Article https://doi.org/10.

1038/s41467-023-38963-y

High-rate and selective conversion of CO2

from aqueous solutions to hydrocarbons

Received: 25 December 2022 Cornelius A. Obasanjo1,3, Guorui Gao1,3, Jackson Crane 1,3
, Viktoria Golovanova2,
F. Pelayo García de Arquer 2 & Cao-Thang Dinh 1
Accepted: 24 May 2023

Electrochemical carbon dioxide (CO2) conversion to hydrocarbon fuels, such

Check for updates as methane (CH4), offers a promising solution for the long-term and large-scale
storage of renewable electricity. To enable this technology, CO2-to-CH4 con-
1234567890():,;
1234567890():,;

version must achieve high selectivity and energy efficiency at high currents.
Here, we report an electrochemical conversion system that features proton-
bicarbonate-CO2 mass transport management coupled with an in-situ copper
(Cu) activation strategy to achieve high CH4 selectivity at high currents. We
find that open matrix Cu electrodes sustain sufficient local CO2 concentration
by combining both dissolved CO2 and in-situ generated CO2 from the bicar-
bonate. In-situ Cu activation through alternating current operation renders
and maintains the catalyst highly selective towards CH4. The combination of
these strategies leads to CH4 Faradaic efficiencies of over 70% in a wide current
density range (100 – 750 mA cm-2) that is stable for at least 12 h at a current
density of 500 mA cm-2. The system also delivers a CH4 concentration of 23.5%
in the gas product stream.

Electrochemical carbon dioxide (CO2) reduction (ECR) to chemical
fuels offers a large-scale solution to the long-term storage of renewable
electricity1. Among many possible ECR products, methane (CH4) is an
appealing target for energy storage applications because of its high
energy density (55.5 MJ kg−1), widespread use, and large market size
(responsible for ~23% of the global energy use)2. CH4 produced from
CO2 and renewable electricity could be readily integrated into the
existing natural gas infrastructure, providing a direct path to its
decarbonization3.
Practical ECR systems must operate at high current densities
(often >300 mA cm−2) with high selectivity and energy efficiency1,4. To
reduce separation costs, high product concentrations and avoiding
CO2 loss to carbonate crossover are desirable5. Gas-phase ECR sys-
tems, including both flow-cell and membrane electrode assembly
(MEA) configurations, have been used as platforms to achieve selective
CO2 conversion at high current densities6–10. Using gas-phase systems,
CO2-to-CH4 conversion with high Faradaic efficiencies (FEs) (60–70%)
at relatively high current densities (200–300 mA cm−2) have been
demonstrated11–22. However, gas-phase systems often use alkaline
electrolytes or anion exchange membranes to achieve these high
selectivities. In these configurations, carbonate formation (with alka-
line electrolytes) or crossover (with anion exchange membranes) is
significant, requiring additional energy to recycle the electrolyte or
CO25,23,24.
Carbonate and bicarbonate-fed systems have recently been
developed to integrate CO2 capture and conversion steps within a
single system25–28. In this architecture, CO2 is generated in situ inside
the reactor when protons generated from a bipolar membrane (BPM)
react with (bi)carbonate. This system offers effective carbon utiliza-
tion that bypasses the energy-intensive step of extracting CO2 from a
CO2 captured solution25,29. It also allows the production of gas pro-
ducts with high concentrations, as the gaseous products do not mix
with the CO2 feedstock29. Using (bi)carbonate fed systems, high car-
bon monoxide (CO) and formate selectivity (FEs over 70%) at rela-
tively high partial current densities (70–150 mA cm−2) have been
demonstrated25,26,28,30.

1
Department of Chemical Engineering, Queen’s University, Kingston, ON K7L 3N6, Canada. 2ICFO–Institut de Ciències Fotòniques, The Barcelona Institute of
Science and Technology, Barcelona 08860, Spain. 3These authors contributed equally: Cornelius A. Obasanjo, Guorui Gao, Jackson Crane.
e-mail: caothang.dinh@queensu.ca

Nature Communications | (2023)14:3176 1

Article https://doi.org/10.1038/s41467-023-38963-y

Selectivity for CH4 production from recent bicarbonate-fed sys- bicarbonate-fed ECR system, we used one-dimensional multiphysics
tems is relatively low compared to gas-phase systems, however. The modeling. The cation exchange layer and porous copper catalyst sub-
current state-of-the-art for CH4 production from bicarbonate exhib- systems are considered in the model (Fig. 1a, b). Key physics are con-
ited a FE of 30% and a partial current density of 120 mA cm−2 ref. 27. We sidered including species and charge transport, electrocatalytic
hypothesized that this limited performance stems from a yet uncon- reactions, CO2 phase transfer, and buffer equilibrium reactions,
trolled reaction environment, including both local CO2 concentration with constants derived from past work35–39. A proton flux from water
and local pH, and the lack of selective catalysts, possibly because Cu dissociation within the BPM, proportional to the total integrated cur-
reconstruction at high current density typically favors the hydrogen rent density, is imposed on the anion exchange layer/cation exchange
evolution reaction31–34. layer interface. A mass flux boundary condition is imposed on
In this work, we unveil the critical role of electrode pore size on the catalyst layer/flow channel interface from the bulk electrolyte.
local CO2 availability in bicarbonate-fed systems using BPMs. We found In this system, CO2 comes from two sources: dissolved CO2 in the
that large Cu pores enhance the transport of dissolved CO2 and favor KHCO3 electrolyte and CO2 generated during the reaction. Under
bicarbonate conversion into CO2 leading to high local CO2 con- reverse bias, protons (H+) generated from the surface of the BPM react
centration. We further develop an in-situ catalyst activation strategy with bicarbonate ions (HCO3−) in the electrolyte, forming CO2 in the
that generates and maintains highly specific Cu surfaces for selective solution.
CH4 production from CO2 dissolved in aqueous solutions. We imple-
ment this catalyst in large pore electrodes to report selective and HCO3
+ H + ! CO2 + H2 O ð1Þ
concentrated CH4 generation, at high current densities. Our aqueous-
fed system achieves a CO2-to-CH4 conversion with over 70% FE at a The CO2 formed in situ diffuses to the surface of the catalyst, and
wide current density range of 100–750 mA cm−2, with a record-high when formed in excess of the solubility limit, forms bubbles on the
CH4 partial current density of over 500 mA cm−2. The system is also surface of the membrane (Fig. 1b). CH4 is generated in an 8-electron
stable, maintaining its high CH4 FE for at least 12 h at a current density reaction, consuming CO2 and producing OH−.
of 250–500 mA cm−2. Our aqueous-fed system also outperforms pre-
vious systems in terms of full-cell energy efficiency and delivers a CO2 + 4H2 O + 8e
! CH4 + 8OH
ð2Þ
record CH4 concentration of up to 23.5% in the gas outlet stream.
State-of-the-art bicarbonate electrolyzers rely on catalysts within
Results a dense porous matrix with the CO2 source in-situ generation from
Modeling the reaction environment and CO2 availability bicarbonate26–28.
Electrochemical CO2 reduction relies on high CO2 availability within Due to the critical importance of CO2 availability on electrolyzer
the catalyst domain. To study the role of CO2 availability in a performance40, we examined CO2 availability in the catalyst domain for a

a b Modeled domain
H2O/ CH4/ Bulk catholyte
KOH KHCO3 flux
CO2 phase transfer
Proton
flux

Electrocatalytic
reactions

Fixed space e-
charge
Buffer Diffusion and
reactions electromigration
Half-cell
O2/ CO2/
potential
KOH KHCO3
Porous Ni BPM Porous Cu CEL Porous Cu Flow plate
Anode Cathode Catalyst

c Flow channel flux 750

d
10 70
In-situ generation 500
CO2 to CH 4 250
Bubbling 60 CO2 sparging
CH 4 Faradaic efficiency (%)

750
CO 2 flux (mmol m −2 s−1)

5 500
250 50 Open matrix
250 500 750 250 500 750
40 N2 sparging
0

30

-5 Dense matrix
20
CO2 sparging

10 N2 sparging
-10

0
N2 CO 2 N2 CO 2 100 200 300 400 500 600 700 800

Dense matrix Open matrix Current density (mA cm −2 )

Fig. 1 | Modeling reaction environment and CO2 availability. a Schematic illus- 750 mA cm−2 for dense matrix and open matrix catalysts with N2 sparging or CO2
tration of the aqueous solution-fed ECR system using porous Cu cathode, Ni foam sparging. 0.3 M KHCO3 was used as an electrolyte for all modeling. d Modeled CH4
anode and bipolar membrane (BPM). b Schematic of modeled domain with key FE as a function of current density for a dense matrix catalyst (solid lines) and open
physics annotated. c CO2 flux components at three current densities, 250, 500, and matrix catalyst (dashed lines) with CO2 sparging and N2 sparging.

