Homogeneous electron transfer catalysis of the electrochemical reduction of carbon dioxide. Do aromatic anion radicals react in an outer-sphere manner?

Article abstract of DOI:10.1021/ja960605o

Electrochemically generated anion radicals of aromatic nitriles and esters possess the remarkable property to reduce carbon dioxide to oxalate with negligible formation of carboxylated products. They may thus serve as selective homogeneous catalysts for the reduction of CO₂ in an aprotic medium. The catalytic enhancement of the cyclic voltammetric peaks of these catalysts is used to determine the rate constant of the electron transfer from these aromatic anion radicals to CO₂ as a function of the catalyst standard potential. Substituted benzoic esters allowed a particularly detailed investigation of the resulting activation-driving force relationship. Using 14 different catalysts in this series made it possible to finely scan a range of reaction standard free energies of 0.4 eV. Detailed analysis of the resulting data leads to the conclusion that the reaction is not a simple outer-sphere electron transfer. It rather consists in a nucleophilic addition of the anion radical on CO₂, forming an oxygen (or nitrogen for the nitriles) - carbon bond, which successively breaks homolytically, generating the parent ester (or nitrile) and the anion radical of CO₂, which eventually dimerizes to oxalate.

Full text of DOI:10.1021/ja960605o

+
+
7190
J. Am. Chem. Soc. 1996, 118, 7190 7196
Homogeneous Electron Transfer Catalysis of the
Electrochemical Reduction of Carbon Dioxide. Do Aromatic
Anion Radicals React in an Outer-Sphere Manner?
Armando Gennaro,1a Abdirisak A. Isse,1a Jean-Michel Save´ant,*,1b
Maria-Gabriella Severin,1a and Elio Vianello*,1a
Contribution from the Dipartimento di Chimica Fisica, Uni ersita` di Pado a, Via Loredan 2,
35131 Pado a, Italy, and Laboratoire d’Electrochimie Mole´culaire de l’Uni ersite´ Denis Diderot
(Paris 7), 2 place Jussieu, 75251 Paris Cedex 05, France
Recei ed February 26, 1996. Re ised Manuscript Recei ed April 17, 1996X
Abstract: Electrochemically generated anion radicals of aromatic nitriles and esters possess the remarkable property
to reduce carbon dioxide to oxalate with negligible formation of carboxylated products. They may thus serve as
selective homogeneous catalysts for the reduction of CO2 in an aprotic medium. The catalytic enhancement of the
cyclic voltammetric peaks of these catalysts is used to determine the rate constant of the electron transfer from these
aromatic anion radicals to CO2 as a function of the catalyst standard potential. Substituted benzoic esters allowed
a particularly detailed investigation of the resulting activation-driving force relationship. Using 14 different catalysts
in this series made it possible to finely scan a range of reaction standard free energies of 0.4 eV. Detailed analysis
of the resulting data leads to the conclusion that the reaction is not a simple outer-sphere electron transfer. It rather
consists in a nucleophilic addition of the anion radical on CO2, forming an oxygen (or nitrogen for the nitriles)
carbon bond, which successively breaks homolytically, generating the parent ester (or nitrile) and the anion radical
of CO2, which eventually dimerizes to oxalate.
The continuous attention the chemistry of carbon dioxide
such as alkyl halides, aryl alkyl sulfides,5 and also CO2. In a
number of cases, the reaction with CO2 does not give rise to
catalysis but rather leads to carboxylation of the electron donor,
a reaction that may be of synthetic interest.2a Monocarboxylates
and/or dicarboxylates have thus been obtained from activated
olefins6 and polycyclic aromatic hydrocarbons.7 N-heteroaro-
matic molecules similarly yield the corresponding dihydrocar-
boxy derivatives8 while -hydroxy and -amino acids are
produced from carbonyl compounds9 and imines, respectively.10
In these reactions, CO2 has been considered to react with the
anion radical either as an electron pair acceptor or as a single
electron acceptor.
(5) (a) For alkyl halides see refs 5b and 5c c and references therein. For
aryl alkyl sulfides see ref 5d and references therein. (b) Save´ant, J.-M. Single
Electron Transfer and Nucleophilic Substitution. In Ad ances in Physical
Organic Chemistry; Bethel, D., Ed.; Academic Press: New York, 1990;
Vol. 26, pp 1 130. (c) Bertran, J.; Gallardo, I.; Moreno, M.; Save´ant, J.-
M. Submitted for publication. (d) Severin, M. G.; Arevalo, M. C.; Maran,
F.; Vianello, E. J. Phys. Chem. 1993, 97, 150.
(6) (a) Tyssee, D. A.; Wagenknecht, J. H.; Baizer, M. M.; Chruma, J. L.
Tetrahedron Lett. 1972, 47, 4809. (b) Tyssee, D. A.; Baizer, M. M. J. Org.
Chem. 1974, 39, 2819. (c) Lamy, E.; Nadjo, L.; Save´ant, J.-M. Nou . J.
Chim. 1979, 3, 21. (d) Gambino, S.; Filardo, G.; Silvestri, G. J. Appl.
Electrochem. 1982, 12, 549. (e) Gambino, S.; Gennaro, A.; Filardo, G.;
Silvestri, G.; Vianello, E. J. Electrochem. Soc. 1987, 134, 2172.
(7) (a) Wawzonek, S.; Wearing, D. J. Am. Chem. Soc. 1959, 81, 2067.
(b) Ticianelli, E. A.; Avaca, L. A.; Gonzalez, E. R. J. Electroanal. Chem.
1989, 258, 369. (c) Ticianelli, E. A.; Avaca, L. A.; Gonzalez, E. R. J.
