Modulating Electric Field Distribution by Alkali Cations for CO2 Electroreduction in Strongly Acidic Medium

The reaction of carbon dioxide with hydroxide to form carbonate in near neutral or alkaline medium severely limits the energy and carbon efficiency of CO2 electroreduction. Here we show that by suppressing the otherwise predominant hydrogen evolution using alkali cations, efficient CO2 electroreduction can be conducted in acidic medium, overcoming the carbonate problem. The cation effects are general for three typical catalysts including carbon supported tin oxide, gold, and copper, leading to Faradaic efficiency of as high as 90% for formic acid and CO formation. Our analysis suggests hydrated alkali cations physisorbed on the cathode modify the distribution of electric field in the double layer, which impedes hydrogen evolution by suppress the migration of hydronium ions while at the same time promotes CO2 reduction by stabilizing key intermediates.

Main text: Electrochemical reduction of CO2 to produce chemicals and fuels is widely studied as a potential solution for renewable energy storage and CO2 recycling. 1 Because the hydrogen evolution reaction (HER) dominates in acidic aqueous solutions, CO2 electroreduction is conducted in an alkaline or near neutral medium. 2,3 These reaction media, however, create one of the most important obstacles for high-efficient steady-state CO2 electrolysis: the facile reaction of CO2 with hydroxide (OH -) to form carbonate (CO3 2-) 4,5 . In alkaline medium where the overpotentials of many catalysts are minimized thanks to a rate-influencing protondecoupled one-electron CO2 reduction step, 6,7 the carbonate problem obliges continuous refreshing of the OHelectrolytes in a flow-cell configuration in order to obtain a stable performance. 2,3 The regeneration of CO2 and 2 OHfrom aqueous carbonate is energy demanding and leads to a low or even negative energy efficiency for CO2 electroreduction. 5 In near neutral media such as bicarbonate solutions, steady-state CO2 electrolysis is possible. 8,9 However, CO2 is still consumed by OHanions electrochemically generated in CO2 electroreduction. The (bi)carbonate is protonated near the anode to regenerate CO2, which leads to a low carbon efficiency. 10 In addition, in near neutral media the high resistance of the solution 11 as well as higher overpotential for oxygen evolution reaction (OER) 12 leads to a high cell voltage and low energy efficiency. In strongly acidic media the resistance and overpotential for OER are lower, and the carbonate problem can be solved since carbonate will not cross the medium to the anode. An acidic medium is also essential to obtain formic acid from CO2 reduction. In near neutral and alkaline media, the same reduction leads to formate, which requires energy-intensive downstream processes for separation and conversion. Efficient CO2 electroreduction in an acidic medium is challenging because the HER is normally more facile than CO2 reduction. It was reported that in CO2-saturated 0.1 M HClO4 solutions (pH =1), the Faradaic efficiency of HER was nearly 100% on Fe-N-C 7 and Au 13 catalysts, both of which are efficient CO2 reduction catalysts in near neutral and alkaline media.
The group of Koper showed that CO2 reduction to CO was feasible in an HClO4-NaClO4 mixed solution with a mild acidity (pH ≥ 3). They proposed that the HER was suppressed by OHanions generated from CO2 reduction. 14 During the preparation of our publication, Huang et al. reported efficient CO2 electroreduction on Cu catalysts in strongly acidic solutions (pH = 0.67) with H3PO4-KCl mixed electrolytes. 15 In both studies, alkali cations are essential for CO2 reduction, but the mechanism of cation-promotion was unclear.
Here we demonstrate efficient CO2 electroreduction with Faraday efficiency as high as 90% in strongly acidic media (pH =1) by suppressing HER with alkali cations. This approach can be applied for three representative classes of catalysts, namely carbon supported SnO2, Au, and Cu nanoparticles (SnO2/C, Au/C and Cu/C), which give formic acid, CO, and hydrocarbons as main CO2 reduction products, respectively. Our simulation and analysis indicate that the alkali cations in the double layer of the cathode effectively shield the electric field in the diffusion layer and suppress the migration of hydronium ions towards the cathode, which lowers the concentration of hydronium ions in outer Helmholtz plane (OHP) and thus suppresses HER.
