Deciphering the selectivity descriptors of heterogeneous metal phthalocyanine electrocatalysts for hydrogen peroxide production

Abstract

The electrocatalytic 2e⁻ oxygen reduction reaction (2e⁻ ORR) provides an appealing pathway to produce hydrogen peroxide (H₂O₂) in a decentralized and clean manner, which drives the demand for developing high selectivity electrocatalysts. However, current understanding on selectivity descriptors of 2e⁻ ORR electrocatalysts is still insufficient, limiting the optimization of catalyst design. Here we study the catalytic performances of a series of metal phthalocyanines (MPcs, M = Co, Ni, Zn, Cu, Mn) for 2e⁻ ORR by combining density functional theory calculations with electrochemical measurements. Two descriptors (ΔG_*O − ΔG_*OOH and ΔG_*H2O2) are uncovered for manipulating the selectivity of H₂O₂ production. ΔG_*O − ΔG_*OOH reflects the preference of O–O bond breaking of *OOH, affecting the intrinsic selectivities. Due to the high value of ΔG_*O − ΔG_*OOH, the molecularly dispersed electrocatalyst (MDE) of ZnPc on carbon nanotubes exhibits high selectivity, even superior to the previously reported NiPc MDE. ΔG_*H2O2 determines the possibility of further H₂O₂ reduction to affect the measured selectivities. Enhancing the hydrophobicity of the catalytic layer can increase ΔG_*H2O2, leading to selectivity improvement, especially under high H₂O₂ production rates. In the gas diffusion electrode measurements, both ZnPc and CoPc MDEs with polytetrafluoroethylene (PTFE) exhibit low overpotentials, high selectivities, and good stability. This study provides guidelines for rational design of 2e⁻ ORR electrocatalysts.

---

Introduction

Hydrogen peroxide (H₂O₂) has widespread applications in industry, health care, and environmental protection.¹ The electrochemical two-electron oxygen reduction reaction (2e⁻ ORR) is a promising approach to produce this important chemical in a green way.² Its application relies on the development of electrocatalysts with high peroxide selectivity to lower electric energy consumption. Various electrocatalysts for this process have been developed recently, including noble metals, carbon-based non-metal materials, transition metal single-atom catalysts (SACs), metal complexes, and metal oxides.³ Noble metal catalysts with isolated active sites have achieved excellent 2e⁻ ORR performances. For example, the PtHg₄ electrocatalyst exhibited selectivities up to 96% in acid.⁴ Nevertheless, low-cost alternatives but also possessing high selectivities remain to be developed for satisfying requirements of practical applications. Recently, transition metal SACs doped with heteroatoms (i.e., N, O, or P) on carbon matrix (M–X–C) have been found promising for 2e⁻ ORR.^5–12 Co–N–C structures exhibited better 2e⁻ ORR performances than other metal active sites.^5,7 In contrast, Hyeon et al. found that the Co–N₄ catalyst exhibited lower selectivities below 50%, while the Co–N₄ structure with surrounding oxygen species could show enhanced selectivities up to 82%.⁶ A recent study showed that ORR selectivity could be tuned beyond the adsorption site by a molecular confinement strategy.¹³

To further optimize the peroxide selectivities, it is important to understand the structure–property relationship of active sites and reveal selectivity descriptors. However, it is still challenging to define them often due to complicated catalyst structures. For example, Chen et al. used the electronic energy difference between *O₂ and *H₂O₂ intermediates (E_*O2 − E_*H2O2) as a selectivity descriptor in a porphyrin-containing Co SAC and found that the 2e⁻ pathway was dominant.⁹ In contrast, Xia et al. indicated that the 4e⁻ pathway was dominant for a similar structure by using a crystal field stabilization energy descriptor.¹⁴ Thus, model systems with well-defined active centers are highly required for identifying universal selectivity descriptors.

