Review ARticle
https://doi.org/10.1038/s41560-019-0450-y
1
Leiden Institute of Chemistry, Leiden University, Leiden, the Netherlands.
2
Present address: Separation and Conversion Technology, Flemish Institute for
Technological Research, Mol, Belgium.
3
Present address: Department of Sustainable Process and Energy Systems, TNO, Delft, the Netherlands.
4
Present address: Department of Chemical Engineering and Chemistry, Eindhoven University of Technology, Eindhoven, the Netherlands.
5
Present address: Departament de Ciència de Materials i Química Fisica and Institut de Química Teòrica i Computacional, Universitat de Barcelona,
Barcelona, Spain. *e-mail: f.calle.vallejo@ub.edu; m.koper@lic.leidenuniv.nl
W
ith the growing importance and falling prices of
renewable electricity, the issue of electricity storage to
deal with the intermittent nature of renewable energy
sources is becoming urgent. Storing renewable electricity in chemi
-
cal bonds (‘electrofuels’) is particularly attractive, as it allows for
high-energy-density storage and potentially high flexibility. While
hydrogen is the most likely and realistic candidate for electricity
storage in electrofuels, research on the electrochemical conversion
of carbon dioxide and water into carbon-based fuels has intrigued
electrochemists for decades, and is currently undergoing a notable
renaissance
1–4
. In contrast to hydrogen production by water elec-
trolysis, carbon dioxide electrolysis is still far from a mature tech-
nology. Significant hurdles regarding energy efficiency, reaction
selectivity and overall conversion rate need to be overcome if elec
-
trochemical carbon dioxide reduction is to become a viable option
for storing renewable electricity.
Many electrocatalysts have been reported for the production
of different compounds from the electrocatalytic carbon dioxide
reduction reaction (CO
2
RR). Table 1 gives an overview of some of
the most active and selective metal or metal-derived electrocata
-
lysts towards specific products in aqueous media. The two-electron
transfer products, CO and HCOOH, can be produced with low
overpotential and high Faradaic efficiency on suitable electrocata
-
lysts, but substantially higher overpotentials and lower selectivities
are observed for multi-electron transfer products such as methane,
ethylene and alcohols
2
. For a recent discussion about the economic
perspectives of CO
2
RR, the reader is referred to a previous analysis
5
.
The aim of this Review is not to be exhaustive, but rather to
selectively (and subjectively) discuss some recent advances and per
-
tinent challenges in this field, focusing on themes that have recently
witnessed important progress
2,3,6,7
. An overview of some of the
themes covered in this Review is shown in Fig. 1. We also discuss
two important methodologies used to increase fundamental under
-
standing of CO
2
RR: insitu spectroscopic techniques and computa-
tional techniques.
Initial activation of CO
2
The first step in the CO
2
RR is the activation of the CO
2
molecule.
It is often claimed that activation and reduction of CO
2
is dif-
ficult because the first electron transfer to form the CO
2
•–
radical
intermediate has a very negative redox potential (−1.9 V versus
normal hydrogen electrode), or because CO
2
is a very stable mol-
ecule
4
. Neither statement is accurate. Electrocatalysts stabilize the
CO
2
•–
radical or reaction intermediate by forming a chemical bond
between CO
2
and catalyst, leading to a less negative redox potential.
With the right electrocatalyst, it is possible to reduce CO
2
to CO
or HCOOH at low overpotential. This low-overpotential reversible
catalysis is related to the mechanism of a two-electron process, typi
-
cally consisting of only one intermediate, which can be optimized
using an appropriate catalyst
8,9
. Enzymes such as formate dehydro-
genase and carbon monoxide dehydrogenase are indeed effective
reversible catalysts for the CO
2
RR to formate and carbon monoxide,
respectively, exhibiting negligible overpotentials
10
. In addition, the
equilibrium potentials of CO
2
RR are close to 0 V (and sometimes
higher), making the thermodynamic stability of CO
2
similar to that
of water.
