Spatial decoupling of CH₄ oxidation and CO₂ reduction enables near-stoichiometric dry reforming of methane

Abstract

The practical application of dry reforming of methane (DRM) is hindered by catalyst deactivation, primarily due to the deviation of the ideal 1 : 1 H₂ : CO stoichiometry for competitive CH₄ and CO₂ adsorption/activation. Excessive CH₄ decomposition results in H₂ : CO > 1 with carbon deposition, while predominant CO₂ chemisorption leads to H₂ : CO < 1 with the favorable reverse water-gas shift (RWGS) side reaction. Herein, we demonstrate an *O-migration coupling strategy on Pt/CeO₂ featuring Pt clusters and frustrated Lewis pairs (FLPs, consisting of two Ce³⁺ and one lattice oxygen) to achieve near-stoichiometric and durable DRM. The FLP sites on the CeO₂ support, independent of Pt–CeO₂ interfaces, reduce CO₂ to CO while generating *O species. These *O species migrate to Pt clusters, driving the partial CH₄ oxidation. Through this *O-migration-enabled spatial decoupling of CO₂ reduction and CH₄ oxidation, the catalyst delivers a near-stoichiometric H₂ : CO ratio of 0.99 and an unprecedented CH₄ conversion rate of 93.9 mol g_Pt⁻¹ h⁻¹ at 700 °C. Moreover, stable performance is exhibited for over 400 h, with a turnover number exceeding 7 200 000. This work establishes oxygen migration coupling as a potential strategy for spatially decoupled redox catalysis beyond DRM.

---

Introduction

Syngas (CO/H₂), a pivotal platform for chemical synthesis, was projected to reach a global production capacity exceeding 290.9 million Nm³ per h in 2025.¹ Dry reforming of methane (DRM) represents a particularly promising route for syngas production, offering the dual advantages of reducing carbon emissions and supplying valuable feedstock.^2–6 However, using conventional catalysts, competitive adsorption/activation of CH₄ and CO₂ at the catalyst surface invariably leads to preferential conversion of one reactant, with the H₂ : CO ratio deviating from unity.^7–11 As illustrated in Fig. 1a, excessive CH₄ adsorption and decomposition lead to elevated H₂ : CO ratios (>1) and severe carbon deposition, causing catalyst deactivation.^7,12,13 Conversely, predominant CO₂ chemisorption triggers the reverse water-gas shift (RWGS) reaction, resulting in sub-stoichiometric H₂ : CO ratios (<1) and compromised hydrogen utilization.^11,14–17 Therefore, the rational design of catalysts that can balance CH₄ and CO₂ activation is essential to realize near-stoichiometric DRM, enabling durability and industrial-scale application.

Fig. 1 Spatially decoupled CO₂ reduction and CH₄ oxidation. (a) Schematic diagram highlighting the inherent limitations of conventional DRM catalysts. (b) Schematic diagram of chemical looping DRM, achieving separation of CO₂ reduction and CH₄ oxidation. (c) Conceptual illustration of the spatially decoupled CH₄ oxidation and CO₂ reduction on the Pt clusters and FLP sites, respectively, mediated through oxygen spillover.

A temporal decoupling strategy, chemical looping DRM, has been developed to address competitive adsorption by separating CH₄ oxidation and CO₂ reduction into alternating steps over oxygen-storage materials (Fig. 1b).^18–20 This approach effectively suppresses coking formation and avoids RWGS, enabling near-stoichiometric H₂ : CO ratios (∼1). Despite this promise, the practical implementation of chemical looping DRM suffers from inherent drawbacks, including energy-intensive temperature swings, reliance on inert gas purging for phase separation, and sluggish kinetics of oxygen mobility in oxygen storage materials, collectively hindering scalability and operational flexibility.

Herein, we report a rationally designed catalyst featuring dual-active sites of Pt clusters and frustrated Lewis pairs (FLPs) (Fig. 1c), where CH₄ oxidation and CO₂ reduction proceed concurrently yet remain spatially segregated on distinct catalytic sites, subsequently coupling through *O migration across the CeO₂ support. The FLP sites on CeO₂(110), composed of adjacent Ce³⁺ Lewis acids paired with a neighbouring lattice O²⁻ Lewis base (Fig. 1c),²¹ exhibit exceptional CO₂ adsorption and activation, outperforming conventional oxygen vacancies and metal–CeO₂ interfaces. This enables spatial decoupling of CO₂ reduction from CH₄ dissociation on Pt clusters. Mechanistically, CO₂ reduction at FLP sites directly generates CO and reactive *O species; the *O then migrates rapidly across the CeO₂ surface to Pt clusters, where it drives the partial oxidation of CH₄ to H₂ and CO (Fig. 1c). This dynamic coupling affords a near-stoichiometric H₂ : CO ratio of 0.99, which in turn delivers exceptional catalytic stability (>400 h) with a turnover number exceeding 7 200 000. Moreover, the catalyst achieves a remarkable CH₄ conversion rate of 93.9 mol g_Pt⁻¹ h⁻¹ at 700 °C, surpassing recently reported state-of-the-art systems.

Results and discussion

Theoretical investigation and catalytic performance

Ceria (CeO₂) serves as an ideal oxygen storage component owing to its exceptional redox properties and defect-engineered surface chemistry, which facilitates facile oxygen mobility within its lattice.^22–24 However, the nonpolar C–H bonds of CH₄ resist activation on the Lewis acid/base pairs of CeO₂, necessitating metal sites where d-orbitals facilitate the requisite electron transfer for C–H scission.^17,25–28 In contrast, CO₂ adsorption is efficiently promoted at the metal–CeO₂ interface.^29–31 To achieve spatially decoupled catalysis, we reasoned that the CeO₂ supports must possess sites with exceptionally strong and selective affinity for CO₂, thereby isolating its activation from CH₄ oxidation on metal sites and preventing *H intermediates from triggering the deleterious RWGS reaction.