Nature Communications | (2023)14:3176 2

Article https://doi.org/10.1038/s41467-023-38963-y

H2 FE CH4 FE Current

a 100 b100
0 0

Current density (mA cm )

80

Current density (mA cm )

-2
80

-2
Cyclic

Faradaic efficiency (%)

Faradaic efficiency (%)

voltammetry
60 -100 60 -100

40 40
-200 -200
20 Constant current
Constant current 20

0 -300 0 -300
0 10 20 30 40 50 60 0 20 40 60 80 100
Time (min) Time (min)

c 100 d 2
0
Oxidation 1

Current density (mA cm )

80

Current density (mA cm )

-2

-2
Faradaic efficiency (%)

0
60 -100
-1

-200
40

-200
20 -250
Reduction
Alternating current
0 -300 -300

0 50 100 150 200 0 20 40 60 80

Time (min) Time (s)

Fig. 2 | Electrocatalytic CO2 reduction with constant and alternating current density of 250 mA cm−2. c Product distribution of Cu mesh operated using alter-
operations. a Product distribution of Cu mesh over time operated at a constant nating negative and positive currents (oxidation current density of 0.5 mA cm−2 and
reduction current density of 250 mA cm−2. b Product distribution of Cu mesh after oxidation time of 3 s; reduction time of 25 s and CO2 reduction current of
catalyst surface activation using cyclic voltammetry (CV) between −20 and 250 mA cm−2). d Magnification in time of the square-wave alternating reduction and
10 mA cm−2 for 10 cycles, followed by operation at a constant reduction current oxidation current density in (c).

dense porous matrix catalyst. We found that most CO2 in-situ generation increasing CO2 concentration from the flow channel by sparging
via Eq. 1 occurs near the catalyst-membrane interface (Fig. S1), while the electrolyte with CO2 instead of N2. CO2 flux tracking shows
dissolved CO2 from the bulk electrolyte enters via the flow channel. To that when sparging with CO2, the CO2 influx from the flow
generate substantial quantities of C1+ products via ECR, CO2 must readily channel improves substantially, leading to CH4 FEs of up to 65% at
diffuse from the boundaries into the catalyst domain. The model is used 500 mA cm-2. Meanwhile, CO2 sparging makes only a minor dif-
to perform a detailed CO2 accounting at different current densities to ference in the dense matrix cases (Fig. 1d) presumably because
understand CO2 dynamics within the system. CO2 fluxes are sub-divided the low diffusivity and mass transport prevent CO2 from ade-
into four components: (1) flux from the flow channel, (2) in-situ gen- quately penetrating the domain. Similar trends are observed for
eration (although CO2 can also be consumed, on net, via fixing into (bi) other catholyte concentrations from 0.1 to 1 M KHCO3 (Fig. S3).
carbonate), (3) conversion of CO2 to CH4, and (4) aqueous to vapor These results suggest a path forward and design principles to
phase transfer advected to the flow channel via bubbling. In the dense realize bicarbonate-fed electrolyzers with increased performance: a
matrix catalysts with N2 sparging at high current density, in-situ CO2 highly porous matrix enables efficient utilization of both dissolved CO2
generation is severely restricted, while flow-channel flux is nearly non- from electrolyte and in-situ formed CO2 from bicarbonate.
existent (Fig. 1c), leading to low CH4 selectivity (Fig. 1d).
Given the dominant role of CO2 diffusivity on ECR rates, we In-situ generation of selective catalysts
hypothesized that an open matrix design, which leads to higher To implement the open matrix strategy, we selected a cathode con-
diffusion and boundary mass transfer, would improve perfor- sisting of a copper mesh with large pores (average pore diameter of
mance. Modeling suggests that when switching from a dense 150 µm and average Cu wire of 100 µm) and a pore density of 100 pores
matrix to an open matrix, both with N2 sparging, performance at per inch (Fig. S4). Electrochemical CO2 reduction was performed using
high current densities improves substantially (Fig. 1d). For an aqueous-fed system coupled with a BPM operated in reverse bias
example, at 500 mA cm−2 total current density, CH4 FE increases mode (Fig. S5). A 0.3 M KHCO3 solution saturated with CO2 was used as
from 13% to 37% when switching from the dense to the open both the electrolyte and CO2 source and a 1 M KOH solution was used
matrix catalyst. Mechanistically, modeling suggests that high as the anolyte (Fig. 1a).
diffusivity from the open matrix facilitates much higher in-situ We first performed the electrochemical reaction using as-received
CO2 generation (Fig. 1c). The CO2 transport from the flow channel Cu mesh at a current density of 250 mA cm−2. In this configuration, H2
in the open matrix N2 sparging case remains, however, low. This is the major product with Faradaic Efficiency (FE) over 90% throughout
leads to adequate CO2 availability on the membrane side of the the test while the FE for CH4 is less than 0.5% (Fig. 2a).
catalyst domain, but poor CO2 availability on the flow-channel To improve CO2 reduction selectivity and to suppress H2 evolu-
side (Fig. S2). We further hypothesized that it would be possible tion reaction, we sought to reconstruct the surface of Cu catalyst
to improve the utilization of the entire catalyst domain by using an electrochemical oxidation−reduction process to form Cu

Nature Communications | (2023)14:3176 3

Article https://doi.org/10.1038/s41467-023-38963-y

100 100
-2 1s
0.1 mA cm
0.5 mA cm
-2 3s
5s

CH4 Faradaic Efficiency (%)

-2

CH4 Faradaic efficiency (%)

80 2.5 mA cm 80
-2 10 s
5.0 mA cm

60 60

40 40

20 20

0 0
100 250 500 750 100 250 500 750
-2 -2
Current density (mA cm ) Current density (mA cm )

100 500
5s

Effective CH4 current density (mA cm )

5s

-2
15 s 15 s
25 s
CH4 Faradaic efficiency (%)

80 400 25 s
50 s 50 s

60 300

40 200

20 100

0 0
100 250 500 750 100 250 500 750
-2 -2
Current density (mA cm ) Current density (mA cm )

Fig. 3 | Optimization of oxidation and reduction conditions. a Effect of oxidation
current on CH4 selectivity. Oxidation and reduction times were fixed as 5 s and 25 s,
respectively. b Effect of oxidation time. The oxidation current density was fixed at
2.5 mA cm−2 and the reduction time was 25 s. c Effect of reduction time. Oxidation
current density and time were fixed as 2.5 mA cm−2 and 5 s, respectively. d Effect of
reduction time on effective CH4 partial current density. The effective CH4 current
was calculated as follows: effective current = (total current) × (Faradaic effi-
ciency) × (time for reduction cycle)/(time for reduction cycle + time for oxidation
cycle). Data for FE of other gas products are shown in Figs. S8–S10. The error bars
represent SD (n = 3 independent replicates).