Electroanal. Chem. 1989, 258, 379.
attracts derives mostly from its abundance on earth as a carbon
source and from the central role it plays in life processes. At
the same time, the chemistry of CO2 raises fundamental issues
related to its single electron and electron pair acceptor properties,
whatever the probabilities of its practical use as a cheap carbon
source in the near future. Direct electrochemical reductive
activation of CO2 runs into the difficulty that a very negative
potential (beyond 2.2 V vs SCE in an aprotic medium) is
required. This fact has been an incentive for many attempts to
catalyze the electrochemical reduction of CO2. Molecules of
particular interest in this respect are reduced states of transition
metal complexes where electron transfer to CO2 and the ensuing
chemical steps are anticipated to take place within the metal
4
coordination sphere.2
Anion radicals of unsaturated organic compounds, generated
electrochemically in aprotic media, are expected to be single
electron donors. They have been reacted with several acceptors
X Abstract published in Ad ance ACS Abstracts, July 1, 1996.
(1) (a) Universita` di Padova. (b) Universite´ Denis Diderot.
(2) (a) For reviews see refs 2b and 2c and also the introduction of refs
3b d. (b) Silvestri, G. In Carbon Dioxide as a Source of Carbon; Aresta,
M., Forti, G., Eds.; NATO ASI Series C; Reidel: Dordrecht, 1987; p 339.
(c) Collin, J. P.; Sauvage, J. P. Coord. Chem. Re . 1989, 93, 245.
(3) (a) Hammouche, M.; Lexa, D.; Save´ant, J.-M.; Momenteau, M. J.
Electroanal. 1988, 249, 347. (b) Hammouche, M.; Lexa, D.; Momenteau,
M.; Save´ant, J.-M. J. Am. Chem. Soc. 1991, 113, 8455. (c) Bhugun, I.;
Lexa, D.; Save´ant, J.-M. J. Am. Chem. Soc. 1994, 116, 5015. (d) Bhugun,
I.; Lexa, D.; Save´ant, J.-M. J. Am. Chem. Soc., in press.
(8) (a) Hess, U.; Fuchs, P.; Jacob, E.; Lund, H. Z. Chem. 1980, 20, 64.
(b) Fuchs, P.; Hess, U.; Holst, H. H.; Lund, H. Acta Chem. Scand., Ser. B
1981, 35, 185.
(4) (a) The transition metal CO2 reduction catalysts that have received
the most active attention are Ni and Co cyclams, rhenium carbonyl, and
rhodium, iridium, osmium, and ruthenium bipyridine complexes (see the
introduction of refs 3b d) as well as cobalt(I) salen complexes.4b,c Iron(0)
porphyrins are particularly efficient catalysts when combined with Bro¨nsted
or Lewis acid synergists.3 (b) Gennaro, A.; Isse, A. A.; Vianello, E.; Floriani,
C. J. Mol. Catal. 1991, 70, 197. (c) Gennaro, A.; Isse, A. A.; Vianello, E.
in Electrochemitstry of Inorganic, Bioinorganic and Organometallic
Compounds, Pombeiro, A. J. L., McCleverty, J. A., Eds.; Kluwer:
Amsterdam, 1993; p 311.
(9) (a) Wawzonek, S.; Gunderson, A. J. Electrochem. Soc. 1960, 107,
537. (b) Harada, J.; Y. Sakakibara, Y.; Kunai, A.; Sasaki, K. Bull. Chem.
Soc. Jpn. 1984, 57, 611. (c) Ikeda Y.; Manda, E. Bull. Chem. Soc. Jpn.
1985, 58, 1723. (d) Silvestri, G.; Gambino, S.; Filardo, G. Tetrahedron
Lett. 1986, 27, 3429. (e) Bulhoes L. O. D. S.; Zara, A. J. J. Electroanal.
Chem. 1988, 248, 159. (f) Mcharek, S.; Heintz, M.; Troupel, M.; Perichon,
J. Bull. Soc. Chim. Fr. 1989, 95. (g) Chan, A. S. C.; Huang, T. T.;
Wagenknecht, J. H.; Miller, R. E. J. Org. Chem. 1995, 60, 742.
(10) (a) Hess, U.; Thiele, R. J. Prakt. Chem. 1982, 324, 385. (b) Silvestri,
G.; Gambino, S.; Filardo, G. Gazz. Chim. Ital. 1988, 118, 643.
S0002-7863(96)00605-1 CCC: $12 00 © 1996 American Chemical Society

+
+
Electrochemical Reduction of Carbon Dioxide
J. Am. Chem. Soc., Vol. 118, No. 30, 1996 7191
Table 1. Homogeneous Catalysts: Designation, Standard
A
CO2 h ACO2
a
Potentials, and Rate Constants of the Reaction with CO2
E°cat
log k
1
A
CO2 h A CO2
catalyst
designation (V vs SCE) (M 1 s
)
dimethyl phthalate
diisobutyl phthalate
dibutyl phthalate
methyl 4-phenylbenzoate
phenyl benzoate
phenyl 3-methylbenzoate
ethyl 3-fluorobenzoate
methyl 3-phenoxybenzoate
phenyl 4-methylbenzoate
methyl benzoate
1E
2E
3E
4E
5E
6E
7E
8E
1.928
1.948
1.958
1.962
2.026
2.036
2.043
2.085
2.094
2.202
2.221
2.237
2.276
2.290
1.957
2.260
2.297
0.50
0.82
0.90
1.06
2.15
2.46
2.43
3.01
3.08
4.19
4.21
4.44
4.74
5.04
1.24
5.31
5.60
In the former case, the carboxylated anion radical thus formed
is further reduced to the dianion which is then protonated or
carboxylated a second time:
2
ACO2
e (and/or A ) h ACO2 ( A)
2
2
9E
ACO2
CO2 (H ) f A(CO2)2 (AHCO2 )
10E
11E
12E
13E
14E
1N
2N
3N
ethyl benzoate
2
methyl 3-methylbenzoate
methyl 2-methylbenzoate
methyl 4-methylbenzoate
4-cyanobiphenyl
benzonitrile
2-tolunitrile
In the second case, ACO2 would be formed by coupling
between the two anion radicals
2
CO2
A
h ACO2
a In DMF 0.1 M n-Bu4NClO4. Temperature 25 °C.
before undergoing a second carboxylation or protonation as in
the preceding case. Whatever its exact mechanism, mono- and
dicarboxylation reactions correspond to a stoichiometry of two
electrons per A molecule.