Density functional theory (DFT) calculations were employed to guide this design. As shown in Fig. 2a and S1–S5, CO₂ adsorption at the oxygen vacancy (O_V) on CeO₂(110) (CeO₂(110)-O_V) is significantly stronger than that on CeO₂(111)-O_V, with an adsorption energy of −2.51 eV. Although this value is more negative than that for CO₂ adsorption on Pt₆ clusters, it remains comparable to that at the Pt₆–CeO₂ interface, indicating that an isolated O_V is insufficient to achieve the full spatial decoupling. Notably, constructing adjacent O_V pairs on CeO₂(110) creates the FLP sites (Ce³⁺…Ce³⁺, O²⁻; Fig. 2b), which dramatically enhance CO₂ adsorption, with an energy of −3.30 eV (Fig. 2a). In contrast, creating more O_Vs on CeO₂(111) merely increases the number of O_Vs without creating a distinct adsorption site (Fig. 2b). These results confirm that a FLP on CeO₂(110), coupled with Pt clusters, provides an ideal platform for spatially decoupled catalysis toward efficient and stoichiometric DRM.

Fig. 2 Theoretical investigation and catalytic performance. (a) The CO₂ adsorption behaviors on the CeO₂(110) surface with various active sites. (b) Optimized structure of one O_V and two adjacent O_Vs on CeO₂(110) and CeO₂(111) surfaces. (c) HAADF-STEM image of Pt_cluster/CeO₂-FLP. (d) Comparison of CH₄ conversion rates over Pt_cluster/CeO₂-FLP and other state-of-the-art catalysts (see Table S2 for more details). (e) Comparison of CH₄ conversion rates and H₂ : CO ratios over Pt_cluster/CeO₂-FLP and other state-of-the-art catalysts under similar reaction conditions (see Table S2 for more details).

Guided by the above analysis, we synthesized porous CeO₂ nanorods (denoted as CeO₂-FLP) designed to host FLP sites through a sequential low- and high-pressure hydrothermal method.^21,32 Dark-field transmission electron microscopy (TEM) images revealed a well-defined porous architecture with an average pore size of 2–3 nm (Fig. S6a). High-resolution TEM showed lattice fringes with a lattice fringe spacing of 0.19 nm, corresponding to the CeO₂ [220] planes (Fig. S6b), indicating the preferential growth along the [110] direction. X-ray photoelectron spectroscopy (XPS) analysis of Ce 3d and O 1s peaks indicated the abundance of surface defects in CeO₂-FLP, as evidenced by the Ce³⁺ (29.4%, Fig. S7a) and Ce³⁺-O (52.4%, Fig. S7b) fractions. The abundance of O_V, along with the exposure of (110) facets, suggests the effective formation of FLP sites on the CeO₂-FLP surface.^21,33,34

Subsequently, Pt clusters were deposited on CeO₂-FLP (Pt_cluster/CeO₂-FLP) using H₂PtCl₆·6H₂O as the precursor via a photo-assisted reduction process. The actual Pt loading was quantified to be 0.9 wt% using inductively coupled plasma optical emission spectroscopy (ICP-OES). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of Pt_cluster/CeO₂-FLP revealed brightness variation on CeO₂-FLP (Fig. 2c), indicating the presence of Pt clusters with an average size of 0.9 ± 0.1 nm (Fig. S8). CO chemical adsorption determined a Pt dispersion of 43.2% (Table S1). Importantly, Pt_cluster/CeO₂-FLP retained high Ce³⁺ (30.9%) and Ce³⁺-O (55.2%) fractions (Fig. S7), confirming the preservation of the FLP-rich surface.

The DRM performances of Pt_cluster/CeO₂-FLP were evaluated in a fixed-bed reactor at a feed ratio of CH₄ : CO₂ : N₂ (2 : 2 : 1). At 700 °C, Pt_cluster/CeO₂-FLP achieved a CH₄ conversion of 75.7% (Fig. S9), approaching the thermodynamic equilibrium (76.0%, at 700 °C), along with a CH₄ conversion rate of 93.9 mol g_Pt⁻¹ h⁻¹. This activity substantially exceeds that of state-of-the-art catalysts (Fig. 2d and Table S2). Importantly, the H₂ and CO production rates reached 187.7 mol g_Pt⁻¹ h⁻¹ and 189.5 mol g_Pt⁻¹ h⁻¹, respectively, giving a near-stoichiometric H₂ : CO ratio of 0.99. Therefore, Pt_cluster/CeO₂-FLP delivers an unprecedented combination of high activity and ideal product stoichiometry for DRM (Fig. 2e and Table S2).

Critical functions of the FLP sites

As a catalyst design concept (Fig. 1c), the FLP sites on the CeO₂(110) surface, rather than Pt clusters or Pt–CeO₂ interfaces, are responsible for the strong CO₂ adsorption and activation. This capability is absent on the CeO₂(111) facet, theoretically leading to inferior DRM performance. We synthesized nano-octahedral CeO₂ (denoted as CeO₂-O_V) exclusively exposing (111) facets, which are structurally incompatible with FLP formation and thus possess only conventional O_V (Fig. 2b and S10).²¹ Consistent with this morphology, CeO₂-O_V exhibited a lower concentration of surface defects, with Ce³⁺ and Ce³⁺-O_V fractions of 19.3% and 40.3%, respectively (Fig. S11 and Table S1). Using the same photo-assisted deposition method, Pt clusters were loaded onto CeO₂-O_V (Pt_cluster/CeO₂-O_V) at 1.1 wt% loading. HAADF-STEM image confirmed the Pt clusters with an average size of 0.9 ± 0.3 nm (Fig. 3a and S12). Meanwhile, Pt deposition did not substantially alter the surface defect of CeO₂-O_V, with Ce³⁺ and Ce³⁺-O fractions remaining at 20.4% and 42.4%, respectively (Fig. S11 and Table S1). Under identical reaction conditions, Pt_cluster/CeO₂-O_V, despite possessing comparable Pt clusters, exhibited markedly inferior DRM performance. Specifically, CH₄ conversion was 35.1% (Fig. 3b), 2.2-fold lower than that of Pt_cluster/CeO₂-FLP (75.7%) at 700 °C. Meanwhile, the H₂ : CO ratio plummeted to 0.5, indicating severe deviation from stoichiometric product ratios (Fig. 3b). These comparative results unequivocally establish that the FLP sites are essential for achieving exceptional DRM activity and stoichiometry.