Metal macrocyclic complexes with well-defined structures have been studied as 2e⁻ ORR electrocatalysts in homogeneous systems and showed high selectivities. A series of N₄-ligated and N₂O₂-ligated cobalt complexes exhibited over 90% selectivities in homogeneous organic electrolytes.^15,16 However, heterogeneous macrocyclic complex electrocatalysts, which can show higher reduction current densities and easy catalyst recovery for practical applications, are still rarely reported with high selectivities.¹⁷ Metal phthalocyanines (MPcs) have been mainly considered as electrocatalysts for 4e⁻ ORR before.^18–23 Recently, we found that the molecularly dispersed electrocatalyst (MDE) of NiPc anchored on carbon nanotubes (CNTs) showed over 80% selectivities on rotating ring-disk electrodes (RRDEs), which was superior to the aggregated NiPc sample.²⁴

In this work, we study a series of MPc MDEs with various metal centers (M = Co, Ni, Zn, Cu, Mn) toward 2e⁻ ORR and examine the performance descriptors. The difference in the binding energy of *O and *OOH intermediates (ΔG_*O − ΔG_*OOH) and the *H₂O₂ binding energy (ΔG_*H2O2) are found as the selectivity descriptors. ZnPc and CuPc MDEs with more positive ΔG_*O − ΔG_*OOH showed even higher selectivities than NiPc MDE. A hydrophobic environment can further enhance the measured selectivities by increasing ΔG_*H2O2. In the gas diffusion electrode (GDE) measurements, CoPc and ZnPc MDEs can produce HO⁻₂ stably and rapidly with high faradaic efficiency of H₂O₂ (FE(H₂O₂)).

Results and discussion, experimental

Theoretical ORR performances

A series of MPcs with different metal centers (M = Co, Ni, Zn, Cu, Mn) were first studied using the computational hydrogen electrode model.²⁵ As depicted in Fig. 1a, oxygen molecules are absorbed on the metal centers of MPcs as active sites and are further reduced through the 2e⁻ or 4e⁻ ORR pathway.²⁶ The calculated free energy profiles in these MPcs are plotted in Fig. 1b at 0.7 V (theoretical equilibrium potential for 2e⁻ ORR) to analyze the ORR selectivities and activities of MPcs.

Fig. 1 (a) Schematic illustration of the 2e⁻ and 4e⁻ pathway on MPcs (M = Co, Ni, Zn, Cu, Mn). Blue, green, grey, and white spheres represent metal, nitrogen, carbon, and hydrogen, respectively. (b) DFT-calculated free energy profiles of the 2e⁻ ORR pathway to H₂O₂ (highlight in a dashed region) and the 4e⁻ ORR pathway to H₂O on MPcs at 0.7 V. (c) Correlation between ΔG_*O − ΔG_*OOH and the measured selectivities at 0.7 V and 0.3 V. (d) Correlation between ΔG_*OOH and the measured 2e⁻ mass activity at 0.7 V.

Further reactions of the *OOH intermediate determine which pathway proceeds. The *–O bond breaks to directly generate H₂O₂ for the 2e⁻ pathway, while the O–O bond breaks to produce H₂O for the 4e⁻ pathway (details are given in ESI Methods†). Therefore, the next proton-coupled electron transfer processes of *OOH determine the selectivities. The catalysts with high selectivities should be more inclined thermodynamically to form H₂O₂ relative to H₂O. Several possible selectivity descriptors are extracted according to the trade-off relationship between the 4e⁻ and the 2e⁻ pathway. Higher ΔG_*O − ΔG_*OOH (Fig. 1c) and ΔG_*O − ΔG_*+H2O2 (Fig. S2a†) could inhibit the 4e⁻ pathway to achieve higher peroxide selectivities. The values of ZnPc and CuPc are higher than the previously reported NiPc, suggesting their higher preference for the 2e⁻ pathway. On the other hand, a higher value of ΔG_*OOH − ΔG_*+H2O2 (Fig. S2b†), which lowers the free energy required for H₂O₂ formation, could also facilitate the 2e⁻ pathway.

