Journal of Applied Electrochemistry (2006) 36:161–172 Ó Springer 2005

DOI 10.1007/s10800-005-9058-y

Calculation for the cathode surface concentrations in the electrochemical reduction of

CO2 in KHCO3 solutions

N. GUPTA, M. GATTRELL* and B. MACDOUGALL

National Research Council, Institute of Chemical Process and Environmental Technology, 1200 Montreal Road,
Ottawa, ON, K1A 0R6, Canada
(*author for correspondence, Tel.: +1-613-990-3819; e-mail: michael.gattrell@nrc-cnrc.gc.ca)

Received 1 May 2005; accepted in revised form 16 August 2005

Key words: boundary layer, carbon dioxide electroreduction, carbonate buffer, CO2 reduction, electrode surface
concentrations

Abstract

This article presents a mathematical model that predicts the chemical conditions at the electrode surface during the
electrochemical reduction of CO2. Such electrochemical reduction of CO2 to valuable products is an area of interest
for the purpose of reducing green house gas emissions. In the reactions involved, CO2 acts as both a reactant and a
buffer, consequently the estimation of local concentrations at the electrode surface is not trivial and a numerical
approach is required. The necessary partial differential equations (PDEs) have been set-up and solved using
MATLAB. The results show the local concentrations at the electrode surface to be significantly different from the
bulk concentrations under typical reported experimental conditions. The importance of buffer strength and a careful
quantification of the degree of mixing produced in the experimental apparatus is demonstrated. The model has also
been used to re-examine previously published data, showing that the Tafel slopes in CO2 reduction are consistent
with those reported for the simpler CO reduction system. Further, the effect of pulsed electroreduction was also
modeled, showing that pulsing causes corresponding swings in local pH and CO2 concentrations.

Nomenclature D0OH
diffusion coefficient for hydroxyl ions

at 25 °C at infinite dilution (m2 s)1)
Symbol Description (Units)
DCO2 diffusion coefficient for carbon dioxide
cefCH4 current efficiency for methane forma- in water at 25 °C and given electrolyte
tion (dimensionless) concentration (m2 s)1)
cefC2 H4 Current efficiency for ethylene forma- DCO
3 diffusion coefficient for carbonate ions
tion (dimensionless) in water at 25 °C and given electrolyte
cefCO current efficiency for carbon monoxide concentration (m2 s)1)
formation (dimensionless) DHCO3) diffusion coefficient for bicarbonate
cefH2 current efficiency for hydrogen forma- ions in water at 25 °C and given elec-
tion (dimensionless) trolyte concentration (m2 s)1)
cefHCOO
current efficiency for formate forma- DOH
diffusion coefficient for hydroxyl ions
tion (dimensionless) at 25 °C and given electrolyte concen-
CO2 consumption rate of CO2 consumption at cathode tration (m2 s)1)
surface (kmol m)2 s)1) F FaradayÕs constant (96486) (C mol)1)
D0CO2 diffusion coefficient for carbon dioxide j current density at the Cu electrode
in water at 25 °C at infinite dilution (A m)2)
(m2 s)1) k1f rate constant for forward reaction (3b)
D0CO
diffusion coefficient for carbonate ions (M)1 s)1)
3
in water at 25 °C at infinite dilution k1r rate constant for reverse reaction (3b)
(m2 s)1) (M)1 s)1)
D0HCO
diffusion coefficient for bicarbonate k2f rate constant for forward reaction (4)
3
ions in water at 25 °C at infinite dilu- (s)1)
tion (m2 s)1) k2r rate constant for reverse reaction (4) (s)1)
162
KH equilibrium constant for reaction (1) zeffC2 H4 electrons exchanged in reaction (14)
(dimensionless) (dimensionless)
K1a equilibrium constant for reaction (3a) zeffCO electrons exchanged in reaction (12)
(M) (dimensionless)
K1b equilibrium constant for reaction (3b) zeffH2 electrons exchanged in reaction (15)
(M)1) (dimensionless)
K2 equilibrium constant for reaction (4) zeffHCOO
electrons exchanged in reaction (11)
(M)1) (dimensionless)
K3 equilibrium constant for reaction (5)
(dimensionless)
OH

formation rate of OH) formation at cathode sur- Greek Description (Units)
face (kmol m)2 s)1) d boundary layer thickness (m)
zeffCH4 electrons exchanged in reaction (13) l viscosity of electrolyte solution (mPa s
(dimensionless) or cP)