Fig. 3 Investigations on the critical functions of FLP. (a) HAADF-STEM image of Pt_cluster/CeO₂-O_V. (b) CH₄ conversions and H₂/CO ratios at 700 °C with a WHSV of 30 000 mL g_cat⁻¹ h⁻¹. (c) Kinetic measurements of Pt_cluster/CeO₂-FLP and Pt_cluster/CeO₂-O_V. In situ DRIFTS spectra of (d) Pt_cluster/CeO₂-FLP and (e) Pt_cluster/CeO₂-O_V under CO₂ flow. In situ DRIFTS spectra of (f) Pt_cluster/CeO₂-FLP and (g) Pt_cluster/CeO₂-O_V under CH₄ flow after CO₂ flow. Note: all of the catalysts were pre-treated by Ar flow for 30 min. Initially, a flow of 50 vol.% CO₂/Ar was introduced and the DRIFTS signals were collected at 350 °C every 1 min. Subsequently, the 50 vol.% CO₂/Ar gas was switched to a flow of 50 vol.% CH₄/Ar. The DRIFTS signals were further collected at 350 °C every 1 min. (h) Pulsed reactions of CO₂ and CH₄ over Pt_cluster/CeO₂-FLP and Pt_cluster/CeO₂-O_V.

Kinetic analysis at low conversions (<20%) further elucidates the critical role of FLP. The CO₂ reaction order over Pt_cluster/CeO₂-FLP was significantly lower than that over Pt_cluster/CeO₂-O_V (Fig. 3c), indicating higher CO₂ surface coverage on the FLP-containing catalyst. Arrhenius plots (ln k vs. 1/T, Fig S13 and S14) revealed that the activation energy (E_a) for CO₂ conversion on CeO₂-FLP (59.6 kJ mol⁻¹) was substantially lower than that on CeO₂-O_V (79.1 kJ mol⁻¹). Thus, the higher CO₂ coverage on Pt_cluster/CeO₂ results in the stronger conversion capacity, directly distinguishing FLP from O_Vs. This trend was further supported by CO₂ temperature-programmed desorption (CO₂-TPD, Fig. S15), in which CeO₂-FLP displayed an additional strong desorption peak at ∼500 °C, confirming enhanced CO₂ adsorption at the FLP sites.

More importantly, introducing Pt clusters onto CeO₂-O_V significantly reduced the E_a for CO₂ conversion from 79.1 kJ mol⁻¹ (CeO₂-O_V) to 71.5 kJ mol⁻¹ (Pt_cluster/CeO₂-O_V, Fig. S14). In the absence of FLP sites on CeO₂ supports, CO₂ adsorption and activation are influenced by Pt or the Pt–CeO₂ interface, consistent with previous reports. Conversely, depositing Pt onto CeO₂-FLP exhibited a negligible effect on the E_a for CO₂ conversion, as evidenced by the nearly identical values for Pt_cluster/CeO₂-FLP (55.7 kJ mol⁻¹) and CeO₂-FLP (59.6 kJ mol⁻¹; Fig. S14). Different form conventional O_V sites, these kinetic results experimentally demonstrate that CO₂ activation occurs predominantly on FLP sites rather than at Pt clusters or the Pt–O–Ce interfaces, with Pt clusters playing a negligible direct role in CO₂ transformation.

Subsequently, in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was employed to monitor the roles of FLP under the given reaction conditions. Both Pt_cluster/CeO₂-FLP (Fig. 3d) and Pt_cluster/CeO₂-O_V (Fig. 3e) exhibited characteristic vibrational signatures of bicarbonate (*HCO₃, at 1602 cm⁻¹) and formate (*HCOO, 1505 cm⁻¹) intermediates, confirming the CO₂ chemisorption on CeO₂ supports.^35,36 However, Pt_cluster/CeO₂-FLP exhibited strong capacity of transformation of b-CO₃²⁻, as evidenced from their much weaker adsorption peaks at ∼1300 cm⁻¹. Furthermore, Pt_cluster/CeO₂-FLP displayed an obvious *CO adsorption band (∼2046 cm⁻¹) associated with Pt clusters, which was not detected on Pt_cluster/CeO₂-O_V. This critical distinction confirms that CeO₂-FLP enables a direct CO₂-to-CO reduction pathway.^37,38 Moreover, during CO₂ exposure on Pt_cluster/CeO₂-FLP, the Ce³⁺-OH stretching mode (∼3656 cm⁻¹) attenuated while Ce⁴⁺-OH features (∼3733 cm⁻¹ and ∼3698 cm⁻¹) intensified (Fig. 3d and e),^39–41 indicating the accumulation of *O species within the CeO₂-FLP lattice during the CO₂-to-CO reduction, accompanied by the Ce³⁺-to-Ce⁴⁺ oxidation.

When a CH₄ flow was introduced into the in situ reactor, Pt_cluster/CeO₂-FLP exhibited a significantly enhanced intensity of *CO adsorption on Pt clusters, indicating that the CH₄ oxidation occurred on the Pt sites (Fig. 3f). Meanwhile, the vibrational features corresponding to the *CH₃ (∼1342 cm⁻¹), *CH₃O (∼1471 cm⁻¹), *CH₄O (∼1583 cm⁻¹) and Pt-bound *HCOO (∼1517 cm⁻¹) species emerged during CH₄ flow, along with the disappearance of CeO₂-bound *HCOO bonding (∼1509 cm⁻¹) and *CO₃²⁻ (∼1397 cm⁻¹) intermediates. Simultaneously, the Ce⁴⁺ species (Ce⁴⁺-OH, 3733 cm⁻¹) was reduced to Ce³⁺ (Ce³⁺-OH, 3658 cm⁻¹) along with these transformations. Thus, these findings collectively demonstrate that CH₄ oxidation is mediated by lattice *O species originating from CeO₂-FLP supports.