oxide-derived catalysts, which have been found to be selective cata-
lysts for CO2 reduction to hydrocarbons41–45. To form the Cu-oxide-
derived catalyst, we performed repeated cyclic voltammetry (10
cycles) in the current density range of −20 to +10 mA cm−2 before
performing CO2 reduction (Fig. 2b). We found that the CH4 FE
increased to 25% while the H2 FE decreased to 50% when Cu surface is
reconstructed via reduction–oxidation cycling (Fig. 2b). However, the
CH4 FE rapidly decreased to 3% after 60 min of reduction time. These
results suggest that the oxide-derived Cu surface can be selective but
unstable for CH4 formation.
From the results above, we reasoned that periodic Cu surface
reconstruction would enable selective and stable CH4 formation. To
this end, we applied a square-wave alternating current in which the
electrical current alternated between a negative and a positive current.
Using this operation mode, the catalyst is periodically oxidized during
CO2 reduction (Fig. 2c, d). The oxidation current density and time were
fixed at 0.5 mA cm−2 and 3 s, respectively, and the reduction time was
25 s. We found that, under these conditions, CH4 was the major pro-
duct with an FE of 55–60% maintained throughout the reaction.
Meanwhile, H2 formation is suppressed, with a stable FE around 30%
(Fig. 2c). Other minor products include C2H4 (1–3%) and CO (<1%).
During the oxidation cycle, H2 produced from the reduction cycle may
be oxidized, contributing to the observed low H2 FE. However, because
the number of charges during the oxidation cycle (0.0015 C cm−2) is
much smaller than those in the reduction reaction (6.25 C cm-2), this
possible contribution is insignificant.
Optimizing oxidation and reduction conditions
Having identified a strategy for the selective production of CH4, we
sought to further optimize the oxidation-induced Cu surface
modification by varying the oxidation and reduction current and
time. We first studied the effect of oxidation current on CH4 selec-
tivity. The oxidation time was fixed at 5 s while we varied the oxidation
current density between 0.1 and 5 mA cm−2. The reduction time was
fixed at 25 s and the reduction current density varied in the range of
100–750 mA cm−2. We found that higher oxidation current density
generally improves CH4 selectivity (Fig. 3a). Particularly, with an oxi-
dation current density of 2.5 mA cm-2, CH4 FE of over 70% was
achieved at 250 mA cm−2 and 500 mA cm−2. At a current density of
750 mA cm−2, we achieved a CH4 FE of 67.8%, corresponding to a
partial CH4 current density of 508.5 mA cm−2. C2H4 is the other main
product with an FE in the range of 2–6%. The total FE for liquid pro-
ducts is less than 5% (Fig. S6). The full cell voltage was 3.52 V at a
current density of 100 mA cm−2 and increased to 4.25 V, 5.4 V, and
7.57 V as the current density increased to 250 mA cm−2, 500 mA cm-2,
and 750 mA cm−2, respectively (Figs. S6, S7). When the oxidation
current density is higher than 2.5 mA cm−2, further increases in oxi-
dation current do not improve CH4 selectivity (Fig. 3a).
To study the effect of oxidation time, we fixed the oxidation
current density at 2.5 mA cm−2 and varied the oxidation time in the
range of 1 to 10 s. As with oxidation current, high oxidation time favors
CH4 production, especially at high current densities (Fig. 3b). To
investigate the effect of reduction time, we fixed oxidation current
density and time at 2.5 mA cm−2 and 5 s, respectively, and varied the
reduction time in the range of 5 to 50 s. Reducing the reduction time
increases CH4 selectivity substantially (Fig. 3c). With a reduction time
of 5 s, CH4 FE higher than 70% was achieved at all current densities in
the range of 100–750 mA cm−2. Notably, CH4 FE reached 77% at
500 mA cm−2 and a CH4 partial current of 525 mA cm−2 was achieved at
current density of 750 mA cm−2. We reason that highly selective Cu

