A broad parameter range for selective methane production with bicarbonate solution in electrochemical CO₂ reduction

Abstract

This study demonstrated the previously unrecognised capability of potassium bicarbonate (KHCO₃) aqueous solutions to assist in the selective generation of methane (CH₄) over a wide range of reaction parameters during electrochemical CO₂ reduction over a bulk polycrystalline Cu catalyst. Compared with the results obtained with potassium chloride (KCl) electrolytes, a 0.5 M KHCO₃ solution maintained a higher faradaic efficiency (FE) for CH₄ production at higher current densities. This trend was further emphasised upon increasing the CO₂ pressure from near-ambient to 3 atm and incorporating stirring. The result was a FE above 50% over a range of current densities from 90 to more than 330 mA cm⁻². The maximum FE reached 80% at 0 °C, while maintaining a broad peak structure when plotted against current density, even over 300 mA cm⁻². The buffering ability of KHCO₃ appears to play an important role in increasing both the reaction rate and the selectivity for CH₄, especially when combined with an optimised CO₂ supply to the electrode and an ideal reaction temperature.

---

Introduction

Recently, the efficiency of direct CH₄ production, as indicated in the context of power-to-gas,¹ has become important. This process offers a practical means of using surplus energy from renewable sources and can take advantage of conventional facilities already used for natural gas storage and distribution. Electrochemical CO₂ reduction is one of the most promising candidates for the direct production of renewable fuels, such as CH₄ from waste CO₂ and water. This reduction proceeds under moderate reaction conditions^2,3 and has been attracting the interest of researchers for decades.⁴

Among the various catalysts for electrochemical CO₂ reduction, only copper (Cu) catalyses the reaction of CO₂ to produce CH₄ as the major product.^5–8 However, as several reports and reviews have pointed out, CO₂ reduction to CH₄ over Cu is not readily controllable and numerous by-products interfere with the selective production of CH₄ at high reaction rates under low overpotentials.^4,8–12 Many studies have found that catalytic CO₂ reduction can be significantly affected not only by modifying the Cu catalyst,^13–19 but also by varying the reaction parameters, such as the applied potential (or current density), electrolyte, temperature and CO₂ pressure, among other parameters.^6,7,20,21 Therefore, many factors must be considered when optimising the catalytic process, even when using conventional CO₂ reduction catalysts such as polycrystalline Cu. The reaction parameters must be carefully controlled and synergistic effects between the factors must be assessed. This is especially vital when designing practical applications involving the efficient conversion of CO₂ to CH₄. Despite this, only a few studies have focused on this aspect of the process.^22,23

Recently, we developed a combinatorial screening system for the analysis of CO₂ reduction reactions, which offers a means of making rapid systematic measurements while controlling multiple reaction parameters.²⁴ Using the power of this system to rapidly ascertain the effects of specific parameters, the roles of CO₂ pressure, stirring speed and temperature have been characterised with regard to the impact on the reaction rate and CH₄ selectivity in potassium chloride (KCl) solutions.

Herein, extending the scope of our research, we report on the use of a potassium bicarbonate (KHCO₃) solution to enhance both the reaction rate and the selectivity during CH₄ production over a bulk polycrystalline Cu electrode, incorporating the other parameters optimised in our previous work.²⁴ Traditionally, KHCO₃ solutions have been used as the de facto standard electrolyte in this research field.^4,8,9 Despite the general consensus that product selectivity could be drastically changed by applying different electrolytes or varying the electrolyte concentration,⁴ the specific role of the electrolyte has often been overlooked in the studies of CO₂ reduction. Recently, Varela et al. discussed the effect of the electrolyte with regard to the buffering ability of KHCO₃ on the local pH at the catalyst surface, and discussed the catalytic activity of planar Cu during the synthesis of hydrocarbons including CH₄ and ethylene (C₂H₄).²⁵ In this study, we demonstrate that the buffering effect of KHCO₃ can be further emphasised when employed in conjunction with specific CO₂ pressures, stirring speeds and reaction temperatures. We also compare these results with those obtained in 0.5 M KCl. Highly selective CH₄ production, considerably more selective than that obtained when using 0.5 M KCl, is demonstrated at elevated current densities in 0.5 M KHCO₃. This result shows the benefits of using a KHCO₃ solution in CH₄ production, especially when coupled with other optimised reaction parameters that maximise the catalytic activity of a conventional Cu electrode.