As *OOH is the key intermediate in the 2e⁻ ORR pathway, the binding energy of *OOH (ΔG_*OOH) should be neither too strong nor too weak.⁷ Hence, ΔG_*OOH is used as an activity descriptor to calculate limiting potentials (details are given in ESI Methods†) for the 2e⁻ ORR pathway.²⁷ We constructed an activity-volcano plot based on the limiting potential as a function of ΔG_*OOH in Fig. 1d. It predicts that CoPc has the lowest overpotential for the 2e⁻ ORR pathway, indicating its highest 2e⁻ activity. It can also be seen that the overpotential of MnPc is the largest among these five MPcs.

Electrochemical ORR performances

We prepared a series of MPc MDEs (M = Co, Ni, Zn, Cu, Mn) by anchoring MPc molecules on CNTs according to our previous approach.^28–31 The metal contents of all MPc MDEs were adjusted to around 0.6 wt% (Table S2†). MPc MDEs show nanotubular structures of CNTs and no noticeable aggregations of MPc molecules are observed in transmission electron microscope (TEM) (Fig. S3a–e†). Discrete bright spots can be seen on the side walls of CNTs in HAADF-STEM (Fig. S3f†), suggesting the dispersion of MPc molecules in MDEs.

Then, we evaluated their electrochemical ORR performances by using RRDEs in 0.1 M KOH. The ORR polarization curves of MPc MDEs vary with the metal centers and are significantly different from that of CNT (Fig. 2a), suggesting MPc molecules are catalytically active in MDEs. The disk currents mainly compose two parts of currents originating from 2e⁻ and 4e⁻ ORR pathway, while the ring currents are produced by the oxidation of HO⁻₂ on the ring electrodes of RRDEs. The peroxide selectivities can be calculated from the disk currents and the ring currents (details are given in ESI Methods†), which are plotted in Fig. 2b. The selectivities of ZnPc and CuPc MDEs reach over 90% in a wide potential range from 0.5 to 0.75 V, which are superior to the previously reported NiPc MDE. CoPc MDE shows around 78% selectivities above 0.5 V, but it drops sharply below 0.5 V and reaches 35% at 0.2 V. In contrast, MnPc MDE exhibits large limiting disk reduction current density and small ring currents, indicating that it mainly facilitates the 4e⁻ pathway. The highest selectivity of MnPc MDE is only 36% at 0.75 V, and a sharp decline is observed with the increase of overpotentials.

Fig. 2 (a) ORR polarization plots for MPc MDEs (M = Co, Ni, Zn, Cu, Mn) with the loading of 0.2 mg cm⁻² on RRDEs at 1600 rpm in 0.1 M KOH saturated with O₂. (b) Corresponding selectivities of MPc MDEs calculated from RRDE measurments.

To reveal the selectivity descriptor, we plotted the selectivities of MPc MDEs at a high potential of 0.7 V and a low potential of 0.3 V in Fig. 1c, which show similar tendencies as the theoretical predictions of ΔG_*O − ΔG_*OOH. The sharply reduced selectivities of MnPc MDE compared to others are attributed to the very negative ΔG_*O − ΔG_*OOH of MnPc. However, the other descriptors showed certain contradiction with the experiment data. For example, there is a huge difference in ΔG_*O − ΔG_*+H2O2 between NiPc and CoPc, suggesting the CoPc is supposed to have substantially lower selectivities than NiPc (Fig. S2a†). And NiPc is supposed to exhibit higher selectivities than ZnPc deduced from the descriptors of ΔG_*OOH − ΔG_*+H2O2 (Fig. S2b†). CoPc would show much lower selectivities based on the previously reported descriptor of E_*O2 − E_*H2O2 (Fig. S2c†).⁹ Therefore, the ΔG_*O − ΔG_*OOH descriptor can better describe the experimental selectivities of these MPc MDEs.

Onset potential is often used to describe ORR activities. CoPc MDE exhibits the most positive onset potential (defined as the potential at the reduction current density of 0.1 mA cm⁻²) of 0.85 V (Fig. 2a) compared to other MPc MDEs. However, to verify the 2e⁻ activity descriptor, it should be noted that the experimental activities cannot be simply extracted from ORR onset potentials due to the interference of the 4e⁻ pathway. Therefore, 2e⁻ mass activity calculated from the product of mass activity and FE(H₂O₂) is proposed to evaluate the activity of 2e⁻ ORR. As shown in Fig. 1d, the trend of experimental 2e⁻ mass activity at 0.7 V is consistent with the theoretical activity volcano plot. Moreover, after a two-hour chronoamperometric measurement under 0.6 V, the UV-vis adsorption profiles show that the molecular structure in CoPc MDE remains unchanged, suggesting its good structural stability (Fig. S5†).