We consider four redox reactions related to the activation of CO
2
(equations (1)–(4)):
þCO
2
þ H
þ
þ e
�
! *COOH ð1Þ
Advances and challenges in understanding
the electrocatalytic conversion of carbon dioxide
to fuels
Yuvraj Y. Birdja
1,2
, Elena Pérez-Gallent
1,3
, Marta C. Figueiredo
1,4
, Adrien J. Göttle
1
,
Federico Calle-Vallejo
1,5
* and Marc T. M. Koper
1
*
The electrocatalytic reduction of carbon dioxide is a promising approach for storing (excess) renewable electricity as chemical
energy in fuels. Here, we review recent advances and challenges in the understanding of electrochemical CO
2
reduction. We dis-
cuss existing models for the initial activation of CO
2
on the electrocatalyst and their importance for understanding selectivity.
Carbon–carbon bond formation is also a key mechanistic step in CO
2
electroreduction to high-density and high-value fuels. We
show that both the initial CO
2
activation and C–C bond formation are influenced by an intricate interplay between surface struc-
ture (both on the nano- and on the mesoscale), electrolyte effects (pH, buffer strength, ion effects) and mass transport condi-
tions. This complex interplay is currently still far from being completely understood. In addition, we discuss recent progress
in insitu spectroscopic techniques and computational techniques for mechanistic work. Finally, we identify some challenges in
furthering our understanding of these themes.
NATURE ENERGY | VOL 4 | SEPTEMBER 2019 | 732–745 | www.nature.com/natureenergy
732
Review ARticle
Nature eNergy
þCO
2
þ H
þ
þ e
�
! *OCHO ð2Þ
þCO
2
þ e
�
! *CO
�
2
ð3Þ
þH
þ
þ 2e
�
! *H
�
ð4Þ
Equations (1) and (2) are so-called concerted proton–electron
transfer (CPET) reactions and have been considered computa
-
tionally for deriving trends in selectivity among (post-)transition
metal surfaces
11
. The authors argued that *COOH is the more likely
first intermediate for CO formation, and *OCHO the more likely
intermediate for formic acid production (an assertion that is gen
-
erally agreed on in the literature). Using calculated binding ener-
gies, they generally find good agreement between their predictions
and experiment: post-transition metals such as Pb and Sn prefer to
bind CO
2
via oxygen and are selective towards formic acid, whereas
transition-metal electrodes prefer to bind via carbon. Interestingly,
in two instances, their calculations deviate from the experimental
observations in Table 1: silver is predicted to be an excellent catalyst
for formic acid production, whereas palladium is predicted to have
the lowest onset potential for CO production. They ascribe these
differences to kinetic effects not included in their calculations.
Interestingly, in the molecular electrocatalysis community, the
views on the initial activation of CO
2
appear to be subtly different.
The initial binding of CO
2
to the catalyst does not involve a CPET
step such as in equations (1) and (2), but rather an electron-transfer
mediated CO
2
binding step, as indicated by equation (3). The CO
2
–
anionic adduct is generally bound to the metal centre of the catalyst.
For instance, for a cobalt-(proto)porphyrin catalyst, CO
2
binding
takes place if the cobalt centre changes oxidation state from Co()
to Co(), with the electronic density flowing back onto the *CO
2
ligand, formally written as in equations (5) and (6).
Co iið Þþe
�
! Co iðÞ ð5Þ
Co iðÞþCO
2
! Co iiðÞ
2
CO
�
2
ð6Þ
Subsequent protonation or CPET steps generate *COOH or
*COOH
–
intermediates
12
. If equation (3) is rate determining or
potential determining, the pH dependence of CO
2
activation may
differ from the pH dependence of the competing hydrogen evolu
-
tion reaction (HER). This different pH dependence of the CO
2
RR
and HER pathways was used to explain the strong pH depen
-
dence of the overall product selectivity on graphite-immobilized
Co-protoporphyrin, with H
2
being the primary product at pH 1 but
CO being the primary product at pH 3 (refs.
13,14
). A similar mecha-
nistic model was proposed for gold-catalysed CO
2
RR, suggesting
that also on gold the *CO
2
–
intermediate, and not *COOH, is the
relevant activated form of CO
2
(ref.