In contrast, Pt_cluster/CeO₂-O_V exhibited negligible activity for this reaction pathway (Fig. 3g). The CH₄ exposure on Pt_cluster/CeO₂-O_V induced the rapid emergence of *CH₃ (at ∼1342 cm⁻¹) and *CH₄ (at ∼1557 cm⁻¹) species, indicating that CH₄ adsorption and activation occurred on Pt sites.^42,43 However, CO production remained consistently low, demonstrating limited further oxidation of activated *CH₃ and *CH₄ species to *CO. Furthermore, the partial retention of Ce⁴⁺ species (Ce⁴⁺-OH, 3733 cm⁻¹) indicated the restricted *O migration during CH₄ flow, owing to the poor capacity for CO₂ activation/reduction of CeO₂-O_V. Concomitantly, the disappearance of *HCO₃, *HCOO, and b-CO₃^2- species could be attributed to the *H spillover from Pt into the CeO₂-O_V supports.

Sequential CO₂/CH₄ pulse experiments further quantified the FLP functions. Pt_cluster/CeO₂-FLP showed distinct CO₂ consumption over 14 pulse cycles, corresponding to an *O storage capacity of 0.136 mmol g_cat⁻¹. This observation contrasted sharply with Pt_cluster/CeO₂-O_V, which exhibited a negligible *O storage capacity (0.008 mmol g_cat⁻¹), consistent with in situ DRIFTS spectra confirming only adsorption of CO₂ on CeO₂-O_V (Fig. 3h). Subsequent CH₄ pulses revealed drastically higher consumption on Pt_cluster/CeO₂-FLP (0.129 mmol_CH4 g_cat⁻¹) than that on Pt_cluster/CeO₂-O_V (0.023 mmol_CH4 g_cat⁻¹). The close match between CH₄ consumption and *O storage on Pt_cluster/CeO₂-FLP confirms efficient transfer of the FLP-generated *O species to Pt sites for CH₄ oxidation.

Investigation into the functions of Pt clusters

Notably, the pronounced reduction in E_a for CH₄ conversion upon introducing Pt clusters (Pt_cluster/CeO₂-FLP and CeO₂-FLP, 64.1 kJ mol⁻¹ and 112.6 kJ mol^—1, respectively; Pt_cluster/CeO₂-O_V and CeO₂-O_V, 79.2 kJ mol⁻¹ and 117.2 kJ mol^—1, respectively; Fig. S14) unambiguously underscores the essential roles of Pt in CH₄ adsorption and C–H bond activation. Moreover, the comparable CH₄ reaction orders observed on both Pt_cluster/CeO₂-FLP and Pt_cluster/CeO₂-O_V indicated that CH₄ activation kinetics were weakly influenced by CeO₂ (Fig. 3c). Taken together, the higher reaction orders and activation energies associated with CH₄ conversion indicated that CH₄ activation constitutes the rate-determining step in the DRM reaction over Pt_cluster/CeO₂-FLP, particularly under preconditions where CO₂ is efficiently activated by FLP.

To further decipher the mechanistic roles of Pt species in CH₄ activation, the single-atom Pt and Pt nanoparticles (average size 2.2 ± 0.1 nm) were deposited on CeO₂-FLP, yielding Pt₁/CeO₂-FLP (Fig. S16) and Pt_NP/CeO₂-FLP (Fig. S17), respectively. ICP-OES determined Pt loadings of 0.5 wt% for Pt₁/CeO₂-FLP and 1.0 wt% for Pt_NP/CeO₂-FLP. Notably, the surface properties of the supports in Pt₁/CeO₂-FLP and Pt_NP/CeO₂-FLP were similar to those of Pt_cluster/CeO₂-FLP (Fig. S18 and Table S1), enabling the systematic investigations of Pt speciation effects on CH₄ activation while excluding the influences of supports.

The Pt L₃ edge X-ray absorption near-edge structure (XANES) was used to investigate the electronic states of Pt. The white line peak of Pt₁/CeO₂-FLP was located at 11 567.4 eV, close to that of PtO₂ (Fig. 4a). The k³-weight Fourier transforms of extended X-ray absorption fine structure (EXAFS) spectra of Pt₁/CeO₂-FLP delivered one prominent peak at ∼1.64 Å, which was labeled as the Pt–O bond (Fig. 4b and S19). The lack of Pt–Pt coordination confirmed the atomically dispersed Pt supported on CeO₂-FLP. The wavelet transform analysis directly revealed the absence of Pt–Pt bonds in Pt₁/CeO₂-FLP (Fig. 4c). In contrast, Pt_cluster/CeO₂-FLP and Pt_NP/CeO₂-FLP exhibited the white line peaks between Pt foil and PtO₂ (Fig. 4a), indicating the mixed valence states of Pt. Meanwhile, both Pt–Pt and Pt–O bonds were clearly observed from the k³-weight Fourier transforms of EXAFS spectra. Compared to Pt_NP/CeO₂-FLP, the higher white line peak revealed more amount of the Pt–O bond in Pt_cluster/CeO₂-FLP, which could be clearly observed from the wavelet transform analysis. Quantitatively, Pt₁/CeO₂-FLP, Pt_cluster/CeO₂-FLP and Pt_NP/CeO₂-FLP exhibited the Pt–O fractions of 100%, 45.4% and 24.2%, respectively, as well as the Pt–Pt fractions of 0%, 54.6% and 75.8%, respectively (Table 1).