We also measured the electrochemical performances of aggregated molecule samples. We prepared the mixture samples of MPc and CNTs by drop-drying from the ethanol dispersion (denoted as MPc + CNT). Significant aggregates of MPc molecules can be observed in MPc + CNT from the scanning electron microscope (SEM) (Fig. S6†). In Fig. S7a,† the onset potentials of CoPc + CNT, NiPc + CNT, and MnPc + CNT are positively shifted compared to their MDE counterparts, whereas those of CuPc + CNT and ZnPc + CNT are negatively shifted. Moreover, the selectivities of NiPc + CNT and CuPc + CNT (Fig. S7b†) show opposite trends when compared to the theoretical predictions. We also prepared neat MPc samples without CNTs. All of them are less active than their MDE counterparts, showing similar onset potentials at ∼0.7 V (Fig. S7c†) and lower reduction current densities. In brief, the performances of MPc electrocatalysts in aggregated states (MPc + CNT and neat MPcs) are inconsistent with the theoretical calculations, revealing the interference of molecular aggregation on studies of intrinsic electrocatalytic performances.

H₂O₂ reduction performances of MDEs

The produced H₂O₂ from 2e⁻ ORR could be trapped inside the catalyst layer and be further reduced.⁶ It is found that the selectivities of all MPc MDEs show sharp drops below certain potentials (Fig. 2b). For example, the selectivity of CoPc MDE decreases dramatically below 0.5 V. When we reduced the catalyst loading amount of CoPc MDE from 0.2 to 0.05 mg cm⁻², the selectivity also became higher below around 0.5 V (Fig. S8†). A similar phenomenon was also observed in the reported Co SAC.⁶ The diffusion pathways of the produced H₂O₂ would be shortened as the catalyst loading amount becomes lower, decreasing its chance for further reduction. The H₂O₂ reduction reaction (H₂O₂RR) performances of MPc MDEs on rotating disk electrodes (RDEs) were evaluated in 0.1 M KOH containing 1 mM H₂O₂ (Fig. 3a). The H₂O₂RR onset potential of MnPc MDE is the most positive, followed by CoPc MDE. The H₂O₂RR activities of the other three samples are much lower. The onset potentials of H₂O₂RR highly coincide with the potentials at which selectivities begin to sharply decrease (Fig. 3b), and the currents of H₂O₂RR are almost positively correlated with the extent of selectivity reduction. These results indicate that the further reduction of H₂O₂ could lower the measured selectivities.

Fig. 3 (a) H₂O₂RR performances of MPc MDEs loaded on RDEs at 1600 rpm in Ar saturated 0.1 M KOH (containing 1 mM H₂O₂). (b) Comparison of H₂O₂RR performances (solid line) and peroxide selectivities (dash line) of CoPc and MnPc MDEs. (c) DFT-calculated free energy profiles of H₂O₂RR on MPcs at 0.7 V. (d) Schematic illustrations of H₂O₂RR on two different types of MPc MDEs.

To investigate the reason for high H₂O₂ reduction abilities of CoPc and MnPc MDEs, we carried out theoretical calculations for H₂O₂RR (Fig. 3c). The adsorption pathway of H₂O₂ molecules on NiPc, CuPc, and ZnPc is endothermic, while that for CoPc and MnPc is exothermic. The differences in the H₂O₂ reduction ability of these two types of MPc MDEs are depicted in Fig. 3d. It indicates that catalysts in real production conditions should have higher ΔG_*H2O2 to hinder H₂O₂RR from diffusive H₂O₂ adsorption for higher measured selectivities. Therefore, ΔG_*H2O2 is another selectivity descriptor.