1. Introduction the cathodic reduction of CO2 at metal electrodes (Hg,

Au, Pb, Zn, Sn and In) in aqueous inorganic salt
With the steady rise in the demand for energy in the solutions. Further, work done by Steven et al. [7] has
industrialized world, the use of fossil based fuels is also shown the reduction of CO2 to CO with high current
on the increase, causing an increase in the CO2 efficiency. Electrodes such as Ti, Ni, or Au have been
emissions. The increased CO2 emissions are a major studied for this application.
concern as regards the green house effect. It is therefore From a greenhouse gas reduction point of view, the
of great interest to develop energy sources that do not most interesting reported results relate to the direct
rely on fossil fuels. While renewable energy can provide reduction of CO2 to methane. At a copper electrode
one solution, it is not always available at the time or Hori et al. [8] reported that CO2 can be directly
location needed. Thus, methods to store such energy in a electrochemically reduced to a mixture of hydrogen,
portable and easily used form are of interest. One methane, ethylene and CO.
approach is to collect CO2 from air or power plant Studies done on CO2 electroreduction use buffered
stacks, etc. and convert it back into fuels. This energy (e.g. K2HPO4), somewhat buffered (viz. 0.1–0.5 M
can then be used within existing infrastructures to KHCO3) and unbuffered (e.g. KCl, KClO4, K2SO4)
displace fossil fuels, resulting in a net decrease of CO2 solutions with the majority using somewhat buffered
emissions. This is especially important because it allows solutions. In unbuffered solutions, the electrode local
renewable energy to be applied to transportation appli- pH can vary widely from the bulk. In the somewhat
cations, which represent a large and growing source of buffered solutions, the equilibrium concentration of
CO2 emissions. H2CO3 is very low and the kinetics of the reactions
Electrochemistry represents an important enabling CO2(aq)+H2O , H2CO3 and CO2 (aq) þ OH
, HCO
3
technology for this approach because of its high are limited. This makes accurate estimation of the pH
efficiency for conversion between electrical and chemical at the electrode surface an interesting problem. This is
energy. Bandi et al. [1] have evaluated the energy further complicated in this particular system by the
demand for a process of CO2 sequestration from air fact that CO2 is not only a part of the buffer, but is
by absorption into KOH solutions followed by its also the reactant. Thus the accurate estimation of the
conversion into fuel. CO2 absorbed as potassium car- electrode surface conditions is not only additionally
bonate is separated by treating the solution with complicated, but is also doubly important for under-
sulphuric acid and could then be electroreduced to standing what is happening in terms of the reaction
useful fuels such as CH4. The potassium sulphate mechanism.
solutions so produced by acidifying the potassium Thus, the effect of pH at the electrode surface is
carbonate could be regenerated by electrodialysis. important in mechanistic studies of CO2 electroreduc-
Tryk and Fujishima [2] have shown the role electro- tion. Hori et al. [9] pointed out this issue of the surface
chemists could play in reducing green house emissions pH variation, and showed some simplified approximate
by making fuels such as CO, CH4 and C2H4 on various calculations to illustrate the importance of the effects.
electrode materials such as Cu, Ni, Ag, etc. There is However, in general, studies in the area fail to address
abundant literature on the electroreduction of CO2 on the issue of electrode surface pH. Therefore, the intent of
various electrode materials [3]. The most recent review this work is to develop a detailed finite difference model
on CO2 electroreduction has been done by Hori [4]. It to properly estimate the surface pH and CO2 (reactant)
has been reported by Teeter and Van Rysselberghe [5] concentration. The model will then be used to quantify
and Ito et al. [6] that formate is exclusively produced in the influence of variable current density and electrolyte
 163
strength on the pH and to re-visit some results from the fast, carbonic acid, H2CO3 never constitutes more
literature with a clearer insight. than 1% of the total CO2(aq) in solution, and so above
a pH of 7.4, reaction (3b) dominates and reactions (1)
and (2) can be ignored in the pH calculations [3].
2. Theory
Also, the kinetics of both the forward and reverse
reactions of Equation (4) are very fast, so that it can be
Various equilibrium reactions in the CO2–KHCO3
assumed that HCO