15
). Recent computational work
has confirmed that on Ag(111) *CO
2
is highly sensitive to the pres-
ence of an electric field, as, for example, modelled by the presence of
a cation
16
. Although the authors do not write the formation of *CO
2
as an electron-transfer step, the resulting *CO
2
is highly polariz-
able, and therefore sensitive to pH and cation effects (see ‘Reaction
and process conditions’). Similar conclusions regarding the impor
-
tance of a *CO
2
δ–
intermediate in the CO
2
RR reduction to CO on a
Ag(110) electrode have been drawn based on kinetic simulations of
CO formation using a multiscale model
17
.
A similar situation exists for formate/formic acid production on
molecular catalysts. Among metalloporphyrins, Rh, Sn and In pro
-
toporphyrins have a high selectivity towards formic acid in aqueous
electrolyte
18
. Density functional theory (DFT) calculations suggest
that the key intermediate is an anionic hydride
19
, formed through
equation (4), as previously suggested in the molecular catalysis
Table 1 | Highly selective/active metal and metal-derived electrocatalysts for synthesis of specific CO
2
RR products
CO
2
RR
product
Electrocatalyst Faradaic
eciency (%)
η (V)
a
j
total
(mA cm
–2
) Electrolyte
(CO
2
saturated)
Ref.
HCOOH Pb 97.4 −1.19 V 5.0 0.1 M KHCO
3
b 8
Sn 88.4 −1.04 V 5.0 0.1 M KHCO
3
b 8
Pd nanoparticles/C 99 −0.15 V 2.4–7.0 2.8 M KHCO
3
c 23
Pd
70
Pt
30
nanoparticles/C 90 −0.36 V 4.0–7.5 0.2 M PO
4
3–
buffer
d 9
CO Au 87.1 −0.64 V 5 0.1 M KHCO
3
b 8
Au nanoparticles 97 −0.58 V 3.49 ± 0.61 0.1 M KHCO
3
b 173
OD-Au nanoparticles >96 −0.25 V 2–4 0.5 M NaHCO
3
e 103
Ag 94 −0.99 V ~5 0.1 M KHCO
3
b 174
CH
4
Cu poly 40.4 −1.34 V ~7 0.1 M KHCO
3
b 55
Cu(210) 64 −1.29 V 5 0.1 M KHCO
3
b 33
C
2
H
4
Cu poly 26.0 −1.13 V 1–2 0.1 M KHCO
3
b 55
O
2
plasma-treated Cu 60 −0.98 V ~15 0.1 M KHCO
3
b 112
Cu-halide 60.5–79.5 −2.11 V 46.1–39.2 3 M KBr
f 83
Graphite/carbon NPs/Cu/PTFE 70 −0.63 V 75–100 7 M KOH
g 31
CH
3
OH Cu
2
O 38 −0.43 V 1–2 0.5 M KHCO
3
h 121
HCl-pretreated Mo 84 −0.33 V 0.12 0.2 M Na
2
SO
4
i 175
C
2
H
5
OH Cu poly 9.8 −1.14 V ~0.6 0.1 M KHCO
3
b 55
Cu
2
O 9–16 −1.08 V 30–35 0.1 M KHCO
3
b 123
CuO nanoparticles 36.1 NA
j
~11.7 0.2 M KI
124
Cu/CNS 63 −1.29 V 2 0.1 M KHCO
3
b 176
a
η = E − E
0
with E
0
versus RHE, where η is overpotential, E is applied potential and E
0
is standard electrode potential
6
.
b
pH ≈ 6.8.
c
pH ≈ 8.2.
d
pH ≈ 6.7.
e
pH ≈ 7.2.
f
pH ≈ 3.
g
pH > 14.
h
pH ≈ 7.6.
i
pH ≈ 4.2.
j
At E
= −1.7 V versus saturated calomel electrode (pH not reported). j
total
, total current density; NA, not available.