Meanwhile the cations strengthen the filed in the Stern layer and stabilizes key intermediates in CO2 reduction. The activity and selectivity of the CO2 reduction reaction are sensitive to alkali metal cations in near neutral bicarbonate solutions. 13,[16][17][18][19] Several hypotheses such as local pH effect 17 and electrostatic interaction 16,18 have been made to explain these cation effects. Inspired by these studies, we decided to explore a possible influence of alkali cations for CO2 reduction visà-vis the HER in strongly acidic medium. We first probed the effect of K + for CO2 reduction on SnO2/C at pH = 1. We used a three-electrode flow cell with a gas diffusion electrode (GDE) as working electrode ( fig. S2) for electrocatalytic tests. . This combination of electrolytes is more practical than the combination of HOTf + KOTf, even though addition of K + ions slightly changes the pH of H2SO4 solutions. For SnO2/C, formic acid was the major product of CO2 reduction. The maximum Faradaic efficiency and partial current density were 88% and 314 mA/cm 2 , respectively ( fig. 1B). These performance metrics are comparable to state-of-theart results of formic acid production from a solid-state electrolyzer 20 and formate formation in near neutral solutions 21 . Aqueous solution of formic acid could be separated from the electrolyte solution by distillation ( fig. S4). This result demonstrated the advantage of an acidic reaction medium for formic acid generation compared to a near neutral or alkaline medium where only formate would be generated. For Au/C, CO was the major product with the maximum Faradaic efficiency and partial current density as 91% and 227 mA/cm 2 , respectively ( fig. 1C). For Cu/C, formic acid, CO, methane, ethylene, propene, acetic acid, ethanol and 1-propanol were detected as the products of CO2 reduction ( fig. S5). The minimum Faradaic efficiency of HER was 16%, lower than that of Cu/PFSA catalyst (PFSA = perfluorosulfonic acid) in H3PO4-KCl medium (Faradaic efficiency of HER = 36%) reported by Huang et al. 15 Among the products of > 2e reduction, ethylene was the major product with the partial current density of 136 mA/cm 2 ( fig.   1D). Its Faradaic efficiency was 25%, similar to the highest Faradaic efficiency for ethylene production on Cu/PFSA in H3PO4-KCl medium by Huang et al. 15 The maximum partial current densities of CO formation on Au/C and ethylene formation on Cu/C in acidic media are comparable to those on state-of-the-art catalysts in near neutral and alkaline media. 2,3,22,23 We also directly compared the performances of Au/C in CO2 electroreduction in acidic, near neutral (0.8 M KHCO3) and alkaline (0.8 M KOH) media in a two-electrode cell ( fig. S6).
As expected, the carbonate problem was severe in an alkaline medium. While the initial cell voltage was low, it increased substantially (by 34%) in 2.5 hours ( fig. 2A and fig. S6D).
Meanwhile, the pH of the solution decreased from 13. As mentioned above, even in near neutral medium, carbonate forms from the reaction between CO2 and OHnear the cathode, crosses the electrolyte solution, and is protonated near the anode to regenerate CO2 ( fig. S9A). For CO2 reduction to CO, 50% of CO2 consumption is captured by the electrolyte to form carbonate. 5 Seger et al. experimentally showed that only 30% of CO2 consumption was involved in CO2 reduction when Cu catalyst was used to form deepreduction products. 10 In our steady state electrolysis in near neutral medium, the amount of CO2 in the gas mixture generated at the anode compartment was higher than that of O2 ( fig. S9C).
In contrast, in acidic medium, release of CO2 from (bi)carbonate at the anode compartment was not observed ( fig. S9C), indicating no crossover of (bi)carbonate anions through the acidic medium. This result supports the notion that in acidic medium a high carbon efficiency of CO2 reduction can be achieved by avoiding the carbonate problem. We next probed whether other alkali cations such as Li + , Na + and Cs + have the similar effect on the competition between HER and CO2 reduction as K + . Indeed, in 0.1 M HOTf + 0.4 M MOTf (M = Li, Na, Cs), plateaus of current densities corresponding to the diffusion-limited reduction of hydronium ions were observed (fig. S10D-F), indicating that these alkali cations also suppressed the migration of hydronium ions. We then measured CO2 electroreduction on SnO2/C and Cu/C in acidic solutions containing these alkali cations (fig. S11-S13). All alkali cations promoted CO2 reduction by inhibiting HER, but the effects are variable. On SnO2/C the Faradaic efficiency of CO2 reduction increased in the order Li < Na < K < Cs for both catalysts ( fig. 3B), and the partial current densities of formic acid and CO increased in the same order ( fig. 3C). On Cu/C the partial current density of ethylene increased in a similar order of Li < Na < K ≈ Cs (fig. S13). In the study of Huang et al on CO2 reduction in strongly acidic medium with Cu catalysts, 15 they proposed that a high cathodic current density would lead to hydronium depletion and pH increase near the cathode. Water instead of hydronium reduction then contributes to the majority of HER currents. Alkali cations were then proposed to favor CO2 reduction over water reduction, although the details of this promotion were not studied. This mechanistic hypothesis does not take into account the migration of hydronium ions under an electric field during HER. It cannot explain why in alkali cation-free solutions the HER current density can exceed the diffusion limit of hydronium ions and why the addition of alkali cations can suppress the HER current density to a limiting value, as observed in our study.
