An environmentally adaptive gold single-atom catalyst with variable valence states

Abstract

Single-atom catalysts revolutionize catalysis by maximizing atomic efficiency and enhancing reaction specificity, offering high activity and selectivity with minimal material usage, which is crucial for sustainable processes. However, the unique properties that distinguish single-atom catalysts from other forms, including bulk and nanoparticle catalysts, as well as the physical mechanisms behind their high activity and selectivity, remain unclear, limiting their broader application. Here, through first-principles calculations, we have identified an environmentally adaptive gold single-atom catalyst on a CeO₂(111) surface capable of adjusting its valence state in response to different environmental conditions. This adaptability enables the catalyst to simultaneously maintain high stability and activity. In a CO gas atmosphere, numerous oxygen vacancies form on the CeO₂(111) surface, where Au single atoms stably adsorb, exhibiting a negative oxidation state that deactivates the catalyst. In an O₂ atmosphere, these vacancies are filled, causing the Au single atoms to adsorb onto lattice oxygen and become oxidized to a positive oxidation state, thereby reactivating the catalyst. Under CO oxidation reaction conditions, the Au single atoms oscillate between these positive and negative oxidation states, effectively facilitating the CO oxidation process. These findings provide new insights into the unique properties and high performance of single-atom catalysts, contributing to a better understanding and utilization of these catalysts in various applications.

---

Introduction

Since Zhang et al.¹ successfully prepared Pt single-atom catalysts supported on FeO_x using a co-precipitation method in 2011, single-atom catalysts (SACs) have gained significant attention. These catalysts feature isolated single atoms as active centers, uniformly dispersed and anchored on various supports.² Due to their exceptional properties, such as high activity and selectivity, SACs are widely applied in diverse catalytic reactions including CO oxidation,^1,3–7 CO₂ reduction,^8–18 and water-gas shift reactions.^19–27 For example, Liu et al.²⁸ employed a straightforward photochemical method to prepare a stable Pd/TiO₂ SAC at room temperature, which demonstrated high activity and stability in hydrogenation reactions. Xie et al.²⁹ reported an atomically dispersed Co and N co-doped carbon (Co–N–C) catalyst that exhibited oxygen reduction reaction activity comparable to an Fe–N–C catalyst but with four-fold greater durability. Moreover, Kyriakou et al.,³⁰ through desorption measurements and high-resolution scanning tunneling microscopy, found that isolated Pd atoms on a Cu surface significantly reduced the energy barriers for hydrogen adsorption and desorption, facilitating selective hydrogenation of styrene and acetylene, which is not achievable with pure Cu or Pd metals. While these studies have advanced our understanding of SACs, key mechanistic aspects—particularly those involving dynamic interactions between single atoms, supports, and reaction environments—have not yet been fully resolved.

First-principles calculations are crucial for elucidating the reaction mechanisms of SACs. For example, Tong et al.³¹ used density functional theory (DFT) to investigate the efficient oxygen reduction reaction (ORR) mechanism of a bimetallic SAC consisting of Cu and Zn on a nitrogen-doped carbon support. They found that the influence of Zn on the Cu's d-orbital electron distribution facilitates the stretching and cleavage of O–O bonds at the Cu active site, thus accelerating the rate-determining step involving OOH*. Similarly, DFT studies by Riley et al.³² proposed a mechanism for the selective hydrogenation of alkynes catalyzed by ceria, which involves the heterolytic dissociation of H₂ at oxygen vacancies on CeO₂(111), facilitated by a suppressed Lewis pair composed of spatially separated O and Ce sites. Based on this mechanism, they suggested using Ni-doped ceria to generate oxygen vacancies, a principle that was confirmed experimentally to enhance activity in the selective hydrogenation of acetylene. However, SACs are inherently complex systems. To stabilize single atoms, substrates that can be reduced, such as CeO₂, are typically used. This can lead to multiple oxidation states of the single atoms, the presence of oxygen vacancies, and redox processes involving substrate metal atoms. For example, in the Au–CeO₂ SAC, Au atoms can exist in different oxidation states and CeO₂ is reducible, exhibiting oxygen vacancies along with transformations between Ce⁴⁺ and Ce³⁺ ions. These dynamic interactions complicate the precise control and mechanistic interpretation of SAC behavior under operando conditions. For instance, theoretical calculations by Camellone et al.³³ suggested that, in a CO atmosphere, Au atoms positioned above the oxygen vacancies of ceria adopt a negative charge state, thus losing the ability to activate CO further. In contrast, experimental X-ray photoelectron spectroscopy (XPS) studies by Guo et al.³⁴ reported that single Au atoms exhibit a high oxidation state with notable activity, a finding that conflicts with theoretical results. This high oxidation state is not explained by Au doping at Ce sites, as calculations indicate that such doping is energetically unfavorable.^35,36 Such discrepancies highlight the need for deeper insights into how environmental adaptability governs SAC performance.