3 is in equilibrium with CO3 at all
electrolyte system are (with the equilibrium constants
times.
at 25 °C from Sullivan et al. [3]):
Figure 1 shows the electrode–electrolyte boundary
CO2 (aq) þ H2 O , H2 CO3 KH ¼ 2:63  10
3 (1) layer in the CO2 electroreduction system. Film theory is
assumed to be applicable where, in the concentration
boundary layer, the velocity gradients or convective
H2 CO3 , HCO

3 þH
þ
(2) effects are assumed to be negligible. The following
balances would then occur within a slice of solution
Combining (1) and (2) gives: from x to x + Dx in Figure 1:
CO2 (aq) þ H2 O , HCO

3 þH
þ
K1a ¼ 4:44  10
7 M
@½CO2 ðaqÞ @ 2 ½CO2 ðaqÞ
(3a) ¼ DCO2
@t @x2
Or in basic solutions where pH > 7, Equation (3) is
½CO2 ðaqÞ½OH
k1f þ ½HCO

3 k1r ð6Þ
written as:
CO2 (aq) þ OH
, HCO

3 K1b ¼ 4:44  107 1=M
@½HCO
3 @ 2 ½HCO
3
(3b) ¼ DHCO
3 þ ½CO2 ðaqÞ½OH
k1f
@t @x2  
Bicarbonate ion may be neutralized by OH) ion
½HCO



3 k1r
HCO3 ½OH k2f
generated on the cathode surface by:  2

þ CO3 k2r ð7Þ
HCO

2

3 þ OH ¼ CO3 þ H2 O K2 ¼ 4:66  103 1=M
(4)
@½CO2
3  @ 2 ½CO2
3 
 
Combining (3b) and (4) gives: ¼ DCO
3 2
þ HCO


3 ½OH k2f
@t @x
 


CO2
3 k2r ð8Þ
CO2 (aq) þ CO2

3 þ H2 O , 2HCO3 K3 ¼ 9:52  103
(5)
In the system under study, CO2(gas) is in equilibrium @½OH
 @ 2 ½OH

¼ DOH

½CO2 ðaqÞ½OH
kf
with the liquid at all times, therefore CO2(aq) concentra- @t @x2  
tion is a constant at a given pressure and temperature. So, þ ½HCO


3 kr
HCO3 ½OH k2f

for the electrolyte in equilibrium with CO2 gas at a partial  2


þ CO3 k2r ð9Þ
pressure of 101.3 kPa and a temperature of 25 °C, CO2(aq)
is 0.03416 M (i.e. ignoring the effect of ionic strength). The rate constants for the forward and reverse reactions
Initial concentrations (at time t=0) of other species viz. (3b) and (4) are given in Table 1 and diffusion coeffi-
HCO
2

3 , CO3 and pH may be determined by considering cients at infinite dilution in water for various species at
Equation (5) to be in equilibrium. While reaction (2) is 25 °C are listed in Table 2.

Fig. 1. Electrode–electrolyte boundary layer.

164
 
Table 1. Rate constants for reactions (3b) and (4) at 25 °C [3] j cefHCOO
cefCO
CO2 consumption ¼ þ
Reaction Forward rate constant/M)1 s)1 Reverse rate constant/s)1 F zeffHCOO
zeffCO
 
cefCH4 cefC2 H4
3b k1f ¼ 5:93  103 k1r ¼ 1:34  10
4 þ þ2 ð15Þ
4 k2f ¼ 1  108 (assumed) k2r = k1f/K2 = 2.15104 zeffCH4 zeffC2 H4

   
Table 2. Diffusion coefficients for CO2, HCO
2

3 , CO3 at infinite dilu-
j cefHCOO
cefCO
OH
¼ þ2
tion and OH) at 25 °C in water [11] formation
F zeffHCOO
zeffCO
   