Results and discussion

The details of the experimental conditions and procedures are summarised in the ESI.† For all experimental conditions (using various electrolytes, CO₂ pressures, stirring speeds and temperatures), we ascertained the variation in faradaic efficiency (FE) for each reaction product with current density, the results of which are summarised in Fig. S2, S3 and Table S1 (ESI).† Meanwhile, the results confirming the reproducibility and time-dependence of the product selectivity are shown in Fig. S4 and S5(a).†

Fig. 1 shows the CH₄ FE values for reactions carried out over strip-shaped polycrystalline Cu plates as functions of current density when using 0.5 M KCl, 0.25 M KHCO₃ or 0.5 M KHCO₃. All these data were acquired at a CO₂ pressure of 1.3 atm in unstirred reactors. The variation of current density shows a clear maximum CH₄ FE for each electrolyte (corresponding to a more negative applied potential), which is consistent with previous observations over Cu.^6,8,21,24 In the case of 0.5 M KCl, the peak efficiency was over 40% at 7.5 mA cm⁻², but FE drastically decreased with further increases in current density. Above 30 mA cm⁻², CH₄ production ceased almost completely and was replaced by H₂ generation (see Table S1 in the ESI†). A different trend was observed when KHCO₃ was used as the cathode electrolyte. In 0.25 M KHCO₃, the peak FE was lower than that observed for 0.5 M KCl. However, the decrease in the amount of CH₄ produced at higher current densities was not as pronounced, with an efficiency of approximately 10% at 60 mA cm⁻². The tail-off in CH₄ selectivity with higher current densities was more gradual when 0.5 M KHCO₃ was used; a CH₄ selectivity of about 4% was observed even at 160 mA cm⁻². The figure also shows that the CH₄ selectivity plots for the two electrolytes have significantly different structures. The peak CH₄ FE was lower but the tail-off was more gradual as the KHCO₃ concentration was increased.

Fig. 1 CH₄ faradaic efficiencies as functions of current density in different electrolytes at 1.3 atm, 0 rpm and 25 °C. Pink squares: 0.5 M KCl, light blue circles: 0.25 M KHCO₃ and light green triangles: 0.5 M KHCO₃.

The different electrolytes also generated dissimilar responses upon increasing the CO₂ pressure. Fig. 2(a)–(c) show the effects of CO₂ pressure on the CH₄ FE values when using 0.5 M KCl or 0.5 M KHCO₃. As seen in Fig. 2(a) and (b), in 0.5 M KCl, the peak CH₄ FE shifted to a higher current density upon increasing the pressure from 1.3 to 3 atm, whereas the peak efficiency increased and the peak became broader at higher current densities and with increasing pressure in 0.5 M KHCO₃. Fig. 2(c) summarises the pressure response of the peak CH₄ FE under each experimental condition. The peak efficiency was evidently not sensitive to pressure in 0.5 M KCl, which is consistent with our previous observations.²⁴ In contrast, when using KHCO₃, the peak efficiency clearly increased with pressure, indicating that CO₂ pressure can be used to tune the CH₄ selectivity, especially with 0.5 M KHCO₃ as the electrolyte.

Fig. 2 CH₄ faradaic efficiencies as functions of current density in (a) 0.5 M KCl (pink squares: 1.3 atm, red triangles: 3 atm) and (b) 0.5 M KHCO₃ (light green squares: 1.3 atm, green triangles: 3 atm). (c) Peak CH₄ faradaic efficiencies as functions of CO₂ pressure in 0.5 M KCl (red squares) and 0.5 M KHCO₃ (green triangles). All experiments were performed at 0 rpm and 25 °C. The dotted lines are intended solely as visual aids.