In this study, we conducted a detailed examination of the thermodynamic and kinetic behaviors of single-atom Au catalysts under CO, O₂, and CO oxidation reaction conditions using first-principles calculations. We discovered an environmentally adaptive gold SAC on the CeO₂(111) surface, which uniquely adjusts its valence state in response to varying environmental conditions. Furthermore, by contrasting the performance of the single-atom Au₁ catalyst with larger Au₄ and Au₁₀ catalyst systems, we underscored the distinctive ability of single-atom catalysts to efficiently modulate their valence states, enabling them to maintain both high stability and activity simultaneously.

Computational details

All electronic structure calculations were performed using density functional theory. Valence electron densities were expanded using plane-wave basis sets, and the projected augmented-wave (PAW) method³⁷ was used to describe the interactions between the core and the valence electrons. We conducted these calculations using the Vienna ab initio simulation package (VASP 5.4.4).³⁸ The spin-polarized Kohn–Sham calculations were based on the generalized gradient approximation, employing the Perdew–Burke–Ernzerhof (PBE)³⁹ exchange–correlation functional. Additionally, the Hubbard-type on-site Coulomb interaction term (PBE+U), according to Dudarev's approximation, was applied to the f-electrons of Ce atoms^40,41 with a Hubbard U value set at 5.0 eV to describe the localization of excess charges in Ce 4f states, consistent with studies recommending U values in the range of 4.5–6.0 eV.^40,42,43 In all these calculations, we applied the projector augmented-wave⁴⁴ method and an energy cut-off of 400 eV.

The CeO₂ support was modeled using a 4 × 4 supercell comprising a nine-layer slab with a 15 Å vacuum space to avoid interactions between periodic images. The bottom three layers (O–Ce–O) were fixed at their bulk positions, while the remaining layers were allowed to relax during geometry optimization. The force convergence criterion was set at 0.05 eV Å⁻¹. A gamma-centered 1 × 1 × 1 k-point mesh was used for the Brillouin zone integration. Testing a 2 × 2 × 1 k-point mesh for the (4 × 4) CeO₂ supercell revealed only minor variations (approximately 10 meV) in CO adsorption energy, confirming the convergence of our gamma-point calculations. The transition states were located using the climbing-image nudged elastic band^45,46 and dimer methods.⁴⁷ The adsorption energies were calculated using the following equation:

E_ad = E_mol/surf − (E_surf + E_mol) (1)

where E_mol/surf is the total energy of the surface with the adsorbed molecules, E_surf is the energy of the substrate, and E_mol is the energy of the isolated molecules in the gas phase.

Results and discussion

Deactivation of the Au single atom in the CO atmosphere

To determine the most stable configuration of the Au₁ atom on the CeO₂(111) surface, different high-symmetry adsorption sites were considered (Fig. S1, ESI†). The most energetically favorable configuration was found when the Au atom adsorbed at the bridge site of surface oxygen, with an adsorption energy of −1.41 eV, aligning with previous literature.³³ This configuration also exhibited a +1 oxidation state for the Au atom, corroborated by spin charge density analysis which revealed the presence of one Ce³⁺ ion, indicating electron transfer from Au to the substrate (Fig. S2, ESI†). In the CO atmosphere, the exceptional oxygen buffering capacity of CeO₂ enables CO oxidation through interaction with lattice oxygen, forming CO₂ and creating an oxygen vacancy. The Au atom strongly interacts with this newly formed vacancy, positioning itself above it as shown in Fig. 1a. The Au atom is positioned 1.21 Å above the oxygen vacancy, with an Au–Ce bond length of 3.12 Å, aligning with reported results.³³ Calculations indicate that while aggregation of negatively charged Au atoms into clusters is thermodynamically feasible, a high kinetic barrier (approximately 1.2 eV) effectively prevents such aggregation under typical experimental conditions. Bader charge analysis reveals a net charge of −0.29 e for the Au atom, classifying it as Au^δ−. We also tested other charge calculation methods, which showed a minimal effect on the charge distribution trends. Differential charge density analysis (Fig. 1b) shows electron accumulation at the Au atom, confirming the negatively charged nature of Au^δ−. The removal of one lattice oxygen results in the generation of two excess electrons localized at the cerium site, which reduce two Ce⁴⁺ to Ce³⁺ ions. For Au₁ adsorbed on the CeO₂(111) system with one oxygen vacancy [Au₁@CeO₂(111)-O_V], the presence of one Ce³⁺ ion was observed from the spin charge density analysis, indicating an oxidation state of −1 for Au. Additionally, we assessed the impact of Ce³⁺ ions positioning on the energy of various structures during these processes, finding that the position of Ce³⁺ does not affect the overall reaction energy (Fig. S2–S5, ESI†).