Carboxylation of A is not necessarily the only result of the
reaction of the anion radicals of unsaturated organic compounds
with CO2. In several instances, the enhancement of the one-
electron wave for the generation of the anion radical upon
addition of CO2 has been observed to overpass a 2e/molecule
stoichiometry. In such cases, the current response takes a
catalytic character, implying that products stemming from CO2
alone, such as oxalate, formate, CO ( carbonate), are formed
instead of (or in addition to) carboxylation products. Such
catalytic currents have been observed with chrysene,7b,11a
picene,7b and benzonitrile.11b However, with the exception of
benzonitrile where the formation of oxalate has been observed,11b
CO2 self-derived products have not been investigated.
The photochemistry of aromatic molecules in the presence
of CO2 has also been investigated as another means of triggering
electron transfer. Photocarboxylation of several aromatic
hydrocarbons, in the presence of various amines used as
sacrificial electron donors, have thus been described.12 Pho-
tocatalysis has been reported to occur to a small extent with
anthracene, phenanthrene, and pyrene, and more efficiently with
oligo(p-phenylenes), containing two five monomer units, where
formate is the main reaction product (the proton necessary for
the formation of the C H bond presumably comes from the
oxidation of the sacrificial electron donor, usually an amine).13
We have found that substituted benzonitriles as well as a large
variety of substituted and unsubstituted alkyl or phenyl benzoates
catalyze the electrochemical reduction of CO2 in N,N-dimeth-
ylformamide (DMF) with exclusive formation of oxalate. Under
such circumstances, the analysis of the kinetics of electron
transfer forming the CO2 anion radical is greatly simplified.
The use of an extended series (of 14 members) of benzoates
then allowed a detailed investigation of the dynamics of the
reaction. Whether electron transfer between these anion radicals
and CO2 is of the outer-sphere type or involves more intimate
Figure 1. Cyclic voltammetry of 10E (2.6 mM) in DMF
0.1 M
n-Bu4NClO4 in the absence (dotted line) and presence (solid line) of
CO2 (26 mM). The current scale is normalized toward the cathodic
peak current in the absence of CO2, ip°. Scan rate 1 V/s. Temperature
25 °C.
chemical interactions is a question that we will address on the
basis of the experimentally established activation driving force
relationship.
Results
The various ester and nitrile catalysts that we have investi-
gated are listed in Table 1 together with the conventional
designations that will be used throughout the paper. In the
absence of CO2, these compounds all give rise to a reversible
cyclic voltammetric wave. The standard potentials, E°cat, that
can be obtained from the midpoint between the cathodic and
anodic peaks are important parameters for the following
discussion. Their values, referenced to the aqueous SCE, are
listed in Table 1.
Figure 1 shows a typical example of the reversible wave
obtained with one of the ester catalysts, 10E, and of the catalytic
current observed upon addition of CO2. Catalysis results in a
loss of reversibility and an enhancement of the cathodic peak
current.
(11) (a) Lund, H.; Simonet, J. J. Electroanal. Chem. 1975, 55, 205. (b)
Filardo, G.; Gambino, S.; Silvestri, G.; Gennaro, A.; Vianello, E. J.
Electroanal. Chem. 1984, 177, 303.
(12) (a) Tazuke, S.; Ozawa, H. J. Chem. Soc., Chem. Commun. 1975,
237. (b) Tazuke, S.; S. Kazama, S.; Kitamura, N. J. Org. Chem. 1986, 51,
4548. (c) Ito, Y.; Uozu, Y.; Matsuura, T. J. Chem. Soc., Chem. Commun.
1988, 562.
(13) (a) Matsuoka, S.; Kohzuki, T.; Pac, C.; Ishida, A.; Takamuku, S.;
Kusaba, M.; Nakashima, N.; Yanagida, S. J. Phys. Chem. 1992, 96, 4437.
(b) Matsuoka, S.; Kohzuki, T. Pac, C.; Yanagida, S. Chem. Lett. 1990, 2047.
Constant current electrolyses were carried out with several
of these catalysts in the presence of CO2 at potentials where

+
+
7192 J. Am. Chem. Soc., Vol. 118, No. 30, 1996
Gennaro et al.
