You are on page 1of 19

Journal of

Electroanalytical
Chemistry
Journal of Electroanalytical Chemistry 594 (2006) 1–19
www.elsevier.com/locate/jelechem

Review

A review of the aqueous electrochemical reduction of CO2 to


hydrocarbons at copper
M. Gattrell *, N. Gupta, A. Co
National Research Council, Institute of Chemical Process and Environmental Technology, 1200 Montreal Road, Ottawa, Ont., Canada K1A 0R6

Received 3 February 2006; accepted 16 May 2006


Available online 30 June 2006

Abstract

A review is provided on the aqueous reduction of CO2 to hydrocarbons at copper electrodes, covering the literature since the first report
of the reaction in 1985. This reaction is of interest as a potential component of a carbon energy cycle (i.e. CO2 + energy ! methane !
CO2 + energy). The synthesis of hydrocarbons from CO2 is a complex multistep reaction with adsorbed intermediates, most notably
adsorbed CO. The exact reaction mechanisms leading to the various products are not clear from the literature data to date and likely
change over the range of conditions at which data has been reported. The reaction product distribution is also very sensitive the surface
crystal structure copper electrode. The influences of various reaction conditions (potential, buffer strength and local pH, local CO2 con-
centration, stirring, and CO2 pressure) are discussed and some relationships between reaction conditions and products formed are
presented.
 2006 Elsevier B.V. All rights reserved.

Keywords: Carbon dioxide; Copper; CO2 reduction; Reaction mechanism

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2. Reaction intermediates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3. CO reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4. Single crystal work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5. The CO2 reaction step . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.6. Other factors – the electrode and electrolyte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.7. Other factors – temperature and pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1. Introduction

*
Corresponding author. Tel.: +1 613 990 3819; fax: +1 613 991 2384. The electrochemical reactions of CO2 are of interest
E-mail address: michael.gattrell@nrc-cnrc.gc.ca (M. Gattrell). for the synthesis of chemicals and for approaches to

0022-0728/$ - see front matter  2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jelechem.2006.05.013
2 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

decrease global warming. Work has been done looking at 2. Review


synthesising specialty chemicals such as formate [1] and
urea [2]. However, a far greater impact on decreasing 2.1. Background
atmospheric CO2 levels might be obtained by synthesising
fuels. In these carbon-based energy cycles, energy from Copper is unique among the metals tried as an electrode
renewable or nuclear power would be used to synthesis for CO2 reduction in its producing hydrocarbons at signif-
fuels from captured CO2, which, when used, would not icant current densities. High hydrogen overvoltage elec-
release additional CO2 to the atmosphere (i.e. ‘‘CO2 neu- trodes with negligible CO adsorption (such as Hg, Cd,
tral fuels’’). As such, these processes represent the storage Pb, Tl, In and Sn) can reduce CO2 with a high current effi-
of these energy sources as chemical energy, which allows ciency. However, these metals are poor catalysts in the
them to be more widely used, particularly for transporta- sense that the primary product is formate (i.e. there is no
tion applications. The main advantages of such carbon breaking of the carbon–oxygen bond of CO2). At the low
based fuels, vs. storing electricity in batteries or as hydro- hydrogen overvoltage metals with a high CO adsorption
gen, are the ease of use within existing infrastructures and strength (such as Pt, Ni, Fe and Ti) CO2 is reduced to form
the higher energy density (see for example [3]). Because tightly adsorbed CO [15,16]. Because of the low turnover of
this involves converting electrical energy to chemical the adsorbed CO, when the electrode is pushed to (for
energy, electrochemistry is an important enabling technol- example) 5–10 mA cm2, the principle product is hydro-
ogy [4]. gen. The electrode materials with a medium hydrogen over-
In one carbon-based energy cycle concept, electrolytic voltage and a weak CO adsorption, catalyse the breaking
hydrogen would be reacted with CO2 to form methanol of the carbon–oxygen bond in CO2, but allow the CO to
and water, which, following distillation, yields liquid meth- desorb. These include Au, Ag, Zn and Cu, with Au, Ag
anol [3,5]. Another concept would react electrolytic hydro- and Zn producing CO with high current efficiencies, but
gen and CO2 to form methane (with condensation of water with Cu being able to further react CO to more reduced
to shift the equilibrium and achieve good yields) [6]. How- species in significant amounts [7,14,17,18].
ever, direct electrochemical reduction of CO2 would allow The primary reactions that occur at the copper electrode
for a simpler process and also, by avoiding high tempera- during the reduction of CO2 are listed below (with the stan-
ture reactors, would enable the process production rate dard potentials calculated using formation energies from
to be quickly varied to follow the time of day availability [19]):
of surplus electricity. A key discovery along the road to this
goal was the direct reduction of CO2 to hydrocarbons 2Hþ þ 2e ¢ H2 E0 ¼ 0:0 V vs: SHE ð1Þ
þ 
(methane and ethylene) with reasonable current density 2CO2 þ 12H þ 12e ¢ C2 H4 þ 4H2 O E0 ¼ 0:079 V vs: SHE
(5–10 mA cm2) and current efficiency (up to 69% at ð2Þ
0 C) at a copper foil electrode, reported by the group of CO2 þ 8Hþ þ 8e ¢ CH4 þ 2H2 O E0 ¼ 0:169 V vs: SHE
Prof. Hori [7,8]. While this discovery has not, to date, led
ð3Þ
to the production of a pure hydrocarbon product, a sur-
þ 
prising range of hydrocarbons is formed (including ethanol CO2 þ 2H þ 2e ¢ CO þ H2 O E0 ¼ 0:103 V vs: SHE
and propanol [9,10]). Further, the gaseous products mix ð4Þ
(hydrogen, methane, ethylene and CO), if the selectivity þ
CO2 þ H þ 2e ¢ HCOO  
E0 ¼ 0:225 V vs: SHE ð5Þ
for hydrocarbons could be improved, would be tantalis-
ingly close in fuel properties to ‘‘hythane’’, a clean burning The equilibrium potentials, over the range of pH values
hydrogen/natural gas blend that has been promoted as an around where the reactions are typically carried out, are
alternative fuel for existing vehicles [11]. Thus, these reac- shown in Fig. 1. While, thermodynamically, methane and
tions are of interest as one possible tool for combating glo- ethylene should occur at a less cathodic potential than
bal warming. hydrogen, kinetically this does not occur.
Taniguchi wrote a review of electrochemical and pho- A set of polarisation curves (from the work of Prof.
toelectrochemical reduction of carbon dioxide in 1989 Hori’s group [20]), measured at the copper electrode using
[12]. Sullivan, Krist and Guard edited a thorough review different gases and electrolytes is shown in Fig. 2. Curves
of the general area of electrochemical reduction of CO2 in (a) and (b) were measured in a 0.1 M KH2PO4 + 0.1 M
1993, with a chapter by Frese that looks specifically at K2HPO4 buffer (pH 6.8) under Ar and CO, respectively.
CO2 electroreduction at solid electrodes [13]. A more Under Ar, hydrogen evolution begins at around 0.8 V
recent overview of CO2 reduction at metal electrodes vs. NHE reaching ca. 2.5 mA cm2 around 1.0 V. Under
has been done by Hori [14]. However, it is aim of this a CO atmosphere, the current is greatly suppressed, not
paper to provide an in depth review focused on specifi- reaching 2.5 mA cm2 until around 1.2 V. For testing
cally the reduction of CO2 at copper electrodes in aque- with CO2, a 0.05 M KH2PO4 + 0.15 M K2HPO4 buffer
ous electrolytes, and also to try to see how what is was used, which equilibrated at pH 6.7 after saturating
known can lead to a better understanding of the possible with CO2 (curve (c)). The reaction of CO2 starts off near
reaction mechanism. that of hydrogen, with the current exceeding that of hydro-
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 3

-0.1

-0.2

Potential / V vs. SHE CO2/CH4


-0.3 CO2/C2H4
H+/H2
CO2/HCOO -
-0.4

-0.5
CO2/CO

-0.6

-0.7
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10
pH

Fig. 1. The equilibrium potentials as a function of pH for the principal overall CO2 reduction reactions (at 25 C).

CO produced by the reduction reaction. If the buffer is


changed to 0.1 M KHCO3 (pH 6.8 after saturating with
CO2), the results in curve (d) are obtained. The current
starts around the same potential, but only increases slowly
with a slight shoulder near the peak in curve (c). This
shows the sensitivity of the reaction to the electrolyte (in
this example, the anion). Because of these complexities,
and in particular the suppression of hydrogen evolution
(i.e. the reactions are not independent), polarisation curves
need to be combined with other techniques to gain insight
into the CO2 reduction reactions.
The product distribution for CO2 reduction as a func-
tion of potential has been reported by the groups of Frese
[21], Ito [10] and Hori [20]. Frese’s group found that the
onset potential for CH4 was around 1.5 to 1.6 V vs.
SCE (at pH 7.6 and 22 C), with maximum efficiency (of
50% current efficiency) at about 1.95 V. At less negative
potentials (above about 1.6 to 1.7 V), CO production
exceeds CH4 production. The exact point of the change
of the dominant product from CO to CH4 was also influ-
enced by the electrode preparation and by the amount of
mass transport, with higher stirring resulting in more CO.
One suggestion for this was that higher mass transport
removes CO more effectively before it can be further
reduced to CH4. The authors also felt that CO was an
Fig. 2. Voltammograms obtained in phosphate solutions with Ar: (a), CO
(b), and CO2 (c) at pH 6.8, and with 0.1 M KHCO3 saturated with CO2 (d)
intermediate in the reduction of CO2 to methane, with
at pH 6.8. (Reproduced from [20] by permission of the Royal Chemical the electrochemical dissociation of CO as the slowest step
Society.) in the overall reaction. A larger set of products was mea-
sured by Ito’s group, with different products dominating
gen briefly (near 0.9 V), then after a peak current at at different potentials. Starting at 1.35 V vs. Ag/AgCl/
1.01 V the reaction current decays to slightly less than saturated KCl, the main product was formate (at about
that for CO reduction. As will be discussed later, this sup- 30% current efficiency), transitioning to CO at 1.47 V,
pression of hydrogen evolution during CO2 reduction is then ethylene below 1.54 V, and methane below
generally agreed to be due to the presence of adsorbed 1.65 V (peaking at 38% efficiency at 1.70 V before
4 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