D0CO2 /m2 s)1 D0HCO
3 /m2 s)1 D0CO3
/m2 s)1 D0OH
/m2 s)1 cefCH4 cefC2 H4
þ8 þ 12 ð16Þ
zeffCH4 zeffC2 H4
1.9110)9 9.2310)10 1.1910)9 5.2710)9
where j is the planar current density on the electrode
surface in A m)2.
Table 3. Current efficiencies for different products in CO2 electrore- Equations (6)–(9) are second order time dependent
duction on Cu in 0.5 M KHCO3 at 100 A m)2 and 25 °C [10] partial differential equations which are to be solved
Product Electrons exchanged (zeff) Current efficiency (cef) under the following boundary conditions: The initial
values of the concentrations (at t=0, before current
CH4 8 0.25 flows) are assumed to be the same in the bulk solution
C2H4 12 0.20
and are listed in Table 4 for different electrolyte
CO 2 0.05
HCOO) 2 0.10 concentrations. The diffusion coefficients in Table 2
H2 2 0.40 were corrected for the viscosity change with varying
electrolyte concentration (see Table 4) using Stokes–
EinsteinÕs equation (Dl/T = constant at T = 298 K)
CO2 is reduced to a variety of products with different and these values of diffusion coefficients were used in
current efficiencies (see Table 3) according to overall Equations (6)–(9).
reactions (10) to (13): At time t > 0 and x = 0 (i.e. at the interface of bulk
CO2 þ H2 O þ 2e
, HCOO
þ OH
(10) solution and the boundary layer):
CO2 ðaqÞ ¼ ½CO2 ðaqÞi (17)

CO2 þ H2 O þ 2e
, CO þ 2OH
(11)

 
HCO


3 ¼ HCO3 i (18)
CO2 þ 6H2 O þ 8 e
, CH4 þ 8OH
(12)
 2

CO2

3 ¼ CO3 i (19)



2CO2 þ 8H2 O þ 12e , C2 H4 þ 12OH (13)
Also water could be reduced to H2 on Cu according to OH
¼ ½OH
i (20)
the following reaction )
where [CO2 (aq)]i, [HCO
2

3 ]i, [CO3 ]i, and [OH ]i are the
2H2 O þ 2e
, H2 þ 2OH
(14) equilibrium values in the bulk solution, and so the same
as the values given in Table 4. The boundary conditions
Table 3 lists the current efficiencies for different at time t > 0 and x = d (i.e. at the electrode surface)
products obtained by the electroreduction of CO2 on are related to the reaction fluxes.
Cu. Based on these products, the current for CO2 d½CO2 ðaqÞ
consumption and OH) formation at the Cu electrode DCO2 ¼
CO2 consumption (21)
dx
may be calculated as

)
Table 4. Initial equilibrium values (at t=0) for CO2, HCO
2

3 , CO3 , OH and pH at different electrolyte (KHCO3) concentrations at 25 °C
and 101.3 kPa CO2 partial pressure (Note: correction was not made for the effect of ionic strength on CO2 solubility)

Electrolyte concentration/M CO2(aq)/M HCO

3 /M CO2

3 / M OH)(M) Ph Viscosity (l)/mPa s

0.05 0.0342 0.050 7.710)6 3.310)8 6.5 1.009

0.1 0.0342 0.099 3.110)5 6.610)8 6.8 1.015
0.2 0.0342 0.199 1.210)4 1.310)7 7.1 1.027
0.5 0.0342 0.499 7.610)4 3.310)7 7.5 1.067
1 0.0342 0.994 0.003 6.610)7 7.8 1.145
2 0.0342 1.976 0.012 1.310)6 8.1 1.319
 165

d½HCO
3. Results and discussion

3
DHCO
3 ¼0 (22)
dx
3.1. General results

d½CO2

3  Hori et al. [10] used Cu cathodes in 0.5 M KHCO3
DCO
3 ¼0 (23)
dx electrolyte for CO2 electroreduction. For HoriÕs condi-
tions, including the product distribution shown in
Table 3, and based on their assumption of boundary
d½OH
 layer thickness as 0.01 cm, the pH at the electrode surface
DOH
¼ OH

formation (24)
dx may be calculated at different current densities by solving
time-dependent partial differential equations (6)–(9).
With all the boundary conditions and constants For example, at a current density of 50 A m)2 in
known, the partial differential equations given in (6)– 0.5 M KHCO3 solution and a boundary layer thickness
(9) were solved using MATLAB 7.0. of 0.01 cm, the calculated pH and CO2, HCO
3 and

Fig. 2. pH profile in the boundary layer at 50 A m)2 (KHCO3 = 0.5 M, d = 0.01 cm).