Fig. 1 Configuration and electronic structure analysis of Au₁@CeO₂(111)-O_V. (a) System configuration and Bader charge of the Au₁ atom. (b) Differential charge density between Au₁ and CeO₂(111)-O_V, and an isosurface displayed at 0.01 e Å⁻³. (c) System configuration with adsorbed CO and Bader charge of the Au₁ atom. (d) Differential charge density between CO and the Au₁@CeO₂(111)-O_V system, and an isosurface displayed at 0.01 e Å⁻³.

We positioned a CO molecule above the Au atom to evaluate the CO adsorption capacity of Au^δ−, resulting in the configuration shown in Fig. 1c. The Au–C bond length measures 2.27 Å, and the C–O bond length remains at 1.15 Å, identical to that in a free CO molecule. The adsorption energy of CO on Au^δ− is approximately zero, which is significantly higher than the −2.54 eV calculated for CO adsorption on an Au atom on the stoichiometric CeO₂(111) surface. This indicates that Au^δ− possesses negligible CO adsorption capacity. Although the thermodynamic energy for CO to displace the Au atom is about 1.36 eV lower than that of the Au atom remaining at the vacancy, a substantial kinetic barrier of around 1 eV hinders this displacement, particularly at lower temperatures, suggesting that CO is unlikely to displace the Au atom from the vacancy site under typical reaction conditions. Bader charge analysis of this configuration shows a net charge of −0.28 e on the Au atom, nearly unchanged from the pre-adsorption state, confirming the persistence of the Au^δ− state. The differential charge density (Fig. 1d) illustrates minimal charge transfer between the CO molecule and Au₁@CeO₂(111)-O_V. This minimal charge transfer and the Bader charge analysis collectively demonstrate that Au^δ−, in the presence of an oxygen vacancy, lacks the capacity to adsorb and activate CO. This finding is consistent with previous calculations.³³ However, experimental evidence suggests the existence of high-valence-states and highly active Au single atoms,³⁴ which contradicts these theoretical calculations. We propose that the discrepancy may arise from the interactions of the Au₁@CeO₂(111)-O_V system with oxygen molecules.

Reactivation of Au single atoms in the O₂ atmosphere

In the Mars–van Krevelen (MvK) mechanism, the catalyst undergoes reduction by forming oxygen vacancies during the initial oxidation step. For subsequent oxidation reactions, dissociatively adsorbed oxygen is needed to fill these vacancies and reoxidize the catalyst. To determine if the Au₁@CeO₂(111)-O_V catalyst follows this mechanism for CO oxidation in an oxygen atmosphere and thus reactivates the deactivated Au₁@CeO₂(111)-O_V, we evaluated the potential for molecular oxygen to fill the surface oxygen vacancies. This process would transition the Au^δ− state to Au^δ+, thereby restoring the catalyst's ability to oxidize CO effectively.