To probe how alkali cations suppressed the reduction of hydronium ions, we conducted a simulation based on the Poisson-Nernst-Planck model (PNP) which includes migration as one of the mechanisms for mass tranport. 24 Hydronium ions, K + and OTfwere considered, and the reduction of hydronium ions was regarded as the only source of HER in strongly acidic media.
The HER current density was assumed proportional to the concentration of hydronium ions in the OHP and exponential to the electrode potential. The simulation reproduced the features of HER at potentials more positive than the onset of reduction of water (fig. S14): in a K + -free solution, the current density of hydronium reduction increased without any limitation as potential went cathodically; in a K + -containing solution, a plateau of current density higher than the limiting diffusion current density based on the Levich equation was observed. For CO2 reduction in aqueous media, the adsorbed CO2 (CO2ad) is regarded as a key intermediate. 7,25 Stabilization of CO2ad on the surface of catalysts promotes the production of CO and formate. 22,26,27 The two C=O bonds of CO2ad bend away from the surface, endowing CO2ad with a large dipole moment oriented outwards. 16 16,29 and is stabilized by the electric field in Stern layer. Therefore, K + ions not only suppressed HER by impeding the migration of hydronium ions in the diffuse layer, but also promoted CO2 electroreduction due to the interaction between electric field and dipole moment of adsorbed intermediates.
The effect of different alkali cations can also be explained by the model in fig.4C. The size of hydrated alkali cations decreased from Li to Cs. As the size of hydrated cations decreases, more alkali cations can accumulate in the OHP, 16 leading to a stronger electric field in Stern layer and a weaker electric field in diffuse layer. 30 This explains how the size of hydrated alkali cations affects the competition between HER and CO2 reduction in strongly acidic media. It is noteworthy that the partial current density of CO2 reduction could significantly exceed the diffusion limitation of hydronium ions in the presence of K + (fig. S16), implying that water molecules are the proton source for CO2 reduction. Thus, CO2 reduction leads to the formation of OHions, which further react with hydronium ions near the electrode and suppress HER. 14,31 Due to this effect, Faradaic efficiency of CO2 reduction reached 90% for SnO2/C and In summary, by using alkali cations to suppress hydronium reduction and promote CO2 reduction, we demonstrated efficient CO2 electroreduction in strongly acidic medium. We showed that this approach is universal for various catalysts and cations, and we revealed cationinduced modulation of electric field as the origin of the cation effects. This work provides a promising strategy to avoid the carbonate problem in CO2 electroreduction, which is one of the main road blockers for low-temperature CO2 electrolysis.     The partial current density of product p was calculated according to the equation: In this equation, j is the current density normalized to the area of working electrode. Two-electrode flow cell: Fig. S6 shows the scheme of the two-electrode flow cell.

Production of aqueous solution
CeTech W1S1009 carbon cloths were used as both cathode and anode. Au/C and IrO2 were used as catalysts for cathode and anode, respectively. Titanium plate was used as current collector for both electrodes. An EPDM plate with the thickness of 1.5 mm was used as the chamber of electrolyte solution. The effective window for electrolysis was a circle with the diameter of 1.13 cm (area = 1 cm 2 ). Kapton tapes with circular windows with the same size was pasted on both electrodes to control the effective area exposed to the electrolyte. No membrane was used between cathode and anode and two electrodes shared the same electrolyte. CO2 and He streams were supplied behind cathode and anode, respectively. The volume of electrolyte solution was 10 mL, which  S17 and movie S1). The reference electrode was put in a Luggin capillary and the distance between the tip of the capillary and RDE was 5 mm. A platinum wire was used as the counter electrode. The electrolyte was saturated with N2.
Linear sweeping voltammetry curves were collected with a scan rate of 5 mV/s. The resistance of electrolyte was determined by high-frequency impedance measurements and IR compensation was done after experiment.
The limiting diffusion current density of the reduction of hydronium ions was and ω is the rotating speed of the RDE (unit: rad/s).

S9
Simulation procedure: The simulation was to solve the governing equations in a 1-dimensional domain from the surface of cathode to the bulk electrolyte during HER in strongly acidic media.
The domain for simulation was divided into two regions. The first region was between the surface of cathode and outer Helmholtz plane (OHP), which is called Stern layer.
The second region was between OHP and bulk electrolyte, which contained diffuse layer and diffusion layer. Boundary conditions at the surface of cathode, OHP and the bulk-electrolyte side were used to solve the equations (scheme S1).