Fig. 4 Investigation on the functions of Pt clusters. (a) XANES spectra of Pt foil, Pt_cluster/CeO₂-FLP, Pt₁/CeO₂-FLP, Pt_NP/CeO₂-FLP and PtO₂. (b) The k³ weighted Fourier-transformed spectra derived from the EXAFS spectra. (c) Wavelet-transform plots by using the k³ space. (d) Direct correlation between Pt–O bond contents and CH₄ conversion rates/H₂ : CO ratios. (e) Plots of Pt–O bond contents vs. calculated E_a values of Pt_cluster/CeO₂-FLP, Pt₁/CeO₂-FLP, and Pt_NP/CeO₂-FLP for DRM. In situ DRIFTS spectra of Pt_NP/CeO₂-FLP under (f) CO₂ flow and (g) CH₄ flow after CO₂ flow. Note: all of the catalysts were pre-treated by a flow of Ar for 30 min. Then, a flow of 50 vol.% CO₂/Ar was introduced and the DRIFTS signals were collected at 350 °C every 1 min. Subsequently, the 50 vol.% CO₂/Ar gas was switched to a flow of 50 vol.% CH₄/Ar. The DRIFTS signals were further collected at 350 °C every 1 min. (h) In situ mass spectrometry analysis of the temperature-programmed CH₄ oxidation of Pt_cluster/CeO₂-FLP upon pre-treatment with H₂¹⁸O.

Table 1 Fitting results of EXAFS for Pt₁/CeO₂-FLP, Pt_cluster/CeO₂-FLP and Pt_NP/CeO₂-FLP^a

Sample	Paths	N^b	R (Å)^c	σ^2 (×10^−3 Å^2)^d	S0^2	E0 (eV)	R Factor
a The data ranges used in the fit are 2.0 ≤ k ≤ 13.0 Å^−1 and 1.2 ≤ R ≤ 3.5 Å, and depend on the quality of data. The number of fitted variable parameters was 10, which was smaller than the total number of independent data points, approximately 16.1. R-Factors for these fittings are all below 0.02.b Average coordination number. The half path length.c The paths for Pt–O, Pt–Pt are from the crystal structure of PtO2 (P63mc) and Pt (Fm3̄m).d Debye–Waller factor.
Pt foil	Pt–Pt	12.0	2.77	—	—	—	—
PtO2	Pt–O	6.0	2.07	—	—	—	—
Pt1/CeO2-FLP	Pt–O	3.4 ± 0.1	2.04 ± 0.07	5.2 ± 0.9	0.8	7.4	0.018
Ptcluster/CeO2-FLP	Pt–O	4.4 ± 0.4	2.07 ± 0.02	2.1 ± 0.2	0.8	8.5	0.012
	Pt–Pt	5.3 ± 0.5	2.82 ± 0.16	5.2 ± 0.1
PtNP/CeO2-FLP	Pt–O	3.0 ± 0.7	2.10 ± 0.03	7.0 ± 2.9	0.8	8.7	0.008
	Pt–Pt	9.4 ± 0.2	2.78 ± 0.14	5.8 ± 0.6

---

For the DRM reaction, Pt_cluster/CeO₂-FLP, featuring the co-existence of Pt–O and Pt–Pt bonds, delivered the highest CH₄ conversion rate compared to Pt₁/CeO₂-FLP and Pt_NP/CeO₂-FLP (Fig. 4d and S20), as well as the lowest E_a of 64.1 kJ mol⁻¹ for CH₄ conversion (Fig. 4e and S21). These observations confirm that the highest intrinsic activity of Pt clusters originates from a synergistic interplay between Pt–Pt and Pt–O bonds in CH₄ activation. This synergy arises from the presence of coordinatively unsaturated metal and oxygen on the Pt surface, which promotes the formation of the adsorbed CH₄ σ-complexes and then facilitates C–H bond cleavage in the complexes.^25,44 Furthermore, due to the similar surface properties of supports, all Pt/CeO₂-FLP catalysts exhibited comparable E_a values for CO₂ conversion (Fig. 4e), further confirming the strong capacity of FLP for CO₂ adsorption/activation and the negligible influence of Pt active sites on this process.

The CH₄ activation on Pt/CeO₂-FLP was further investigated by CH₄ temperature-programmed reduction (CH₄-TPR). After the pre-treatment with Ar purging, a CH₄ flow was introduced to probe the catalyst surface. The CeO₂-FLP supports alone exhibited no detectable activation peaks (Fig. S22), further confirming that the CH₄ adsorption and activation occur exclusively at Pt sites. Importantly, Pt_cluster/CeO₂-FLP exhibited the strongest peak intensity and the lowest initiation temperature compared to both Pt₁/CeO₂-FLP and Pt_NP/CeO₂-FLP (Fig. S22). These observations directly confirmed the highest capacity of Pt_cluster/CeO₂-FLP for the CH₄ adsorption and activation in DRM.

Subsequently, in situ DRIFTS experiments were performed to compare the behaviors of Pt clusters and nanoparticles under the given reaction conditions. Under CO₂ flow, Pt_NP/CeO₂-FLP exhibited spectral changes nearly identical to those of Pt_cluster/CeO₂-FLP, consistent with their similar CeO₂ surface properties for CO₂ adsorption (Fig. 4f). Upon switching to CH₄ flow, characteristic peaks corresponding to adsorbed *CH₄ (1557 cm⁻¹) appeared on Pt_NP/CeO₂-FLP, confirming CH₄ adsorption (Fig. 4g). Unlike Pt_cluster/CeO₂-FLP, no clear signals for *CH_xO intermediates were detected on Pt_NP/CeO₂-FLP. Moreover, the Ce⁴⁺-OH peak persisted even after 9 min CH₄ flow, indicating restricted *O migration from the support to the Pt nanoparticles. Together with CH₄-TPR results, these observations demonstrate that Pt clusters uniquely promote the migration of *O species, which in turn enhances CH₄ activation and overall DRM activity.