As shown in Fig. 2a, molecular oxygen from the atmosphere overcomes a low energy barrier (0.34 eV) to fill the oxygen vacancy beneath the deactivated Au^δ− atom (Fig. S6, ESI†), with an adsorption energy of −1.39 eV. Additionally, our calculations indicate that CO does not occupy the oxygen vacancy, further supporting the proposed mechanism (Fig. S7, ESI†). One oxygen atom from the O₂ molecule occupies the vacancy, while the other forms a bond with lattice oxygen, resulting in an O–O bond length of 1.48 Å, which is longer than the 1.23 Å bond length in a free O₂ molecule. Bader charge analysis indicates that the Au atom acquires a +1 e charge, suggesting substantial charge transfer between the Au atom and the substrate, induced by the adsorbed O₂. Differential charge density analysis (Fig. 2b) confirms this significant charge transfer. This transformation reactivates the Au₁@CeO₂(111)-O_V catalyst for CO oxidation. To confirm the catalytic capability of the reactivated Au₁@CeO₂(111)-O_V–O₂ structure, we calculated the adsorption of a CO molecule and identified the most stable configuration, as shown in Fig. 2c. After CO adsorption, the bond length between the original peroxide oxygen and the Au atom elongates from 2.07 Å to 3.18 Å, while the bond length between the Au atom and the lattice oxygen shortens from 2.05 Å to 1.96 Å. The bond length between C and Au measures 1.87 Å, and the C–O bond length extends slightly to 1.15 Å from 1.14 Å in a free CO molecule. The adsorption energy of CO on this activated structure is −1.68 eV. Bader charge analysis shows a +0.73 e charge on the Au atom, indicating that in both structures (Fig. 2a and c), Au is positively charged and thus catalytically active for CO oxidation. Differential charge density analysis (Fig. 2d) demonstrates obvious charge transfer between the CO molecule and the Au₁@CeO₂(111)-O_V–O₂ catalyst, confirming the catalyst's capability to oxidize CO effectively. Additionally, our calculations indicate that the attempt to adsorb a second CO molecule on Au₁@CeO₂(111) leads to an increased distance of 3.44 Å from the Au atom, which is significantly larger than the typical CO adsorption distance (Fig. S8, ESI†). This suggests that further CO adsorption is energetically unfavorable in this configuration.

Fig. 2 Configuration and electronic structure analysis of Au₁@CeO₂(111)-O_V–O₂. (a) Configuration of Au₁@CeO₂(111)-O_V–O₂ and Bader charge of the Au₁ atom. (b) Differential charge density between Au₁ and CeO₂(111)-O_V–O₂, and an isosurface displayed at 0.01 e Å⁻³. (c) Configuration of Au₁@CeO₂(111)-O_V–O₂ with the adsorbed CO molecule and Bader charge of the Au₁ atom. (d) Differential charge density between CO and the Au₁@CeO₂(111)-O_V–O₂ system, and an isosurface displayed at 0.01 e Å⁻³.

Through analyses including the Bader charge and differential charge density, we have shown that the Au^δ− state on an oxygen vacancy lacks the capacity to adsorb and activate CO, consistent with previous findings.³³ However, the introduction of molecular oxygen into the oxygen vacancy transforms Au^δ− into Au^δ+, effectively reactivating the Au₁@CeO₂(111)-O_V catalyst for CO oxidation. This reactivation mechanism conforms to the MvK mechanism and offers new insights into the behavior of the Au₁@CeO₂(111) SAC.

Valence state oscillations of the environmentally adaptive gold single-atom catalyst under CO oxidation reaction conditions

To further elucidate the deactivation and reactivation mechanisms of the Au₁@CeO₂(111) catalyst, we primarily focused on Au atoms adsorbed on three specific sites: the bridge site of surface oxygen (B_O), the top site of the surface oxygen vacancy (O_V), and the site where the oxygen vacancy is filled by an O₂ molecule (O_V–O₂) (Fig. 3(a)). We conducted a projected density of states (PDOS) analysis of Au atoms at these three distinct sites (Fig. 3b).

Fig. 3 (a) Configuration of B_O, O_V and O_V–O₂. (b) Density of states (DOS) for Au atom adsorption at various sites including the bridge site of surface oxygen (B_O), the top site of the surface oxygen vacancy (O_V), and the site with the oxygen vacancy filled by an O₂ molecule (O_V–O₂).

The PDOS for the Au atom on O_V shows orbitals located farther from the Fermi level, suggesting a more stable electronic state that is less reactive with other atoms. This stability is also reflected in the significantly higher integrated density, indicative of greater electron density, consistent with earlier Bader charge and valence state analyses. Conversely, the valence electrons of the Au atoms in the other two configurations (O_V and O_V–O₂) are closer to the Fermi level, indicating higher energy and reactivity, therefore a greater likelihood of interacting with other atoms. As a result, the Au^δ+ in these configurations is catalytically active for CO activation. The Bader charges and CO adsorption energies for Au atoms on these different structures are collected in Table 1.