H₂O₂ production in gas diffusion electrodes

To further explore practical application of MPc MDEs toward 2e⁻ ORR, we measured MPc MDEs in GDE devices (Fig. S10†), which could afford higher reduction current densities by increasing oxygen concentration on catalyst layers.³² The catalyst inks of MPc MDEs with polytetrafluoroethylene (PTFE) were loaded on gas diffusion layers and tested in flowed 1 M KOH. CoPc MDE can still exhibit the highest activity among all MPc MDEs (Fig. 4a). The chronoamperometry measurements at 10 to 225 mA cm⁻² were conducted (Fig. S11†), and the determined FE(H₂O₂)s at varied applied voltages are shown in Fig. 4b. NiPc, CuPc, and ZnPc MDEs can exhibit over 95% FE(H₂O₂)s in the whole current density range, which are even higher than the RRDE measurements (Fig. S12†). Interestingly, the FE(H₂O₂)s of CoPc MDE are around 90%, which are substantially higher than those of CoPc MDE without PTFE. The H₂O₂RR performances of CoPc MDE with and without PTFE are compared in Fig. S13.† CoPc MDE with PTFE shows much smaller H₂O₂ reduction current densities than the one without PTFE. These results suggest that H₂O₂RR is inhibited in CoPc MDE with PTFE, leading to higher FE(H₂O₂)s.

Fig. 4 (a) Linear sweep voltammetry of MPc MDEs (M = Co, Ni, Zn, Cu) with PTFE and CoPc MDE* without PTFE in 1 M KOH tested by GDEs. (b) The measured FE(H₂O₂)s at different applied potentials. (c) Schematic illustrations of the hydrophobic effect of PTFE on inhibiting further reduction of H₂O₂ and contact angles of CoPc MDE with and without PTFE loaded on carbon paper. (d) DFT-calculated free energy profiles of the H₂O₂ adsorption process on CoPc with different relative permittivity. (e) Long-term stability tests of CoPc and ZnPc MDEs at the reduction current density of 100 mA cm⁻². (f) Comparison of the FE(H₂O₂)s and production rates of CoPc and ZnPc MDEs with other reported electrocatalysts.

As shown in Fig. 4c, PTFE increases the hydrophobicity of catalyst layers. For simulating the influence of the hydrophobic environment with PTFE in the implicit solvation model, we assumed that the access of water around active sites is limited and the relative permittivity (ε_r) in the DFT calculations becomes lower than that of bulk water (78.4).^33,34 When ε_r decreased from 78.4 to 2–5, the calculated ΔG_*H2O2 at 0.7 V increases from 1.97 eV to 2.26–2.18 eV (Fig. 4d), respectively, suggesting the inhibition of H₂O₂ adsorption. However, the value of another selectivity descriptor ΔG_*O − ΔG_*OOH remains basically unchanged (less than 0.05 eV) with decreasing relative permittivity (Table S4†). These results indicate that the hydrophobic effect (low ε_r) increases FE(H₂O₂)s by decreasing H₂O₂RR.

In the stability test for 10 hours (Fig. 4e), CoPc MDE can produce HO⁻₂ stably with a high reduction current density of 100 mA cm⁻² at ∼0.7 V. The FE(H₂O₂) maintained 89% and the production rate is calculated to be 8200 mmol g_cat⁻¹ h⁻¹ on the average (calculation details are provided in ESI†). ZnPc MDE can also deliver ∼98% FE(H₂O₂) and a production rate of 9100 mmol g_cat⁻¹ h⁻¹ at around 0.65 V. CoPc and ZnPc MDEs outperform the reported catalysts with relatively high FE(H₂O₂)s under higher production rates (Fig. 4f and Table S5†). The cathodic energy efficiency (calculation details are provided in the ESI†) was also analyzed. Although ZnPc MDE has lower activity than CoPc MDE, it exhibits similar cathodic energy efficiency of around 80% to CoPc MDE owning to its higher FE(H₂O₂)s.