NATURE ENERGY | VOL 4 | SEPTEMBER 2019 | 732–745 | www.nature.com/natureenergy
733
Review ARticle
Nature eNergy
literature
20,21
. The anionic hydride performs a nucleophilic attack
on the carbon atom of CO
2
, yielding HCOO
–
. Again, the reaction
is triggered by a potential-induced change in the oxidation state
of the catalyst, either of the metal centre or the ligand, with the
hydride residing on the metal or the ligand (the latter is the most
likely pathway for In and Sn porphyrins). The stability of the result
-
ing species is crucial to the subsequent elementary step and the
formation of either CO or HCOOH/HCOO
–
. Interestingly, such a
(lattice-)hydride mechanism was recently proposed for nanostruc
-
tured copper-hydride catalysts with a much enhanced selectivity for
formic acid (‘normal’ copper yields primarily CO as a two-electron
product)
22
. The importance of a hydride-mediated pathway has also
been suggested for formate production on palladium electrodes, as
well as in solution
23,24
.
The relevance of considering equations (3) and (4) in the activa
-
tion of CO
2
as opposed to equations (1) and (2), is that the inter-
mediates involved are charged (or strongly polarizable) and hence
sensitive to pH and cation effects. A summary of the discussed acti
-
vation routes for CO
2
RR is given in Fig. 2.
Carbon–carbon bond formation
One of the most interesting observations in the CO
2
RR is the for-
mation of species with one or more carbon–carbon bonds, mostly
on copper-based electrodes. Apart from copper-based catalysts,
a few others, for example, NiGa, PdAu, NiP and N-doped carbon
catalysts, produce compounds with two or more carbon atoms, but
not with the same efficiency as copper
25–28
. There have been some
remarkable achievements recently in terms of improving selectivity
to C
2
compounds by tuning copper structure (for example, using
high-surface-area oxide-derived (OD) copper), by adjusting elec
-
trolyte composition and by employing organic films on copper
29–32
.
The interest in making more reduced (liquid) products from CO
2
lies in the fact that they have higher energy density and economic
value
5
. Therefore, elucidating the pathway(s) from CO
2
or CO to
Initial activation
C–C bond formation
H
−
C
OO
−
(2) Sequential
proton
–electron
transfer
(1) Concerted
proton–electron
transfer
(3) Hydride transfer
HCOOH
CO
H
2
Other
+
nH
+
me
−
C
C
HO O
H
O
C
C
H H
Selectivity-determining
intermediate
??
Experimentally
observed
intermediate
n-propanol
EthyleneCO
Ethanol
C
O
Cu
C
O
Cu
C
3+
products
∙ Rarely formed
∙ Proposed to form by
polymerization of -CH
x
-
or by CO insertion
(
1
)
,
(
2
)
o
r
(
3
)
?
Electrode morphology
∙ Subsurface atoms
∙ Grain boundaries
- Particle size
- Crystal facets
∙ Shape effects
∙ Interparticle distance
∙ Pore size effects
C
HO
O
O
C
O
H
C
OO
H
H
C
OO
CO
2
+
Selectivity and activity of
CO
2
RR
c
a
b
d
Electrocatalyst
a
b
d
c
Local
pH
Nanospheres
Nanocubes
Nanorods
Nanowires
Reaction and process conditions
Fig. 1 | Overview of themes discussed in this review. a, Initial activation of CO
2
, which can proceed via CPET, SPET or hydride-transfer mechanisms.
b, Influence of the mesostructure of the electrode on the selectivity and activity of CO
2
RR. c, Recent views on C–C bond formation from CO
2
RR, specifically
the formation of C
2
H
4
and C
3+
products. d, The important role of reaction and process conditions, such as local pH, on CO
2
RR activity and selectivity.
NATURE ENERGY | VOL 4 | SEPTEMBER 2019 | 732–745 | www.nature.com/natureenergy
734
Review ARticle
Nature eNergy
C
2
products, of which ethylene and ethanol are the most typical
examples, has been the subject of several experimental
33–38
and theo-
retical
39–43
studies. Two separate pathways for ethylene production
have been observed experimentally: a high-overpotential pathway
involving a shared intermediate with the methane pathway taking
place primarily on Cu(111), and a low-overpotential pathway tak
-
ing place primarily on Cu(100) and which does not yield C
1
prod-
ucts
36
. The two pathways also differ in the their pH dependence: the
‘Cu(111) pathway’ is pH dependent (on the normal hydrogen elec
-
trode potential scale), whereas the ‘Cu(100) C
2
pathway’ is not. As a
result, the latter pathway is favoured over the former pathway, and
over hydrogen evolution in alkaline media, making the selectivity of
the CO
2
and CO reduction on copper highly sensitive to (local) pH.