Table 2. Preparative Scale Electrolysis of CO2 Catalyzed by
over the whole temperature range explored (Table 2) since
practically no CO is found among the electrolysis products even
when the CO2 concentration is much higher than the steady-
state concentration of CO2 , thus favoring the occurrence of
reaction ad. The competition between steps ad r and step d
may be described quantitatively by the dimensionless parameter
Aromatic Ester and Nitrile Anion Radicalsa
faradaic yields
e
E°cat
catalyst (V vs SCE) [cat] (mM) [CO2] (mM) oxalate
CO
none
noneb
5E
152
102
114
101
51
102
152
262
102
102
101
21
0.672
0.419
0.790
0.840
0.969
0.937
0.930
0.997
0.928
0.902
0.740
0.800
0.253
0.483
0.001
0.001
0.001
0.005
0.008
0.003
0.004
0.009
0.001
0.007
Kadkrkd 1/2k 3/4[CO2]1/4(I/FD1/2 1/2
(I is the current density and
)
2.026
2.202
2.202
2.202
2.202
2.202
2.202
2.202
2.260
2.260
1.14
1.63
1.53
1.54
1.47
1.39
1.33
1.54
1.84
2.10
D the diffusion coefficient of CO2).15a This should be less than
10 2 for the yield of CO to be smaller than 1%.15a In the results
displayed in Table 2, the most favorable conditions for CO
production involve catalyst 5E. The experimental conditions
(Table 2 and the Experimental Section) were such that I/FD1/2
219 M 1 dm3 s1/2; k 1.4 × 102 M 1 s 1 (Table 1). kd can
be estimated as 5 × 108 M 1 s 1 from previous high scan rate
cyclic voltammetric experiments on the direct reduction of CO2
in DMF at a mercury electrode.15b,c Thus, Kad kr e 103 s 1. A
rough estimate of kr can be derived from earlier electron
photoinjection experiments in acetonitrile.16a The reduction
potential of the CO2 CO2 adduct thereof appears to be ca.
240 mV positive of the reduction potential of CO2,16b,c implying
that kr is roughly 2 orders of magnitude larger than k. Reaction
ad would thus be slightly uphill (Kad e 0.07 M 1).
10E
10Eb
10Eb
10Eb
10Eb
10Ec
10Ed
2N
2N
a In DMF
0.1 M n-Bu4NCLO4 at a mercury pool electrode.
Current density 1.6 mA/cm2. Temperature is 25 °C unless otherwise
stated. b Temperature 0 °C. c Temperature 10 °C. d Temperature 10
°C. e At 25 °C.
Scheme 1
A• –
A + e–
• –
A• – + CO2
A + CO2
(et)
(d)
The kinetics of reaction et may be derived from the relative
enhancement of the peak current of the catalyst triggered by
the addition of CO2 to the solution, ip/ip° as exemplified in Figure
1. Figure 2 summarizes, in a typical case, the procedure that
we followed for measuring the rate constant k of reaction et.
The theoretical variations of ip/ip° were computed as a function
of the rate dimensionless parameter (RT/F)(k/ ) by a finite
difference method for the mechanism depicted in Scheme 1
where the et step was regarded as rate determining.17 For given
values of the catalyst concentration [cat] and of the ratio [CO2]/
[Cat], the parameter (RT/F)(k/ ) was varied by changing the
scan rate. The value of k was determined from the best fitting
of the data points with the theoretical curve. For each catalyst,
several sets of such experiments were carried out with different
values of [cat] and of the ratio [CO2]/[cat] as illustrated in Figure
2. The experimental values of ipip° and the curve-fitting
procedure are given in the supporting information for the whole
series of ester (14) and nitrile (3) anion radicals. The resulting
average values of log k ( 0.1) are listed in Table 1.
O
O
• –
2CO2
C
C
O
O
only the catalytic reduction takes place (the effective reduction
potential was found to be slightly more positive than the standard
potential of the catalyst). The results are displayed in Table 2.
Analysis of products showed that oxalate is practically the only
product formed whatever the temperature and the concentrations
of catalyst and CO2. Cyclic voltammetric examination of the
solution after electrolysis revealed that the consumption of the
catalyst after 30 turnovers is small, less than 30% (see the
Experimental Section).
These results suggest the mechanism shown in Scheme 1.
The CO2 anion radicals formed from the ester or nitrile anion
radicals in the et step undergo a fast dimerization (d) to oxalate.
The formation of a carbon oxygen adduct between CO2 and
CO2 (ad) should however be envisaged in view of the Lewis
The rate constant k diminishes as the standard potential of
the catalyst becomes less and less negative as a consequence
of the attending increase of the standard free energy of the
reaction. A parallel increase of the reverse et reaction is
expected. It is thus conceivable that, on going to less and less
C•
O
O
CO2• – + CO2
O
C
(ad)
O–
acid properties of CO2 and the basic properties of CO2 . This
intermediate has been previously invoked14d to explain the
f
(15) (a) Isse, A. A.; Save´ant, J.-M.; Severin, M. G. J. Electroanal. Chem.
1995, 399, 157. (b) In the original publication15c a rough estimate of 107
competitive formation of oxalate and CO in direct electrochemi-
cal reduction at inert electrodes such as mercury and lead.14
Formation of CO and of an equimolar amount of carbonate was
deemed to occur through reduction of this carbon oxygen
1
1
M
s
was given. A more accurate simulation of the cyclic voltammo-
grams, taking account of the large double layer charging current and of the
fact than the scan is reverted not far from the foot of the cathode, gives a
1
value of 5 × 108
M
s
1. (c) Lamy, E.; Nadjo, L.; Save´ant, J.-M. J.
adduct.14d While this reaction may take place at the electrode
f
Electroanal. Chem. 1977, 78, 403.
(16) (a) Vassiliev, Y. B.; Bagotsky, V. S.; Khasova, O. A.; Mayorova,
N. A. J. Electroanal. Chem. Interfacial Electrochem. 1985, 189, 295. (b)
This estimation of the standard potential may seem at variance with an
assumption often made in molecular electrochemistry, namely, that a radical
deriving from the association of the primary anion radical with a Lewis or
Bro¨nsted acid is much easier to reduce than the starting molecule. The
reduction rate constant should then be close to the diffusion limit. In fact,
it may well be that the secondary radical is only moderately easier to reduce
than the starting molecule or even harder to reduce. A striking example of
the latter case is provided by the reduction of benzaldehyde in ethanolic
media where it has been shown that the protonated anion radical is more
difficult to reduce than benzaldehyde.16c (c) Andrieux, C. P.; Grzeszczuk,
M.; Save´ant, J.-M. J. Am. Chem. Soc. 1991, 113, 8811.
surface, its counterpart in the solution (r), where the electron
O
O
C•
2–
O
C
+ A• –
A + CO + CO3
(r)
O–
donor would be A seems unable to compete with dimerization
(14) (a) Kaiser, U.; Heitz, E. Ber. Bunsen-Ges. Phys. Chem. 1973, 77,
818. (b) Gressin, J. C.; Michelet, D.; Nadjo, L.; Save´ant, J.-M. Nou . J.