Table 1 The Bader charges and CO adsorption energies for Au atoms on different sites

Adsorption site	Bader charge/e	CO adsorption energy/eV
BO	+0.55	−2.54
OV	−0.29	+0.05
OV–O2	+1	−1.68

---

The valence changes of the Au atom in the Au₁@CeO₂(111) catalyst, which lead to catalyst deactivation and reactivation, are crucial for understanding the CO oxidation mechanisms on this catalyst. As shown in Fig. 4, the CO oxidation mechanisms include three steps: (I) initial CO oxidation on stoichiometric Au₁@CeO₂(111), resulting in the formation of an oxygen vacancy and CO₂ release (Fig. 4(a)), (II) molecular O₂ adsorption at the vacancy, leading to the formation of surface oxygen species (Fig. 4(b)), and (III) interaction of another CO molecule with these oxygen species, driving the formation of the second CO₂ molecule (Fig. 4(c)) and regenerating the catalyst. The rate-limiting step occurs when the second adsorbed CO molecule reacts with surface oxygen species, requiring an activation energy of just 0.55 eV (Fig. 4j and k), highlighting the high activity of the Au₁@CeO₂(111) catalyst toward CO oxidation. Throughout these processes, the Au atom oscillates between negative and positive oxidation states, where negative states contribute to high stability and positive states contribute to high activity. Under the CO reaction conditions, the Au atoms can adaptively change their oxidation states in response to the adsorbed species to achieve both high stability and activity simultaneously.

Fig. 4 Computed reaction pathway diagram of Au₁@CeO₂(111) catalyzed CO oxidation. (a) First CO oxidation, (b) O₂ adsorption at O_v, and (c) second CO oxidation. The structure represented by the red number is the transition state configuration.

Thermodynamic analysis of Au valence states and oxygen vacancy formation under reaction conditions

To comprehensively address the environmental adaptability of the Au₁@CeO₂(111) catalyst, we performed ab initio thermodynamic analyses to systematically investigate the effects of reaction temperature and partial pressures of CO and CO₂ on the valence state of Au single atoms and the formation of oxygen vacancies on the CeO₂(111) surface (Fig. 5). Specifically, we analysed two distinct pathways for the transition from the initial structure (Fig. 4a) to the oxygen vacancy-containing structure (Fig. 4f), where the Au atom shifts from a positively charged (Au^δ+) state to a negatively charged (Au^δ−) state.

Fig. 5 Thermodynamic analysis of the Au₁@CeO₂(111) catalyst under varying environmental conditions. (a) Gibbs free energy change (ΔG) for CO oxidation (pathway I) at 574 K as a function of CO and CO₂ partial pressures. (b) ΔG profiles for pathway I at 874 K, illustrating the shift of the equilibrium toward lower CO₂ pressures. (c) ΔG analysis at 1174 K, highlighting enhanced spontaneity of oxygen vacancy formation at elevated temperatures. (d) Thermodynamic equilibrium (ΔG = 0) for direct oxygen vacancy generation (pathway II), plotted against temperature and O₂ partial pressure. All contours correspond to calculated free energy landscapes.

Pathway I corresponds to the CO oxidation process (Fig. 4a–f), where CO reacts with lattice oxygen to form CO₂ and generate an oxygen vacancy. Pathway II involves the direct formation of an oxygen vacancy without CO oxidation. For pathway I, the Gibbs free energy change (ΔG) was calculated under varying CO and CO₂ partial pressures at three representative temperatures: 574 K, 874 K, and 1174 K (Fig. 5a–c). At 574 K, the ΔG = 0 contour intersects at CO partial pressures of ∼1 × 10⁻¹⁶⁻ mbar (CO₂: ∼1 × 10⁴ mbar) and ∼1 × 10⁻² mbar (CO₂: ∼1 × 10¹⁸ mbar), indicating that higher CO pressures and lower CO₂ pressures thermodynamically favour the reaction (ΔG < 0). As the temperature increases to 874 K and 1174 K, the ΔG = 0 contour shifts toward lower CO₂ pressures, demonstrating enhanced spontaneity of oxygen vacancy formation at elevated temperatures.