decaying to 10% efficiency at 1.75 V as hydrogen evolu- appearing. These reactions accelerate (with methane pro-
tion began to dominate). Solution phase products identi- duction showing the stronger potential dependence), dom-
fied included ethanol (with 14% efficiency at 1.65 V) inating over CO and formate at around 1.35 V.
along with propanal, propanol and ethanal (peaking It is important to consider in these reactions the local
around 1.58 V with around 5%, 4.5% and 2% efficiencies, concentrations of pH and CO2 at the electrode surface.
respectively). Similar results were reported by Hori’s group The reason for the increased pH at the electrode surface
[20], along with the current densities. From the data of is the generation of OH ions (or the consumption of
Hori’s group, partial currents could be calculated for vari- H+) from the various reactions (Eqs. (1)–(5)). This is com-
ous products and are plotted in Fig. 3 along with estimates plicated in this system, because CO2 is both a reactant and
of the local pH and CO2 concentrations at the electrode a buffer and because some of the buffer reactions are kinet-
surface (taken from [22] and discussed in more detail later). ically limited. In the CO2 and bicarbonate system the key
Initially, the current is predominantly for hydrogen evo- equilibria are:
lution. However, the rate of increase of hydrogen evolution
CO2 þ H2 O ¢ ðH2 CO3 Þ ¢ Hþ þ HCO
3 pK a ¼ 6:4 ð6Þ
with overpotential begins to decrease around 0.95 V vs.
SHE, where CO begins to form. The hydrogen evolution HCO
3 ¢H þ þ
CO2
3 pK a ¼ 10:3 ð7Þ
reaction eventually begins to recover below 1.26 V but Eq. (6) results in the CO2 saturated bulk solutions hav-
at a level roughly two orders of magnitude less than that ing pH values from around pH 6–8 (for bicarbonate con-
obtained by extrapolating from the initial data (around centrations of 0.05–2 M, respectively). However, in CO2
0.9 V). This suggests, as has been indicated elsewhere saturated bicarbonate solutions, the equilibrium concentra-
[23], that greater than 90% of the electrode surface is tion of H2CO3 is very low and the kinetics of its formation
blocked by adsorbed CO. Initially, CO2 reduction produces are slow [24,25].
CO and formate, until below 1.12 V, where hydrocarbon
products begin to form, with first ethylene, then methane CO2ðaqÞ þ H2 O ¢ H2 CO3 k f ¼ 6:2  102 s1 ð8Þ

10
Total

Partial current density / mA cm -2


1

H2

-
HCOO 0.1

CH4
0.01
CO
C2H4

0.001
-0.8 -0.9 -1 -1.1 -1.2 -1.3 -1.4 -1.5
Potential / V vs SHE

1.0E-08 1.0E-01
[CO2]
CO2 Concentration / M
+ Concentration / M

1.0E-09 [H+] 1.0E-02

1.0E-10 1.0E-03
-0.8 -0.9 -1 -1.1 -1.2 -1.3 -1.4 -1.5
Potential / V vs SHE

Fig. 3. Partial current data from Hori et al. [20]. (Conditions: 0.1 M KHCO3, 19 C, CO2 bubbled, bulk [H+] = 1.55 · 107 M, bulk
[CO2] = 3.41 · 102 M. Estimated local [H+] and [CO2] values for polarisation measurements from [22].)
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 5

Thus, near the electrode surface where hydroxide is Table 1


forming, most of the buffer capacity results from reaction Infrared adsorption peak frequencies (cm1) for CO adsorbed on various
copper crystal faces under UHV conditions (approximately 77 K) (from
(7) (with a much higher pKa) and from the direct reaction [31])
of CO2 with hydroxide:
Crystal face Low coverage Saturation coverage
CO2 ðaqÞ þ OH ¢ HCO
3 k f ¼ 5:93  103 M1 s1 ð9Þ 100 2079 2088
111 2080 2070
This can result in a significantly higher local pH. These 110 2088, 2104 2094
equations have been numerically solved by Gupta et al. 211 2095, 2109 2110
[22], and the results of these estimations are included in 311 2093, 2109 2104
Fig. 3. It can be seen that the local pH is predicted to be 755 2111 2106
around pH 9 for most of the measurements, then increasing
to around pH 10 when the total current exceeds 2 mA cm2. When CO2 is reduced at a copper cathode, it was shown
This likely limits the recovery of hydrogen evolution at by the group of Prof. Hori, that the infrared adsorption of
higher overpotentials. The decreased local concentration CO is detected (at 2085 cm1 at 0.9 V vs. NHE and
of CO2 will slow the formation of formate and CO, with 2081 cm1 at 1.0 V) [33]. These results have been con-
the more rapid drop in CO product recovery likely due to firmed by other workers using both infrared and surface
the increased rate of its further reaction to CH4 and C2H4. enhanced Raman spectroscopy (SERS). In the work of
Other products reported from CO2 reduction at the cop- Oda et al. [34], CO2 reduction was studied at electrodepos-
per electrode include [9,10,14]: ethanal, ethanol, acetic acid ited copper in 0.05 M Na2SO4 at 0.5 and 0.7 V vs. SHE
[26], propanal, 2-propen-1-ol, propanol. Interestingly, (using 0.3 V for the reference spectra). CO adsorption
methanol, formaldehyde and ethane are not reported. was noted both for the reference and the measured poten-
tials (shifting with potential), but new peaks were noted at
2.2. Reaction intermediates 1370 and 1542 cm1 assigned to carbonate (discussed fur-
ther below). SERS measurements were also made using
In Fig. 3, CO and formate are initially formed before 3.5 M KCl solution. These spectra showed a peak at
hydrocarbons form at more cathodic potentials. Early 283 cm1, characteristic of Cu–Cl stretching at 0.6 and
work found that the reduction of CO at Cu leads to a sim- 1.0 V. At 1.4 V this was replaced by a pair of peaks
ilar product distribution, suggesting that CO was an inter- at 2046 and 1998 cm1, characteristic of CO stretching.
mediate in CO2 reduction [27,21,28]. Tests starting from At 1.6 V, these peaks shifted to 2082 and 2037 cm1,
formate do not result in any measurable products [20,29] and peaks appeared at 360 cm1 attributed to Cu–CO
(though from concentrated formic acid, deposits on the stretching and at 1054 cm1 attributed to symmetric
electrode and trace methane have been reported [28,29]). stretching of carbonate. The latter presumably appearing
It was also found that CO directly suppressed hydrogen due to the high overpotential resulting a high local pH,
evolution, while CO2 required some charge to be passed and hence carbonate from reaction (7).
before hydrogen evolution was suppressed. The charge A SERS study of CO2 reduction was also carried out by
required at different current densities was found to be con- Smith et al. [35]. On polarisation of an electropolished cop-
sistent with the reduction of CO2 to form an adsorbed CO per electrode to 1.06 V vs. NHE in a CO2 or CO satu-
layer [23]. In both cases, hydrogen evolution activity could rated 0.1 M NaHCO3 solution, a number of adsorption
be recovered if the reaction was interrupted and the solu- peaks were observed in the spectra. These included the
tion bubbled with Ar, suggesting the material blocking vibrations for on-top adsorbed CO at 2090 cm1 (CO
the electrode surface was reversibly adsorbed. stretch), 358 cm1 (Cu–CO stretch) and 280 cm1 (Cu–CO
Optical experiments with CO have shown it is adsorbed frustrated rotation). A broad flat band at 2900 cm1 and a
on Cu once the potential is sufficiently cathodic to allow weak band at 1450 cm1 are attributed to CHx stretching
CO to displace adsorbed anions [30]. The observed absorp- and CHx deformation, respectively, and a peak at
tion peak was found at 2040 cm1 at less cathodic poten- 1050 cm1 related to carbonate stretching. Bands were also
tials (around 0.6 to 0.7 V vs. NHE), shifting to found after stepping the electrode anodic (to 1.3 V to clean
2080 cm1 at more negative potentials (around 1.0 V). the surface and maintain the required roughness for SERS
This peak corresponds to the vibration of the CO molecule activity). On the return steps in the cathodic direction at
adsorbed on-top (i.e. at a single copper atom) and is similar 0.16 and 0.36 V, new peaks were observed at 1580
to, though at a slightly lower wavenumber than, typical and 1340 cm1, as well as a large increase in the intensity
results measured for CO adsorption on Cu under of the 1050 cm1 peak. These were assigned to graphitic
ultra-high vacuum (UHV) conditions (see Table 1). (A dis- carbon. However, the suggested assignments are for crys-
cussion of the different effects of UHV vs. solution environ- talline graphite vibrations, which appear unlikely, as one
ments on the vibrations of adsorbed CO and NO has been would expect carbon formed under these conditions to
presented by Weaver et al. [32].) One can also see in Table 1 more amorphous. Others have reported similar vibrations
that the vibration varies with the copper crystal face and during the oxidation of formate using silver, which they
with the degree of coverage. attributed to adsorbed carboxy species [36]. Adsorbed
6 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