Nature Communications | (2023)14:3176 4

Article
Cu 2p
Intensity (arb. units)
970 960 950 940
Binding energy (eV)
Fig. 4 | Surface changes induced by alternating reduction–oxidation current.
a, b Scanning electron microscopy (SEM) images of Cu mesh electrode before (a)
and after (b) CO2 reduction using fixed reduction current. Samples SEM image after
electrolysis with alternating reduction–oxidation times of 25 s–5 s (c), 50 s–5 s (d),
15 s–5 s (e), and 5 s–5 s (f). Cu 2p X-ray photoelectron spectroscopy (XPS) spectra of
Cu mesh electrode after CO2 reduction using fixed currents (g); and alternating
species are generated during the oxidation step. During the reduction
cycle, the Cu surface is gradually reconstructed leading to a decrease
in CH4 selectivity. Thus, shortening reduction time enables the effi-
cient use of highly selective Cu species, leading to higher CH4 selec-
tivity. It is also possible that as the reduction timescale becomes
sufficiently short, replenished CO2 within the catalyst domain enables
higher CH4 selectivity.
While decreasing reduction time increases instantaneous CH4
selectivity, it also reduces the effective duty cycle for CO2 reduction
(reduction time/(reduction time + oxidation time)). To evaluate the
effective production rate of CH4, we normalized CH4 partial current
density to total operation time (reduction and oxidation time). The
highest effective CH4 partial current recorded was over 400 mA cm-2,
achieved with a reduction time of 25 s (Fig. 3d). Longer reduction times
lower CH4 selectivity while shorter reduction times reduce effective
operating time. Our results show a tradeoff between CH4 selectivity
and effective partial current density towards reduction time.
Catalyst surface changes induced by oxidation
To understand the effect of different operating procedures on the
surface changes of the catalysts, we performed scanning electron
microscopy (SEM) characterization of the catalyst before and after CO2
Nature Communications | (2023)14:3176
https://doi.org/10.1038/s41467-023-38963-y
Cu 2p
Intensity (arb. units)
930 970 960 950 940 930
Binding energy (eV)
25 s–5 s reduction–oxidation current (h). The samples for SEM and XPS were col-
lected after being tested at 100, 250, 500, and 750 mA cm−2 reduction current
densities for 40 min at each current regime (total reaction time of 160 min). The
oxidation current density and time were 2.5 mA cm−2 and 5 s, respectively. The
reduction time considered for samples (c)–(f) are 25 s, 50 s, 15 s, and 5 s, respec-
tively. Scale bar in Fig. a–f = 200 nm.
reduction reactions. We found that CO2 reduction conditions have a
significant impact on the surface of Cu catalysts. SEM characterizations
after ECR reactions with constant reduction current reveal the for-
mation of Cu nanoparticles with irregular shapes (Figs. 4b, S12, S13) as
opposed to a smooth surface of the sample before the reaction
(Figs. 4a, S11). This observation is consistent across different locations
of multiple samples. Previous studies using operando characteriza-
tions of catalysts under ECR conditions suggested that the Cu surface
is dynamic and undergoes significant morphological changes due to
multiple processes, including dissolution/redeposition, agglomera-
tion, fragmentation and reshaping33,34,46–48. Formation of Cu nano-
particles on the Cu surface during ECR has been observed in a previous
study at a current density below 50 mA cm−2 ref. 12. Because the sur-
face changes are accelerated at higher reaction rates or more negative
applied potentials33, formation of Cu nanoparticles at current densities
of 500 to 750 mA cm−2 in our system is reasonable. In sharp contrast to
the constant reduction operation, alternating reduction–oxidation
currents were found to induce the formation of a porous Cu surface
(Fig. 4c). The porous structure becomes more pronounced when the
reduction time is reduced from 50 s to 5 s (Figs. 4d–f, S14), suggesting
that oxidation cycle frequency is an important factor governing the
morphology of the catalyst.
5
Article
Alternating
Constant
current
current
80
Faradaic efficiency (%)
60
40
20
0 1
600 Fresh
After continuous
After alternating
550 Thermally treated
Roughness factor
500
10
5 20
0 0
Sample
Fig. 5 | Roles of operation mode and surface roughness factor. a Product dis-
tribution of Cu mesh over time operated using alternating currents (oxidation
current density of 2.5 mA cm−2 and oxidation time of 5 s; reduction time of 25 s and
CO2 reduction current of 250 mA cm−2), followed by constant reduction current at
250 mA cm−2 and then alternating currents. b Roughness factors of fresh Cu mesh,
Cu mesh after constant reduction current operation, Cu mesh after alternating
current operation, and thermally treated Cu mesh after alternating current
To study the changes in the surface composition of the catalysts
induced by the different operating protocols, we performed X-ray
photoelectron spectroscopy (XPS) analyses of the samples before and
after CO2 reduction reactions (Figs. 4g, h, S15). All samples show the
presence of both metallic copper (Cu), characterized by the peak at a
binding energy of 932 eV, and copper oxide (CuO), indicated by the
peak at a binding energy of 933.5 eV (Fig. 4g, h). The ratio of the two
deconvoluted peaks at 932 and 933.5 eV are similar between the two
samples, indicating a similar ratio of copper oxide and metallic copper.
The presence of copper oxide on the Cu electrodes after the reaction
are likely due to Cu oxidation during exposure to air.
Origin of high CH4 selectivity
Previous studies of ECR operating at low current densities of
10–50 mA cm−2 in aqueous systems have suggested three key factors
that affect hydrocarbon selectivity, including the nature of Cu active
sites44,49,50, catalyst surface roughness or loading51–53, and the
electrolyte54. In our system, we found that catalyst morphology and
alternating current operation are two crucial factors for high CH4
production. To understand the effect of each factor, we performed a
series of control experiments to decouple their roles. To investigate
the contribution of operating mode, we performed ECR at a current
density of 250 mA cm−2 using alternating current until a stable CH4 FE
of over 70% was achieved for 1 h (Fig. 5a) and then switched the
reaction to the constant current mode at the same current density. The
CH4 FE drops quickly to 20% after 1 h of reaction while H2 FE increases
significantly. When the system was switched back to alternating
Nature Communications | (2023)14:3176
https://doi.org/10.1038/s41467-023-38963-y
Alternating
current
CH4
H2
2 3 4
Reaction time (h)
100 CH4
H2
80 CO
Faradaic efficiency (%)
C2H4
60
40
100 250 500 750
-2
Current density (mA cm )
operation. c Product distribution of thermally treated Cu mesh using alternating
current operation. Samples after reaction were collected after being tested at 100,
250, 500, and 750 mA cm−2 current densities for 40 min at each current density
(total reaction time of 160 min). For samples with alternating current operations,
the oxidation current density was 2.5 mA cm−2, the oxidation and reduction times
were 5 s and 25 s, respectively.
operating mode again, CH4 FE recovered to the initial FE of over 70%.
These results confirm that even with oxidation-induced activated Cu
surface, alternating current operation is needed to maintain high CH4
FE in aqueous fed system at high current densities.
To study the possible effect of impurities such as Ni and Fe ions,
which could leach from the Ni foam anode and be transported to the
cathode during alternating current operation55, we performed ECR
using stable IrOx supported on titanium felt as the anode. This yielded
similar CH4 FEs compared to Ni-based anodes (Fig. S16). We also
intentionally added Ni and Fe ions to the catholyte with concentrations
ranging from 0.1 to 1 ppm and found that Ni and Fe impurities both
show detrimental effects on CH4 production (Figs. S17, S18). These
results confirm that potential Ni and Fe impurities do not contribute to
the high CH4 FE obtained in our systems.
To study the effect of surface morphology, we estimated the
surface roughness factors of the catalysts from double-layer capaci-
tance measurements. Cu electrodes after both constant and alternat-
ing current operations show similar roughness factors of 6-8 which is
ca. 3 times higher than that of fresh Cu mesh (Figs. 5b, S19). We reason
that a relatively low surface roughness factor is the main reason for the
high CH4 FE observed in our study. This is consistent with what has
been demonstrated in aqueous ECR systems using Cu mesh and Cu
foils as electrodes, suggesting that a balance between surface rough-
ness and applied potential is crucial for high CH4 selectivity53. To fur-
ther confirm the effect of surface roughness, we prepared a high
surface area Cu mesh electrode using a thermal treatment process56.
With a high surface roughness factor of 590, the electrode produced
6
Article
C2H4 as the main hydrocarbon product (FE of 15–25%) while CH4
production is suppressed (FE less than 1%) in current densities of 250
to 750 mA cm-2 using alternating current operation (Figs. 5c, S20).
These results suggest that both low roughness factors and alternating
current operation are needed to achieve high CH4 FE and production
rates in our aqueous solution ECR system.
Finally, the local pH in the catalyst layer, estimated from numerical
simulation for the 0.3 M KHCO3 catholyte, is between 10 and 12 at
current densities of 100–500 mA cm-2. This suggests that water proton
source for CH4 formation (Fig. S21). Our results are consistent with
previous works showing an optimal local pH of 10.5–11 for CH4 forma-
tion on a relatively flat Cu surface53. Experiments using lower KHCO3
catholyte concentration show a higher C2H4:CH4 ratio (Fig. S22), further
supporting the role of pH within the catalyst layer to tune hydrocarbon
selectivity. Lower KHCO3 concentrations have reduced buffering
effects, leading to higher local pH near the flow channel (Fig. S21),
presumably outside of the optimal pH range for CH4 formation.
Experimental validation of CO2 sources
As discussed above, there are two sources of CO2 in the aqueous
solution-fed system: CO2 dissolved in the electrolyte and CO2 gener-
ated in-situ from protons reacting with bicarbonate. To understand the
role of each source, we performed a series of controlled experiments.
First, we performed ECR using CO2 saturated KHCO3 electrolyte and an
anion exchange membrane (AEM). In this configuration, dissolved CO2
is the only CO2 source because the in-situ pathway is blocked due to
lack of protons. We found that dissolved CO2 enables high CH4
selectivity (up to 52%) at low current densities of 100 mA cm−2 and 250
mA cm−2 (Fig. S23). At the higher current densities of 500 mA cm−2 and
750 mA cm−2, CH4 FE is decreased to below 40%, which is much lower
than those obtained using CO2-saturated KHCO3 and a BPM (Fig. S23).
These results suggest that dissolved CO2 can serve as the main CO2
source for CH4 production at low current densities (below
250 mA cm−2) but is not sufficient to maintain high CH4 selectivity at
higher current densities (above 500 mA cm−2).
To explore the role of in-situ generated CO2 gas, we performed
the reaction using an N2 saturated KHCO3 electrolyte and a BPM. In this
configuration, the contribution of dissolved CO2 is limited to the
equilibrium dissolved CO2 concentration of the electrolyte (~1.5 mM
vs. 33 mM in a CO2-saturated case57). While in-situ generated CO2
enables the formation of CH4, its FE is relatively low. The highest
recorded CH4 FE was 27%, achieved at 500 mA cm−2, corresponding to
a CH4 partial current density of 135 mA cm−2 (Fig. S23), which is com-
parable to previous reported data using in-situ generated CO2 gas from
bicarbonate electrolyte27. These results confirm that the presence of
dissolved CO2 is crucial for high CH4 FE and both CO2 sources are
required for high CH4 selectivity at high current densities.
We also studied the effect of catholyte KHCO3 concentration on
ECR performance. KHCO3 concentration can influence CO2 solubility,
the amount of in-situ CO2 generated, and the local pH on the catalyst
surface. Electrolyte with low KHCO3 concentration has higher CO2
solubility and yields higher local pH on the catalyst surface during ECR
due to reduced buffering. Meanwhile, high KHCO3 concentration
increases the availability of the HCO3− ion at the membrane surface for
in-situ CO2 generation. At the current density of 100 mA cm−2, the CH4
FEs with 0.1 M and 0.3 M KHCO3 electrolyte are much higher than
that of 1 M KHCO3, suggesting the important role of high local pH
at low current densities (Fig. S22). At high current densities
(500–750 mA cm−2), 0.3 M KHCO3 yields a substantially higher CH4 FE
compared to either 0.1 M or 1 M KHCO3 (Fig. S22). These results sug-
gest that a balance between dissolved CO2, in-situ generated CO2, and
local pH is needed to achieve high CO2-to-CH4 conversion. Compared
to the dense electrodes used in previous studies27, the open matrix
catalyst used in this work effectively utilizes dissolved CO2 as the
carbon source. Therefore, highly concentrated KHCO3 electrolyte is
Nature Communications | (2023)14:3176
https://doi.org/10.1038/s41467-023-38963-y
not needed to provide enough in-situ generated CO2. As a result, the
open matrix electrode allows the utilization of KHCO3 electrolyte with
relatively low concentration, which produces favorable local pH for
CO2 conversion.
To validate the experimental trends and the conclusions from the
multiphysics model, the model was tested against various experi-
mental configurations, including AEM operation (Fig. S24), various
sparging gases (Fig. S24), and at different electrolyte concentrations
(Fig. S25). The model can replicate cell performance trends in all cases.
Product concentration and system stability
To analyze product concentration in the outlet stream, we designed a
system in which the gas product is separated from CO2 bubbling
solution. Gas and liquid electrolyte go through a separator. Gas pro-
ducts were collected for analysis while liquid electrolyte is recycled to
the CO2-saturating solution (Fig. 6a). CH4, H2, and CO2 are the main
components in the gas products. At 100 mA cm−2, both CH4 and H2
concentrations are relatively low, while CO2 makes up the bulk of the
gas stream (Fig. 6b). At higher current densities, CH4 and H2 con-
centration increases, reaching a maximum concentration of 23.5% for
CH4 at 500 mA cm−2. The molar ratio of CH4 produced to unreacted
CO2 gas was 40.7%, exceeding the highest previously reported of
34%27. During the reaction, CO2 gas is produced from the reaction of
bicarbonate and protons from the BPM (Eq. 1). At the same time, CO2 is
consumed via the formation of CH4 (Eq. 2) and the reaction with
hydroxyl ions (OH−) during the reaction.
CO2 + OH
! HCO3
ð3Þ
In principle, all CO2 produced from Eq. (1) can be converted to
CH4 or HCO3− via Eqs. (2) and (3) because the amount of H+ and OH−
produced are equal. Excess amount of CO2 observed in the gas stream
could be due to the low efficiency of the reaction (3) or slow absorp-
tion from gas phase CO2 bubbles into the bulk electrolyte. We reason
that a well-designed gas/liquid separator that allows efficient CO2
conversion to bicarbonate would significantly reduce the amount of
CO2 gas in the product stream.
To evaluate the stability of the system, we performed the reaction
at current densities of 250 and 500 mA cm−2 and tracked the gas pro-
ducts over time (Figs. 6c, d, S26). At 250 mA cm−2, a CH4 FE in the range
of 70–75% was maintained over a period of 12 h. H2 and CO FEs were
relatively constant while ethylene FE increased slightly from 2% to 4%
after 12 h of reaction. A similar trend was observed at the current
density of 500 mA cm−2 with CH4 FE being stable at around 65 - 70%
throughout the experiment of 12 h in duration (Fig. 6d).
Discussion
High CH4 selectivity at relatively high partial current density can be
obtained using an alkaline flow cell system (Fig. 7a and Table S1)11–17.
However, carbonate formation in this system requires a significant
amount of energy for electrolyte regeneration. Neutral-pH electrolytes
can reduce carbonate formation but often lead to lower CH4 selectivity
at relatively lower partial current density compared to alkaline flow
cells (Fig. 7a and Table S1)18–22. While high CH4 selectivity has been
achieved with flow cells, the flow cell platform typically exhibits a large
cell resistance, requiring high cell voltages to achieve high current
densities. Thus, energy efficiency has not been reported for flow cell
systems11–22. MEA cells using an anion exchange membrane can pro-
duce CH4 at relatively high selectivity and energy efficiency (Fig. 7a, b)2.
However, CO2 crossover with anion exchange membranes requires
additional CO2 separation. In both flow cell and MEA systems, CH4
produced is diluted with the unreacted CO2 gas stream leading to low
CH4 product concentration (Fig. 7c). Previous bicarbonate-fed systems
produced CH4 with relatively high concentration and minimized CO2
crossover. However, they exhibit low energy efficiency, FE, and partial
7
Article https://doi.org/10.1038/s41467-023-38963-y