Several experimental
34,37
and theoretical
39–41
studies have suggested
a CO dimer intermediate as the key species in this C
2
pathway. DFT
calculations show that at more negative potentials, CO dimerization
is disfavoured and that an alternative reaction pathway based on the
coupling of *CO and *CHO has a lower activation energy
41,44
. Both
experiments and DFT calculations suggest that the rate-determining
step in C–C coupling involves a decoupled proton–electron trans
-
fer
39
, explaining its pH sensitivity. DFT calculations also show that
the CO dimer is energetically favoured in the presence of a water
layer, a local electric field and alkali cations, and that its formation
energy is most favourable on Cu(100) (refs.
40,42
), in agreement with
the experimental observation that it is a structure-sensitive and
pH-sensitive reaction. Recently, evidence has been provided for a
hydrogenated CO dimer intermediate (OCCOH) at low overpo
-
tentials in alkaline LiOH electrolyte during CO reduction using
Fourier transform infrared spectroscopy, supported by DFT cal
-
culations
45
. Its formation was only observed on Cu(100) and not
on Cu(111), in agreement with previous studies
35,42
. Interestingly,
ethylene formation is favoured in Cs
+
-containing electrolyte, both
during CO
2
and CO reduction, although the alleged dimer interme-
diate was not observed in CsOH electrolyte, presumably because it
is more reactive. Copper modified with certain organic layers also
shows improved selectivity to C
2
products
29,30
. These effects are not
fully understood yet but may be related to local pH effects. It is well
known that ‘designing’ an electrolyte that allows for a high near-
electrode pH during CO
2
RR, such as a low buffer strength electro-
lyte, leads to enhanced C
2
production
46
.
Besides ethylene, other C
2
products observed during CO
2
and
CO reduction on copper are acetaldehyde and ethanol. It is nor
-
mally assumed that these C
2
species are formed through common
intermediates on Cu(100) up to a selectivity-determining inter
-
mediate, the hydrogenation of which leads to acetaldehyde and
subsequently ethanol, and the hydrogenolysis of which leads to eth
-
ylene
39
. We have proposed that *CH
2
CHO could be such an inter-
mediate
47,48
. The structure sensitivity of the further reduction of this
intermediate then determines the dominant C
2
product, ethylene
or ethanol
47–49
. However, recent computational work suggested
that the selectivity-determining intermediate is formed earlier in
the mechanism
44
. An alternative pathway for ethanol formation
was suggested, claiming that on CuZn or CuAg an insertion-type
mechanism (CO reacting with Cu–CH
2
species) is operative
50
. We
note that the CO dimerization mechanism can also explain why sil
-
ver produces small amounts of ethanol (and no ethylene)
47
. Acetate
is another frequently observed C
2
product. It has been proposed
to form as a by-product in the reaction path towards ethylene via
isomerization of a three-membered ring species adsorbed on the
surface (*OCH
2
COH) as well as through the coupling of differ-
ent adsorbed intermediates
51,52
. However, acetate has been shown
experimentally to form in solution as a result of a base-catalysed
Methanol (y=3) and formaldehyde (y=1)
or
_
+
−
or
2e
−
or
_
Formic acid
(x=1)
or formate (x=0)
Glyoxal
CO
Ethylene
1-propanol and
propionaldehyde
Hydrogen
Ethanol (z=3) and
acetaldehyde (z=1)
H
+
y(H
+
+ e
−
)
5(H
+
+ e
−
)
methane
x(H
+
+ e
−
)
H
+
+ e
−
H
+
+ e
−
H
+
+ e
−
H
+
+ e
−
H
+
+ e
−
e
−
4(H
+
+ e
−
)
3(H
+
+ e
−
)
H
+
+ e
−
e
−
H
+
+ e
−
H
+
+ e
−
z(H
+
+ e
−
)
C
1
products
C
2+
product
s
H
2
CO
2
Fig. 2 | Overview of reaction pathways for CO
2
RR towards different products. Black spheres, carbon; red spheres, oxygen; white spheres, hydrogen; blue
spheres, (metal) catalyst. The arrows indicate whether proton, electron or CPETs take place.