The differences between KHCO₃ and KCl may be further emphasised by incorporating a stirring effect. Several papers have reported that stirring can increase the supply of CO₂ to the catalyst surface by decreasing the boundary layer thickness between the catalyst surface and the bulk electrolyte.^21,24Fig. 3(a) plots the CH₄ FE values as functions of current density in 0.5 M KCl and KHCO₃, at 3 atm and 0 rpm. Here, the peak CH₄ FE is higher in 0.5 M KCl than that in 0.5 M KHCO₃, but CH₄ is produced across a wider current density range in 0.5 M KHCO₃. Fig. 3(b) shows the dependence of the CH₄ FE on current density for the two electrolytes under stirred conditions. In 0.5 M KCl, there is a sharp CH₄ FE maximum at approximately 120 mA cm⁻², although the efficiency rapidly drops below 40% at densities above 150 mA cm⁻². In contrast, the CH₄ FE in 0.5 M KHCO₃ continues to increase even when density is increased above 150 mA cm⁻², and maintains a value in the range of 50 to 60% even at 300 mA cm⁻². Fig. 3(c) plots the CH₄ production rates against current density, and clearly demonstrates the enhanced CH₄ production in 0.5 M KHCO₃. In each electrolyte, the CH₄ production rate increased monotonically throughout the low current density region (below 100 mA cm⁻²). However, in 0.5 M KCl, the CH₄ production rate abruptly plateaued at 80 nmol cm⁻² s⁻¹, whereas the CH₄ production in 0.5 M KHCO₃ continued to increase, reaching over 240 nmol cm⁻² s⁻¹. The incorporation of stirring can thus further increase the CH₄ selectivity over a broad range of current densities, resulting in a CH₄ production rate in KHCO₃ more than three times that in KCl.

Fig. 3 (a) CH₄ faradaic efficiencies at 3 atm, 0 rpm and 25 °C. (b) CH₄ faradaic efficiencies and (c) CH₄ production rates as functions of current density in different electrolytes at 3 atm, 500 rpm and 25 °C. Red squares: 0.5 M KCl and green triangles: 0.5 M KHCO₃.

In addition to the CO₂ pressure and stirring speed, temperature can affect the product selectivity, and it is known that the CH₄ FE is improved at lower temperatures.^20,24Fig. 4 shows the current density dependence of the CH₄ FE at 3 atm, 500 rpm and 0 °C. In these trials, we used a 0.5 M KHCO₃ + 1 M KCl solution to avoid decreasing the electrolyte conductivity upon lowering the temperature; otherwise this might have affected the current density range. The effect of adding KCl is evident from Fig. S6 (ESI),† in which similar CH₄ FEs are observed both with and without KCl. Remarkably, upon decreasing the temperature to 0 °C, the CH₄ FE was elevated, especially for a current density of less than 240 mA cm⁻², where the maximum FE reached 80.0% at 80 mA cm⁻². To the best of our knowledge, this is one of the highest selectivity values yet reported for CH₄ production in an aqueous electrolyte using a bulk, polycrystalline Cu electrode.^20,26–28 Moreover, CH₄ FEs over 65% were achieved from 80 to 240 mA cm⁻², demonstrating selective production over a broad current density range as a result of tuning various reaction parameters in the 0.5 M KHCO₃ solution.

Fig. 4 CH₄ faradaic efficiencies as functions of current density at different temperatures and at 3 atm and 500 rpm in 0.5 M KHCO₃ at 25 °C and 0.5 M KHCO₃ + 1 M KCl (to prevent reduced electrolyte resistance, see Fig. S2 in the ESI for more details†) at 0 °C. Green triangles: 25 °C and orange circles: 0 °C. The red dotted line is intended solely as a visual aid.