formate on silver is also reported to give vibrations at carbons [21,20]. Therefore, studies have been done focusing
1550 cm1 (O–C–O antisymmetric stretch) and 1350 cm1 on the reduction of CO. This has the advantage of having
(O–C–O symmetric stretch) [37], and at 1560 and fewer reaction steps to consider. It also avoids the compli-
1360 cm1 in UHV [38]. Prof. Hori’s group has recorded cation found with CO2 reduction studies, where the reac-
a vibration at 1545 cm1 in 0.2 M K2CO3 solution assigned tant is also part of the buffer, thus simplifying the
to adsorbed carbonate [30]. As discussed above, Oda et al. estimation of the local CO concentration at the electrode
assigned vibrations at 1370 and 1542 cm1 measured at surface.
0.5 and 0.7 V vs. SHE in CO2 saturated 0.05 M Na2SO4
to carbonate. However, in both Oda and Smith’s experi- 2.3. CO reduction
ments, because they were saturated with CO2, carbonate
would only be expected at high overpotentials where the A detailed study of the reduction of CO at copper elec-
local pH becomes high. Therefore, at the low overpoten- trodes was carried out by Prof. Hori’s group [40]. Reac-
tials involved, the assignment of the peaks around 1560 tions were carried out at various pHs using phosphate
and 1350 cm1 to carbonate (and also the increased inten- buffer, borate buffer, or K2HPO4–KOH with a total elec-
sity of the 1050 cm1 peak in Smith et al.’s work) is unsat- trolyte concentration of 0.2 M. The cell was stirred with
isfying. And so, while we feel the assignment of these peaks a magnetic stirrer and the mass transport characterised
to graphite is unlikely, we can at best say they are associ- using ferricyanide. Potentiostatic and galvanostatic runs
ated with some type of adsorbed carboxy species. In the were carried out for 30 min or more under conditions
work of Smith et al., there is also peak (around below 20% of the estimated limiting current for CO. From
525 cm1) that grows with time and is associated with a these data, potential and pH dependencies of the CO
slow loss of activity for CO2 reduction (a poisoning reduction could be determined. The results reported for
species). However, their solutions were not purified by the partial current for ethylene formation are shown in
pre-electrolysis [39], making any interpretation uncertain Fig. 4. The data show a distinct Tafel slope corresponding
(discussed more in later). to a transfer coefficient of 0.35 without any apparent effect
These studies clearly show that the electrode surface is of pH. The data found for the partial current for methane
dominated by adsorbed CO during the CO2 reduction formation, however, did not result in Tafel relationship
and that CO is an intermediate in the production of hydro- unless the results were first corrected for the pH. The best
carbons. The latter point is especially nicely shown in the straight line relationship on a semi-logarithmic plot was
work of Smith et al., where both CO2 and CO saturated found when the reaction was assumed to be first order in
solutions result in similar spectra. Because of the high proton concentration (see Fig. 5). From the corrected data,
implied coverage of CO during the reduction of CO2, it a transfer coefficient of 1.33 was determined.
was felt that the reduction of adsorbed CO was the key rate These results raised some interesting questions about the
determining step for the overall reaction of CO2 to hydro- hydrocarbon formation mechanism. In prior papers, it had

pH 6.0-6.3
pH 7.1-7.7
pH 8.0-8.6
pH 8.7-8.9
pH 9.0-9.3
Partial current / mA cm-2

pH 10.5-11.3
pH 12.2

0.1

0.01
-1.05 -1.1 -1.15 -1.2 -1.25 -1.3 -1.35 -1.4
Potential / V vs. SHE

Fig. 4. The partial current densities for ethylene formation from CO in various buffer solutions (data taken from [40]).
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 7

11

pH 6.0-6.3
pH 7.1-7.7 10
pH 8.0-8.6
pH 8.7-8.9
9
pH 9.0-9.3

Log(imethane / mA cm-2) + pH
pH 10.5-11.3
pH 12.2 8

3
-1.05 -1.1 -1.15 -1.2 -1.25 -1.3 -1.35 -1.4
Potential / V vs. SHE

Fig. 5. The partial current densities for methane formation from CO in various buffer solutions assuming a first order reaction with respect to [H+] (data
taken from [40]).

been hypothesised by analogy to Fischer–Tropsch synthe- O


O
-
O
- -
sis and based on the methane and ethylene products, that C e C C
carbonyl (Cu@CH2) was a key intermediate leading to C u Cu Cu Cu C u C u Cu Cu Cu
both products [41,28,20]. However, the very different Tafel
Fig. 6. The expected resonant structures following an initial electron
slopes found indicate that the pathways to methane and transfer to adsorbed CO.
ethylene are different (also their appearance at different
potentials in Fig. 3). Further, the transfer coefficient of
0.35 for ethylene formation implies that the rate determin- [43]). There is also a proton involved in the reaction, and
ing step is the first electron transfer, indicating that the this must occur reversibly before the second electron trans-
mechanisms may differ right at the first electron transfer fer rate determining step (hence an ECE reaction) to yield
to CO. As well as the products which are formed, another the observed transfer coefficient. As the formation of a
important criterion in considering the mechanism is the C–H bond would not be reversible, this implies an acid–
lack of formation of methanol. This implies that the car- base reaction at the oxygen. This leads to mechanisms such
bon–oxygen bond of CO is broken early and consistently as shown in Fig. 7 (with the right hand product similar to
in the mechanism. that suggested by Kim et al. [21]).
For the case of methane formation, the transfer coeffi- (Note, for simplicity in this and subsequent examples,
cient of greater than 1 suggests that there is an initial elec- the reaction is written indicating on top bonding to a single
tron transfer in equilibrium before the rate determining copper. Bridge bonding might also occur.) Another possi-
step. This would result in a CO anion radical as shown in bility would be the reaction with adsorbed hydrogen (see
Fig. 6. Fig. 8). Depending on the potential and pH dependency
The state of the adsorbed CO anion radical on copper of the adsorbed hydrogen coverage, this might also lead
has been evaluated through ab initio calculations by to the observed reaction order and transfer coefficient.
Watanabe et al. [42]. They estimate a slight decrease of Once the adsorbed CO has been electrochemically split,
the Cu–C bond and an increase of the C–O bond to about the product of this reaction would then go through hydro-
1.25 Å, thus predicting a mostly double bond character gen additions leading to methane. These hydrogen addi-
(e.g. CO2 1.16 Å, H2C@O 1.21 Å, and H3C–OH 1.42 Å tions could occur either through reaction with adsorbed

- - -
O OH OH
C H
+
C H
C e
- C O
Cu Cu Cu Cu Cu Cu Cu Cu Cu or C u C u C u C u

Fig. 7. A reaction mechanism that would lead to the observed transfer coefficient and reaction order with pH.
8 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

- 2.4. Single crystal work


O H -
H C C O
Cu Cu Cu Cu Cu Cu In the review by Frese [13], work is presented that inves-
Fig. 8. A reaction mechanism that might also lead to the observed transfer tigates the influence of the copper surface structure on the
coefficient and reaction order with pH. (Note that the adsorbed oxygen product distribution (in particular the rate and current effi-
might also occur as Cu2O.) ciency for methane production). Methane formation rates
found at 1.85 V vs. SCE in 0.5 M KHCO3 were 0.5 ·
105 mol cm2 h1 on Cu(1 0 0), 1.0 · 105 mol cm2 h1
hydrogen and/or through electron transfer followed by
on Cu(1 1 0) and 5.8 · 105 mol cm2 h1 on Cu(1 1 1).
protonation.
The maximum rates of methane formation, presumably
The reaction to form ethylene begins at a lower potential
due to mass transfer limitations of the CO2, also varied at
(see Fig. 3), without a pH dependence and with the first
each crystal face, indicating different product distributions.
electron transfer being the rate determining step. As the
Variations were also found using electrodeposited copper
first electron transfer begins a reaction pathway that subse-
with different plating baths, deposition current densities
quently results in a two carbon product, it is reasonable to
and substrates. One trend that was noted was a decrease
assume that some type of bond formation occurs in this
in methane production with higher copper deposition cur-
reaction step. A reaction is proposed that involves a ‘‘prior
rent (possibly due to less crystalline deposits). In fact very
association’’ of two adsorbed COs (see Fig. 9). The postu-
low currents, including simply holding a copper foil at open
lation of a starting ‘‘associated pair’’ of adsorbed COs also
circuit, resulted in improved methane yields, attributed to a
would be consistent with the lower activation potential vs.
gentle electropolishing of the copper surface.
the formation of CO (and hence methane). This hypoth-
A detailed study of the effect of copper crystal faces on
esis will be discussed more in the following section.
product distributions, including the influence of step edges,
It can also be shown that the polarisation curves for CO
has been done by the group of Prof. Hori, with a summary
are consistent with those for CO2. From the data shown in
of their results shown in Table 2. It can be seen that the
Fig. 3, one can obtain a transfer coefficient of 0.48 for eth-
product distributions are strongly affected by the copper
ylene production (obtained by excluding the data at
crystal structure. Generally, the crystal faces dominated
1.46 V), similar to the value of 0.35 for CO reduction.
by Cu(1 0 0) tend to result in a significant current efficiency
For methane production a transfer coefficient of 0.86 is
for ethylene at relatively low overpotentials. The crystal
obtained, but assuming the reaction from CO2 is also first
faces dominated by Cu(1 1 1) tend to be polarised to more
order in hydrogen ion, we can correct the partial current
negative potentials and favour methane production.
density by [H+]bulk/[H+]local. This results in a transfer coef-
Finally, the Cu(1 1 0) type surfaces polarise to the most neg-
ficient of 1.19, which is similar to that reported for methane
ative potentials and result in other 2 carbon and 3 carbon
for CO reduction. Thus, the mechanisms appear to be the
products (including acetic acid [26]). Using the Tafel slopes
same with the two different starting reactants. (While we
from the previous section and the data from Table 2, par-
also have estimates for [CO2]local, correcting the partial cur-
tial currents for methane and ethylene can also be norma-
rents would be complicated by needing to be able to relate
lised to a common potential. This is shown in Fig. 10,
this to the coverage of adsorbed CO.)
where one can see that at the chosen relatively low overpo-
In the work of Kim et al. [21], CO2 was reduced in
tential, overall, ethylene production is favoured vs. meth-
0.5 M KHCO3 at a mechanically polished and HCl
ane (consistent with Fig. 3). This formation of ethylene
cleaned copper foil. They obtained a Tafel slope of
at low overpotentials is favoured on Cu(1 0 0) type surfaces,
110 mV decade1 (transfer coefficient of 0.55) for methane
preferably with reasonably wide terraces. If the potential is
and 200 mV decade1 (transfer coefficient of 0.30) for eth-
made more negative, the methane reaction (because of its
ylene. However, they used a higher range of current den-
higher transfer coefficient) rapidly dominates, most notably
sities than those in Fig. 3 and found methane formation to
at the 3 1 1 and 2 1 1 crystal faces. One possible reason for
be mass transport controlled negative of 1.9 V vs. SCE.
the favouring of ethylene formation on Cu(1 0 0) type sur-
Because some of the data was close to the limiting current,
faces might be related to the hypothesised rate determining
and the correction for local pH will likely be large, these
step in Fig. 9. Such a step would require the p orbitals of
results represent minimum possible values for the transfer
the adsorbed CO to interact simultaneously with vibra-
coefficients and so are not conflicting with those taken
tional motions bringing the two oxygens close to copper
from Fig. 3.
surface atoms. Such an intermediate state would be more
likely with a right angle arrangement of surface copper
atoms (see Fig. 11).
O O - The tilting motion of adsorbed CO brings the oxygen
O O
C C C C C C
e- O O closer to the copper surface, which is a likely prerequisite
C u Cu C u C u Cu Cu C u C u Cu C u Cu Cu for CO dissociation. This vibration is described as a frus-
Fig. 9. A reaction mechanism that would lead to the observed transfer trated translational motion or T-mode vibration where
coefficient and produce a two carbon product. most of the motion occurs at the oxygen [45]. The displace-
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 9