Fig. 3. CO2 profile in the boundary layer at 50 A m)2.

166

)2
Fig. 4. HCO

3 profile in the boundary layer at 50 A m .

)2
Fig. 5. CO2

3 profile in the boundary layer at 50 A m .

CO2
3 concentration profiles in the boundary layer may as the product distribution might be expected to be
be observed in Figures 2–5, respectively. One can also affected by the relative amounts of CO2 and H+ present
see that the system takes roughly 10–12 s to reach steady at the electrode surface (discussed later). The stabiliza-
state after the current is turned on (i.e. when the cathode tion of the local pH at the electrode surface around
is polarised). (Note that because pH is a logarithmic pH 9.3 is due the actions of both CO2 (reaction 3b,
quantity, it appears to stabilise sooner). Before this which produces bicarbonate) and the buffering action of
period the concentrations of each of the species, i.e. bicarbonate (reaction 4, which consumes bicarbonate
CO2, HCO
2

3 , CO3 and the pH at any given x (>0) in and produces carbonate). This buffer action within the
the boundary layer, is transient (varies with time). boundary layer can be seen in Figures 3–5.
Figure 2 further illustrates that the electrode surface One can also see, in Figure 3, the non-linearity of
pH at steady state (t > 6 s) is 9.3 which is considerably the steady-state profile for the CO2 concentration.
different from the bulk solution pH of 7.5. The reason This is due to CO2 reacting at both the electrode
for this increased pH is the generation of OH) ions at surface, in electrochemical reactions, and in the
the electrode surface. This increased surface pH may be boundary layer, in acid–base reactions. It is this
critical to the mechanism of CO2 electroreduction to complexity that creates the need for numerical calcu-
various products viz. CH4, C2H4, CO, HCOO) and H2 lation methods.
 167

Fig. 6. Cathode surface pH vs. electrolyte concentration at d = 0.01 and 0.001 cm.

Figure 6 shows the electrode surface pH profiles at 50 The profiles in Figure 6a are calculated for a bound-
and 150 A m)2 and the bulk pH profiles as a function ary layer thickness of 0.01 cm (equivalent to an rpm of
of the electrolyte concentration with boundary layer 75 at a rotating disc electrode), correspond to the
thicknesses (d) of 0.01 and 0.001 cm. It is observed that estimated value of boundary layer thickness used by
the electrode surface pH at 50 and 150 A m)2 Hori et al. [9]. However, the boundary layer thickness
approaches the bulk pH value with increasing electro- depends on the stirring rate or the flow profile in a given
lyte strength. This pH approach occurs due to a rise in electrochemical system (and/or the CO2 bubbling rate).
buffer capacity of electrolytes with increased concen- If the electrolyte is stirred at a higher rate or if the flow
trations. Specifically, this is due to a higher starting velocity is increased in a given flow cell, the boundary
concentration of bicarbonate as can be seen in Table 4. layer thickness will decrease and thereby alter the
On the other hand, the cathode surface pH may electrode surface pH (the effect on boundary layer due
increase to 9.87 at 150 A m)2 from 9.3 at 50 A m)2 in to viscosity variation with electrolyte concentration has
a 0.5 M KHCO3 solution with d = 0.01 cm. This been neglected in the present calculations).
increase in pH occurs due to an increased generation Figure 6b illustrates a case where the boundary layer
rate of OH) on the cathode surface with increasing thickness has been decreased by an order of magnitude
current density. This is obviously something that should from that in Figure 6a (equivalent to increasing the
be considered when comparing data measured at rotation rate of an RDE by two orders of magnitude).
different current densities. With this thinner boundary layer, d = 0.001 cm, it is
168
observed from Figure 6b that electrode surface pH Figure 7 along with the current efficiency values from
approaches the bulk pH more closely than it does at HoriÕs work.
d = 0.01 cm. Physically the closer approach may be In plotting these data, one can see a clear shift to less
explained by the fact that a better mixing of the hydrogenated products with higher local pH. This effect
electrolyte causes a more uniform solution pH. This is dominates over the slight decrease that occurs in the
an important consideration because many papers study- local CO2 concentration. On going from the 1.5 M buffer
ing the electrochemical reduction of CO2 do not clearly to the 0.03 M buffer, the local CO2 concentration
define the degree of mixing used in their work, making decreases by two times while the H+ concentration
the local pH and CO2 concentrations at the electrode decreases by about 15 times. This decrease in local CO2
surface difficult to estimate. For example, for 0.5 M concentration is driven by a higher current efficiency for
bicarbonate and 50 A m)2, the local pH varies from CO2 consuming reactions and a higher rate of reaction
9.3 when d=0.01 cm, to 8.5 when d = 0.001 cm. 3b as the pH increases.
We also examined polarisation measurements made
3.2. Data analysis by Hori et al. [9]. The data has been re-plotted along
with estimates for the local H+ and CO2 concentrations
To show the value of solving the equations and in Figure 8. Initially, the results show hydrogen evolu-
estimating the true conditions at the electrode surface, tion, which results in a local pH shifting from the bulk
we have re-examined some data that has appeared in the value of 6.81 to close to pH 9 by )1 V (10 A m)2). At
literature. Hori et al. [9] have carried out experiments this potential, adsorbed CO begins to form, suppressing
using a range of electrolyte strengths from 0.03 to 1.5 M, the hydrogen evolution, resulting in a slight decrease in
which resulted in different product distributions. In their current and an increase in local pH [13]. Below )1.12 V,
paper, it is felt that the change in local pH values with hydrocarbon products begin to form, with first ethylene,
the different buffer strengths plays a role in the obtained then methane appearing. These reactions accelerate,
product distributions. We have used our model to more causing a rapid rise in current at potentials below
accurately estimate the exact pH and CO2 concentra- )1.26 V (8.5 A m)2). Along with the increased reac-
tions, with the results of our calculations shown in tions, there is a slight (about 2.1 times) decrease in local