Conclusions

With the advantages of well-defined active sites and no influence from molecular aggregation, MPc MDEs are beneficial for studying intrinsic performances. Two selectivity descriptors toward 2e⁻ ORR are uncovered. The first descriptor ΔG_*O − ΔG_*OOH emphasizes the importance of O–O bond breaking for intrinsic selectivity. The second descriptor ΔG_*H2O2 determines the extent of further H₂O₂ reduction that affects measured selectivity. The hydrophobic environment in the catalyst layer can increase ΔG_*H2O2 to inhibit H₂O₂ reduction. In GDE devices, ZnPc and CoPc MDEs can continuously produce HO⁻₂ at over 8200 mmol g_cat⁻¹ h⁻¹ with 98% FE(H₂O₂) and 89% FE(H₂O₂), respectively. This work can benefit the development of high-performance electrocatalysts toward 2e⁻ ORR, and facilitate the rational design of noble-metal-free electrocatalysts.

Data availability

All supporting data can be found in the ESI.†

Author contributions

Y. L. conceived the project and designed the experiments. Y. Y. carried out the DFT calculations and electrochemical measurements. H. L. and Z. J. performed the material characterizations. Y. L. and Y. Y. analyzed the data and prepared the manuscript with input from all the authors. All authors discussed the results and commented on the manuscript.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The work was supported by National Natural Science Foundation of China (22075125).