The CH₄ selectivity of 80.0% was observed for an applied cathodic potential of approximately −2.02 V vs. Ag/AgCl (see Table S1 in the ESI†) and −1.38 V vs. the RHE without IR compensation. We calculated energy efficiency with regard to the cathode electrochemical reaction to produce CH₄ at the cathode.²⁹ This yielded an energy efficiency of 36.4%, which is also among leading values reported for CH₄ production at a high current density value of 80 mA cm⁻². We also note that this value includes IR resistance. Thus, further optimization of the system could yield even more efficient production of CH₄, and, indeed, of the entire reaction (including oxygen evolution).

At this point, it is helpful to consider why CH₄ production over a broad current density range in KHCO₃ was higher than that observed in KCl. We therefore calculated the limiting current density (LCD) for CO₂ reduction under the three experimental conditions shown in Fig. 1, and these are plotted in Fig. 5. The LCD data for all experimental conditions tested in this study are summarised in Table S2.† The LCD values were determined based on the maximum of partial current density for CO₂ reduction under each condition. Generally, LCD relates to the mass transport properties of the reactant, and in this case LCD is believed to be closely related to the mass transport of CO₂ in the system. Interestingly, Fig. 5 clearly shows that the LCD values increased when increasing the KHCO₃ concentration to 0.5 M but keeping all other conditions the same in all three electrolytes. This result suggests that the extent of CO₂ reduction can be varied based on the choice of electrolyte and its concentration.

Fig. 5 Limiting current densities for CO₂ reduction in different electrolytes at 1.3 atm, 0 rpm and 25 °C.

One possible explanation for this phenomenon is obtained by considering the concentrations of dissolved CO₂ in various electrolytes. In electrochemical CO₂ reduction, the reactant is generally considered to be CO₂ dissolved in the electrolyte.³⁰ In addition, the level of dissolved CO₂ is known to vary slightly based on the choice of electrolyte and electrolyte concentration.^23,31 According to Zhong et al., the saturated CO₂ concentration in 0.5 M KHCO₃ at ambient pressure is 36.3 mmol l⁻¹, which is approximately 1.16 times that in 0.5 M KCl (31.3 mmol l⁻¹).³¹ This partially explains the enhanced LCD in KHCO₃ solutions, because the amount of CO₂ available to participate in the reaction would have been higher. However, the LCD in 0.5 M KHCO₃ was more than twice the value in 0.5 M KCl at 1.3 atm. Thus there must be another mechanism responsible for the enhancement of the LCD values in KHCO₃ solutions.

The most important difference between these electrolytes is the concentration of HCO₃⁻. In general, a CO₂-saturated KHCO₃ solution has a much stronger buffering ability than a CO₂-saturated KCl solution, which means it is better able to compensate for consumed reactants like H⁺ and CO₂. Hori et al. pointed out that the pH varies between the bulk electrolyte (bulk pH) and the pH at the catalyst surface (local pH).⁶ During electrochemical CO₂ reduction, the local pH tends to be higher than the bulk pH owing to the consumption of H⁺ by the reaction, which also limits the concentration of CO₂ as well as H⁺ (for a pH over 9, CO₂ concentration becomes much lower by equilibrium).^23,31 To mitigate this, Gupta et al. suggested that the local pH can be decreased by increasing the KHCO₃ concentration (i.e., by increasing the buffer capacity) to compensate for H⁺ consumed by the reaction.³⁰ Indeed, recent simulation results showed that LCD can be increased by increasing the electrolyte concentration of HCO₃⁻.²³ As shown in Table S3 and discussed further in the ESI,† this quantitatively and qualitatively supports our experimental data in suggesting that higher LCD values would be anticipated in conjunction with increased HCO₃⁻ concentrations, which in turn enhance the buffering ability of the solution.