Fig. 7. The influence of buffer strength on the product distribution (Current efficiency data from Hori et al. [9], 50 A m)2, potential )1.28 V
vs. SHE at 1.5 M varying to )1.42 V at 0.03 M KHCO3 buffer, CO2 bubbled).
 169

Fig. 8. Estimated local [H+] and [CO2] values for polarisation measurements (Partial current data from Hori et al. [9], 0.1 M KHCO3, CO2
bubbled, bulk [H+] = 1.5510)7 M, bulk [CO2]=3.4110)2 M).

CO2 concentration and a large decrease in local H+ the partial current densities by [H+]bulk/[H+]local. This
concentration (about a 12 times decrease leading to a results in a charge transfer coefficient of 1.19 (50 mV
local pH of 9.96 vs. 6.8 in the bulk solution). decade)1 Tafel slope), which is similar to that reported
It is interesting to note that, while high pH favoured for methane for CO reduction, supporting the possibility
ethylene in Figure 7, in this data set the increased of a similar mechanism. While we also have data for
potential seems to be able to overcome the effect. A key [CO2]local, correcting the partial currents would be
study on the influence of potential on these types of complicated by the near saturation adsorption of the
reactions was carried out by Hori et al. [12] using CO as CO intermediate [13]. Because of the associated iso-
the starting reactant. This system was studied because it therm, this would be expected to lead to a non-linear
is simpler, both in having fewer reaction steps and relationship between [CO2]local and the partial currents.
because CO is not involved in acid–base reactions, thus However, knowing the local concentrations is also
simplifying the estimation of the local CO concentration important and useful in any attempt to clarify the
at the electrode surface. In that work, they found the details of the reaction mechanisms.
reaction to form ethylene first order with respect to CO
and having a charge transfer coefficient of 0.35 (171 mV 3.3. Pulsed CO2 reduction
decade)1 Tafel slope) and the reaction to form methane
first order in CO, first order in hydrogen ion and having A common problem reported associated with the
a charge transfer coefficient of 1.33 (45 mV decade)1 electroreduction of CO2 on Cu is the ‘‘poisoning’’ of
Tafel slope). From the data shown in Figure 8, one the electrodes, which is characterised not by a significant
would obtain a charge transfer coefficient of 0.48 change in total current (or potential), but by a shift in
(124 mV decade)1 Tafel slope) for ethylene production the product distribution over time to favour hydrogen
and 0.86 (69 mV decade)1 Tafel slope) for methane evolution. Various theories have been put forward
production (obtained by excluding the data at )1.46 V). including electrolyte trace impurity deposition, accumu-
However, having the estimated local concentrations at lation of adsorbed or insoluble reaction by-products,
the electrode surface should allow one to correct the and copper surface morphological changes [14].
data. For the methane data, assuming the reaction from However, the real cause for the poisoning is still not
CO2 is also first order in hydrogen ion, we can correct clear from the literature. In the recent review by
170