CO2
100

Product concentration (%)

80
CO2

60
CO2 saturated
electrolyte To GC Pump
40
CH4

20
H2

Pump Gas / liquid separator Cathode Anode 0

100 250 500 750
-2
Current density (mA cm )

80 80

Faradaic efficiency (%)

Faradaic efficiency (%)

60 60
CH4 CH4
-2
250 mA cm 500 mA cm
-2
H2 H2
40 C2H4 40 C2H4
CO CO

20 20

0 0

0 2 4 6 8 10 12 0 2 4 6 8 10 12
Reaction time (h) Reaction time (h)

Fig. 6 | Product concentration and system stability. a Schematic illustration of (d) shows the stability of the system. The operation conditions in Fig. (b)–(d) were
the experimental setup for analyzing product concentration. b H2, CH4, and CO2 as follows: 25 s reduction time, 2.5 mA cm−2 oxidation current density, and 5 s
concentrations in the outlet gas stream at different current densities. Gas product oxidation time.
distribution over time at current densities of 250 mA cm−2 (c) and 500 mA cm−2

current density compared to those obtained with flow cell and MEA anode, respectively, and were separated by a bipolar membrane
systems (Fig. 7)27. (Fumasep) or an anion exchange membrane (Fumasep). The exposed
We developed a bicarbonate-fed system which achieves the high sizes of both the cathode and anode electrodes were 1 cm × 2 cm (a
CH4 selectivity (>70%) typically found in flow cell configurations using geometric area of 2 cm2 was used for all current density calculations).
gas-phase CO2 electrolysis. The obtained CH4 partial current density of The copper mesh was configured to be in direct contact with the
over 400 mA cm−2 exceeds most of what has been previously achieved membrane. The polytetrafluoroethylene (PTFE) gasket was used to
in both flow cell and MEA systems (Fig. 7a). Our system delivers an prevent contact between anode and cathode titanium flow plates. A
energy efficiency of 15–18% in the current range of 250–500 mA cm−2 PTFE spacer was used to avoid direct contact between the Cu mesh and
which are comparable to previous MEA systems while minimizing the cathode flow plate.
carbonate crossover (Fig. 7b). Particularly, our dissolved CO2 and For the electrolysis system, a potentiostat (Autolab PGSTAT204)
bicarbonate-fed system also delivers the highest CH4 product con- with a current booster (Metrohm Autolab, 10 A) was used for all
centration compared to all previous CO2-to-CH4 systems (Fig. 7c). The experiments. Two peristaltic pumps were used to circulate the anolyte
performance of the architecture is facilitated by in-situ surface acti- (1 M KOH; 200 mL) and catholyte (0.1 M, 0.3 M, or 1 M KHCO3; 1000 mL)
vation of the copper catalyst combined with facile liquid transport between the reservoirs and the electrochemical cell at a flow rate of
enabled by an open matrix structure. An optimized surface activation 30 mL min−1. Before the measurement, CO2 gas (Praxair, 99.99%) or N2
strategy is presented which balances surface reconstruction with gas (Praxair, 99.99%) was purged in the KHCO3 aqueous solution for at
maximizing effective current density. The large pore structure of the least 30 min. CO2 was continuously purged into the catholyte
catalyst allows both dissolved CO2 and CO2 generated in situ to be throughout the experimental process (during both oxidation and
efficiently utilized, as confirmed by both experiment and multiphysics reduction cycles) at a flow rate of 50 standard cubic centimeters per
modeling. Further improvements to the system performance are minute (sccm). Electrolysis was carried out for 2500 s at each tested
possible through the suppression of CO2 bubbling and optimization of current density, with the reaction gaseous product being analyzed
the gas/liquid separation scheme. every 15 min. Gas products coming out from the cell were carried by the
CO2 or N2 stream bubbled through the catholyte reservoir to a 6-way
Methods valve where it was injected to a gas chromatography (GC, PerkinElmer
Electrochemical CO2 reduction Clarus 590) for quantification. The GC is equipped with a thermal
Electrochemical measurements were performed in a two-electrode conductivity detector (TCD) operated at 180 °C and a flame ionization
MEA flow cell. Details of the electrolysis cell and the electrolysis system detector (FID) operated at 250 °C. A molecular sieve (5A) packed col-
employed in this work are described in Fig. S5. In the MEA flow cell, umn (Supelco) connected to the TCD was used to analyze CO, CO2, and
copper mesh (100 pores per inch (ppi)) and nickel foam (1.5 mm H2 products while a Carboxen-1000 packed column (Supelco) con-
thickness; 80–100 ppi, MTI Corp.) were used as the cathode and nected to the FID was employed to quantify CH4, C2H4 and other

Nature Communications | (2023)14:3176 8

Article https://doi.org/10.1038/s41467-023-38963-y

a
80

CH Faradaic efficiency (%)

60

40 Flow cell (akaline)

Flow cell (neutral)
MEA
4

20 Bicarbonate fed
This work (Total CH4 current density)
This work (Effective CH4 current density)
0
100 200 300 400 500 600
-2
CH4 partial current density (mA cm )
b c
20 25