As well as considering the H⁺ equilibrium, we also considered the direct compensation of CO₂ consumed during the reaction. However, the interconversion between HCO₃⁻ and CO₂ was estimated to be small given the lack of available catalyst.³² Nonetheless, we note that recent work used isotope labelling to experimentally propose the possibility of conversion from HCO₃⁻ to CO₂.^33,34 Thus, although several factors remain unclarified, we suggest that the enhancement of LCD in KHCO₃ solution is likely to be connected to the interplay between the H⁺ and CO₂ equilibria that is derived from KHCO₃'s buffering ability.

The present system favours CH₄—rather than H₂—production with increasing CO₂ pressure and stirring speed in 0.5 M KHCO₃. However, as shown in Fig. S7 (ESI†), based on the current density dependence data obtained in 0.5 M KHCO₃, the minimum FE for H₂ is significantly affected by the pressure and stirring speed, accompanied by the increase of the CH₄ FE. Interestingly, this tendency was not observed in the case of 0.5 M KCl (see Fig. S8†). This result indicates that the buffering ability of the solution plays a role in suppressing H₂ production upon the increase of CO₂ pressure and stirring speed, possibly because the regenerated H⁺ can be more effectively coupled with CO₂ on the surface of the catalyst. This synergetic effect between the electrolyte and the supply of CO₂ may have contributed to the broad current density range over which CH₄ production was observed, as well as the increased LCD values in the 0.5 M KHCO₃ solution.

We are also keen to note the effect of reaction parameters on product selectivity during the reduction of CO₂. In Fig. S3,† the FE of CO increases with pressure in both 0.5 M KCl and KHCO₃ at 0 rpm and 25 °C, especially for low current densities (below 30 mA cm⁻²). Moreover, the selectivity values for the C₂ and C₃ products (C₂H₄, alcohols and aldehydes) also increase with pressure, as demonstrated by other experimental studies.^35,36 This is also in-line with recent theoretical studies that showed that increasing surface coverage of CO intermediates—which is likely promoted by increasing CO₂ pressure—accelerates CO desorption³⁷ and C–C coupling.³⁸ With regard to the effect of temperature on product selectivity, this work and several other studies have found that CH₄ selectivity increases with decreasing temperature, accompanying the decrease in the FE for other CO₂-reduction products.^7,20,24 However, the mechanism by which low temperatures preferentially promote CH₄ production remains unclear. Generally, decreasing the temperature suppresses the thermal excitation of electrons in the catalyst, increasing the selectivity for the most energetically favourable product. Recent computational studies suggest that a CO intermediate, which is considered a step in the reaction pathway toward forming hydrocarbons and alcohols, is more energetically favourable than HCOO⁻.^10,39 In addition, the adsorbed CO on Cu is known to be more energetically favourable than gaseous CO.¹² Furthermore, starting from CO, CH₄ is predicted to be the most energetically favourable product (rather than methanol, C₂ or C₃ products) at a near-neutral pH.⁴⁰ These findings could explain why CH₄ production increases at lower temperatures in experimental studies.

Finally, we note that the Cu morphology is a key to the product selectivity. Recent work has shown that introducing nanostructures or adding nanoparticles to the Cu catalyst can alter the selectivity between C₂H₄ and CH₄. This suggests that a smoother surface may favour the production of CH₄ over C₂H₄.^14,15,19 We therefore also compared the surface roughness of the Cu electrodes using SEM analysis, as shown in Fig. S9(a) and (b).† The degree of surface roughness was less than for other reported structures, perhaps contributing to the increase in selectivity for CH₄ observed here. Moreover, an electron backscattering pattern (EBSP) study showed that the majority of the surface was considered more favourable for CH₄ production, than for that of other products, such as C₂H₄.²⁷ Together with the effects of parameters, these catalyst aspects may have also contributed to the observed selective CH₄ production over a broad parameter range.