Building on the foregoing analysis, the spatially separated Pt clusters and FLP sites serve distinct roles in CH₄ oxidation and CO₂ reduction, respectively, with *O migration acting as the key coupling step, as illustrated in Fig. 1c. To further probe the origin and transfer of *O species, we performed ¹⁸O-isotope labeling experiments. Because labeled CO₂ can lead to ambiguity from residual adsorption or desorption of ¹⁸O-containing species, we designed the experiment using H₂¹⁸O pretreatment, which readily dissociates on CeO₂ but not on Pt clusters.^32,45,46 Specifically, after pretreating Pt_cluster/CeO₂-FLP with H₂¹⁸O at 150 °C for 30 min, purging with Ar for 30 min was performed to remove physically adsorbed species. The programmed temperature increase of CH₄ oxidation from 200 °C to 750 °C (10 vol.% CH₄/He, 50 mL min⁻¹) revealed the appearance of the C¹⁸O/CO signal (Fig. 4h), which directly demonstrated that *O species stored in the CeO₂-FLP support participate in CH₄ oxidation through oxygen transfer to Pt clusters. Furthermore, the H₂ signal appears simultaneously with the CO/C¹⁸O signals, indicating that CH₄ dissociates on Pt sites, with CO and H₂ formation. Notably, the observed C¹⁸O signal indicates that the oxygen involved in CH₄ oxidation originated predominantly from H₂¹⁸O-derived *¹⁸O species stored in the CeO₂-FLP support rather than from the lattice oxygen. Together, these results strongly support a spatially decoupled reaction pathway in which *O migration couples CO₂ reduction on the support with CH₄ oxidation on Pt clusters.

Activity matching between Pt clusters and FLP sites

Achieving activity balance between the spatially decoupled CO₂ reduction and CH₄ oxidation is critical for high-performance DRM under the *O migration coupling process, as it ensures near stoichiometric H₂ : CO ratios (1 : 1) and long-term catalytic stability. As shown in Fig. 5a, this activity balance is inherently determined by the quantitative relationship between FLP sites (responsible for CO₂ reduction) and Pt clusters (mediating CH₄ oxidation). While the FLP sites remain fixed within the CeO₂-FLP supports, we precisely adjusted the number of Pt centers by controlling the Pt loadings. HAADF-STEM images revealed that increasing Pt loadings from 0.5 wt% to 2.0 wt% enhanced the densities of Pt clusters while preserving their similar cluster dimensions (an average diameter of 0.9–1.0 nm, Fig. S23). This controlled variation in Pt cluster population enabled systematic modulation of the FLP-to-Pt site ratios, a crucial factor governing overall DRM performance.

Fig. 5 Activity matching. (a) Scheme of activity matching between Pt clusters and FLPs for the DRM reaction. (b) Influences of Pt loadings on the CH₄/CO₂ generation rates and H₂ : CO ratios of the Pt_cluster/CeO₂-FLP catalysts for DRM. (c) Influence of Pt loadings on the E_a values of Pt_cluster/CeO₂-FLP for CO₂ and CH₄ conversions. (d) Pulsed reactions of CO₂ and CH₄ over Pt_cluster/CeO₂-FLP with various Pt loadings. (e) Catalytic stability of various catalysts with a WHSV of 30 000 mL g_cat⁻¹ h⁻¹ (CH₄ : CO₂ : N₂ = 2 : 2 : 1) at 700 °C. CO₂ treatment conditions: CO₂/N₂ at 700 °C for 10 h with a WHSV of 18 000 mL g_cat⁻¹ h⁻¹. (f) Comparison of TON and TOF between Pt_cluster/CeO₂-FLP and other state-of-the-art catalysts under similar reaction conditions (see Table S2 for more details).

For DRM, Pt loading significantly influenced reaction kinetics and product stoichiometry. At 0.5 wt% Pt loading, the CH₄ conversion rate of 115.7 mmol g_cat⁻¹ h⁻¹ was lower than the CO₂ conversion rate of 169.7 mmol g_cat⁻¹ h⁻¹, resulting in a significantly reduced H₂ : CO ratio of 0.81 at 700 °C (Fig. 5b and S24). Elevating the Pt loading to 1 wt%, accompanied with a higher density of Pt clusters, drastically improved the enhanced CH₄ conversion rate (405.4 mmol g_cat⁻¹ h⁻¹), matching the CO₂ conversion rate of 405.6 mmol g_cat⁻¹ h⁻¹. Simultaneously, the H₂ : CO ratio approached 0.99, nearly achieving a stoichiometric ratio of 1 (Fig. 5b). Therefore, increasing the Pt loading facilitated the optimal matching between Pt clusters and FLP sites.

Further increasing the Pt loading to 2 wt% resulted in CH₄ and CO₂ conversion rates similar to those observed for Pt_cluster/CeO₂-FLP with 1 wt% Pt loading (Fig. 5b and S24). This observation demonstrated that the CH₄ activation and conversion depended on the migration of *O species generated from the CO₂ reduction on FLP of CeO₂-FLP, rather than the availability of Pt clusters. Consequently, CH₄ conversion rates plateau at the 1 wt% Pt loading, as additional Pt clusters cannot compensate for the limited *O supply from the FLP. The extra CH₄ conversion would lead to carbon deposition on the catalyst surface (Fig. 5a). Carbon balance analysis before and after the reaction revealed greater carbon imbalance for Pt_cluster/CeO₂-FLP with higher Pt loadings, indicating direct carbonization of CH₄ on Pt (Fig. S25). Raman spectroscopy of the spent Pt_cluster/CeO₂-FLP catalysts proved the direct evidence of carbonaceous species (Fig. S26). After 45 h reaction at 700 °C, thermogravimetric analysis (TGA) quantified a carbon deposition rate of 2.69 mg g_cat⁻¹ h⁻¹ for the 2 wt% Pt_cluster/CeO₂-FLP catalysts, which was ∼7 times higher than that for 1 wt% and 0.5 wt% catalysts (0.39 and 0.38 mg g_cat⁻¹ h⁻¹, respectively; Fig. S27). Collectively, these results reveal the critical role of the balanced FLP-to-Pt site ratios in achieving efficient DRM performance with stoichiometric H₂ : CO output.