For pathway II (direct oxygen vacancy generation, Fig. 5d), the reaction is endothermic (ΔE = 1.13 eV) with O₂ as a product. The equilibrium condition (ΔG = 0) requires extremely low O₂ partial pressures (<1 × 10⁻¹⁶ mbar) at temperatures above 400 K, while at higher O₂ pressures (e.g., 1 × 10⁴ mbar), temperatures exceeding 1274 K are needed to drive the reaction. This stark contrast highlights the critical role of CO oxidation (pathway I) in facilitating oxygen vacancy formation under experimentally relevant conditions.

These thermodynamic insights elucidate how environmental parameters (temperatures and gas pressures) dynamically regulate the valence state of Au atoms and oxygen vacancy populations on the CeO₂(111) surface. The adaptive switching between Au^δ+ and Au^δ− states, coupled with the temperature- and pressure-dependent vacancy formation, underpins the dual stability and activity of the Au₁@CeO₂(111) catalyst during CO oxidation.

Valence state oscillations of the supported gold cluster catalysts under CO oxidation reaction conditions

To elucidate the unique properties of the Au₁@CeO₂(111) catalyst, we analyzed the CO oxidation processes in the Au₄@CeO₂(111) and Au₁₀@CeO₂(111) catalysts (Fig. S9–S11 and Tables S1, S2, ESI†). The stable configurations of Au₄ and Au₁₀ clusters on CeO₂(111) were referenced from Zhang et al.'s literature,³⁶ and the structure search method was not used to determine the structure of Au clusters. Bader charge analysis throughout the reaction processes showed that the charge changes in gold atoms for Au₄@CeO₂(111) and Au₁₀@CeO₂(111) are less significant compared to those in Au₁@CeO₂(111). This indicates that valence state oscillations are more pronounced in Au₁@CeO₂(111), while in the larger clusters of Au₄@CeO₂(111) and Au₁₀@CeO₂(111), the changes are less significant and distributed among all gold atoms in the cluster, resulting in less noticeable shifts per atom. This non-adaptive behavior with respect to environmental changes has less impact on the catalytic activity of these larger gold clusters during CO oxidation. Consequently, the reaction barriers are higher for Au₄@CeO₂(111) and Au₁₀@CeO₂(111), at 1.59 eV and 0.82 eV, respectively, compared to 0.55 eV for Au₁@CeO₂(111). In contrast, the Au₁@CeO₂(111) catalyst can effectively oscillate its valence state between positive and negative under varying atmospheric conditions, adapting its valence to suit different reaction environments and maintaining high activity and stability throughout the CO oxidation process.

Conclusions

This study provides a comprehensive analysis of the catalytic activity and mechanisms of CO oxidation on Au₁@CeO₂(111) catalysts, using DFT+U calculations, and compares these results with those obtained for Au₄@CeO₂(111) and Au₁₀@CeO₂(111) catalysts. Our findings highlight the crucial role of the valence state of the Au atom in influencing the performance of the catalyst. In a CO atmosphere, CO reacts with surface lattice oxygen to form an oxygen vacancy, stabilizing the Au atom in a catalytically inactive Au^δ− state. In an O₂ atmosphere, the deactivated Au atom is reactivated by molecular oxygen filling the oxygen vacancy, converting the Au^δ− state to a catalytically active Au^δ+ state. The detailed electronic structure analysis, including density of states calculations and Bader charge analysis, confirms that the Au^δ− state is electronically stable, while the Au^δ+ state is active and essential for CO oxidation. Unlike Au₄@CeO₂(111) and Au₁₀@CeO₂(111) catalysts, where the Au valence state is less responsive to environmental changes, the Au₁@CeO₂(111) catalyst effectively adjusts its valence state during the reaction, maintaining both high activity and stability for CO oxidation. This study advances the fundamental understanding of the unique properties of single-atom catalysis and paves the way for practical manipulation of single-atom catalyst behaviors.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We acknowledge the financial support from the National Key Research and Development Program of China (2022YFA1505500 and 2023YFA1506904), the National Natural Science Foundation of China (22302173, 52025011, 92045301, 51971202, and 52171019), and the Fundamental Research Funds for the Central Universities.

