+ to H2, as well as the many possible coupling
reactions which can occur on the electrode surface.[1] The vast
majority of work on CO2 photoreduction has focused on
proton-coupled electron-transfer (PCET) reactions,[2] which
usually require the participation of multiple electrons and
protons as well as long-lived intermediates, and are therefore
inefficient. A much more direct and more efficient pathway
would be the one-electron reduction of CO2 into CO2Cꢀ. This
pathway has been little studied photochemically because the
one-electron reduction, eꢀ + CO2!CO2Cꢀ, has a standard
redox potential (E0) of ꢀ1.9 V versus the standard hydrogen
electrode (SHE).[1] This value lies above the conduction band
edge of nearly all semiconductors,[3] thus rendering them
unable to initiate the one-electron reduction process.[1,4]
Herein, we demonstrate a new approach to CO2 reduction
using aqueous solvated electrons, and show that solvated
electrons produced by illuminating diamond surfaces selec-
tively reduce CO2 into CO with high product selectivity and
minimal proton reduction. To enable this new pathway, we
build on recent work showing that inexpensive, commercial-
grade diamond samples can act as facile solid-state sources of
solvated electrons in water when illuminated with ultraviolet
light. The possible role of solvated electrons in photochemical
Figure 1. a) Energy diagram of H-terminated diamond compared with
relevant redox potentials on absolute energy scale (left) and electro-
chemical energy scale (right).[8,11] E0 values are shown for production
of solvated electron (eꢀ(aq)), hydrogen atom (HC) and CO2 radical anion
(CO2Cꢀ). (E(SO42ꢀ/SO32ꢀ) for counter-electrode is shown for pH 3.2.
b) Solvated-electron yield versus wavelength from transient absorption
measurements, normalized to yield at 210 nm.
band typically lies at ꢀ5.2 V versus SHE, thus placing it above
the one-electron reduction potential of CO2, and even higher
than the energy of a free electron in space (the vacuum
level).[9] When H-terminated diamond is illuminated with
above-bandgap (Egap = 5.5 eVor l < 225 nm) light, absorption
of light by the diamond (1) excites electrons to the conduction
band. These electrons are readily ejected into the adjacent
aqueous phase (2), where they quickly relax (3) to form
solvated electrons[6,10] and other high-energy species.
We can directly observe solvated electrons by measuring
the transient absorption of a l = 705 nm laser beam grazing
a diamond surface, which is then exposed to a brief pulse of
ultraviolet light, as described in the Supporting Information.
Figure 1b shows measurements of the solvated electron yield
relative to the value at l = 210 nm. These data show that the
electron yield is much greater using above-bandgap excitation
(l < 225 nm) than that at longer wavelengths, thus confirming
that absorption of light in the bulk diamond and subsequent
electron emission from the diamond conduction band is more
[*] L. Zhang, D. Zhu, Prof. G. M. Nathanson, Prof. R. J. Hamers
Department of Chemistry, University of Wisconsin-Madison
1101 University Avenue, Madison, WI 53706 (USA)
E-mail: rjhamers@wisc.edu
Homepage: http://hamers.chem.wisc.edu
[**] This work was supported by the Air Force Office of Scientific
Research under AFOSR Award FA9550-12-1-0063. L.Z. thanks Fei
Meng for assistance of X-ray diffraction measurements, Steve
Strebel, LeRoy Dobson, and Christopher McSweeny at Wisconsin
State Laboratory of Hygiene for the assistance of ion chromatog-
raphy measurements.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201404328.
Angew. Chem. Int. Ed. 2014, 53, 9746 –9750
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 9747

.
Angewandte
Communications
efficient at creating solvated electrons compared with other
possible sub-bandgap excitation processes.