Kinetic analysis further clarified the activity matching requirements between Pt clusters and FLP sites (Fig. S28). The comparable E_a values of Pt_cluster/CeO₂-FLP for CO₂ conversion indicated that CO₂ adsorption and transformation predominantly occurred on FLP sites (Fig. 5c). Theoretically, similar Pt clusters in Pt_cluster/CeO₂-FLP should yield comparable E_a values for CH₄ conversion. However, when the Pt loading was 0.5 wt%, the relatively large distance between Pt clusters and FLP sites resulted in a higher E_a value for CH₄ conversion. In contrast, sufficient Pt loading (>1.0 wt%) reduces the average spatial distance between Pt clusters and FLP sites, leading to sufficient migration and supply of *O species and thus delivering comparable E_a values for CH₄ conversion. When the number of Pt clusters matched that of FLP sites, the E_a values for CH₄ conversion and CO₂ conversion were also comparable, thereby facilitating the DRM reaction with a near-stoichiometric ratio of H₂ : CO ≈ 1. This kinetic deconvolution highlights the critical roles of the available Pt clusters and FLP sites in balancing *O supply (from FLPs) and CH₄ activation (at Pt) for DRM.

The CO₂ pulse experiments revealed that Pt_cluster/CeO₂-FLP with varying Pt loadings exhibited comparable *O storage capacities (0.129–0.136 mmol g_cat⁻¹, Fig. 5d). These findings further confirm that FLP sites act as active centers for CO₂ reduction, generating reactive *O species that subsequently participate in CH₄ oxidation pathways. Specially, the Pt loading of 0.5 wt% yielded a low rate of 0.043 mmol g_cat⁻¹ due to the insufficient amount and density of Pt clusters. When the Pt loadings were elevated to 1.0 wt% and 2.0 wt%, the CH₄ consumption increased to 0.129 mmol g_cat⁻¹ and 0.138 mmol g_cat⁻¹, respectively. Excessive Pt loading did not enhance DRM activity, as the FLP sites imposed a kinetic bottleneck of the *O supply from the CO₂ reduction.

Notably, the observed CH₄ consumption over Pt_cluster/CeO₂-FLP did not scale with the population of Pt clusters as the Pt loading increased from 1.0 wt% to 2.0 wt%, contradicting the scenario where only interfacial oxygen is involved in the reaction. This deviation indicates that the *O species are not confined to the Pt–CeO₂ interface. Instead, the facile migration of oxygen across the CeO₂-FLP support enables the entire oxygen reservoir of the carrier to participate in CH₄ oxidation. Consequently, these findings further demonstrate that CO₂ activation occurs extensively on the FLP sites of CeO₂-FLP, thereby achieving spatial decoupling from CH₄ oxidation at the Pt clusters.

Additionally, the optimal synergy of Pt clusters and FLP sites directly enhances catalytic stability. Pt_cluster/CeO₂-O_V, lacking FLP sites, exhibited inferior stability, as evidenced by the substantially declined conversions of CH₄ and CO₂ during a period of 45 h (Fig. 5e). Similarly, Pt_cluster/CeO₂-FLP with either insufficient (0.5 wt%) or excessive (2.0 wt%) Pt loading exhibited compromised durability, attributable respectively to limited CH₄ activation capacity and inadequate *O-migration kinetics. In contrast, the optimally matched Pt_cluster/CeO₂-FLP (1 wt% Pt loading) catalyst delivered exceptional long-term stability, maintaining nearly constant CH₄ conversion rates (Fig. 5e) and a near-stoichiometric H₂ : CO ratio (Fig. S29) for over 400 h. During the initial 100 h, when the H₂/CO ratio remained near 1, no detectable water appeared in the product stream, indicating effective suppression of the RWGS reaction. However, XPS analysis of the spent Pt_cluster/CeO₂-FLP catalyst revealed both reduced surface defects and accumulated carbonaceous species (Fig. S30), which together mask Pt clusters and FLP sites. Notably, a slight 15% loss in activity during long-term testing was completely reversed by a simple CO₂ treatment (Fig. 5e). This remarkable stability reflected in a turnover number (TON) exceeding 7 200 000 per exposed Pt site for CH₄ conversion (Fig. 5f and S31). To the best of our knowledge, Pt_cluster/CeO₂-FLP represents the first to simultaneously achieve such a record-high TON and unprecedented activity, with its TON value nearly double the previously reported maximum (Fig. 5f).

Finally, we propose a catalytic process for DRM on Pt_cluster/CeO₂-FLP that spatially decouples CO₂ reduction and CH₄ oxidation, as illustrated in Fig. 1c. The CO₂ reduction step follows a Mars–van Krevelen (MvK) mechanism: CO₂ is adsorbed and activated at FLP sites, directly forming CO and leaving *O species on the CeO₂-FLP support. Subsequently, the *O species migrate from FLP sites to Pt clusters. On Pt clusters, CH₄ is converted to CO and H₂ via *CH_xO intermediates, facilitated by the migrated *O species. Thus, while the individual CO₂ reduction step obeys the classical MvK redox cycle, the overall process represents a modified MvK-type pathway enabled by spatial decoupling and *O migration between distinct active sites. This design separates CO₂ reduction (FLP sites) from CH₄ oxidation (Pt clusters) and couples them through oxygen spillover. Through precise matching of Pt clusters and FLP sites, the two half-reactions are efficiently coupled, leading to a near-stoichiometric H₂/CO ratio and sustained DRM activity.