Notes and references

1. B. Qiao, A. Wang, X. Yang, L. F. Allard, Z. Jiang, Y. Cui, J. Liu, J. Li and T. Zhang, Nat. Chem., 2011, 3, 634–641.
2. H. He, H. H. Wang, J. Liu, X. Liu, W. Li and Y. Wang, Molecules, 2021, 26, 6501.
3. B. Han, T. Li, J. Zhang, C. Zeng, H. Matsumoto, Y. Su, B. Qiao and T. Zhang, Chem. Commun., 2020, 56, 4870–4873.
4. F. Li, Y. Li, X. C. Zeng and Z. Chen, ACS Catal., 2015, 5, 544–552.
5. Y. Lu, J. Wang, L. Yu, L. Kovarik, X. Zhang, A. S. Hoffman, A. Gallo, S. R. Bare, D. Sokaras, T. Kroll, V. Dagle, H. Xin and A. M. Karim, Nat. Catal., 2018, 2, 149–156.
6. B. Qiao, J.-X. Liang, A. Wang, J. Liu and T. Zhang, Chin. J. Catal., 2016, 37, 1580–1586.
7. J. Xu, Y. Wang, K. Wang, M. Zhao, R. Zhang, W. Cui, L. Liu, M. S. Bootharaju, J. H. Kim, T. Hyeon, H. Zhang, Y. Wang, S. Song and X. Wang, Angew. Chem., Int. Ed., 2023, 62, e202302877.
8. B.-H. Lee, E. Gong, M. Kim, S. Park, H. R. Kim, J. Lee, E. Jung, C. W. Lee, J. Bok, Y. Jung, Y. S. Kim, K.-S. Lee, S.-P. Cho, J.-W. Jung, C.-H. Cho, S. Lebègue, K. T. Nam, H. Kim, S.-I. In and T. Hyeon, Energy Environ. Sci., 2022, 15, 601–609.
9. Y. Zhu, S. F. Yuk, J. Zheng, M.-T. Nguyen, M.-S. Lee, J. Szanyi, L. Kovarik, Z. Zhu, M. Balasubramanian, V.-A. Glezakou, J. L. Fulton, J. A. Lercher, R. Rousseau and O. Y. Gutiérrez, J. Am. Chem. Soc., 2021, 143, 5540–5549.
10. S. Chen, B. Wang, J. Zhu, L. Wang, H. Ou, Z. Zhang, X. Liang, L. Zheng, L. Zhou, Y.-Q. Su, D. Wang and Y. Li, Nano Lett., 2021, 21, 7325–7331.
11. Y. Guo, S. Mei, K. Yuan, D.-J. Wang, H.-C. Liu, C.-H. Yan and Y.-W. Zhang, ACS Catal., 2018, 8, 6203–6215.
12. M.-M. Millet, G. Algara-Siller, S. Wrabetz, A. Mazheika, F. Girgsdies, D. Teschner, F. Seitz, A. Tarasov, S. V. Levchenko, R. Schlögl and E. Frei, J. Am. Chem. Soc., 2019, 141, 2451–2461.
13. B. Han, X. Ou, Z. Deng, Y. Song, C. Tian, H. Deng, Y. Xu and Z. Lin, Angew. Chem., Int. Ed., 2018, 130, 17053–17057.
14. F. J. Caparrós, L. Soler, M. D. Rossell, I. Angurell, L. Piccolo, O. Rossell and J. Llorca, ChemCatChem, 2018, 10, 2365–2369.
15. B. Wu, X. Yu, M. Huang, L. Zhong and Y. Sun, Chin. J. Chem. Eng., 2022, 43, 62–69.
16. Y. Wang, H. Arandiyan, J. Scott, K.-F. Aguey-Zinsou and R. Amal, ACS Appl. Energy Mater., 2018, 1, 6781–6789.
17. S. Dong, W. Liu, S. Liu, F. Li, J. Hou, R. Hao, X. Bai, H. Zhao, J. Liu and L. Guo, Mater. Today Nano, 2022, 17, 100157.
18. Y. Wang, Z. Chen, P. Han, Y. Du, Z. Gu, X. Xu and G. Zheng, ACS Catal., 2018, 8, 7113–7119.
19. J. Liang, J. Lin, J. Liu, X. Wang, T. Zhang and J. Li, Angew. Chem., Int. Ed., 2020, 59, 12868–12875.
20. X. Chen, C. Qu, Y. Xiao, W. Wang, J. Zhang, X. Zheng and Q. Ye, Catal. Today, 2024, 434, 114687.
21. X. Sun, J. Lin, Y. Zhou, L. Li, Y. Su, X. Wang and T. Zhang, AIChE J., 2017, 63, 4022–4031.
22. L. Sun, J. Xu, X. Liu, B. Qiao, L. Li, Y. Ren, Q. Wan, J. Lin, S. Lin, X. Wang, H. Guo and T. Zhang, ACS Catal., 2021, 11, 5942–5950.