20

CH4 concentration (%)

Energy efficiency (%)

15

15
Flow cell
10
MEA
10 Bicarbonate fed
This work
5
MEA 5
Bicarbonate fed
This work
0 0

100 200 300 400 500 100 200 300 400 500
-2 -2
Effective CH4 partial current density (mA cm ) Effective CH4 partial current density (mA cm )

Fig. 7 | Performance comparison between ECR systems. Comparison of our work a CH4 Faradaic efficiency. b Energy efficiency. c CH4 product concentration at the
with previous studies on CO2-to-CH4 conversion with partial current densities over outlet stream. Effective CH4 partial current density was used for Fig. 7b, c. Detailed
100 mA cm−2 using a flow cell (with alkaline electrolyte11–17 and pH-neutral data are shown in Table S1.
electrolyte18–22), an MEA cell with AEM2, 11, and bicarbonate fed using BPM27.

potential hydrocarbons. Both columns were operated at a fixed tem- calculated as follows:
perature of 180 °C and with Ar carrier gas (flowrate of 20 sccm).
Aliquots of the liquid products were analyzed using nuclear 4t r + 4t o
xk = x0 k 
magnetic resonance spectroscopy (NMR)1.H NMR spectra of freshly 4t r
collected liquid products were acquired on an Auto-400 ultrashield
Bruker Instrument operating at denoted spectrometer frequency where x’k is the diluted concentration (mixing of gas production in
given in megahertz (MHz) at 25 °C in D2O using water suppression both oxidation and reduction cycle) measured by the Gas chromato-
mode, with dimethyl sulfoxide as the internal standard. graphy. 4t r and 4t o are the reduction and oxidation time in each
The FE of ECR gas products was calculated as follows: reduction–oxidation cycle.
nk  F  x k  F m
FE k = × 100% The FE of ECR liquid products were calculated as follows:
I

n  F  4δ l
Where FE k is the FE of product k , nk is the number of electrons (for k= FE l = × 100%
4Q
CH4, nk is 8) transferred to form gaseous product k , F is Faraday’s
constant (96,485 C mol−1), F m is the molar flow rate of the gas outlet
stream in mol Where n is the number of electrons transferred to form liquid product
s . x k is the molar fraction of the gas product k in the gas
outlet stream during reduction cycle. I is the total (applied) current in l, F is the Faraday’s constant (same as above), Δδl is the accumulated
Amperes (A ) during reduction cycle. number of moles of the corresponding product l. 4Q is the total
Because the reduction and oxidation cycles are relatively short charge transfer during the electrolysis.
and the CO2 is constantly flowing through the system (during both CH4 partial current density was calculated in similar way as:
reduction and oxidation cycles), gas produced during the reduction
cycle is diluted by CO2 flow during the oxidation cycle. Thus, the actual I  FE CH 4
j CH4 =
molar fraction of the gas product during the reduction cycle is a

Nature Communications | (2023)14:3176 9

Article
Where j CH 4 is the partial current density of CH4. a is the geometric
surface area (a = 2 cm2 in our system). I is the total current during the
reduction cycle.
Effective CH4 current density was calculated as follows:


4t r
J CH 4 ,ef f = I  FE CH 4 
4t r + 4t o
where J CH 4 ,ef f is the effective CH4 current density. The effective current
density reflects the effective operation time of the electrolyzer con-
sidering both the time for CO2 conversion (reduction) and the time for
catalyst regeneration (oxidation). 4t r and 4t o are the reduction and
oxidation times in each reduction–oxidation cycle. The factor 4t 4t+ r4t is
r o
used in the expression to weight operation (conversion) time to the
total time.
The energy efficiency, EE, was calculated as follows:
! the concentration-dependent Butler–Volmer equation57. Only the H2
E o  FE CH 4
EE =
V cell
Where V cell is the applied cell potential for a given geometric current
density. E o = 1.06 is the equilibrium potential for the reaction:
CO2 + 2H2O = CH4 + 2O2.
IrOx/Ti felt preparation
The IrOx/Ti felt was prepared using a drop-casting method. A mixture
of 40 mg of IrOx (Fuel Cell Store) and 160 µL of 5 wt% Nafion per-
fluorinated resin solution (Sigma-Aldrich) in 5 ml methanol was soni-
cated for 30 min. Then 1 mL of the resulting mixed solution was drop-
casted onto a Ti felt substrate (1 cm × 2 cm; Fuel Cell Store). The
sample was allowed to dry overnight in ambient conditions.
High surface area Cu electrode preparation
The high surface area Cu electrode was prepared by thermal treat-
ment. The Cu mesh was kept in the furnace and heated to 500 °C at a
ramp rate of 5 °C min−1. The furnace temperature was maintained at
500 °C for 3 h before it was cooled down to room temperature.
The resulting high surface area Cu mesh was pretreated and electro-
reduced at a current density of 20 mA cm−2 for 300 s in H-cell with a
0.3 M KHCO3 catholyte solution before performing an alternating
reduction/oxidation reaction.
Electrochemical double-layer capacitance measurement
Electrochemical double-layer capacitance was determined by mea-
suring cyclic voltammetry (CV) in an H-cell, made up of two different
chambers and separated by an anion exchange membrane. A three-
electrode setup was used, comprised of the working electrode and a
KCl saturated reference electrode (Ag/AgCl electrode) in the cathode
chamber and a platinum counter electrode placed in the anode
chamber. All potentials for this measurement were determined against
the Ag/AgCl reference electrode. The dimension of the working elec-
trode which was immersed in the electrolyte is 1 cm × 1 cm. First, the
potential range of the non-Faradaic current regime was determined
from CV. CV measurements were then conducted in a CO2 saturated
0.3 M KHCO3 electrolyte solution by sweeping the potential in the non-
faradaic region between −0.55 and −0.65 V vs. Ag/AgCl with 3 cycles for
each scan rate of 0.1, 0.2, 0.4, 0.6 0.8, 1, 1.2, and 1.4 V s−1. For the
thermally treated Cu mesh, scan rates of 0.005, 0.01, 0.02, 0.04, and
0.06 V s-1 were used. Roughness factors of Cu electrodes were esti-
mated by comparing their capacitances against that of an ideally
smooth Cu surface (0.029 mF cm−2)56.
Characterizations
Catalyst surface morphology was characterized by a ZEISS Auriga FE-
SEM operated at 3 kV. XPS measurements were performed using a
Nature Communications | (2023)14:3176
https://doi.org/10.1038/s41467-023-38963-y
Thermo Scientific K-Alpha spectrophotometer with a monochromated
Al Kα X-ray radiation source.
Multiphysics modeling
A one-dimensional multiphysics model was developed to describe the
catalyst layer and cation exchange membrane transport dynamics. The
transport and reaction of aqueous HCO3–, K+, CO32–, OH–, H+, and CO2
are considered in the model, treated with dilute solution theory. The
transport of species is governed by the Nernst-Planck equation, with
the convection term neglected and electroneutrality imposed. The
effective diffusion coefficient is determined with the Bruggeman cor-
rection. The porosity of the open matrix catalyst is assumed to be 0.99,
based on the free transport expected in a mesoscale copper mesh, and
the dense matrix 0.838), matching previous works, with the gaseous
fraction set to 0.2 in both cases38. The current distribution is deter-
mined using Ohm’s law. The electrochemical kinetics are described by
and the CH4 evolution reactions are considered due to the low
experimentally recorded rates of other competing reactions. The
electrochemical rates are fit to the experimental performance of the
copper mesh catalyst under alternating current (5 s oxidation, 25 s
reduction) for CO2 sparging and N2 sparging cases. The model per-
formance was tested against the CO2 sparging with an AEM case (Fig.
S24). We assume that the dense matrix and open matrix catalysts share
the same electrochemical rates because it is not expected that mac-
roscale porosity would affect catalyst morphology and performance38.
The electrode-specific surface area is assumed to be 104 m–1 for open
matrix catalyst and 105 m–1 for the dense matrix to account for reduced
specific surface area when comparing a copper mesh to a copper foam.
The Donnan equilibrium boundary condition was used to describe the
charge discontinuity between the catalyst layer and the cation
exchange layer (CEL). The CEL has a fixed space charge density of
–1.75 C m−3 ref. 58. The water hydration (mol H2O per mol SO3–) in the
CEL was determined to be 659, which defines CEL diffusivity, as in
ref. 60 Finite-rate carbonate buffer kinetics were included throughout
the catalyst and CEL domains to model HCO3–, CO32–, OH–, H+, and CO2
equilibrium37. Henry’s law was used to model CO2 phase transfer from
liquid to vapor phase, and it was assumed that any CO2 bubbled away
could not re-dissolve into the solution due to the fast timescales
associated with bubble advection38,39.
The boundary condition at the end of the catalyst layer corre-
sponds to the interface between the catalyst layer and flow plate. A
mass flux boundary condition corresponding to finite mass flux from a
channel with a constant Sherwood number was imposed. An assumed
Sherwood number of 36.6 was used for the open matrix case and 3.66
(corresponding to laminar channel flow) for the dense matrix case to
account for unsteady convection increasing catholyte flux in the open
matrix case. The catholyte used was 0.1 M, 0.3 M, or 1.0 M KHCO3
where indicated. CO2 sparging was imposed by setting the CO2
concentration to the CO2 saturation concentration (33 mM), and N2
sparging by setting the CO2 concentration to the equilibrium con-
centration (1.5 mM for 0.3 M KHCO3)59. All other species concentra-
tions are set to their equilibrium values, shown in Table S2. The solid-
phase electric potential on the flow-plate interface is set to the half-cell
potential (versus SHE), varied from 0 V to −2.1 V. The boundary con-
dition on the other side of the domain corresponds to the interface
between the cation exchange layer and anion exchange layer in a BPM.
To account for the proton flux from the BPM, a boundary flux of
protons equal to the integrated current density divided by the Faraday
constant times the transference number is imposed as in ref. 38 The
net ionic current is from K+ across the membrane. All other species
(HCO3–, CO32–, OH–, and CO2) have a no-flux condition at the CEL/AEL
boundary. A transference number of 0.75 is used for the BPM cases
(somewhat lower than in Lees et al. due to the comparatively lower
catholyte K+ concentration compared to the anolyte), and 0 for the
10
Article
AEM case. The electrolyte potential at the CEL/AEL boundary is
set to 0 V.
The model is implemented in COMSOL Multiphysics version 6.0
and solved, assuming steady state, using the PARDISO solver. The 1D
domain is discretized into elements of maximum size 0.5 µm, with
element sizes of 0.02 µm close to boundaries. The results were found
to be independent of further mesh refinement (Fig. S27). Additional
model parameters and definitions are included in Table S2 of the SI.
Data availability
All the data supporting the findings of this study are available within
the article and its Supplementary Information and Source Data
file. Source data are provided in this paper.
References
1. De Luna, P. et al. What would it take for renewably powered elec-
trosynthesis to displace petrochemical processes? Science 364,
eaav3506 (2019).
2. Xu, Y. et al. Low coordination number copper catalysts for elec-
trochemical CO2 methanation in a membrane electrode assembly.
Nat. Commun. 12, 2932 (2021).
3. Crane, J. & Dinh, C.-T. Strategies for decarbonizing natural gas with
electrosynthesized methane. Cell Rep. Phys. Sci. 3, 101027
(2022).
4. Kibria, M. G. et al. Electrochemical CO2 reduction into chemical
feedstocks: From mechanistic electrocatalysis models to system
design. Adv. Mater. 31, 1807166 (2019).
5. Sisler, J. et al. Ethylene electrosynthesis: a comparative techno-
economic analysis of alkaline vs membrane electrode assembly vs
CO2–CO–C2H4 tandems. ACS Energy Lett. 6, 997–1002 (2021).
6. García de Arquer, F. P. et al. CO2 electrolysis to multicarbon pro-
ducts at activities greater than 1 A cm−2. Science 367, 661 (2020).
7. Dinh, C.-T. et al. CO2 electroreduction to ethylene via hydroxide-
mediated copper catalysis at an abrupt interface. Science 360,
783 (2018).
8. Weekes, D. M., Salvatore, D. A., Reyes, A., Huang, A. & Berlinguette,
C. P. Electrolytic CO2 reduction in a flow cell. Acc. Chem. Soc. 51,
910 (2018).
9. Nguyen, T. N. & Dinh, C. T. Gas diffusion electrode design for
electrochemical carbon dioxide reduction. Chem. Soc. Rev. 49,
7488–7504 (2020).
10. Wakerley, D. et al. Gas diffusion electrodes, reactor designs and key
metrics of low-temperature CO2 electrolysers. Nat. Energy 7,
130–143 (2022).
11. Wu, Y. et al. Enhancing CO2 electroreduction to CH4 over Cu
nanoparticles supported on N-doped carbon. Chem. Sci. 13,
8388–8394 (2022).
12. Han, Z. et al. Steering surface reconstruction of copper with elec-