Chim. 1979, 3, 545. (c) Fisher, J.; Lehmann, T.; Heitz, E. J. Appl.
Electrochem. 1981, 11, 743. (d) Amatore, C.; Save´ant, J.-M. J. Am. Chem.
Soc. 1981, 103, 5021. (e) Amatore, C.; Save´ant, J.-M. J. Electroanal. Chem.
1981, 125, 22. (f) Amatore, C.; Nadjo, L.; Save´ant, J.-M. Nou . J. Chim.
1984, 8, 565.
(17) (a) Andrieux, C. P.; Save´ant, J.-M. Electrochemical Reactions. In
In estigations of Rates and Mechanisms; Bernasconi, C. F., Ed.; Wiley:
New York, 1986; Vol. 6, 4/E, Part 2, p 305. (b) Andrieux, C. P.; Hapiot,
P.; Save´ant, J.-M. Chem. Re . 1990, 90, 723.

+
+
Electrochemical Reduction of Carbon Dioxide
J. Am. Chem. Soc., Vol. 118, No. 30, 1996 7193
Figure 3. Variation of the rate constant of electron transfer between
CO2 and ester (O) and nitrile (1) anion radicals with the standard free
energy of the reaction (in DMF 0.1 M n-Bu4NCLO4). Temperature
25 °C. For the definition of lines a and b, see the text.
Discussion
Figure 3 shows how the rate constant of the electron transfer
reaction varies with the Figure 3 & standard free energy of the
reaction (∆G° (eV) E°cat E°CO (V) with E°CO
2.21 V
2
2
vs SCE) in the ester and nitrile series. Concerning the ester
series, we note that on the right-hand side of the plot, log k
varies with ∆G° by ca. 1/(120 mV) (line a) as expected for a
simple electron transfer reaction in a driving force range around
zero where the symmetry factor is close to 0.5. However, on
the left-hand side, the average slope (line b) is closer to 1/(60
mV). The backward et reaction may be controlled jointly by
activation and diffusion. Thus, the rate constant for the forward
step would be expressed by eq 1 (kD is the diffusion-limited
Figure 2. Cyclic voltammetry of 10E in DMF 0.1 M n-Bu4NClO4.
Determination of the rate constant, k, of reaction et from the ratio of
the peak currents in the presence (ip) and absence (ip°) of CO2.
Temperature 25 °C. The solid lines are the theoretical curves with a
value of k corresponding to the best fitting. The experimental conditions
corresponding to the data points are as defined below:
1
k
1
1
(1)
[CO2]/[cat] data pts [cat] (mM)
scan rate (V/s)
kact KkD
4
O
4
1
O
4
1
1.27
2.60
4.62
1.27
2.60
4.62
50, 35, 20, 10, 5, 3.5
50, 35, 20, 10, 5, 3.5, 2, 1
50, 35, 20, 10, 5, 3.5, 2, 1, 0.5
50, 35, 20, 10, 5, 3.5
50, 35, 20, 10, 5, 3.5, 2, 1
50, 35, 20, 10, 5, 3.5, 2
bimolecular rate constant). Such mixed activation diffusion-
controlled kinetics have been observed previously in reactions
of aromatic anion radicals with aryl halides5b and aryl methyl
sulfides.5d A possible interpretation of the observed variation
of k with ∆G° in Figure 3 could thus be that, in the high driving
10
negative values of the standard potential of the catalyst, the
kinetics pass from control by forward reaction et to partial
control by reaction d. More precisely, the passage from the
former situation to the latter is governed by the increase of the
dimensionless parameter K2(kd/k) where k is the forward rate
constant for the et step, K its equilibrium constant (RT/F) ln K
force zone (a), activation control predominates (k
kact),
whereas, in the low driving force zone (b), the reverse reaction
is under diffusion control. Within this framework, line a in
Figure 3 was drawn as the best fitting Marcus parabola (eq 2)
q
q 2
log k log A (F/RT)∆G0 (1 ∆G°/4∆G0 )
(2)
E°CO E°cat and kd the dimerization rate constant. Simulation
2
of the theoretical ip/ip° variations showed that when the
dimensionless parameter is smaller than 0.025, the interference
of the dimerization in the kinetics ceases to be negligible. Under
these conditions, the fitting of the data points was found to be
less satisfactory than it is in the case where forward et is
regarded as the rate-determining step. The fact that the forward
et reaction is the rate-determining step implies that the overall
dimerization rate constant is equal to or larger than 3 × 108
through the right-hand side points with a preexponential factor
q
A
3 × 1011
M
1 s 1 and an intrinsic barrier free energy ∆G0
0.439 eV.5a The latter value is qualitatively reasonable in
view of solvent reorganization and of the change in geometry
(from linear to bent18) attending electron transfer to the CO2
molecule.
Under such conditions, it is also predicted that the linear
extrapolation of the 1/(60 mV) line should reach k kD for K
1
1
M
s
s
1. This condition is fulfilled since kd 5 × 108 M
1, i.e., for ∆G°
0, meaning that within this endergonic
1
region the reverse reaction would be diffusion-controlled. As
as discussed earlier.
seen in Figure 3, the value of k for ∆G° 0 is only 1.8 × 105
1
(18) Ovenall, D. W.; Whiffen, D. H. Mol. Phys. 1961, 4, 135.