Table 2
CO2 reduction products at various copper single crystal electrodes (data from [44]) (5 mA cm2, 0.1 M KHCO3)
Test conditions Current efficiency/%
Crystal series n Crystal Potential CH4 C2H4 CO H2 HCOO Other C2+
[n(1 0 0) · (1 1 1)] inf 100 1.40 30.4 40.4 0.9 6.8 3.0 17.4
6 1111 1.37 8.9 50.2 1.8 8.8 3.2 24.2
5 911 1.36 5.7 50.9 0.0 12.7 3.5 28.6
4 711 1.34 5.0 50.0 1.1 15.6 4.6 21.5
3 511 1.36 11.4 39.0 1.9 18.1 8.8 22.3
2 311 1.37 36.0 23.8 2.6 13.3 14.0 9.2
111 1.55 46.3 8.3 6.4 16.3 11.5 7.5
[n(1 0 0) · (1 1 0)] inf 100 1.40 30.4 40.4 0.9 6.8 3.0 17.4
8 810 1.38 6.4 45.1 1.4 8.7 1.5 32.4
6 610 1.37 7.6 44.7 0.9 9.0 1.4 33.6
5 510 1.38 8.1 42.3 2.1 10.5 2.8 37.2
3 310 1.42 17.7 34.6 0.0 12.8 2.7 38.6
2 210 1.52 64.0 13.4 2.2 7.0 5.5 9.5
[n(1 1 1) · (1 0 0)] inf 111 1.55 46.3 8.3 6.4 16.3 11.5 7.5
10 1199 1.48 62.4 10.2 4.9 7.2 7.9 9.0
6 755 1.43 62.9 11.5 4.4 6.9 12.3 8.9
4 533 1.42 62.9 13.0 3.0 4.7 9.7 4.6
3 211 1.38 45.6 17.8 2.1 11.2 13.6 8.2
2 311 1.37 36.0 23.8 2.6 13.3 14.0 9.2
[n(1 1 1) · (1 1 1)] inf 111 1.55 46.3 8.3 6.4 16.3 11.5 7.5
6 332 1.51 39.6 9.9 6.1 10.3 9.4 16.3
3 331 1.55 13.8 16.6 7.7 5.7 9.1 31.1
2 110 1.58 6.9 13.5 13.9 3.1 10.1 52.5
[n(1 1 0) · (1 0 0)] 6 650 1.59 10.5 16.2 14.5 2.5 6.1 48.6
3 320 1.52 52.4 13.7 5.4 5.3 5.8 15.8
2 210 1.52 64.0 13.4 2.2 7.0 5.5 9.5
Other C2+ includes: ethanal, ethanol, acetic acid, propanal, 2-propen-1-ol, propanol.

CH4
5
C2 H4
-2
Partial current / mA cm

0
inf 6 5 4 3 2 0 inf 8 6 5 3 2 inf 10 6 4 3 2 6 3 2 inf 6 3 2
[n(100)x(111)] [n(100)x(110)] [n(111)x(100)] [n(110)x(100)] [n(111)x(111)]
Crystal face

Fig. 10. Data from Table 2 normalised to 1.39 V vs. SHE using transfer coefficients of 0.4 for ethylene and 1.3 for methane.

ments of the carbon and oxygen within the four main of ±5.14 · 109 V cm1 [46]). The T-mode vibration has
vibrational modes for adsorbed CO have been estimated been studied in the gas phase using He scattering and a
to be relatively independent of electric fields (in the range strong influence of the copper surface structure has been
10 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

O 2.5. The CO2 reaction step


C O
C The initial stage of the reaction of CO2 to form hydro-
carbons involves the reaction of CO2 to form CO. This
reaction competes with the formation of formate. Refer-
ring to Fig. 3, hydrogen, formate and CO are the dominant
Fig. 11. One explanation for the catalysis of ethylene formation on reactions at less negative potentials. However, the under-
Cu(1 0 0) type copper crystals. standing of the initial reactions of CO2 is complicated by
the blocking of electrode surface by CO and the further
reactions of CO to hydrocarbons. Thus, some clearer
found (see Table 3) [47]. The results show a clear increase in
insight might be gained by examining the literature for
the amplitude of such vibrations as the copper surface
CO2 reduction on similar metals to copper, such as silver
becomes more defective, suggesting that defects and step
and gold, where CO is more easily desorbed and so does
edges should help catalyse the dissociation of CO.
not further react to any significant extent. On these metals
An increase in step edges on the Cu(1 0 0) face does seem
the predominant reaction is CO2 to CO, with a small
to result in increased activity for ethylene (see Fig. 10). How-
amount of formate and hydrogen.
ever, a maximum activity occurs at Cu[4(1 0 0) · (1 1 1)]
At the gold electrode, the reaction has been found to be
(which also corresponds to the lowest overpotential in Table
first order in CO2 with a Tafel slope of around 120 mV
2). As the Cu(1 0 0) terraces become smaller than 4 atoms
decade1, and so the rate determining step is felt to be the ini-
wide, the activity begins to drop. For methane, the highest
tial reduction of CO2 [49,50].
partial current occurs for Cu[2(1 1 1) · (1 0 0)]. It should be
noted, however, that the estimated partial currents in CO2 þ e ¢ CO
2ads ð10Þ
Fig. 10 are for the complete CO2 reduction to hydrocarbons Results using single crystal silver electrodes show that
reaction, and so they include both dissociating CO2 and con- the reaction of CO2 to CO is favoured on Ag(1 1 0) (see
verting CO to hydrocarbons. Thus, the best crystal face for Fig. 12) [51]. The reaction to form formate may also show
the conversion of CO2 to hydrocarbons needs to provide some enhancement at Ag(1 1 0), but the reaction slows
sites for CO2 dissociation and for further reaction of CO. below 1.5 V vs. SHE, likely due to depletion of the local
In some of Prof. Hori’s group’s earlier copper single crystal CO2 concentration by the CO forming reaction.
studies, the product distributions of galvanostatic reactions The adsorption and reaction of CO2 on copper has been
of both CO2 and CO were measured [48]. Because of the studied in the gas phase because of interest in the methanol
lower solubility of CO, experiments were carried out at synthesis reaction (H2 + CO + CO2 ! CH3OH + H2O),
2.5 mA cm2 vs. 5.0 mA cm2 for CO2 reduction. The prod- which is normally carried out on a copper-zinc oxide cata-
uct partial currents for methane and ethylene at polycrystal- lyst. CO2 is found to both physisorb (as a linear molecule)
line copper, Cu(1 0 0), Cu(1 1 0) and Cu(1 1 1) electrodes and chemisorb (as a bent COd 2 molecule) [52]. The activa-
followed generally the same trends. Though, overall, the tion of physisorbed CO2 to chemisorbed form has been
CO2 reduction required a slightly higher overpotential, reported to be promoted by surface defects [53], alkali
and a greater difference in overpotentials was found between metal promoted surfaces [54–56] and possibly through
the different crystal faces. the action of X-rays or photoelectrons during measure-
In any case, the results of the single crystal studies ments [52]. Using SERS the symmetric stretch mode and
clearly show why the electrode preparation can have such deformations modes were reported at 1368 and 650 cm1,
a strong influence on the results obtained. respectively, for physisorbed CO2 and at 1182 and
768 cm1 for chemisorbed CO2 [53]. The dissociative
Table 3 adsorption of CO2 has been reported on stepped copper
The energy related to the frustrated translational vibration of adsorbed
CO on copper (data from [47]) surfaces [57,58], thought to occur through a chemisorbed
COd2 intermediate and leading to CO and adsorbed oxy-
Surface T-modea Relative CO adsorption
energy/meV amplitudeb energy/kJ mol1 gen. The adsorbed COd 2 is also thought to be stabilised
by formation of a T-shaped [OCOCO2] structure with
Cu(1 1 1) 4.07 1.00 47.3
Cu(1 0 0) 3.94 1.03 51.1
neighbouring physisorbed CO2 [52]. As well as dissociation
Cu(1 1 0) parallel 3.75 1.09 54.0 to CO and O, the COd 2 has been reported on alkali metal
to step edge activated copper surfaces to lead to oxalate (thought to be
Cu(1 1 0) perpendicular 3.40 1.20 54.0 via dimerisation of CO 2 ) and CO + carbonate products
to step edge (thought to be via [OCOCO2] or possibly oxalate)
Cu(2 1 1) 3.00 1.36 58.4
Cu(5 1 1) 3.05 1.33 57.7
[52,54]. When COd 2 is co-adsorbed with hydrogen, again
Sputtered on Cu(1 0 0) 3.20–2.52 1.27–1.62 58.2 on alkali metal activate copper surfaces, formate is
a
The frustrated translational vibration of CO parallel to the surface,
reported [55,56].
which results in a tilting vibration. The exact adsorption geometry of COd 2 is not clear,
b
Inversely proportional to the T-mode vibration energy. with the postulated structures shown in Fig. 13 based on
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 11