Conclusions

We have experimentally demonstrated selective CH₄ production over a broad current density range in 0.5 M KHCO₃. In aqueous solutions of KHCO₃, the peak CH₄ FE in a plot against current density exhibited a more gradual tail-off toward higher current density values than in 0.5 M KCl. Moreover, we have shown that CH₄ production over bulk polycrystalline Cu can be further increased by raising the CO₂ pressure and stirring speed in 0.5 M KHCO₃, achieving CH₄ selectivity values greater than 50% over a broad current density range with a maximum exceeding 300 mA cm⁻². Tuning the reaction temperature, the maximum FE increased to 80%, while maintaining a peak efficiency across a broad current density range. The buffering ability of HCO₃⁻, together with other optimised reaction parameters, creates an environment favourable for selective CH₄ production over this broader current density range. This study has demonstrated the advantages of a concentrated KHCO₃ solution in terms of obtaining selective CH₄ production over a current density range of several hundred mA cm⁻² for stirred reaction mixtures under pressures of only several atm. This approach could be especially beneficial in terms of its practical aspects, because it allows a wide parameter window in conjunction with the use of concentrated KHCO₃ solutions and relatively low pressures. Our study also provides guidelines for optimising the reaction parameters to maximise the catalytic properties of conventional Cu electrodes for rapid and selective CH₄ production. These findings suggest that, when developing catalysts for CO₂ reduction, one should keep in mind that there is still much room to improve the activity of the catalyst by carefully considering the effects of reaction parameters. This could also be important during the evaluation of novel catalysts for practical applications in efficient CO₂-reduction systems.

Conflicts of interest

There are no conflicts to declare.