M M
1 s 1, much smaller than the diffusion limit (kD 1010

+
+
7194 J. Am. Chem. Soc., Vol. 118, No. 30, 1996
1). However, it may be conceived that the diffusion-controlled
formation of the successor complex from the separated products
would be slowed by an unfavorable work term. We will discuss
this point starting from the scheme below (A and B represent
the ester or nitrile molecules and their anion radicals, respec-
tively, and P stands for CO2 and Q for CO2 ):
Gennaro et al.
s
even if the formation of the successor complex from the
separated products involves an unfavorable work term.
The upper pathways for the formation of the precursor and
successor complexes correspond to long distance interactions
(the interactions remain the same over the whole (mutual)
diffusion layer). This is the situation where the effect of
unfavorable work terms on the formation of the precursor and
R
P
kD
ke
kD
R
successor complexes would be maximal. Then kD
KRkD and
kD for
A
Q
AQ
BP
B
P
P
R
R
P
P
k
kD
kD
e
kD
kD for the reactants and kD
KPkD and kD
the products. In other words, an unfavorable work term slows
the decomposition of the precursor and successor complexes
into the separated reactants and products, respectively, while it
leaves unchanged the rate of formation of the precursor and
successor complexes from the separated reactants and products,
respectively. Thus, the rate constant, k , for the endergonic
conversion of A Q into B P is expressed as
According to the Debye Smoluchowski model19
R
kD
kD
R
R
kD
with kD
KR
FwR(x)
d
∞exp
dx
∫
[
]
x2
d
RT
P
kD
kD
F
RT
P
P
k
kDKRKe kDKP exp
(E°cat E°
)
CO2
kD
with kD
[
]
KP
FwP(x)
d
∞exp
dx
x2
∫
[
]
d
RT
while k
kDkP (the rate constant for the reverse reaction is
smaller than kD if the work term for the formation of the
successor complex from the separated products is unfavorable).
The interpretation of the 1/(60 mV) straight line in Figure 3
along those lines would thus imply a long-distance repulsive
interaction worth ca. 0.3 eV between the ester molecule and
CO2 . This is extremely unlikely; if any, the only interaction
one could envision is a slight attractive interaction between the
negative charge in CO2 and the polarizable ester molecule.
These observations rule out the possibility that the 1/(60 mV)
line in the small driving force region could be due to diffusion
control of the reverse et reaction. This is a first indication that
reaction et cannot be regarded as a simple outer-sphere electron
transfer process.
x is the distance between the centers of the (hard sphere) reactant
or product molecules, and d is the closest approach value of x.
wR(x) and wR(x) are the distance dependent work terms, and
KR and KP are defined by
RT
F
ln KR wR(d) wR
RT
F
ln KP wP(d) wP
Since the nature of the interactions and of their variation with
distance are not precisely known, we may consider that they
are bracketed by the two limiting situations represented in the
scheme below:
Furthermore, it appears that there is no possibility to fit the
whole set of data points with a single Marcus correlation line.
Indeed, forced application of eq 2 for fitting all the points leads
to obviously unrealistic values of A, 1.9 × 105 M 1 s 1, and of
(A) + (Q)
(B) + (P)
KR
kD
kD
q
∆G0 , 0.058 eV. The impossibility to interpolate the whole set
kD
kD
KR
KP
kD
ke
A + Q
of k values with a single Marcus parabola is a second indication
that the reaction cannot be regarded as a simple outer-sphere
process. These observations suggest that the reaction possesses
an inner-sphere character or, equivalently, that it may be viewed
as the transfer of an electron pair rather than of a single electron.
In this framework, the results can be rationalized by a mech-
anism involving two successive steps as depicted in the
following reaction scheme (or similar reactions in the nitrile
case):
AQ
BP
B + P
k–e
kD
kD
KP
kD
(AQ)
(BP)
The lower pathways for the formation of the precursor and
successor complexes correspond to short distance interactions
(the interactions vanish within a distance much shorter than the
R
R
size of the (mutual) diffusion layer). Then kD
kD and kD
kD/kP for the
P
P
kD for the reactants and kD
kD and kD
products. In other words, an unfavorable work term accelerates
the decomposition of the precursor and successor complexes
into the separated reactants and products, respectively, while it
leaves unchanged the rate of formation of the precursor and
successor complexes from the separated reactants and products,
respectively. Thus, the rate constant, k , for the endergonic
conversion of A Q into B P is expressed as
O
kn
k–n
•
•
C
(n)
(c)
C
C
+ CO2
O
O–
RO O–
RO
O
kc
• –
•
+ CO2
C
C
C
O–
O
RO
RO
O
KRKe
KP
F
RT
k
kD
kD exp
(E°cat E°CO )
2
In the framework of such a mechanism, the overall rate constant
k is related to the rate constants of each step according to eq 3.
[
]
while k
kD. It thus appears that the experimental data in
knkc
the endergonic region of Figure 3 cannot be accounted for by
considering that the reverse reaction is under diffusion control
k
(3)
k
kc
n
(19) (a) von Smoluchowski, M. Phys. Z. 1916, 17, 557. (b) von
Smoluchowski, M. Phys. Z. 1916, 17, 585. (c) von Smoluchowski, M. Z.
Phys. Chem. 1917, 92, 129. (d) Debye, P. Trans. Electrochem. Soc. 1942,
82, 265. (e) Andrieux, C. P.; Save´ant, J.-M. J. Electroanal. Chem. 1986,
205, 43.
Both the standard free energy and the activation barrier of
reaction c should not depend much on the substituents on the
phenyl ring. Indeed the effect of conjugation between the
unpaired electron and the phenyl ring in the radical of the left-

+
+
Electrochemical Reduction of Carbon Dioxide
J. Am. Chem. Soc., Vol. 118, No. 30, 1996 7195
hand side are likely to parallel the effect of conjugation between
the carbonyl group and the phenyl ring in the right-hand side
of the reaction. This is also likely to be true for the effect of
R. The variation of Kn ( kn/k n) in the series thus follows the
variation of E°cat. In the less negative end of the catalyst
potentials, Kn is small, and therefore kn is small and k n is large.