Fig. 9. Pulsed pH profile in the boundary layer (50 A m)2, 5 s then 0 A m)2, 5 s, 0.1 M KHCO3, d=0.01 cm).

Fig. 10. Pulsed surface pH and CO2 profiles with bulk and steady state concentrations indicated (50 A m)2, 5 s then 0 A m)2, 5 s, 0.1 M
KHCO3, d = 0.01 cm).

Hori et al. [4], the poisoning is attributed to the be avoided by pulsed potentials [16] or periodic anodic
deposition of trace metal ions present in the KHCO3 sweep treatments [17]. In the work of Shiratsuchi et al.
salts (ppm level) destroying the Cu catalytic activity. He [16], they tested pulsed waveforms stepping between
recommends that the contaminants be removed using )1.8 and 0 V vs. SCE, using a varying anodic pulse time
pre-electrolysis. However, Smith et al. [15] and others from 0 to 10 s in an interval of 10 s. By pulsing the
have claimed copper oxide type species to be the real electrodes, the activity of Cu electrodes did not deteri-
cause of poisoning and according to them pre-electro- orate even after several hours of operation. Because of
lyzing only postpones the inadvertent poisoning of the the complexity of the CO2 solution chemistry, it should
electrodes. be noted that such a pulsing (or sweeping) regime will
Even though the exact cause of poisoning is not clear, not only change the potential of the electrode, but also
it has been found that poisoning of Cu electrodes may the local chemical conditions at the electrode surface.
 171

Fig. 11. A Pourbaix diagram for copper, indicating the regions traversed when the electrode is pulsed (50 A m)2, 5 s then 0 A m)2, 5 s,
0.1 M KHCO3, d = 0.01 cm) at 25 °C.