CO is produced in the absence of UV light (Figure 2; No
illumination) or in an argon-purged solution (Figure 2; No
CO2). A very small amount of CO is observed when a CO2-
saturated solution is illuminated without the diamond sample
(Figure 2; No diamond). This CO arises because CO2 has
a very small photochemical reactivity at short wavelengths.[13]
All features in Figure 2 arise from CO and CO2. The positions
and intensities are reproduced almost exactly using known
spectral constants (see the Supporting Information).[14]
We calibrated our apparatus to establish the absolute CO
yield and determined that a 1 cm2 diamond sample typically
produces about 300 mg of CO, thus corresponding to a con-
centration of CO in the gas phase of about 100 parts per
million (ppm). We attempted to detect H2 by GC and found
that the concentration of H2 product lies below our detection
limit of 5 ppm. This result demonstrates that reduction of H+
to H2 does not compete with reduction of CO2 into CO. IC
measurements of the liquid phase revealed only a small
amount (28 mg) of the formate ion. We also attempted to
detect other liquid-phase products using both GC-MS and
NMR methods, but the amounts were below the detection
limit of less than 1 mglꢀ1 (corresponding to a ca. 80 mg of
product). Our results indicate that the reduction of CO2
occurs with high selectivity for CO, and also with high
(> 95%) selectivity for CO2 reduction over proton reduction.
We evaluated whether the CO produced arose entirely
from excitation across the diamond bandgap or whether sub-
bandgap could induce CO2 reduction, by comparing the CO
production using broad-spectrum light with CO production
when a l = 225 nm cutoff filter was inserted (thereby passing
only sub-bandgap light). This experiment (see the Supporting
Information) showed that no CO was produced using
excitation wavelengths greater than l = 225 nm. This result
is consistent with the solvated electron yield results of
Figure 1b.
To establish how these solvated electrons induce reduc-
tion of CO2 into CO, we conducted experiments using
inexpensive “electrochemical grade” boron-doped diamond
(BDD) substrates (Element Six, Inc., resistivity
< 0.05 ohmcmꢀ1
) supported within a two-compartment
quartz electrochemical cell. As shown previously,[6] the use
of conductive, boron-doped diamond facilitates electron
emission by allowing diamond to act as the electron emitter,
while a platinum counter electrode can be used to perform an
oxidation reaction to maintain charge neutrality. The exper-
imental apparatus and procedures are described in the
Supporting Information. Briefly, the diamond sample in
a Na2SO4 supporting electrolyte is connected to a platinum
counter electrode immersed in a Na2SO4/Na2SO3 solution.
Sulfite acts as a sacrificial hole scavenger by the oxidation
half-reaction SO32ꢀ + 2OHꢀ!SO42ꢀ + H2O + 2eꢀ, (E =
ꢀ0.27 V under reaction conditions). The quartz cell is
housed within a reactor (Parr, Inc.) with a fused silica
window, pressurized to 2.5 MPa (~ 25 times atmospheric
pressure) with CO2, and the diamond sample is illuminated
with broadband light from a HgXe lamp. Under these
reaction conditions the pH is about 3.2.[12] After reaction
the gas-phase headspace is released into a gas cell and
analyzed using Fourier-transform infrared spectroscopy
(FTIR). The liquid is analyzed using ion chromatography
(IC), gas chromatography mass spectroscopy (GC-MS), and
nuclear magnetic resonance (NMR) spectroscopy.
Figure 2 shows FTIR spectra of the gas-phase headspace
after CO2 photoreduction by illumination of a diamond
sample. The spectrum from illuminated diamond in CO2/
To maintain charge neutrality, every electron emitted
from the diamond must be accompanied by an oxidation
reaction of the valence-band holes. In a previous study using
identical samples[6] we found that the valence-band edge lies
at + 0.3 V versus SHE[15] (Figure 1), and is in agreement with
earlier studies.[7] The oxidation of diamond by C + H2O!