Fig. 12. Partial current densities on silver single crystals (0.1 M KHCO3, CO2 saturated, from Hori et al. [51]). (Reproduced by permission of the Journal
of Electroanalytical Chemistry.)

overvoltage electrodes with weak adsorption of hydrogen


also have a high overvoltage for CO2 reduction to CO 2ads
(hence a weak adsorption/stabilisation of CO 2ads ) and pro-
duce predominantly formate (or oxalate in non-aqueous
CO2 electrolysis) (e.g. Hg, Cd, Pb, Tl, In and Sn). Low
hydrogen overvoltage metals also reduce CO2 at low over-
potential to form tightly adsorbed CO (e.g. Pt, Ni, and Fe).
And medium hydrogen overvoltage metals reduce CO2 at
medium overpotential to form more loosely held CO (e.g.
Au, Ag, Zn and Cu) [14,17].
In the electrochemical reduction of CO2 at solid
electrodes it is generally agreed that, at medium to high
Fig. 13. Possible structures for adsorbed COd
overpotentials, the first electron transfer to form a CO 2
2 on metals (adapted from
[52,59] with the metal atom diameter scaled for copper). anion radical is the rate determining step [12–14]. In aque-
ous solution, this intermediate then reacts further to form
formate or CO with the product distribution being deter-
the structures of CO2–metal complexes [52,59]. One reason mined by the relative rates of the different reaction path-
for the uncertainty is that the expected vibrations could be ways. The rates of the different reaction pathways will be
affected by surface geometry and, as discussed above, the determined by the local rate determining step within that
COd2 species is most easily produced at surface defects or pathway. Different possible mechanisms have been sug-
in the presence of submonolayer coverage of an alkali gested for the different pathways [12–14].
metal. There is also the possibility that the COd2 is stabi- The reaction to form formate could occur by proton-
lised by neighbouring physisorbed CO2 in a [OCOCO2] ation then reduction.
structure. On Ni(1 1 0), which has been studied in some
detail [52], it is thought that the predominant form is oxy- COO
 þ
ads þ H ! HCOOads

ð11Þ
gen coordinated. Though, it is also found that, in the pres- HCOOads þ e ! HCOO ð12Þ
ence of adsorbed hydrogen, formate is formed, which is
Such a reaction mechanism, with a proton adding at the
also oxygen coordinated [52]. Note that such a reaction
carbon, would appear to be favoured by CO 2 being
with the adsorbed hydrogen must involve bringing the car-
adsorbed with oxygen coordination. On high hydrogen
bon closer to the metal surface to interact. This might indi-
overpotential electrodes, where CO2 is reduced to formate,
cate that some interchange of coordinated states is
the reaction has been reported to have a Tafel slope char-
possible. It was also found that the COd 2 molecular plane, acteristic of a second electron transfer r.d.s. at low overpo-
while perpendicular to the surface, was randomly oriented
tential, shifting to the first electron transfer r.d.s. at
in relation to the h1 1 
0i azimuth, while formate was aligned
medium to high overpotential. This was suggested to corre-
parallel. This was taken to indicate at least some of the
spond to step 12 at low overpotential and step 10 at med-
COd was carbon coordinated allowing free rotation of
2 ium to high overpotential [12]. Formate could also be
the molecule. For the electrochemical case, one would have
formed by reaction with adsorbed hydrogen similar to
additional influences on the preferred orientation from the
the gas phase results. Adsorbed hydrogen would be present
electrolyte and the electric field.
as an intermediate in the hydrogen evolution reaction.
The adsorbed state of COd 2 would also be influenced by
the type of metal. As discussed previously, high hydrogen

COO
ads þ Hads ! HCOO

ð13Þ
12 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

Similar mechanisms are possible for CO formation. One r.d.s. *-


CO2 + e- CO2
pathway might also be protonation then reduction. ads

Though, in this case, with a proton adding at the oxygen, HCOO


- ?
the mechanism would appear to be favoured by CO 2 being
adsorbed with carbon coordination.

COO þ 
ads þ H ¢ COOHads ð14Þ
slow
 
COOHads þ e ! COads þ OH 
ð15Þ CO COads

And, CO could be formed by reaction with adsorbed ?


hydrogen. hydrocarbons

COO
ads þ Hads ! COads þ OH

ð16Þ r.d.s.
H+ + e- Hads
A third alternative for CO formation involves reaction
H2
between two CO2 molecules. As discussed earlier, such for- Cu electrode
mation of a T-shaped structure between the CO2 anion rad-
ical and the neutral form has been estimated to result in a Fig. 14. The main reaction pathways at the electrode surface, with
adsorbed CO blocking the majority of the surface and hydrocarbon
more stable intermediate species [52].
products being formed by the further reduction of adsorbed CO.

CO
2ads þ CO2 ¢ ½OCOCO2 ads ð17Þ
a case, a full understanding of the mechanism requires an
½OCOCO2 
ads

þ e ! COads þ CO2
3 ð18Þ
understanding of hydrogen evolution on copper. (Note

Note that a reaction route involving direct reaction of that other the possible interactions such as CO
2 or CO
two CO 2ads species would have to overcome the repulsion with hydrocarbon intermediates are not shown here, but
of the negative charges and so would likely be more might be expected to be involved in the production of 2
difficult. and 3 carbon alcohols and acids.)
A schematic overview of the reaction pathways is shown Hydrogen evolution on copper over a range of solution
in Fig. 14. Following CO2 reduction, two pathways are pHs has been studied by Bockris et al. [60]. The results of
shown, one leading to formate and the other to CO and their work have been plotted in Fig. 15. One can see that
hydrocarbon products. Both CO2 reduction and hydrogen the reaction kinetics are pH dependent in the acid region
evolution occur on unblocked areas of the electrode sur- (proton discharge) and pH independent in the alkaline
face, with CO being the primary cause of surface inhibi- region (water discharge). Extrapolating from the two data
tion. The possible involvement of adsorbed hydrogen in sets suggests that the shift in mechanisms occurs around
the reaction pathways is also indicated. If the reaction pH 11–12. Thus, for most CO2 reduction work (see, for
pathways involve hydrogenation, then there will be an example, Fig. 3 where pH < 10), the hydrogen evolution
interaction with the hydrogen evolution reaction. In such reaction can be written in terms of protons.

-0.2
Potential / V vs. SHE

-0.4
Erev

-0.6
0.1 mA cm-2

1 mA cm-2
-0.8

-1

-1.2
0 2 4 6 8 10 12 14
pH

Fig. 15. Hydrogen evolution polarisation data at copper with varying solution pHs. Data from [60].
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 13

Hþ + e ! Hads ð19Þ [20,50]. The results of these studies are shown in Fig. 16.
While, different mass transport intensities were likely used,
This can be followed by recombination or electrochemical
the results are qualitatively similar. In both sets of data, as
desorption.
CO2 reduction begins, formate and CO appear, with for-
2Hads ! H2 ð20Þ mate formation dominating over CO formation. As CO
or formation increases, hydrogen evolution becomes sup-
pressed. At around 1.2 V, CO begins to be converted to
Hads + Hþ + e ! H2 ð21Þ hydrocarbons, with first ethylene then methane appearing.
The exact mechanism has not been conclusively deter- Also around this potential the total current for the path-
mined. A recent impedance study on hydrogen evolution way leading to CO plus its products (pathway shown in
in acidic solutions (pH 0.7–1.5) has concluded that the Fig. 14) begins to dominate over formate formation.
reaction follows steps 19 and 21 [61]. It was also deter- Finally, at potentials negative of 1.4 V, as the steady state
mined that both steps are irreversible and the steady state CO coverage decreases due to its conversion to hydrocar-
coverage of adsorbed hydrogen is small. Another study bons and the local CO2 concentration decrease (see, for
using optical second harmonic generation and closer to example, Fig. 3), hydrogen evolution begins to recover.
neutral solutions (pH 5.5–6.5) determined the reaction fol- Thus, the data from these two groups shows similar trends.
lows steps 19 and 20, with 19 being rate determining [62]. The data from Prof. Hori’s group also includes the total
Steady state polarisation data would be helpful in trying current at each potential, allowing the partial currents for
to determine the mechanism for CO2 reduction at copper. each species to be estimated (as shown in Fig. 3). This data
However, good steady state data requires that the electrode can also be converted to reaction rates and so the rates of
surface does not change. This in turn requires, because of the formate and CO + products pathways can be plotted.
the very negative potentials used, that there can be no trace This is shown in Fig. 17 along with the estimated condi-
impurities in the electrolyte that could deposit [39]. Two tions at the electrode surface.
published studies, one by Prof. Hori’s group (using electro- At low overpotentials the rates of CO2 reduction and
lyte pre-electrolysis) and another by Prof. Ito’s group CO formation have Tafel slopes below 120 mV decade1,
(using re-crystallised KHCO3) fulfil the requirements which then increase to above that value. This appears to

45
H2 CH4 40
35

Current Eff. (%)


30
HCOO-
C2H4 25
CO 20
15
10
5
0
-0.8 -0.9 -1 -1.1 -1.2 -1.3 -1.4 -1.5 -1.6
Potential / V vs. SHE A

45
H2
40
35
CH4
Current Eff. (%)

- 30
HCOO
25
20
CO 15
10
C2H4 5
0
-0.8 -0.9 -1 -1.1 -1.2 -1.3 -1.4 -1.5 -1.6
Potential / V vs. SHE B

Fig. 16. Current efficiencies at different potentials (0.1 M KHCO3, CO2 bubbled). A: data from [20] (19 C), B: data from [50] (25 C).
14 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

1.0E-07

total CO2 reductn.