SN2-type reaction where the oxygen or the nitrogen end of the
anion radical catalyst binds to the carbon of CO2. This reaction
may equivalently be viewed as an inner-sphere process in which
bond formation and electron transfer are concerted. The CO2
anion radical is eventually produced after homolytic cleavage
of the bond thus formed. Such a reaction mechanism bears a
strong resemblance with what happens when aromatic anion
radicals are reacted with alkyl halides, where, in the absence of
steric hindrance, the most favorable pathway is also of the SN2
type.5b,c
Thus k tends to become larger than kc, and thus, k Knkc.
n
Under these conditions, log k varies by one unit when E°cat
varies by 60 mV. When E°cat becomes more and more negative,
Kn and kn increase and k decreases. A point will thus be
n
reached where k
kc, meaning that forward reaction n
n
becomes the rate-determining step. From there on log k is
expected to vary by 1/(120 mV). These predictions fit the data
obtained in the ester series.
The assumption that the standard free energy and the
activation barrier of reaction c do not depend on the overall
driving force for the generation of CO2 might be too simplistic.
In the framework of the above reaction scheme, a more general
interpretation may be as follows. We assume that the standard
free energies for reactions n and c are proportional to the
standard free energy for the overall reaction:
Experimental Section
Chemicals. DMF (Carlo Erba, RPE) was kept over anhydrous Na2-
CO3 for several days and stirred from time to time. It was then
fractionally distilled under reduced pressure under N2 twice and stored
in a dark bottle under N2. In order to remove as much residual water
as possible, the solvent was repeatedly percolated before use through
a column of neutral alumina (Merck, activity grade 1) previously
activated overnight at 360 °C under vacuum. Tetra-n-butylammonium
perchlorate (Fluka purum) was recrystallized twice from a 2:1 water
ethanol mixture and then dried in a vacuum oven at 60 °C. Carbon
dioxide (99.998%) was supplied by SIAD (Italy). The solubility of
the gas in DMF (0.199 M at 25 °C and 1 atm of pressure) at various
temperatures and various partial pressures has been previously re-
ported.21 DMF solutions containing various concentrations of CO2 were
prepared by saturating the solvent with appropriate mixtures of CO2
and argon. The apparatus used for the preparation of the gas mixtures
and the method of calculating CO2 concentrations in the solution from
its partial pressure in the gas phase were previously described.21
∆Gn° γn(E°cat E°CO
)
∆Gc° γc(E°cat E°CO
)
2
2
Thus
γn γc
1
(4)
It may also be assumed that the activation free energies of both
reactions n and c are approximate linear functions of their
respective standard free energies n and c as slopes. Thus, in
the low driving force region
The commercially available catalysts methyl 4-methylbenzoate,
methyl 3-methylbenzoate, methyl 2-methylbenzoate, ethyl benzoate,
phenyl benzoate, dibutyl phthalate, dimethyl phthalate, and dipropyl
phthalate (from Aldrich Chemie), methyl benzoate (Janssen), benzo-
nitrile (Fluka), 2-tolunitrile (Ega-Chemie), methyl 4-phenylbenzoate
(Heraeus), diisobutyl phthalate (Carlo Erba), ethyl 3-fluorobenzoate
(Sigma), and 4-cyanobiphenyl (K & K) were used as received. Phenyl
3-methylbenzoate and phenyl 4-methylbenzoate were prepared from
3-methyl- or 4-methylbenzoic acid and phenol, respectively, as previ-
ously described.22 Methyl 3-phenoxybenzoate was prepared by re-
fluxing 3-phenoxybenzoic acid in methanol in the presence of sulfuric
acid.23
RT
F
RT
F
ln k
ln Knkc ln Ac (γn
cγc)(E°cat E°CO )
2
and in the high driving force region
RT
F
RT
F
ln k
ln kn ln An
( nγn)(E°cat E°CO )
2
Electrochemical Instrumentation and Procedures. The cyclic
oltammetric measurements were made with a PAR Model 173
potentiostat, a programmable function generator Amel Model 568 and
a 2090 Nicolet oscilloscope. An Amel Model 863 X Y recorder was
used for recording the cyclic voltammograms. A mercury microelec-
trode was used as the working electrode and a platinum wire as the
counterelectrode. The working electrode was made from a 2-mm-
diameter platinum sphere coated with mercury after electrolytic
deposition of silver. The reference electrode was Ag/AgI/n-Bu4NI 0.1
M in DMF whose potential was always measured compared to the SCE,
to which all potentials are finally referenced. Both the reference
From the experimental slopes of the log k vs E°cat E°CO plot
(Figure 3), it follows that
2
γn
cγc
1
(5)
and
nγn O.5
(6)
The four variables γn, γc, n, and
are thus related by eqs
c
4 6. Thus, in the framework of the two-step reaction scheme,
γc and are not necessarily equal to zero.
Whatever the exact value of these proportionality coefficients,
the two-step reaction scheme offers a likely explanation of the
observation that the reaction is not a simple outer-sphere process.
The fact that the data points in the nitrile series fall above the
ester correlation line lends further support to this conclusion.