NATURE ENERGY | VOL 4 | SEPTEMBER 2019 | 732–745 | www.nature.com/natureenergy
735
Review ARticle
Nature eNergy
disproportionation of acetaldehyde
53
. As far as we know, this is the
only pathway that can explain why acetate can be formed also dur
-
ing CO reduction. An overview of reaction pathways towards C
2+
products is shown in Fig. 2, with the important caveat that it seems
probable that several competing pathways are operative, in particu
-
lar at large overpotentials.
The formation of higher-order (C
3+
) hydrocarbons has also been
observed on copper, with a Faradaic efficiency of up to 20% on a
suitable copper catalyst.
54
The main C
3
products are propanol and
propionaldehyde, with small amounts of hydroxyacetone, acetone
and allyl alcohol
55
. The elementary steps through which these C
3
products are formed have not been well studied yet. C–C cou
-
pling between CO and C
2
H
4
precursors has been reported to form
n-propanol on agglomerates of OD copper nanocrystals
56
. For this
reaction, defect sites were proposed to be the active sites. Moreover,
polymerization of adsorbed carbon-based species has been pro
-
posed as a mechanism for the formation of higher-order hydrocar-
bons on bimetallic PdAu electrodes
26
. Although CO insertion-type
reaction steps seem likely, there is still a lack of mechanistic work on
the electrochemical formation of C
3+
hydrocarbons.
Reaction and process conditions
The nature of the solvent (aqueous or non-aqueous), pH, the iden-
tity of ionic species, buffering strength and the exact mass-trans-
port conditions all influence the activity and selectivity of CO
2
RR
57
.
However, the interpretation of the individual effects remains poorly
understood, and only recently systematic work was performed to
elucidate and separate these effects. The importance of reaction and
process conditions also implies the importance of proper standard
protocols, as discussed recently
58
.
The CO
2
/HCO
3
–
, H
2
CO
3
/HCO
3
–
and HCO
3
–
/CO
3
2–
equilibria in
water are sensitive to (bulk) pH, electrolyte composition and buffer
capacity, which leads to different concentrations of carbonaceous
species in solution (Fig. 3a)
59
. This has led to some controversy
about the real active form of CO
2
during CO
2
RR. Most authors
acknowledge now that dissolved CO
2
is the active species, though
bicarbonate has also been suggested as the active species, especially
towards formate
60–62
. Enhanced CO
2
RR activity in bicarbonate elec-
trolyte, compared with other electrolytes under similar conditions,
has been proposed to be associated with the formation of a bicar
-
bonate–CO
2
complex. Such a complex is a primary carbon source
during CO
2
RR, leading to an increase in CO
2
concentration in the
vicinity of the electrode
63
. Insitu Fourier transform infrared experi-
ments using isotope labelling have confirmed this role of bicarbon-
ate
64
. However, it was concluded that bicarbonate is not explicitly
involved as an active species in the rate-limiting step of CO forma
-
tion on gold, but rather that bicarbonate is a proton donor in a reac-
tion step after the rate-limiting step
65
. The authors also conclude
that the sluggish kinetics of the reaction converting bicarbonate into
CO
2
must be considered. More generally, buffering anions in the
electrolyte can donate a proton more effectively than water, or com
-
pared with other anions, which leads to an anion effect for the HER
and formation of CH
4
, but not for CO
2
RR towards CO, HCOO
–
,
C
2
H
4
and ethanol
66
.
As soon as Faradaic currents are flowing, the importance of the
local pH and local concentration of carbonaceous species must be
taken into account
46,57,67,68
. The interpretation of experimental data
must invoke the presence of local concentration gradients
68
. During
CO
2
RR, a more alkaline pH than the bulk pH is manifested in the
vicinity of the electrode, as a result of hydroxide formation from both
the HER and CO
2
RR. This local alkaline pH is the result of mass-
transport limitations
57
. This local pH change is proportional to the
current density
69
, but also depends on the nature and buffer capacity
of the electrolyte, and on the morphology of the electrode. Therefore,
modelling concentration gradients near the electrode surface during
CO
2
RR is crucial to understand the overall reactive system. Recent
model calculations of pH gradients and CO
2
transport within a fixed
boundary layer for different electrolytes showed the importance of
buffer identity, buffer capacity, buffer pK
a
and CO
2
pressure on the
reaction kinetics, as well as the importance of the buffer kinetics
70
.