1.0E-08

Reaction rate / mole cm -2 s-1


H2

1.0E-09

CO
1.0E-10
-
HCOO CO+prod.

1.0E-11

120 mV decade-1

1.0E-12
-0.8 -0.92 -1.04 -1.16 -1.28 -1.4 -1.52
Potential / V vs. SHE

1.0E-08 1.0E-01
[CO2]

CO2 Conc. / M
H+ Conc. / M

1.0E-09 1.0E-02
+
[H ]

1.0E-10 1.0E-03
-0.80 -0.92 -1.04 -1.16 -1.28 -1.40 -1.52
Potential / V vs. SHE

Fig. 17. Data calculated from Hori et al. [20] showing the total CO2 reduction reaction rate and the rates for formate and the total rate for the pathway
leading to CO + hydrocarbon products (using a moles-of-carbon basis) (same data as Fig. 3). (Conditions: 0.1 M KHCO3, 19 C, CO2 bubbled, bulk
[H+] = 1.55 · 107 M, bulk [CO2] = 3.41 · 102 M. Estimated local [H+] and [CO2] values from [22] using a boundary layer of 0.01 cm.)

support a mechanism with the rate determining step shift- CO on copper, where adsorption is felt to be initially at
ing from a later step, to the first electron transfer with step edges, and only at higher coverage on terraces [63].
higher overpotentials. This would be similar to the mecha- Thus, any model of the reactions should consider the pos-
nism discussed above for CO2 reduction to formate at high sible effects of surface non-homogeneity.
hydrogen overpotential electrodes. Around 1 V formate In the region from 1.12 to 1.26 V there is relatively
exceeds that of CO. Then by around 1.2 V the rate of little change in the product distribution. The proportion
CO production has reached its maximum rate, most likely of CO2 forming formate vs. CO changes from 1.5 times
limited by desorption. to 1.3 times and hydrogen evolution decreases by 1.6 times.
In this range, the estimated local concentration of CO2
rateCO / hCO k desorption ð22Þ
remains essentially constant and that of H+ slightly
The appearance of a maximum rate for CO production increases. This is interesting because, if two reaction path-
is therefore indicative of essentially complete coverage of ways share a common intermediate and one pathway has
the electrode surface by adsorbed CO (i.e. hCO  1) (in an electrochemical controlling step and the other does
agreement with [23]). However, the system shows an inter- not, one would expect to see a large shift in product distri-
esting feature. While one might expect the hydrogen evolu- bution over this range of 140 mV. Thus for CO 2 , the path-
tion to show its most rapid drop in rate as the coverage of ways to formate or to CO must (in this potential region)
CO nears 1, it actually occurs earlier (from 1.0 to both have controlling steps that are chemical or both elec-
1.16 V) where one would estimate hCO to be 0.33–0.70. trochemical (with similar symmetry factors). Similar con-
One possible reason for this would be surface non-homoge- clusions can be made if adsorbed hydrogen is a key
neities. If the preferred locations for initial CO adsorption shared intermediate, however, because in this case steps
were also the most active sites for hydrogen evolution, one 13 and 16 are chemical steps, this would imply that hydro-
might see the observed strong inhibition with a hCO below gen evolution occurred through step 20 (i.e. also a chemical
0.9. This would be consistent with gas phase adsorption of step).
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 15

At potentials negative of 1.3 V, the increase in subse- trode surface pH and CO2 concentrations, with the results
quent reactions of CO causes a decrease in the steady-state of our calculations shown in Fig. 18 along with calculated
surface coverage of CO as can be seen from the decrease in reaction rates from Hori’s data. The estimated decrease in
observed CO production and considering Eq. (22). This local CO2 concentration at higher pH values is driven by a
decrease in the coverage of CO exposes copper surface higher current efficiency for CO2 consuming reactions and
allowing other reactions to proceed. From 1.26 V to a higher rate of reaction (9) as the pH increases.
1.41 V the reduction of CO2 increases 5 times and hydro- Comparing the data in Fig. 18 with Fig. 17, generally
gen evolution increases 2.4 times. However, as the potential similar reaction rates are observed when comparing data
is becoming more negative, CO2 discharge is increasing and at the same potentials and the estimated local pH and
hydrogen evolution is increasing, one can see that the prod- CO2 concentration values are also similar. However, some
uct distribution shifts away from formate towards differences are notable, with the 1.5 M KHCO3 data at
CO + products. The most likely cause of this is the roughly 5 mA cm2 yielding 1.28 V vs. 0.85 mA cm2 at
12 times decrease in estimated proton concentration at the 1.26 V in Fig. 3. A second difference, not shown in the
electrode surface (the estimated [CO2] decrease is 2.1 data plotted here, is that in the 0.03 M, low strength buffer
times). This suggests that reaction steps 11 and 12 play at 1.42 V, the rate of formation of ethylene exceeds that
an important role in formate formation (and conversely of methane (unlike Fig. 3). This is at least partially due
that steps 14 and 15 are less important in CO formations). to the higher local pH slowing the rate of methane produc-
Hori et al. [20] have also carried out a galvanostatic tion according to the postulated mechanisms shown in
experiment using a range of electrolyte strengths from Fig. 7 or 8. (The lower buffer strength, higher pH data also
0.03 M to 1.5 M KHCO3, which resulted in different prod- showed significant fractions of the current producing etha-
uct distributions. The experiment used a single electrode nol and propanol.) Day-to-day variations in the electrode
and exchanged the electrolyte to avoid possible variations preparation can also contribute to differences between the
due to the electrode preparation [64]. In their paper, it is data in Figs. 3 and 17 vs. Fig. 18.
felt that the change in local pH values with the different In considering the trends shown in Fig. 18, as the esti-
buffer strengths plays a key role in the obtained product mated local concentration of protons decreases by 15.1
distributions. We have used our model to estimate the elec- times, the estimated local CO2 concentration decreases

1.0E-07
Reaction rate / mole cm-2 s-1

-1.28 V -1.4 2 V
H2 total CO2 redn.
1.0E-08

HCOO-
1.0E-09 CO + products

CO

1.0E-10
8.8 9 9.2 9.4 9.6 9.8 10 10.2
surface pH

1.0E-08 1.0E-01
[CO2]
CO2 conc. / M
H+ conc. / M

1.0E-09 1.0E-02
1.5 M [H+]
1.0 M
0.5 M 0.3 M
1.0E-10 1.0E-03
0.1 M 0.05 0.03 M
1.0E-11 1.0E-04
8.8 9 9.2 9.4 9.6 9.8 10 10.2
surface pH

Fig. 18. The influence of buffer strength on the product distribution. Data calculated from Hori et al. [20] showing the total CO2 reduction reaction rate
and CO + products rate (using a moles-of-carbon basis). (Conditions: 5 mA cm2, potential 1.28 V vs. SHE at 1.5 M varying to 1.42 V at 0.03 M
KHCO3 buffer (buffer strength indicated on the lower plot), 19 C, CO2 bubbled.) Estimated local [H+] and [CO2] values from [22] using a boundary layer
of 0.01 cm.
16 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