Notes and references

1. Y. Y. Jiang, P. J. Ni, C. X. Chen, Y. Z. Lu, P. Yang, B. Kong, A. Fisher and X. Wang, Adv. Energy Mater., 2018, 8, 25.
2. K. Jiang, J. J. Zhao and H. T. Wang, Adv. Funct. Mater., 2020, 30, 11.
3. N. Wang, S. Ma, P. Zuo, J. Duan and B. Hou, Adv. Sci., 2021, 8, 2100076.
4. S. Siahrostami, A. Verdaguer-Casadevall, M. Karamad, D. Deiana, P. Malacrida, B. Wickman, M. Escudero-Escribano, E. A. Paoli, R. Frydendal, T. W. Hansen, I. Chorkendorff, I. E. L. Stephens and J. Rossmeisl, Nat. Mater., 2013, 12, 1137–1143.
5. Y. Y. Sun, L. Silvioli, N. R. Sahraie, W. Ju, J. K. Li, A. Zitolo, S. Li, A. Bagger, L. Arnarson, X. L. Wang, T. Moeller, D. Bernsmeier, J. Rossmeisl, F. Jaouen and P. Strasser, J. Am. Chem. Soc., 2019, 141, 12372–12381.
6. E. Jung, H. Shin, B.-H. Lee, V. Efremov, S. Lee, H. S. Lee, J. Kim, W. Hooch Antink, S. Park, K.-S. Lee, S.-P. Cho, J. S. Yoo, Y.-E. Sung and T. Hyeon, Nat. Mater., 2020, 19, 436–442.
7. J. J. Gao, H. B. Yang, X. Huang, S. F. Hung, W. Z. Cai, C. M. Jia, S. Miao, H. M. Chen, X. F. Yang, Y. Q. Huang, T. Zhang and B. Liu, Chem, 2020, 6, 658–674.
8. Q. Zhang, X. Tan, N. M. Bedford, Z. Han, L. Thomsen, S. Smith, R. Amal and X. Lu, Nat. Commun., 2020, 11, 4181.
9. C. Liu, H. Li, F. Liu, J. Chen, Z. Yu, Z. Yuan, C. Wang, H. Zheng, G. Henkelman, L. Wei and Y. Chen, J. Am. Chem. Soc., 2020, 142, 21861–21871.
10. Q. L. Zhao, Y. Wang, W. H. Lai, F. Xiao, Y. X. Lyu, C. Z. Liao and M. H. Shao, Energy Environ. Sci., 2021, 14, 5444–5456.
11. C. Tang, Y. Jiao, B. Shi, J.-N. Liu, Z. Xie, X. Chen, Q. Zhang and S.-Z. Qiao, Angew. Chem., Int. Ed., 2020, 59, 9171–9176.
12. B.-Q. Li, C.-X. Zhao, J.-N. Liu and Q. Zhang, Adv. Mater., 2019, 31, 1808173.
13. X. Li, S. Tang, S. Dou, H. J. Fan, T. S. Choksi and X. Wang, Adv. Mater., 2022, 34, 2104891.
14. C.-Y. Lin, L. Zhang, Z. Zhao and Z. Xia, Adv. Mater., 2017, 29, 1606635.
15. A. Rana, Y.-M. Lee, X. Li, R. Cao, S. Fukuzumi and W. Nam, ACS Catal., 2021, 11, 3073–3083.
16. Y.-H. Wang, M. L. Pegis, J. M. Mayer and S. S. Stahl, J. Am. Chem. Soc., 2017, 139, 16458–16461.
17. Y. Wang, G. I. N. Waterhouse, L. Shang and T. Zhang, Adv. Energy Mater., 2021, 11, 2003323.
18. R. Jasinski, Nature, 1964, 201, 1212–1213.
19. J. Guo, H. Li, H. He, D. Chu and R. Chen, J. Phys. Chem. C, 2011, 115, 8494–8502.
20. R. M. Reis, R. B. Valim, R. S. Rocha, A. S. Lima, P. S. Castro, M. Bertotti and M. R. V. Lanza, Electrochim. Acta, 2014, 139, 1–6.
21. H. He, Y. Lei, C. Xiao, D. Chu, R. Chen and G. Wang, J. Phys. Chem. C, 2012, 116, 16038–16046.
22. R. R. Chen, H. X. Li, D. Chu and G. F. Wang, J. Phys. Chem. C, 2009, 113, 20689–20697.
23. Z. Zhang, S. Yang, M. Dou, H. Liu, L. Gu and F. Wang, RSC Adv., 2016, 6, 67049–67056.
24. Y. Wang, Z. Zhang, X. Zhang, Y. Yuan, Z. Jiang, H. Zheng, Y.-G. Wang, H. Zhou and Y. Liang, CCS Chem., 2022, 4, 228–236.
25. J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bligaard and H. Jónsson, J. Phys. Chem. B, 2004, 108, 17886–17892.
26. G. F. Wang, N. Ramesh, A. Hsu, D. Chu and R. R. Chen, Mol. Simul., 2008, 34, 1051–1056.
27. G.-F. Han, F. Li, W. Zou, M. Karamad, J.-P. Jeon, S.-W. Kim, S.-J. Kim, Y. Bu, Z. Fu, Y. Lu, S. Siahrostami and J.-B. Baek, Nat. Commun., 2020, 11, 2209.
28. X. Zhang, Z. Wu, X. Zhang, L. Li, Y. Li, H. Xu, X. Li, X. Yu, Z. Zhang, Y. Liang and H. Wang, Nat. Commun., 2017, 8, 14675.
29. Z. Jiang, Y. Wang, X. Zhang, H. Zheng, X. Wang and Y. Liang, Nano Res., 2019, 12, 2330–2334.
30. X. Zhang, Y. Wang, M. Gu, M. Wang, Z. Zhang, W. Pan, Z. Jiang, H. Zheng, M. Lucero, H. Wang, G. E. Sterbinsky, Q. Ma, Y.-G. Wang, Z. Feng, J. Li, H. Dai and Y. Liang, Nat. Energy, 2020, 5, 684–692.
31. Z. Lin, Z. Jiang, Y. Yuan, H. Li, H. Wang, Y. Tang, C. Liu and Y. Liang, Chin. J. Catal., 2022, 43, 104–109.
32. Y. L. Wang, R. Shi, L. Shang, G. I. N. Waterhouse, J. Q. Zhao, Q. H. Zhang, L. Gu and T. R. Zhang, Angew. Chem., Int. Ed., 2020, 59, 13057–13062.
33. B. W. Noffke, Q. Li, K. Raghavachari and L.-s. Li, J. Am. Chem. Soc., 2016, 138, 13923–13929.
34. K. Takeyasu, M. Furukawa, Y. Shimoyama, S. K. Singh and J. Nakamura, Angew. Chem., Int. Ed., 2021, 60, 5121–5124.

---

Footnote

† Electronic supplementary information (ESI) available. See https://doi.org/10.1039/d2sc03714a

---