CO + 2H+ + 2eꢀ has an associated reduction potential of
E
red = 0.71 V at pH 3.2 and should therefore not occur.[16] To
prove that our CO product comes from CO2 and not from the
diamond substrate or other sources, we used isotopically
labeled 13CO2 as the source to perform the reaction. Figure 3
shows FTIR spectra of the gas-phase products starting from
natural-abundance CO2 (which we refer to as 12CO2) and
starting from 13CO2. Also shown are simulations showing the
spectra expected from pure 12CO and 13CO based on the
HITRAN database.[14] The experimental FTIR spectra show
clear rotational-vibrational lines expected for isotopically
pure products and exactly match the simulated spectra. Using
12CO2 yields the 12CO product with its band center near
2140 cmꢀ1, and using 13CO2 produces shifted bands centered
near 2100 cmꢀ1 for 13CO. In each case the shifted spikes and
pattern between 2100 cmꢀ1 and 2000 cmꢀ1 are from CO2 (also
simulated in the Supporting Information). These isotope data
prove that the CO product arises from the photoreduction of
Figure 2. FTIR spectra of gaseous headspace demonstrating reduction
of CO2 to CO by illuminated diamond, along with control samples.
water shows distinct absorption lines of CO. The features
shown here are the P(DJ=ꢀ1) and R(DJ=+1) rotational
branches associated with excitation from the lowest (v = 0) to
the first excited (v = 1) vibrational state of CO. No significant
9748 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 9746 –9750

Angewandte
Chemie
CO2Cꢀ is approximately 0.4 eV lower in energy than [HC] (see
Figure 1), HC atoms never accumulate and therefore the
bimolecular reaction HC + HC!H2 (an undesirable side-reac-
tion) is unimportant. Our observation of high selectivity for
CO reduction over proton reduction with no detectable H2
product strongly supports this model and highlights the
importance of CO2Cꢀ as a reservoir for electrons.
Electrochemical studies have hypothesized that the one-
electron reduction of surface-adsorbed CO2 to form adsorbed
CO2Cꢀ(ads) might occur at high overpotentials and then undergo
additional bimolecular coupling or proton-coupled electron-
transfer reactions to form CO and formate.[2b,e,18a,c,d,22] Hori
et al.[18a] hypothesized that protonation of CO2Cꢀ
could
(ads)
Figure 3. FTIR spectra showing CO produced from diamond illumi-
nated in contact with isotopically labeled CO2. The data show that
13CO and 12CO are produced from 13CO2 and 12CO2 starting materials,
respectively. Also shown are calculated spectra for 13CO and 12CO.
form adsorbed HOCO and then be reduced to OH + CO.
Saveant et al.[18c,22] proposed that two CO2Cꢀ
anions could
(ads)
form a transient intermediate which disproportionates by
2ꢀ
2CO2Cꢀ!CO + CO3 in a solution with low proton concen-
trations.[22] On a surface, pathways involving bimolecular
coupling are important because of the locally high concen-
trations of accumulated species, but such reactions are less
likely to occur in homogeneous solution. However, CO2Cꢀ in
large water clusters photodissociates into CO and Oꢀ when
illuminated with light at l = 266 nm or l = 355 nm.[23] This
result suggests that CO production in our studies may involve
initial formation of CO2Cꢀ and subsequent photodissociation
of CO2Cꢀ to produce the final CO product. To ensure that our
CO product did not come from photodissociation of formate
(a common CO2 reduction product),[18a] we tested the photo-
chemical activity of formate-containing solutions. These
experiments yielded very little CO (see Figure S9 in the
Supporting Information), thus showing that CO comes from
the CO2Cꢀ intermediate.
While further studies will undoubtedly be needed to fully
resolve the detailed mechanistic pathways, our results dem-
onstrate that CO2 can be efficiently and selectively reduced to
CO using solvated electrons. The unique electronic structure
of H-terminated diamond allows it to be used as a convenient
solid-state source of electrons at the conduction-band energy.
The high energy of these electrons enables formation of
solvated electrons and the direct one-electron reduction of
CO2 to CO2Cꢀ in solution, a reaction step not accessible with
other water-stable semiconductor photocatalysts. The rapid
reaction of solvated electrons with CO2 at 2.5 MPa allows
CO2 reduction to outcompete proton reduction even at
pH 3.2, thus resulting in selective reduction of CO2 with
only minimal production of H2 as a side product. This work
represents a new approach to catalysis by bringing electrons
as a reagent directly into reactant solutions, thus enabling new
mechanistic pathways which can provide unique and selective
product distributions.