1.9 times. Thus, if formate was formed by reaction steps 11 strate [41]. In this work the electrolysis was started with
and 12 and CO by steps 17 and 18, one would expect the 0.5 M KHCO3 and 0.5 mM CuSO4, and run for 15 min
relative rates of production of CO + products vs. formate at constant current at 0 C prior to analysis of the cell exit
to increase by around 15.1/1.9 or 7.9 times. The actual data gas. High Faradaic efficiencies of 73% for CH4 and 25% for
shows an increase in the ratio of CO + products to formate C2H4 (hence 98% for hydrocarbons) were reported using
of 1.8 times. Thus, while the results in Fig. 17 show that 8.3 mA cm2.
formate production is sensitive to the proton concentra- The electrolyte also needs to be carefully prepared
tion, other factors must also play a role. The small change [14,21,39]. As discussed previously, because of the very
in the ratio is due to the slow increase in CO + products in cathodic potentials required for CO2 reduction, trace impu-
comparison to Fig. 17 over the same potential range. One rities can plate onto the cathode over time changing the
clear difference between the sets of data is the decrease in electrode characteristics. Typical impurities in KHCO3 salt,
hydrogen evolution with more negative potential in such as iron or zinc, even at very low concentrations, have
Fig. 18 vs. Fig. 17. This, along with the previously referred been shown to collect on the cathode surface resulting in a
to shift from methane to ethylene and the greater appear- decrease in the Faradaic efficiency for hydrocarbon forma-
ance of ethanol and propanol (less hydrogenated prod- tion [39]. This effect adds uncertainty to the interpretation
ucts), hints that adsorbed hydrogen plays a role these of much work in the literature where the problem of elec-
reactions. In other words, a lower availability of adsorbed trolyte impurities was not considered.
hydrogen may result in a slower formation of methane
(possibly because of a reaction such as in Fig. 8). This 2.7. Other factors – temperature and pressure
may also result in the appearance of partially oxygenated
products, possibly through a higher availability of CO The reaction temperature also influences the product
(i.e. the reaction in Fig. 6 continues but those in Figs. 7 distribution. In galvanostatic work over the range from 0
and/or 8 are slower). to 40 C, lower temperatures resulted in little change in
In reviewing this data it was found that no simple choice the current efficiency for formate, decreased efficiencies
of rate determining steps from among reactions (10)–(21) for hydrogen evolution and ethylene formation and an
fit the data. It was also found that different relative trends increased efficiency for methane formation (reaching 65%
occur in the data in Figs. 17 and 18. This seems to imply methane current efficiency at 0 C) [8]. Similar results have
shifting and/or parallel mechanisms operate and that the been reported potentiostatically at 2.0 V vs. Ag/AgCl/
relative balance between these mechanisms might vary saturated KCl in 0.65 M NaHCO3 over the range from
depending on the electrode preparation (not surprising in 15 C to 2 C [68] and on comparing results at 2.2 V
light of Figs. 10 and 12) and possibly the electrolyte com- vs. SCE in 0.05 M KHCO3 at 0 C and at 20 C [18]. In
position (i.e. the concentration and types of anions and cat- addition to increased current efficiency for methane, other
ions present). workers have also reported a shift in the apparent Tafel
slope for the methane partial current from 93 mV decade1
2.6. Other factors – the electrode and electrolyte at 22 C to 520 mV decade1 at 0 C [21]. As well as
changes in reaction rates, lower temperatures would also
In the work of Prof. Hori’s group, to obtain a reasonable influence adsorption equilibria and increase CO2 solubility.
degree of reproducibility, the electrodes had to be carefully Work has also been carried out where the CO2 solubility
prepared. High purity 99.999% copper [8] was used and in was raised by increasing the CO2 pressure [69]. In this
some cases electrochemically deposited foil. Typical elec- work, pre-electrolysed solutions were used and the copper
trode preparation involved mechanical polishing followed wire cathode was electropolished. The reactions were run
by electropolishing in 85% phosphoric acid. The electropol- galvanostatically and the IR corrected potential was
ishing involves driving the copper dissolution into its limit- recorded. On increasing from 0 to 60 atm of CO2 at a con-
ing current regime to remove surface irregularities and stant current of 163 mA cm2, the predominant reaction
produce a mirror-like finish surface (see for example [65]). shifted from hydrogen to hydrocarbons, reaching 44% cur-
For preparing copper single crystals mixtures of phosphoric rent efficiency by around 10 atm of CO2. On further
and sulphuric acids (which allow a higher limiting current) increasing the CO2 pressure to around 20 atm, hydrocar-
and current densities around 2 A cm2 in a pulse or pulses bons decreased to around 12% current efficiency and for-
of around 1 s are preferred to produce a good surface mate and CO rise to around 30% current efficiency. At
[66,67]. (Too long a pulse will result in oxygen evolution higher pressures formate and CO are the primary products
and the resulting bubbles can mar the electrode surface with formate production having a maximum of 54% cur-
[64].) It is felt that the electropolishing is an important step rent efficiency at 30 atm. Thus, as the authors point out,
in order to remove the stressed surface layer (produced by as increasing amounts of CO2 are fed to the reactor, less
polishing) and possible impurities from the electrode surface reduced CO2 products are formed. Higher pressure data
[66], with 30 lm thickness or more typically removed [67]. also showed a poorer charge balance (as low as 60% of
Interesting results have also been reported using in situ the applied current accounted for as products at 60 atm),
electrodeposited copper on a roughened glassy carbon sub- suggesting the possible formation of new products.
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 17

Measurements were also done at 30 atm CO2 pressure much higher current densities used (even with the assumed
using varying current densities. The reported potential thinner boundary layer). The high CO2 pressure results in
and current efficiency data has been used to calculate reac- a high local concentration of CO2 at the electrode surface.
tion rates at the various potentials, which are plotted in This is predicted to have only a moderate effect on the
Fig. 19. We have also used our boundary layer model local pH because of the slow kinetics of reaction (9). It
[22] to try to estimate the local pH and CO2 concentrations is interesting to note that even with a very high local
at the electrode surface. Because the reactor was stirred and CO2 concentration and a relatively negative potential,
there is expected to be additional mixing from the high for much of the data the rate of formation of CO + prod-
rates of gas evolution at the current densities used, a smal- ucts is less than that for formate. This would appear unli-
ler boundary layer thickness was used in the calculations. kely if reaction steps 17 and 18 were the main route to
No correction was made for the curvature of the copper CO. In fact, the low rate of CO formation appears more
wire electrode and the CO2 solubility was assumed linear likely to be linked to the low rate of hydrogen evolution
with pressure. The results of these calculations are also (supporting reaction step 16). Thus, the observation by
shown in Fig. 19. the authors of the paper that as the CO2 concentration
The main product at 30 atm and lower current densities is increased, the current needs to be increased to maintain
is formate. As the current density is increased (and the hydrocarbon production, may be at least partially due to a
potential becomes more negative), hydrogen and methane need to have sufficient coverage of adsorbed hydrogen. In
formation rapidly increase and around 600–700 mA cm2 spite of a change of 320 mV in Fig. 19, the production of
(1.57 to 1.61 V) CO + products dominates over for- formate only slightly increases, possibly due to the concur-
mate production. The estimated local concentrations show rent drop in local proton concentration (consistent with
a much higher pH than Figs. 17 and 18 because of the reaction step 11).

1.0E-05

HCOO - CO+prod.
total CO2 reductn.

Reaction rate / mole cm-2 s-1


1.0E-06

1.0E-07

H2

CO
1.0E-08

CH4

1.0E-09
-1.3 -1.42 -1.54 -1.66 -1.78
Potential / V vs. SHE

1.0E-09 1.0E+01

[CO2]
CO2 Conc. / M
H Conc. / M

1.0E-10 1.0E+00
+

1.0E-11 1.0E-01
[H+]

1.0E-12 1.0E-02
-1.3 -1.42 -1.54 -1.66 -1.78
Potential / V vs. SHE

Fig. 19. Product distribution with potential with 30 atm CO2. Reaction rate data calculated from Hara et al. [69]. (Conditions: galvanostatic from 75 to
900 mA/cm2, 0.1 M KHCO3, 30 atm CO2, stirred, 25 C.) Local [H+] and [CO2] values estimated as described in [22] using a boundary layer thickness of
0.002 cm.
18 M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19