(20) (a) There are two types of homogeneous catalysis of electrochemical
reactions. In redox catalysis,17,20b the reduced form of the catalyst couple
is merely an outer-sphere electron donor that shuttles electrons from the
electrode to the substrate. This homogeneous electron transfer is subject to
the same activation/ driving force limitations as the outer-sphere electron
transfer at an inert electrode. The very existence of a catalytic effect thus
derives from a physical rather than a chemical process, namely, the
dispersion of the electrons in the same three-dimensional space as the
substrate instead of the two-dimensional availability of the electrons at the
electrode surface. In chemical catalysis, the interactions between the reduced
form of the catalyst and the substrate are more intimate, involving the
transient formation of an addition product between the reduced form of the
catalyst and the substrate before regeneration of the oxidized form of the
catalyst. (b) Andrieux, C. P.; Dumas-Bouchiat, J.-M.; Save´ant, J.-M. J.
Electroanal. Chem. 1978, 87, 39.
c
Conclusions
Anion radicals of aromatic esters and nitriles are remarkably
selective and fairly persistent homogeneous catalysts of the
electrochemical reduction of carbon dioxide yielding exclusively
oxalate. The generation of CO2 by electron transfer from these
donors is not a simple outer-sphere process, and therefore,
catalysis is not strictly speaking of the outer-sphere type.20 The
reaction involves two successive steps. The first of these is an
(21) Gennaro, A.; Isse, A. A.; Vianello, E. J. Electroanal. Chem. 1990,
289, 203.
(22) Hashimoto, S.; Furukawa, I. Bull. Chem. Soc. Jpn. 1981, 54, 2227.
(23) Ceccon, A.; Gentiloni, M.; Romanin, A.; Venzo, A. J. Organomet.
Chem. 1977, 127, 315.

+
+
7196 J. Am. Chem. Soc., Vol. 118, No. 30, 1996
Gennaro et al.
electrode and the counterelectrode were separated from the working
electrode compartment by glass frits and salt bridges.
eluted with 0.01 N H2SO4, on which conversion of oxalate to oxalic
acid takes place. The validity of this procedure was checked by
comparison with the classical permanganate colorimetric method.26
Possible side products resulting from carboxylation of the catalyst
The diffusion coefficient of CO2 in DMF has been previously found
1 3d
to be 2.7 × 10 5 cm2 s
.
For the catalyst we have found an average
value of 0.8 × 10 5 cm2 s 1. These values were used in the construction
of the working curves for extracting the rate constants from the cyclic
voltammetric data.
ArCO2R CO2 2e f ArCOCO2
or from spontaneous decomposition of the anion radicals27
ArCO2R f ArCO2 RO
RO
Preparati e scale experiments were performed in a gas-tight
electrolysis cell under galvanostatic conditions with a current density
of 1.6 mA/cm2. Before the electrolysis was started, a solution saturated
either with CO2 alone or with a mixture of CO2 and argon was prepared
by bubbling the gas into the catholyte for about 30 min. During
electrolysis, the cell was maintained in contact with a gas reservoir
containing CO2 at the appropriate partial pressure. A mercury pool of
12.4 cm2 was used as the cathode. The same stirring rate was used in
all the experiments in order to achieve reproducible hydrodynamic
conditions.
Electrolysis products present in both the gas and liquid phases were
analyzed by chromatographic techniques. Carbon monoxide and formic
acid were analyzed on a Varian 3700 gas chromatograph equipped with
a thermal conductivity detector, using helium as the carrier gas. CO
was analyzed only in the gas phase using a molecular sieve column.
Since the solubility of CO in dipolar aprotic solvents is very low,24 the
quantity of CO present in solution was considered to be negligible.
Samples were extracted from the gas phase of the working electrode
compartment with a gas syringe through a septum cap. To detect
formate, the solution was first esterified by addition of methyl iodide
to a portion of the electrolyzed solution.25 The resulting solution in
DMF was then analyzed by GC using a GP 10% SP-1200, 1% H3PO4
Supelco column. Oxalate was directly analyzed in HPLC with a Perkin-
Elmer Series 4 liquid chromatograph equipped with a UV detector
operating at 210 nm, using an Aminex HPX-87H ion exchange column
were investigated but could not be detected. It should be emphasized
that any two-electron process destroying the catalyst, even totally, would
consume only a small fraction of the total charge passed as compared
to the catalytic reaction. Since less than 30% of the catalyst is lost in
all cases, such side reactions contribute negligibly to the total amount
of charge. The 10 20% that are lacking in the charge balance (Table
2) are thus more likely due to the diffusion of the small fraction of the
oxalate formed toward the anodic compartment where it is immediately
reoxidized into CO2.
Supporting Information Available: A series of diagrams
showing how the rate constant k was derived from the ip/ip°
data by fitting with the appropriate theoretical working curves
(computed by an implicit finite difference method) for each of
the 14 ester and 3 nitrile catalysts (6 pages). See any current
masthead page for ordering and Internet access instructions.
JA960605O
(26) Vogel, A. I. Textbook of Quantitati e Inorganic Analysis, 2nd ed.;
Longman, Green and Co.: London, 1955; p 275.
(24) (a) Wilhelm, B.; Battino, R. Chem. Re . 1973, 73, 1. (b) Ercolani,
C.; Moncelli, F.; Pennasi, G.; Rossi, G.; Antonini, E.; Acsenzi, P. J. Chem.
Soc., Dalton Trans. 1981, 1120.
(25) Wagenknecht, J. H.; Baizer, M. M.; Chruma, J. L. Synth. Commun.
1972, 2, 215.
(27) (a) Seeber, R.; Magno, F.; Bontempelli, G.; Mazzochin, G. A. J.
Electroanal. Chem. 1976, 72, 219. (b) Wagenknetch, J. H.; Goodin, R. D.;
Kinlen, P. J.; Woodward, F. E. J. Electrochem. Soc. 1984, 131, 1559. (c)
Masnovi, J. J. Am. Chem. Soc. 1989, 111, 9801. (d) Vincent, M. L.; Peters,
D. G. J. Electroanal. Chem. 1992, 327, 121.