23. J. Lin, A. Wang, B. Qiao, X. Liu, X. Yang, X. Wang, J. Liang, J. Li, J. Liu and T. Zhang, J. Am. Chem. Soc., 2013, 135, 15314–15317.
24. L. Sun, L. Cao, Y. Su, C. Wang, J. Lin and X. Wang, Appl. Catal., B, 2022, 318, 121841.
25. Y. Li, Z. Zhao, W. Lu, H. Zhu, F. Sun, B. Mei, Z. Jiang, Y. Lyu, X. Chen, L. Guo, T. Wu, X. Ma, Y. Meng and Y. Ding, Appl. Catal., B, 2023, 324, 122298.
26. W. Zhang, A. Vidal-López and A. Comas-Vives, Front. Chem., 2023, 11, 1144189.
27. J. Li, L. Sun, Q. Wan, J. Lin, S. Lin and X. Wang, J. Phys. Chem. Lett., 2021, 12, 11415–11421.
28. P. Liu, Y. Zhao, R. Qin, S. Mo, G. Chen, L. Gu, D. M. Chevrier, P. Zhang, Q. Guo, D. Zang, B. Wu, G. Fu and N. Zheng, Science, 2016, 352, 797–800.
29. X. Xie, C. He, B. Li, Y. He, D. A. Cullen, E. C. Wegener, A. J. Kropf, U. Martinez, Y. Cheng, M. H. Engelhard, M. E. Bowden, M. Song, T. Lemmon, X. S. Li, Z. Nie, J. Liu, D. J. Myers, P. Zelenay, G. Wang, G. Wu, V. Ramani and Y. Shao, Nat. Catal., 2020, 3, 1044–1054.
30. G. Kyriakou, M. B. Boucher, A. D. Jewell, E. A. Lewis, T. J. Lawton, A. E. Baber, H. L. Tierney, M. Flytzani-Stephanopoulos and E. C. H. Sykes, Science, 2012, 335, 1209–1212.
31. M. Tong, F. Sun, Y. Xie, Y. Wang, Y. Yang, C. Tian, L. Wang and H. Fu, Angew. Chem., Int. Ed., 2021, 60, 14005–14012.
32. C. Riley, S. Zhou, D. Kunwar, A. De La Riva, E. Peterson, R. Payne, L. Gao, S. Lin, H. Guo and A. Datye, J. Am. Chem. Soc., 2018, 140, 12964–12973.
33. M. F. Camellone and S. Fabris, J. Am. Chem. Soc., 2009, 131, 10473–10483.
34. Z. Guo, J. Zhang, Y. Luo, D. Li, R. Zhao, Y. Huang, H. Ren and X. Yao, Nanotechnology, 2023, 34, 285101.
35. C. Zhang, A. Michaelides, D. A. King and S. J. Jenkins, J. Phys. Chem. C, 2009, 113, 6411–6417.
36. C. Zhang, A. Michaelides, D. A. King and S. J. Jenkins, J. Am. Chem. Soc., 2010, 132, 2175–2182.
37. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953–17979.
38. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
39. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
40. S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys and A. P. Sutton, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 57, 1505–1509.
41. M. Cococcioni and S. De Gironcoli, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 035105.
42. M. Nolan, S. Grigoleit, D. C. Sayle, S. C. Parker and G. W. Watson, Surf. Sci., 2005, 576, 217–229.
43. C. W. M. Castleton, J. Kullgren and K. Hermansson, J. Chem. Phys., 2007, 127, 244704.
44. G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 1758–1775.
45. G. Henkelman, B. P. Uberuaga and H. Jónsson, J. Chem. Phys., 2000, 113, 9901–9904.
46. H. Jónsson, G. Mills and K. W. Jacobsen, Classical and Quantum Dynamics in Condensed Phase Simulations, World Scientific, 1998, pp. 385–404.
47. G. Henkelman and H. Jónsson, J. Chem. Phys., 1999, 111, 7010–7022.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5cp00468c

‡ Denotes equal contributions.

---