We have therefore also looked at estimating the local which should be considered when investigating the
concentrations of CO2 and hydrogen ion under pulsing mechanism for the electrode regeneration.
conditions. In Figure 11, a Pourbaix diagram for Cu is shown
In the theoretical framework one needs to solve the with the systemÕs state over the course of the electrode
differential equations (6)–(9) with changing boundary pulsing indicated. The system starts at 0 V (taken to be 0
conditions. However, because the boundary conditions current) then is shifted to )1.8 V (taken to be a current
are written in terms of fluxes, it is simpler mathematically of 50 A m)2). As the electrode is held at )1.8 V the local
to assess the influence of pulsing using an equivalent pH increases from the bulk value of 6.82 to 9.79. On
galvanostatic pulse. Hence, the boundary conditions for returning to 0 V, the pH tends back towards the bulk
CO2 (Equation (21)) and OH) flux (Equation (24)) at the value, reaching 8.31 after 5 s. It can be seen that, over
electrode surface were repeatedly changed from the latter pH drift, the stable form of copper oxide
50 A m)2 for 5 s to 0 A m)2 for 5 s (using the product would change from Cu(OH)2 to Cu2O. This underlines
distribution in Table 2). An electrolyte concentration of the importance of considering the local chemical condi-
0.1 M KHCO3 (bulk pH = 6.8 cf. Table 4) was also tions at the electrode surface in understanding these
used. These conditions roughly correspond to the best complex reactions.
conditions reported in the work of Shiratsuchi et al. [16].
The changing pH profile so obtained is plotted in 4. Conclusions
Figure 9, with the electrode surface hydrogen ion and
CO2 concentrations plotted in Figure 10. A model has been developed for calculating the elec-
The local CO2 concentration is depleted during the trode surface hydrogen ion and CO2 concentrations for
cathodic pulses, but before the steady state concentra- CO2 electroreduction on Cu in KHCO3 electrolyte
tion is achieved, the non-reducing part of the waveform solution. The modelled results show that the electrode
allows fresh CO2 to diffuse to the electrode. This results surface conditions are dependent on the stirring rate,
in an average local CO2 concentration of 0.0263 vs. buffer capacity of the electrolyte and the current density
0.0181 M for the steady state conditions (the 0.1 M case of operation. It has also been shown using the model
in Figure 7). One can also see that the predicted pH estimates that the stirring rate results in a significant
varies over a range of almost 1.5 pH units during the change in the local concentrations (cf. Figure 6). This is
pulsing. In reality the pH variation could be higher important because results from other workers have
because our model does not include the formation of shown that the CO2 reduction product distribution is
copper oxides, which would consume local hydroxide affected by hydrogen ion concentration, potential (cur-
ions. Also, the model uses a constant product distribu- rent density), and CO2 concentration. Thus work in this
tion, which in reality would be changing with local area should try to quantify the level of stirring used.
concentrations and electrode potential. However, even These results also depend strongly on the buffer capacity
this simplified model shows changes in local conditions, (electrolyte strength) used.
172
The developed model was also useful for correcting 3. B.P. Sullivan, K. Krist and H.E. Guard (Ed., Electrochemical and
previously reported literature data, to show that the Electrocatalytic Reactions of Carbon Dioxide (Elsevier, Amster-
dam, 1993).
Tafel slopes observed for the production of methane and
4. Y. Hori, in W. Vielstich and H.A. Gasteiger and A. Lamm (Eds),
ethylene from CO2 are similar to those reported for the Handbook of Fuel CellsÕ, (John Wiley and Sons, Ltd., West Sus-
formation of the same products from CO. Through sex, England 2 2003) p. 720.
adjustment of the boundary conditions, the model can 5. T.E. Teeter and P. Van Rysselberghe, J. Chem. Phys. 22 (1954)
be extended to roughly evaluate the response of local 759.
6. K. Ito, T. Murata and S. Ikeda, Bull. Nagoya Inst. Tech. 27 (1975)
hydrogen ion and CO2 concentrations to pulsed CO2
209.
reduction. This approach might prove useful in better 7. G.B. Stevens, T. Reda and B. Raguse, J. Electroanal. Chem. 526
understanding the role of pulsed operation in preventing (2002) 125.
electrode activity losses. 8. Y. Hori, K. Kikuchi and S. Suzuki, Chem. Lett. (1985) 1695.
Thus, the model developed in this work should prove 9. Y. Hori, M. Akira and R. Takahashi, J. Chem. Soc. Faraday
Trans. 1(85) (1989) 2309.
useful in furthering the understanding of the reactions
10. Y. Hori, K. Kikuchi, A. Murata and S. Suzuki, Chem. Lett. (1986)
involved in the electroreduction of CO2 from carbonate 897.
buffers. In particular, it should be helpful in future 11. D.R. Lide, CRC Handbook of Chemistry and Physics, Internet
mechanistic studies. Version 2005, <http://www.hbcpnetbase.com>, (CRC Press,
Boca Raton, FL, 2005).
12. Y. Hori, R. Takahashi, Y. Yoshinami and A. Murata, J. Phys.
Chem. B 101 (1997) 7075.
Acknowledgements 13. Y. Hori, M. Akira and Y. Yoshinami, J. Chem. Soc. Faraday
Trans. 1(87) (1991) 125.
The authors would like to thank the Innovative Re- 14. R.L. Cook, R.C. MacDuff and A.F. Sammells, J. Electrochem.
search Initiative for Greenhouse Gas Mitigation for Soc. 135 (1988) 1320.
15. B.D. Smith, D.E. Irish, P. Kedzierzawski and J. Augustynski,
their financial support for this work, and Prof. Colin
J. Electrochem. Soc. 144 (1997) 4288.
Oloman of the University of British Columbia for use- 16. R. Shiratsuchi, Y. Aikoh and G. Nogami, J. Electrochem. Soc. 140
ful discussions. (1993) 3479.
17. B. Jermann and J. Augustynski, Electrochim. Acta 39 (1994) 1891.

References

1. A. Bandi, M. Specht, T. Weimer and K. Schaber, Energy Conv.

Manag. 36 (1995) 899.
2. D.A. Tryk and A. Fujishima, Electrochem. Soc. Int. 10 (2001) 32.