CO2 introduced as a gas-phase reactant and not from
diamond or other sources.
The iodide ion (Iꢀ) is also known to generate solvated
electrons upon UV exposure.[17] As a comparison, we placed
a 10 mm KI solution into the pressure vessel, pressurized with
CO2 to 2.5 MPa, and exposed the sample to UV light for
16 hours. Figure S8 in the Supporting Information shows
FTIR spectra of the resulting gas-phase headspace. Again CO
is the major product with very little formation of H2.
Despite much effort, the mechanistic pathways for photo-
chemical CO2 reduction remain only partially understood.[1,18]
Nearly all prior photochemical studies have focused on PCET
reactions such as CO2 + 2H+ + 2eꢀ!CO + H2O.[1] However,
PCET reactions require adsorption of both CO2 and H+ and
multi-electron transfers, thus leading to complex pathways
and multiple products, including H2.[2a,b,18a,b,19] The high
selectivity we observe with almost no H2 suggests the solvated
electron-induced reduction of CO2 occurs by a unique path-
way.
A notable difference between our work and all previous
work on electrocatalytic and photocatalytic reduction of CO2
is that the diamond electrode introduces electrons directly
into the aqueous phase and it is therefore unnecessary (even
undesirable) for reactants, intermediates, or products to bind
to the surface. Instead, once the electrons are emitted,
subsequent reactions occur in homogeneous solution. As
depicted in Figure 1, electrons emitted from the diamond
conduction band into water could react directly with CO2 to
form CO2Cꢀ in free solution (not adsorbed on a surface) or
could relax first and produce solvated electrons eꢀ(aq). These
then react with CO2 to form CO2·ꢀ or with H+ to form neutral
hydrogen atoms (HC). The reaction eꢀ(aq) + CO2!CO2Cꢀ has
a rate constant of k = 7.7 ꢀ 109 mꢀ1 sꢀ1, while the competing
reaction with protons eꢀ(aq) + H+!HC has a rate constant of
k = 2.3 ꢀ 1010 mꢀ1 sꢀ1.[20] At a CO2 pressure of 2.5 MPa and
a temperature of 300 K, [CO2(aq)] = 0.8m[21] and the pH is
3.2.[12] Since [CO2(aq)](0.8m) @ [H+](8ꢀ10ꢀ4 m), the vast major-
ity of solvated electrons react by eꢀ(aq) + CO2!CO2Cꢀ. These
rate equations predict that [HC] will be small and that solvated
electrons predominantly react with CO2 to form CO2Cꢀ. Since
Received: April 15, 2014
Revised: June 6, 2014
Published online: July 14, 2014
Keywords: carbon dioxide · diamond · photochemistry ·
.
reduction · surface chemistry
Angew. Chem. Int. Ed. 2014, 53, 9746 –9750
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 9749

.
Angewandte
Communications
[13] J. A. Schmidt, M. S. Johnson, R. Schinke, Proc. Natl. Acad. Sci.
USA 2013, 110, 17691 – 17696.
[14] L. S. Rothman, C. P. Rinsland, A. Goldman, S. T. Massie, D. P.
Edwards, J.-M. Flaud, A. Perrin, C. Camy-Peyret, V. Dana, J.-Y.
Mandin, J. Schroeder, A. McCann, J. Quant. Spectrosc. Radiat.
Transfer 2010, 111, 1568 – 1613.
[15] D. Takeuchi, H. Kato, G. S. Ri, T. Yamada, P. R. Vinod, D.
Hwang, C. E. Nebel, H. Okushi, S. Yamasaki, Appl. Phys. Lett.
2005, 86, 152103.
[16] a) H.-S. Choo, T. Kinumoto, S.-K. Jeong, Y. Iriyama, T. Abe, Z.
Ogumi, J. Electrochem. Soc. 2007, 154, B1017; b) S. G. Bratsch, J.