It should be noted, however, that simply assuming the Even after 20 years [7] much work is still to be done, but
reactions follow step 11 (i.e. the rate of formate dependent this ‘‘one-pot’’ electrochemical synthesis of hydrocarbons
on the local proton concentration), step 16 (i.e. the rate of directly from CO2 represents an interesting and potentially
CO + products dependent on the adsorbed hydrogen cov- useful reaction. One can still say, as was stated in an early
erage) and step 20 (i.e. the rate of hydrogen evolution paper of Prof. Hori’s, ‘‘it is remarkable that hydrocarbons
dependent on the adsorbed hydrogen coverage squared) are produced from electrochemical reduction of CO2’’ [8].
does not fit all the reported data. The above combination
of reaction pathways should result in the relationship Acknowledgements
shown in the following equation:
pffiffiffiffiffiffiffiffiffiffiffiffiffi The authors acknowledge useful discussions with Profs.
rateCOþproducts rateH
/ þ 2 ð23Þ Hori and Koga of Chiba Univeristy, Dr. Akio Saito for-
rateHCOO ½H local
merly of the Kyushu National Industrial Research Institute
While some regions of data do show such a relationship, and Prof. Colin Oloman and Li Hui of the University of
neither this relationship nor other simple relationships that British Columbia. We also thank Prof. Koji Hashimoto
were tried were able to fit the complete set of reported data for discussions on global warming and his infectious enthu-
(i.e. Figs. 17–19). One reason for this might be a shifting of siasm for finding solutions through green chemistry. Final-
the favoured reaction pathways under different reaction ly, we thank the Canadian Office of Energy Research and
conditions. Development’s Innovative Research Initiative for funding
The reactions also show additional complexities. For this work.
example, in Fig. 19, the rate of methane production slows
its increase after about 1.55 V. In this region the CO cov- References
erage does not appear to greatly change, hydrogen evolu-
tion continues to increase, and while the local proton [1] R.P.S. Chaplin, A.A. Wragg, J. Appl. Electrochem. 33 (2003) 1107.
[2] M. Shibata, K. Yoshida, N. Furuya, J. Electroanal. Chem. 442 (1–2)
concentration is dropping, the increase in potential and
(1998) 67;
the reported transfer coefficient of 1.33 should be sufficient M. Shibata, K. Yoshida, N. Furuya, J. Electroanal. Chem. 387 (1995)
to result in a continuing increase in methane production. 143.
This suggests a new limitation on the reaction at these very [3] (a) T. Weimer, K. Schaber, M. Specht, A. Bandi, Energy Convers.
high rates, one possibility being limited turnover at active Manage. 37 (6–8) (1996) 1351;
(b) M. Specht, F. Staiss, A. Bandi, T. Weimer, Int. J. Hydrogen
sites. Again, issues such as surface heterogeneity should
Energy 23 (5) (1998) 387.
be considered in trying to understand these reactions. [4] D.A. Tryk, A. Fujishima, Electrochem. Soc. Interface 10 (1) (2001)
Indeed, the exact reaction pathways favoured in a given cir- 32.
cumstance may depend on the reaction conditions (poten- [5] D. Mignard, M. Sahibzada, J.M. Duthie, H.W. Whittington, Int. J.
tial, local pH, CO2 concentration, etc.) and on the Hydrogen Energy 28 (2003) 455.
[6] K. Hashimoto, H. Habazaki, M. Yamasaki, S. Meguro, T. Sasaki,
availability of different types of active sites. If this is the
H. Katagiri, T. Matsui, K. Fujimura, K. Izumiya, N. Kumagai, E.
case, polarisation measurements using copper single crys- Akiyama, Mater. Sci. Eng. A 304–306 (2001) 88.
tals (hopefully with a reduced set of types of active sites) [7] Y. Hori, K. Kikuchi, S. Suzuki, Chem. Lett. (1985) 1695.
might yield data that is simpler to interpret mechanistically. [8] Y. Hori, K. Kikuchi, A. Murata, S. Suzuki, Chem. Lett. (1986) 897.
Finally, these results of Hara et al. at high pressures do [9] Y. Hori, A. Murata, R. Takahashi, S. Suzuki, J. Chem. Soc., Chem.
Commun. (1988) 17.
clearly show that hydrocarbons can be produced from CO2
[10] H. Noda, S. Ikeda, Y. Oda, K. Ito, Chem. Lett. (1989) 289.
at close to 60% current efficiency at 600 mA cm2, which is [11] F.E. Lynch, R.W. Marmaro, ‘‘Special purpose blends of hydrogen
important for any practical CO2 conversion reactor. and natural gas’’, US Patent 5,139,002 (1992).
[12] I. Taniguchi, in: J.O’M. Bockris, R.E. White, B.E. Conway (Eds.),
3. Conclusions Modern Aspects of Electrochemistry, vol. 20, Plenum Press, New
York, 1989.
[13] K.W. Frese Jr., in: B.P. Sullivan, K. Krist, H.E. Guard (Eds.),
The reduction of CO2 to hydrocarbon products at cop- Electrochemical and Electrocatalytic Reactions of Carbon Dioxide,
per is a complex multi-step reaction involving shared inter- Elsevier, Amsterdam, 1993, p. 145.
mediates and multiple reaction pathways. The reaction is [14] Y. Hori, in: W. Vielstich, H.A. Gasteiger, A. Lamm (Eds.),
sensitive to the electrode surface structure, local conditions Handbook of Fuel Cells, vol. 2, Wiley, West Sussex, England, 2003,
such as pH and CO2 concentration (which are in turn influ- pp. 720–733 (Chapter 48).
[15] Y. Hori, A. Murata, Electrochim. Acta 35 (11/12) (1990) 1777.
enced by buffer strength, the degree of mass transport and [16] S. Taguchi, A. Aramata, Electrochim. Acta 39 (17) (1994) 2533.
the bulk CO2 concentration/pressure) and temperature. [17] Y. Hori, H. Wakebe, T. Tsukamoto, O. Koga, Electrochim. Acta 39
Other factors, such as the type and amount of other ions (11–12) (1994) 1833.
in solution, may also effect the product distribution. Spe- [18] M. Azuma, K. Hashimoto, M. Hiramoto, M. Watanbe, T. Sakata, J.
cific and significant influence has been noted of the local Electrochem. Soc. 137 (6) (1990) 1772.
[19] J.A. Dean (Ed.), Lange’s Handbook of Chemistry, 15th ed.,
proton concentration on formate production and of hydro- McGraw-Hill, New York, 1999.
gen evolution (presumably through the action of adsorbed [20] Y. Hori, A. Murata, R. Takahashi, J. Chem. Soc., Faraday Trans. 1
hydrogen intermediate) on CO formation. 85 (8) (1989) 2309.
M. Gattrell et al. / Journal of Electroanalytical Chemistry 594 (2006) 1–19 19

[21] J.J. Kim, D.P. Summers, K.W. Frese Jr., J. Electroanal. Chem. 245 K. Watanabe, U. Nagashima, H. Hosoya, Chem. Phys. Lett. 209 (1–
(1988) 223. 2) (1993) 109.
[22] N. Gupta, M. Gattrell, B. MacDougall, J. Appl. Electrochem. 36 [43] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, Internet
(2006) 161. Version, CRC Press, Boca Raton, FL, 2005. Available from: <http://
[23] Y. Hori, A. Murata, Y. Yoshinami, J. Chem. Soc., Faraday Trans. 87 www.hbcpnetbase.com>.
(1) (1991) 125. [44] Y. Hori, I. Takahashi, O. Koga, N. Hoshi, J. Mol. Catal. A 199
[24] F.R. Keene, in: B.P. Sullivan, K. Krist, H.E. Guard (Eds.), (2003) 39.
Electrochemical and Electrocatalytic Reactions of Carbon Dioxide, [45] M. Head-Gordon, J.C. Tully, Phys. Rev. B 46 (3) (1992) 1853.
Elsevier, Amsterdam, 1993, p. 1. [46] M. Head-Gordon, J.C. Tully, Chem. Phys. 175 (1993) 37.
[25] D.A. Palmer, R. Van Eldik, Chem. Rev. 83 (1983) 651. [47] G. Witte, Surf. Sci. 502–503 (2002) 405.
[26] I. Takahashi, O. Koga, N. Hoshi, Y. Hori, J. Electroanal. Chem. 533 [48] Y. Hori, H. Wakebe, T. Tsukamoto, O. Koga, Surf. Sci. 335 (1995)
(2002) 135. 258.
[27] Y. Hori, A. Murata, R. Takahashi, S. Suzuki, J. Am. Chem. Soc. 109 [49] Y. Hori, A. Murata, K. Kikuchi, S. Suzuki, J. Chem. Soc., Chem.
(1987) 5022. Commun. (1987) 728.
[28] D.W. DeWulf, T. Jin, A.J. Bard, J. Electrochem. Soc. 136 (6) (1989) [50] H. Noda, S. Ikeda, A. Yamamoto, H. Einaga, K. Ito, Bull. Chem.
1686. Soc. Jpn. 68 (1995) 1889.
[29] R.L. Cook, R.C. MacDuff, A.F. Sammells, J. Electrochem. Soc. 136 [51] N. Hoshi, M. Kato, Y. Hori, J. Electroanal. Chem. 440 (1997) 283.
(7) (1989) 1982. [52] H.-J. Freund, M.W. Roberts, Surf. Sci. Reports 25 (1996) 225.
[30] Y. Hori, O. Koga, Y. Watanabe, T. Matsuo, Electrochim. Acta 44 [53] W. Akemann, A. Otto, Surf. Sci. 287/288 (1993) 104.
(1998) 1389. [54] J. Krause, D. Borgmann, G. Wedler, Surf. Sci. 347 (1996) 1.
[31] J. Pritchard, Surf. Sci. 79 (1979) 231. [55] P.J. Godowski, J. Onsgaard, S.V. Hoffmann, S. Quist-Sckerl, Vacuum
[32] M.J. Weaver, S. Zou, C. Tang, J. Chem. Phys. 111 (1) (1999) 368. 63 (1–2) (2001) 257.
[33] Y. Hori, O. Koga, H. Yamazaki, T. Matsuo, Electrochim. Acta 40 [56] J. Onsgaard, L. Bech, C. Svensgaard, P.J. Godowski, S.V. Hoffmann,
(1995) 2617. Prog. Surf. Sci. 67 (1–8) (2001) 205.
[34] I. Oda, H. Ogasawara, M. Ito, Langmuir 12 (1996) 1094. [57] I.A. Bonicke, W. Kirstein, F. Thieme, Surf. Sci. 307–309 (1994) 177.
[35] B.D. Smith, D.E. Irish, P. Kedzierzawski, J. Augustynski, J. [58] S.S. Fu, G.A. Somorjai, Surf. Sci. 262 (1992) 68.
Electrochem. Soc. 144 (12) (1997) 4288. [59] H.-J. Freund, R.P. Messer, Surf. Sci. 172 (1986) 1.
[36] A.J. McQuillan, P.J. Hendra, M. Fleischmann, J. Electroanal. Chem. [60] J.O’M. Bockris, N. Pentland, Trans. Faraday Soc. 48 (1952) 833.
65 (1975) 933. [61] V.I. Kichigin, N.I. Kavardakov, Elektrokhimiya 30 (8) (1994) 1008
[37] Y. Ichinohe, T. Wadayama, A. Hatta, J. Raman Spectrosc. (26) (Russian J. Electrochem., p. 911).
(1995) 335. [62] T.D. Hewitt, D. Roy, Chem. Phys. Let. 181 (5) (1991) 407.
[38] J.M. Phillips, F.M. Leibsle, A.J. Holder, T. Keith, Surf. Sci. 545 [63] M.-C. Marinica, H. Le Rouzo, G. Raseev, Surf. Sci. 542 (2003) 1.
(2003) 1. [64] Y. Hori, O. Koga, Private communication.
[39] Y. Hori, H. Konishi, T. Futamura, A. Murata, O. Koga, H. Sakurai, [65] B. Du, I.I. Suni, J. Electrochem. Soc. 151 (6) (2004) C375.
K. Oguma, Electrochim. Acta 50 (2005) 5354. [66] Y. Hori, I. Takahashi, O. Koga, N. Hoshi, J. Phys. Chem. B 106
[40] Y. Hori, R. Takahashi, Y. Yoshinami, A. Murata, J. Phys. Chem. B (2002) 15.
101 (1997) 7075. [67] H. Siegenthaler, K. Juttner, J. Electroanal. Chem. 163 (1984) 327.
[41] R.L. Cook, R.C. MacDuff, A.F. Sammells, J. Electrochem. Soc. 135 [68] S. Kaneco, N.-H. Hiei, Y. Xing, H. Katsumata, H. Ohnishi, T.
(6) (1988) 1320. Suzuki, K. Ohta, Electrochim. Acta 48 (2002) 51.
[42] K. Watanabe, U. Nagashima, H. Hosoya, Appl. Surf. Sci. 75 (1994) [69] K. Hara, A. Tsuneto, A. Kudo, T. Sakata, J. Electrochem. Soc. 141
121; (8) (1994) 2097.

You might also like