trolyte additives for CO2 electroreduction. Nat. Commun. 13,
3158 (2022).
13. Chen, S. et al. Engineering water molecules activation center on
multisite electrocatalysts for enhanced CO2 methanation. J. Am.
Chem. Soc. 144, 12807–12815 (2022).
14. Chen, S. et al. MOF Encapsulating N-heterocyclic carbene-ligated
copper single-atom site catalyst towards efficient methane elec-
trosynthesis. Angew. Chem. Int. Ed. 61, e202114450 (2022).
15. Xiong, L. et al. Geometric modulation of local co flux in Ag@Cu2O
nanoreactors for steering the CO2RR pathway toward high-efficacy
methane production. Adv. Mater. 33, 2101741 (2021).
16. Fan, Q. et al. Manipulating Cu nanoparticle surface oxidation states
tunes catalytic selectivity toward CH4 or C2+ products in CO2
electroreduction. Adv. Energy Mater. 11, 2101424 (2021).
17. Chen, S. et al. Highly selective carbon dioxide electroreduction on
structure-evolved copper perovskite oxide toward methane pro-
duction. ACS Catal. 10, 4640–4646 (2020).
Nature Communications | (2023)14:3176
https://doi.org/10.1038/s41467-023-38963-y
18. Wang, Y.-R. et al. Implanting numerous hydrogen-bonding net-
works in a Cu-porphyrin-based nanosheet to boost CH4 selectivity
in neutral-media CO2 electroreduction. Angew. Chem. Int. Ed. 60,
21952–21958 (2021).
19. Wang, X. et al. Gold-in-copper at low *CO coverage enables effi-
cient electromethanation of CO2. Nat. Commun. 12, 3387 (2021).
20. Wang, X. et al. Efficient methane electrosynthesis enabled by tun-
ing local CO2 availability. J. Am. Chem. Soc. 142, 3525–3531 (2020).
21. Sedighian Rasouli, A. et al. CO2 electroreduction to methane at
production rates exceeding 100 mA/cm2. ACS Sustain. Chem. Eng.
8, 14668–14673 (2020).
22. Li, Y. et al. Promoting CO2 methanation via ligand-stabilized metal
oxide clusters as hydrogen-donating motifs. Nat. Commun. 11,
6190 (2020).
23. Rabinowitz, J. A. & Kanan, M. W. The future of low-temperature
carbon dioxide electrolysis depends on solving one basic problem.
Nat. Commun. 11, 5231 (2020).
24. Salvatore, D. A. et al. Designing anion exchange membranes for
CO2 electrolysers. Nat. Energy 6, 339–348 (2021).
25. Zhang, Z. et al. Conversion of reactive carbon solutions into co at
low voltage and high carbon efficiency. ACS Cent. Sci. 8,
749–755 (2022).
26. Zhang, Z. et al. Porous metal electrodes enable efficient electrolysis
of carbon capture solutions. Energy Environ. Sci. 15, 705–713 (2022).
27. Lees, E. W. et al. Electrolytic methane production from reactive
carbon solutions. ACS Energy Lett. 7, 1712–1718 (2022).
28. Li, T., Lees, E. W., Zhang, Z. & Berlinguette, C. P. Conversion of
bicarbonate to formate in an electrochemical flow reactor. ACS
Energy Lett. 5, 2624–2630 (2020).
29. Li, M., Irtem, E., Iglesias van Montfort, H.-P., Abdinejad, M. & Bur-
dyny, T. Energy comparison of sequential and integrated CO2
capture and electrochemical conversion. Nat. Commun. 13,
5398 (2022).
30. Li, T. et al. Electrolytic conversion of bicarbonate into CO in a flow
cell. Joule 3, 1487–1497 (2019).
31. Kim, Y.-G., Baricuatro, J. H., Javier, A., Gregoire, J. M. & Soriaga, M. P.
The evolution of the polycrystalline copper surface, first to Cu(111)
and then to Cu(100), at a fixed CO2RR potential: a study by oper-
ando EC-STM. Langmuir 30, 15053–15056 (2014).
32. Nguyen, T. N. et al. Catalyst regeneration via chemical oxidation
enables long-term electrochemical carbon dioxide reduction. J.
Am. Chem. Soc. 144, 13254–13265 (2022).
33. Lee, S. H. et al. Oxidation state and surface reconstruction of Cu
under CO2 reduction conditions from in situ X-ray characterization.
J. Am. Chem. Soc. 143, 588–592 (2021).
34. Popović, S. et al. Stability and degradation mechanisms of copper-
based catalysts for electrochemical CO2 reduction. Angew. Chem.
Int. Ed. 59, 14736–14746 (2020).
35. Weng, L.-C., Bell, A. T. & Weber, A. Z. A systematic analysis of Cu-
based membrane-electrode assemblies for CO2 reduction through
multiphysics simulation. Energy Environ. Sci. 13, 3592–3606 (2020).
36. Weng, L.-C., Bell, A. T. & Weber, A. Z. Towards membrane-electrode
assembly systems for CO2 reduction: a modeling study. Energy
Environ. Sci. 12, 1950–1968 (2019).
37. Weng, L.-C., Bell, A. T. & Weber, A. Z. Modeling gas-diffusion elec-
trodes for CO2 reduction. Phys. Chem. Chem. Phys. 20, 16973–16984
(2018).
38. Lees, E. W., Bui, J. C., Song, D., Weber, A. Z. & Berlinguette, C. P.
Continuum model to define the chemistry and mass transfer in a
bicarbonate electrolyzer. ACS Energy Lett. 7, 834–842 (2022).
39. Kas, R. et al. Modeling the local environment within porous elec-
trode during electrochemical reduction of bicarbonate. Ind. Eng.
Chem. Res. 61, 10461–10473 (2022).
40. Wen, G. et al. Continuous CO2 electrolysis using a CO2 exsolution-
induced flow cell. Nat. Energy 7, 978–988 (2022).
11
Article
41. Jouny, M., Luc, W. & Jiao, F. High-rate electroreduction of carbon
monoxide to multi-carbon products. Nat. Catal. 1, 748–755 (2018).
42. Li, C. W., Ciston, J. & Kanan, M. W. Electroreduction of carbon
monoxide to liquid fuel on oxide-derived nanocrystalline copper.
Nature 508, 504–507 (2014).
43. Nitopi, S. et al. Progress and perspectives of electrochemical CO2
reduction on copper in aqueous electrolyte. Chem. Rev. 119,
7610–7672 (2019).
44. Timoshenko, J. et al. Steering the structure and selectivity of CO2
electroreduction catalysts by potential pulses. Nat. Catal. 5,
259–267 (2022).
45. Arán-Ais, R. M., Scholten, F., Kunze, S. & Rizo, R. & Roldan Cuenya, B.
The role of in situ generated morphological motifs and Cu(I) species
in C2+ product selectivity during CO2 pulsed electroreduction. Nat.
Energy 5, 317–325 (2020).
46. Vavra, J., Shen, T.-H., Stoian, D., Tileli, V. & Buonsanti, R. Real-time
monitoring reveals dissolution/redeposition mechanism in copper
nanocatalysts during the initial stages of the CO2 reduction reac-
tion. Angew. Chem. Int. Ed. 60, 1347–1354 (2021).
47. Simon, G. H., Kley, C. S. & Roldan Cuenya, B. Potential-dependent
morphology of copper catalysts during CO2 electroreduction
revealed by in situ atomic force microscopy. Angew. Chem. Int. Ed.
60, 2561–2568 (2021).
48. Huang, J. et al. Potential-induced nanoclustering of metallic cata-
lysts during electrochemical CO2 reduction. Nat. Commun. 9,
3117 (2018).
49. Wei, S. et al. Construction of single-atom copper sites with low
coordination number for efficient CO2 electroreduction to CH4. J.
Mater. Chem. A 10, 6187–6192 (2022).
50. Cai, Y. et al. Insights on forming N,O-coordinated Cu single-atom
catalysts for electrochemical reduction CO2 to methane. Nat.
Commun. 12, 586 (2021).
51. Jiang, K. et al. Effects of surface roughness on the electrochemical
reduction of CO2 over Cu. ACS Energy Lett. 5, 1206–1214 (2020).
52. Ebaid, M. et al. Production of C2/C3 oxygenates from planar copper
nitride-derived mesoporous copper via electrochemical reduction
of CO2. Chem. Mater. 32, 3304–3311 (2020).
53. Ren, D., Fong, J. & Yeo, B. S. The effects of currents and potentials
on the selectivities of copper toward carbon dioxide electro-
reduction. Nat. Commun. 9, 925 (2018).
54. Varela, A. S., Ju, W., Reier, T. & Strasser, P. Tuning the catalytic
activity and selectivity of cu for CO2 electroreduction in the pre-
sence of halides. ACS Catal. 6, 2136–2144 (2016).
55. Klaus, S., Cai, Y., Louie, M. W., Trotochaud, L. & Bell, A. T. Effects of
Fe electrolyte impurities on Ni(OH)2/NiOOH structure and oxygen
evolution activity. J. Phys. Chem. C. 119, 7243–7254 (2015).
56. Li, C. W. & Kanan, M. W. CO2 Reduction at low overpotential on cu
electrodes resulting from the reduction of thick Cu2O films. J. Am.
Chem. Soc. 134, 7231–7234 (2012).
57. Min, X. & Kanan, M. W. Pd-catalyzed electrohydrogenation of car-
bon dioxide to formate: high mass activity at low overpotential and
identification of the deactivation pathway. J. Am. Chem. Soc. 137,
4701–4708 (2015).
58. Pärnamäe, R. et al. Bipolar membranes: a review on principles, lat-
est developments, and applications. J. Membr. Sci. 617,
118538 (2021).
59. Bui, J. C., Kim, C., Weber, A. Z. & Bell, A. T. Dynamic boundary layer
simulation of pulsed CO2 electrolysis on a copper catalyst. ACS
Energy Lett. 6, 1181–1188 (2021).
Nature Communications | (2023)14:3176
https://doi.org/10.1038/s41467-023-38963-y
60. Grew, K. N. & Chiu, W. K. S. A dusty fluid model for predicting
hydroxyl anion conductivity in alkaline anion exchange mem-
branes. J. Electrochem. Soc. 157, B327 (2010).
Acknowledgements
C.A.O., G.G., and J.C. acknowledge financial support from Queen’s
University. C.T.D. acknowledges the financial support from the Natural
Sciences and Engineering Research Council of Canada (NSERC), Canada
Foundation for Innovation (CFI), and Queen’s University. The authors
thank Peter Brodersen and Prof. Oleksandr Voznyy for XPS analyses.
F.P.G.d.A. and V.G. acknowledge funding from CEX2019- 000910-S
(MCIN/AEI/10.13039/501100011033), Fundació Cellex, Fundació Mir-
Puig, Generalitat de Catalunya through CERCA and the La Caixa Foun-
dation (100010434, E.U. Horizon 2020 Marie Skłodowska-Curie grant
agreement 847648).
Author contributions
C.T.D. supervised the project. C.A.O. and G.G. conducted the experi-
ments and data processing. J.C. performed modeling. V.G. and F.P.G.A.
performed SEM analysis. C.A.O., G.G., J.C., and C.T.D. wrote the
manuscript. All authors discussed the results and assisted during
manuscript preparation.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-023-38963-y.
Correspondence and requests for materials should be addressed to
Cao-Thang Dinh.
Peer review information Nature Communications thanks the anon-
ymous, reviewer(s) for their contribution to the peer review of this work.
A peer review file is available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jur-
isdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2023
12