Scheme S1. Governing equations and boundary conditions used for the 1dimensional simulation. From left to right: the cathode, the Stern layer, the diffuse/diffusion layer, and the bulk electrolyte region. OHP is used as origin (x=0).
The transport of three solvated ionic species (K + , H + and OTf -) and the corresponding charge transfer were considered in the simulation. The Poisson-Nernst-Planck equations at steady state 7 are solved in the region between OHP and bulk electrolyte. These equations include the diffusion, migration and convection terms of each species: where is the concentration of species (with i = K + , H + and OTf -), is the diffusion coefficient of species , is the charge of species , is the ideal gas constant, is the temperature, is the Faradaic constant, is the potential and is the velocity of solution in x-direction. For a rotating disk electrode, the velocity in axial direction at different x-locations can be estimated as: 8 where is the rotation speed (unit: rad/s) of the disk electrode, is the kinematic viscosity of water.
The Poisson equation is used to calculate the potential change, given by: where 0 is the permittivity of vacuum and is the relative permittivity of water.
The thickness of Stern layer (dStern) was assumed to be 0.4 nm. 9 The thickness of the region between OHP and bulk solution in simulation was assumed to be 100 μm, ,H + = / (S11) where j is the HER current density and we assumed hydronium ions were the only proton source for HER in strongly acidic media. We assumed the HER current density showed proportional relations with hydronium concentration at OHP and exponential relations with the potential of cathode: where α is the charge transfer coefficient and φcathode is the potential of cathode. ) show the relationship as: The experimental HER current density on a flat Au electrode at -0.4 V vs SHE in an acidic medium with pH = 1 was -1 mA/cm 2 . 10 The PZC of Au in an acidic medium with pH = 1 and containing weakly adsorbed anions (such as SO4 2and ClO4 -) was 0.2 V vs SHE. 11 We assumed the PZC of Au in HOTf-KOTf solution with pH = 1 was also 0.2 V vs SHE. Therefore, we assigned cathode 0 = -0.6 V vs PZC, H + 0 = 0.1 M and j0 = -1 mA/cm 2 in equation S13. We assumed α = 0.5. Then, the HER current density at a certain condition can be estimated as: To confirm that the accuracy of the above assignment did not affect the trends of shifted laterally and the shape of the curves did not change.

S12
In Stern layer, the Poisson equation is given by: The left boundary condition for equation S15 in Stern layer was: Neumann boundary condition was used for equation S8 at OHP (x = 0), namely: Table S1 summarizes the values of the various parameters in the model.     In near neutral medium, water molecule is the proton source for cathodic reaction. As one CO2 molecule is reduced to CO, two OHions are generated, which further react with another CO2 molecule to form one CO3 2ion. At anode, OER generates hydronium ions which protonate CO3 2ions to regenerate CO2. In acidic medium where the pH is significantly lower than pKa1 of H2CO3 (3.6), CO3 2and HCO3ions are not generated from cathode. CO2 molecules are exclusively converted to reduction products. (C) GC curves of gas generated at the anode compartment. Au/C and IrO2 were used as catalysts for cathode and anode, respectively. Helium was supplied behind anode. Blue curve: near neutral medium (0.8 M KHCO3) was used and CO2 was supplied behind the cathode. Red curve: acidic medium (0.1 M H2SO4 + 0.4 M K2SO4) was used and CO2 was supplied behind the cathode. Green curve: acidic medium was used and N2 was supplied behind the cathode. GC samples were taken after electrolysis at 200 mA/cm 2 for 3 h to ensure the system reached a steady state. The small peaks of CO2 from acidic medium were due to the oxidation of carbon cloth.  fig. 3A, the plateau current density was about 6% higher than the limiting diffusion current density. This difference was ascribed to that the steric effect of cations was not considered in PNP model. If the steric effect was considered, the repulsion from K + ions to H + ions near OHP should be stronger, leading to lower concentration of H + near cathode and lower HER current density, closer to the HER current density in our experiment observation. The black dashed line shows the current density when N2 was supplied. The solid lines show the partial current density of H2 (black), CO (red) and formic acid (blue) when CO2 was supplied. In N2 atmosphere, a plateau of current density about 65 mA/cm 2 was observed, corresponding to the diffusion limitation of hydronium ions under this condition. The partial current density of formic acid could be significantly higher than 65 mA/cm 2 , indicating water molecules served as the proton source for CO2 reduction and OHions were generated. The HER current density in CO2 atmosphere was remarkably lower than in N2 atmosphere. value in equation S13 was set to -0.5 V (black curves), -0.6 V (red curves) and -0.7 V (blue curves), respectively.