Phys. Chem. Ref. Data 1989, 18, 1 – 21.
[1] B. Kumar, M. Llorente, J. Froehlich, T. Dang, A. Sathrum, C. P.
Kubiak, Annu. Rev. Phys. Chem. 2012, 63, 541.
[2] a) H. Fujiwara, H. Hosokawa, K. Murakoshi, Y. Wada, S.
Yanagida, Langmuir 1998, 14, 5154 – 5159; b) T. Inoue, A.
Fujishima, S. Konishi, K. Honda, Nature 1979, 277, 637 – 638;
c) B. A. Rosen, A. Salehi-Khojin, M. R. Thorson, W. Zhu, D. T.
Whipple, P. J. Kenis, R. I. Masel, Science 2011, 334, 643 – 644;
d) C. Costentin, S. Drouet, M. Robert, J. M. Saveant, Science
2012, 338, 90 – 94; e) R. Angamuthu, P. Byers, M. Lutz, A. L.
Spek, E. Bouwman, Science 2010, 327, 313 – 315.
[3] H. A. Schwarz, R. W. Dodson, J. Phys. Chem. 1989, 93, 409 – 414.
[4] A. L. Linsebigler, G. Q. Lu, J. T. Yates, Chem. Rev. 1995, 95,
735 – 758.
[17] H. Matsumura, F. Mafune, T. Kondow, J. Phys. Chem. 1995, 99,
5861 – 5864.
[5] L. Boonma, T. Yano, D. A. Tryk, K. Hashimoto, A. Fujishima, J.
Electrochem. Soc. 1997, 144, L142 – L145.
[6] D. Zhu, L. H. Zhang, R. E. Ruther, R. J. Hamers, Nat. Mater.
2013, 12, 836 – 841.
[7] T. N. Rao, D. A. Tryk, K. Hashimoto, A. Fujishima, J. Electro-
chem. Soc. 1999, 146, 680 – 684.
[8] D. D. M. Wayner, V. D. Parker, Acc. Chem. Res. 1993, 26, 287 –
294.
[9] F. J. Himpsel, J. A. Knapp, J. A. Vanvechten, D. E. Eastman,
Phys. Rev. B 1979, 20, 624 – 627.
[10] J. R. Christianson, D. Zhu, R. J. Hamers, J. R. Schmidt, J. Phys.
Chem. B 2014, 118, 195 – 203.
[18] a) Y. Hori, H. Wakebe, T. Tsukamoto, O. Koga, Electrochim.
Acta 1994, 39, 1833 – 1839; b) Y. Izumi, Coord. Chem. Rev. 2013,
257, 171 – 186; c) C. Costentin, M. Robert, J. M. Saveant, Chem.
Soc. Rev. 2013, 42, 2423 – 2436; d) M. Gattrell, N. Gupta, A. Co,
J. Electroanal. Chem. 2006, 594, 1 – 19.
[19] N. Hoshi, M. Kato, Y. Hori, J. Electroanal. Chem. 1997, 440, 283 –
286.
[20] G. V. Buxton, C. L. Greenstock, W. P. Helman, A. B. Ross, J.
Phys. Chem. Ref. Data 1988, 17, 513 – 886.
[21] G. V. L. Wiebe, J. Am. Chem. Soc. 1940, 62, 815 – 817.
[22] C. Amatore, J. M. Saveant, J. Am. Chem. Soc. 1981, 103, 5021 –
5023.
[11] P. Wardman, J. Phys. Chem. Ref. Data 1989, 18, 1637 – 1755.
[12] C. Peng, J. P. Crawshaw, G. C. Maitland, J. P. M. Trusler, D. Vega-
Maza, J. Supercrit. Fluids 2013, 82, 129 – 137.
[23] a) T. Habteyes, L. Velarde, A. Sanov, J. Chem. Phys. 2007, 126,
154301; b) L. Velarde, T. Habteyes, A. Sanov, J. Chem. Phys.
2006, 125, 114303.
9750 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 9746 –9750