We expect that the combination of such models with surface kinet
-
ics models will become increasingly important in designing reactive
electrochemical three-dimensional interphases as shown in recent
work on pH effects to explain C
1
versus C
2
selectivity
71
.
Since the (local) pH near the electrode has such an important role
in some of the crucial steps in CO
2
RR (see the previous sections),
reactivity and selectivity can be tuned by the electrolyte composi
-
tion
46,67,70
. A prime example of this effect happens on Cu electro-
catalysts, for which the selectivity towards ethylene can be enhanced
by lowering the buffer capacity, thereby favouring a higher local
pH
46,72
. In addition, the selectivity can be steered towards ethylene
by increasing the CO
2
pressure, leading to enhanced local CO con-
centration and a higher *CO coverage
71
.
Another parameter influencing the local pH and consequently
the product distribution of CO
2
RR is the electrode morphology
(Fig. 3c). It has been shown that the increased mesoporosity of sil
-
ver and gold catalysts alters the selectivity in favour of CO
2
-to-CO
conversion, compared with H
2
evolution, which is attributed to
enhanced local pH gradients in highly mesoporous films
73,74
. These
trends were corroborated by comparison to rotating (cone) elec
-
trode experiments, showing that enhanced mass transport is advan-
tageous to hydrogen evolution, but has a neutral or inhibitive effect
on CO
2
RR. Using rotating copper cylinder electrodes, it was shown
that the CO
2
RR activity decreases with higher rotation rate, related
to a change in selectivity from CH
4
to CO as CO is more readily
transported away with enhanced convection
75
. Rotating copper disk
electrode experiments in mildly acidic media revealed that water
reduction is the HER pathway in competition with CO
2
RR, instead
of proton reduction
76
. The suppression of water reduction on cop-
per was suggested to be the result of *CO on the copper electrode,
which is more pronounced at low rotation rates (as CO does not dif
-
fuse away). Note that one should be careful with the interpretation
and comparison of experimental data of high-surface-area electro
-
catalysts on the reversible hydrogen electrode (RHE) scale, since the
local pH may be significantly different compared with the bulk pH,
leading to deviations on the RHE scale
58
.
Another consequence of a local alkaline pH as a result of the
concomitant HER during CO
2
RR is the occurrence of homoge-
neously catalysed chemical reactions, which may seriously affect
the overall product spectrum
53,77
. Cannizzaro-type reactions can
take place during CO
2
RR, in which intermediate aldehydes dispro-
portionate into their corresponding carboxylic acid and primary
alcohol
53
. This phenomenon is enhanced by the alkaline pH near
the electrode, and is therefore more important in poorly buffered
or unbuffered electrolytes, and on electrodes with high porosity.
The obtained products (acids and alcohols) should be distinguished
from direct CO
2
RR products.
The nature of the electrolyte cations and anions has been shown
to have a significant effect on the activity and selectivity of the
CO
2
RR. Larger cations usually lead to higher CO
2
RR rates and a
higher C
2
/C
1
ratio (on copper electrodes)
78–80
. A cation hydrolysis
effect has been suggested to lead to an enhanced buffering near the
electrode for larger cations, which in turn increases the local con
-
centration of dissolved CO
2
(refs.
79,81
). An alternative rationaliza-
tion of the cation effect is the lesser stabilization of 2 *CO compared
with *OCCO and *OCCOH (refs.
16,32,82
). The onset potential for
ethylene depends on the cation nature, while for methane no cor
-
relation with cation size was found
82
.
The effect of anions on the CO
2
RR has been investigated,
although less extensively compared with cations, and these studies
generally deal with halide effects on copper
83–87
. Anion effects are
generally explained in terms of halide adsorption on the catalyst
NATURE ENERGY | VOL 4 | SEPTEMBER 2019 | 732–745 | www.nature.com/natureenergy
736