References

1. M. Jentsch, T. Trost and M. Sterner, Energy Procedia, 2014, 46, 254–261.
2. D. T. Whipple and P. J. A. Kenis, J. Phys. Chem. Lett., 2010, 1, 3451–3458.
3. E. V. Kondratenko, G. Mul, J. Baltrusaitis, G. O. Larrazábal and J. Pérez-Ramírez, Energy Environ. Sci., 2013, 6, 3112.
4. Y. Hori, Electrochemical CO₂ Reduction of Metal Electrodes, in Modern Aspects of Electrochemistry, ed. C. G. Vayenas, R. E. White and M. E. Gamboa-Aldeco, Springer, New York, 2008, pp. 89–189.
5. Y. Hori, K. Kikuchi and S. Suzuki, Chem. Lett., 1985, 1695–1698.
6. Y. Hori, A. Murata and R. Takahashi, J. Chem. Soc., Faraday Trans. 1, 1989, 85, 2309–2326.
7. M. Azuma, K. Hashimoto, M. Hiramoto, M. Watanabe and T. Sakata, J. Electrochem. Soc., 1990, 137, 1772–1778.
8. K. P. Kuhl, E. R. Cave, D. N. Abram and T. F. Jaramillo, Energy Environ. Sci., 2012, 5, 7050–7059.
9. M. Gattrell, N. Gupta and A. Co, J. Electroanal. Chem., 2006, 594, 1–19.
10. A. A. Peterson, F. Abild-Pedersen, F. Studt, J. Rossmeisl and J. K. Norskov, Energy Environ. Sci., 2010, 3, 1311–1315.
11. T. Cheng, H. Xiao and W. A. Goddard, J. Phys. Chem. Lett., 2015, 6, 4767–4773.
12. A. A. Peterson and J. K. Nørskov, J. Phys. Chem. Lett., 2012, 3, 251–258.
13. C. W. Li and M. W. Kanan, J. Am. Chem. Soc., 2012, 134, 7231–7234.
14. W. Tang, A. A. Peterson, A. S. Varela, Z. P. Jovanov, L. Bech, W. J. Durand, S. Dahl, J. K. Nørskov and I. Chorkendorff, Phys. Chem. Chem. Phys., 2012, 14, 76–81.
15. R. Kas, R. Kortlever, A. Milbrat, M. T. Koper, G. Mul and J. Baltrusaitis, Phys. Chem. Chem. Phys., 2014, 16, 12194–12201.
16. C. W. Li, J. Ciston and M. W. Kanan, Nature, 2014, 508, 504–507.
17. R. Kas, K. K. Hummadi, R. Kortlever, P. de Wit, A. Milbrat, M. W. Luiten-Olieman, N. E. Benes, M. T. Koper and G. Mul, Nat. Commun., 2016, 7, 10748.
18. Y. Kwon, Y. Lum, E. L. Clark, J. W. Ager and A. T. Bell, ChemElectroChem, 2016, 3, 1013–1019.
19. H. Mistry, F. Behafarid, R. Reske, A. S. Varela, P. Strasser and B. R. Cuenya, ACS Catal., 2016, 6, 1075–1080.
20. Y. Hori, K. Kikuchi, A. Murata and S. Suzuki, Chem. Lett., 1986, 897–898.
21. K. Hara, A. Tsuneto, A. Kudo and T. Sakata, J. Electrochem. Soc., 1994, 141, 2097–2103.
22. M. R. Thorson, K. I. Siil and P. J. A. Kenis, J. Electrochem. Soc., 2012, 160, F69–F74.
23. M. R. Singh, E. L. Clark and A. T. Bell, Phys. Chem. Chem. Phys., 2015, 17, 18924–18936.
24. H. Hashiba, S. Yotsuhashi, M. Deguchi and Y. Yamada, ACS Comb. Sci., 2016, 18, 203–208.
25. A. S. Varela, M. Kroschel, T. Reier and P. Strasser, Catal. Today, 2016, 260, 8–13.
26. R. L. Cook, R. C. MacDuff and A. F. Sammells, J. Electrochem. Soc., 1987, 134, 2375–2376.
27. Y. Hori, I. Takahashi, O. Koga and N. Hoshi, J. Mol. Catal. A: Chem., 2003, 199, 39–47.
28. K. Manthiram, B. J. Beberwyck and A. P. Alivisatos, J. Am. Chem. Soc., 2014, 136, 13319–13325.
29. A. J. Martín, G. O. Larrazábal and J. Pérez-Ramírez, Green Chem., 2015, 17, 5114–5130.
30. N. Gupta, M. Gattrell and B. MacDougall, J. Appl. Electrochem., 2006, 36, 161–172.
31. H. Zhong, K. Fujii, Y. Nakano and F. Jin, J. Phys. Chem. C, 2015, 119, 55–61.
32. Y. Chen, N. S. Lewis and C. Xiang, Energy Environ. Sci., 2015, 8, 3663–3674.
33. T. Sekimoto, H. Hashiba, S. Shinagawa, Y. Uetake, M. Deguchi, S. Yotsuhashi and K. Ohkawa, J. Phys. Chem. C, 2016, 120, 13970–13975.
34. M. Dunwell, Q. Lu, J. M. Heyes, J. Rosen, J. G. Chen, Y. Yan, F. Jiao and B. Xu, J. Am. Chem. Soc., 2017, 139, 3774–3783.
35. C. F. C. Lim, D. A. Harrington and A. T. Marshall, Electrochim. Acta, 2016, 222, 133–140.
36. Y. Huang, A. D. Handoko, P. Hirunsit and B. S. Yeo, ACS Catal., 2017, 7, 1749–1756.
37. C. Shi, H. A. Hansen, A. C. Lausche and J. K. Nørskov, Phys. Chem. Chem. Phys., 2014, 16, 4720–4727.
38. R. B. Sandberg, J. H. Montoya, K. Chan and J. K. Nørskov, Surf. Sci., 2016, 654, 56–62.
39. T. Cheng, H. Xiao and W. A. Goddard, J. Am. Chem. Soc., 2016, 138, 13802–13805.
40. H. Xiao, T. Cheng, W. A. Goddard and R. Sundararaman, J. Am. Chem. Soc., 2016, 138, 483–486.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7se00352h

---