Conclusions

In summary, this work demonstrates that *O migration across a spatially decoupled Pt/CeO₂ catalyst enables near-stoichiometric DRM under continuous operation, delivering durable and high-performance DRM catalysis with a stoichiometric H₂ : CO ratio. This catalyst separates the antagonistic adsorption and activation steps: CO₂ is selectively reduced at FLP sites on the CeO₂(110) facet, while CH₄ is partially oxidized on Pt clusters, with *O shuttling dynamically between the two sites. This *O-transport mechanism intrinsically couples the two half-reactions, maintains a balanced redox cycle, and effectively maintains a H₂ : CO ratio of 0.99 while suppressing coke formation and deactivation pathways. As a result, the catalyst delivers a record CH₄ conversion rate of 93.9 mol g_Pt⁻¹ h⁻¹ at 700 °C with stable operation exceeding 400 h under continuous, isothermal conditions. Although the present study employs Pt/CeO₂(110) as a model platform, the underlying principle is not limited to this specific combination. The key requirements are: (i) a reducible oxide support capable of forming FLP sites (adjacent oxygen vacancies) that strongly activate CO₂ and (ii) metal sites (clusters or nanoparticles) that can activate CH₄ and accept migrating *O species. Many earth-abundant metals (e.g., Ni, Co, and Ru) and other reducible oxides (e.g., TiO₂, WO₃, and In₂O₃) are known to exhibit similar oxygen spillover behavior and have been reported to form frustrated Lewis pair-like defects under reducing conditions. Therefore, we anticipate that the spatial decoupling strategy can be extended to more practical catalyst compositions, guided by the design principles established here.

Author contributions

W. L. performed most of the experiments. J. Y., W. G. and Q. G. participated in data analysis. S. Z. and Y. Q. designed the studies and wrote the paper. All authors discussed the results and commented on the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting the findings of this study are available in the main text and supplementary information (SI). Supplementary information: experimental details, catalyst synthesis procedures, characterization data, catalytic performance data, kinetic analysis, in situ spectroscopic results, computational details, figures and tables. Additional raw data are available from the corresponding author upon reasonable request. See DOI: https://doi.org/10.1039/d6sc03014a.

Acknowledgements

This work was supported by the National Natural Science Foundation of China Young Student Basic Research Project (Doctoral Student) (225B2209) and the Key Project of the Natural Science Basic Research Program of Shaanxi Province (2024JC-TBZC-16).

Notes and references

1. M. Intelligence, Syngas Market Size, Mordor Intelligence, https://www.mordorintelligence.com/industry-reports/syngas-market, accessed 07 July 2025.
2. C. Palmer, D. C. Upham, S. Smart, M. J. Gordon, H. Metiu and E. W. McFarland, Nat. Catal., 2020, 3, 83–89.
3. Q. Zhu, H. Zhou, L. Wang, L. Wang, C. Wang, H. Wang, W. Fang, M. He, Q. Wu and F.-S. Xiao, Nat. Catal., 2022, 5, 1030–1037.
4. Y. Song, E. Ozdemir, S. Ramesh, A. Adishev, S. Subramanian, A. Harale, M. Albuali, B. A. Fadhel, A. Jamal, D. Moon, S. H. Choi and C. T. Yavuz, Science, 2020, 367, 777–781.
5. Y. Liu, D. Deng and X. Bao, Chem, 2020, 6, 2497–2514.
6. L. Zhou, J. M. P. Martirez, J. Finzel, C. Zhang, D. F. Swearer, S. Tian, H. Robatjazi, M. Lou, L. Dong, L. Henderson, P. Christopher, E. A. Carter, P. Nordlander and N. J. Halas, Nat. Energy, 2020, 5, 61–70.
7. T. Zhou, X. Li, J. Zhao, L. Luo, Y. Wang, Z. Xiao, S. Hu, R. Wang, Z. Zhao, C. Liu, W. Wu, H. Li, Z. Zhang, L. Zhao, H. Yan and J. Zeng, Nat. Mater., 2025, 24, 891–899.
8. S. D. Senanayake, J. A. Rodriguez and J. F. Weaver, Acc. Chem. Res., 2020, 53, 1488–1497.
9. D. Pakhare and J. Spivey, Chem. Soc. Rev., 2014, 43, 7813–7837.
10. 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.
11. L. Sandoval-Diaz, D. Cruz, M. Vuijk, G. Ducci, M. Hävecker, W. Jiang, M. Plodinec, A. Hammud, D. Ivanov, T. Götsch, K. Reuter, R. Schlögl, C. Scheurer, A. Knop-Gericke and T. Lunkenbein, Nat. Catal., 2024, 7, 161–171.
12. M. Akri, S. Zhao, X. Li, K. Zang, A. F. Lee, M. A. Isaacs, W. Xi, Y. Gangarajula, J. Luo, Y. Ren, Y. T. Cui, L. Li, Y. Su, X. Pan, W. Wen, Y. Pan, K. Wilson, L. Li, B. Qiao, H. Ishii, Y. F. Liao, A. Wang, X. Wang and T. Zhang, Nat. Commun., 2019, 10, 5181.
13. H. Wang, G. Cui, H. Lu, Z. Li, L. Wang, H. Meng, J. Li, H. Yan, Y. Yang and M. Wei, Nat. Commun., 2024, 15, 3765.
14. Q. Zhu, H. Zhou, L. Wang, L. Wang, C. Wang, H. Wang, W. Fang, M. He, Q. Wu and F.-S. Xiao, Nat. Catal., 2022, 5, 1030–1037.
15. Y. Tang, Y. Wei, Z. Wang, S. Zhang, Y. Li, L. Nguyen, Y. Li, Y. Zhou, W. Shen, F. F. Tao and P. Hu, J. Am. Chem. Soc., 2019, 141, 7283–7293.
16. C. Palmer, D. C. Upham, S. Smart, M. J. Gordon, H. Metiu and E. W. McFarland, Nat. Catal., 2020, 3, 83–89.
17. H. Lv, X. Dong, R. Li, C. Zeng, X. Zhang, Y. Song, H. Liu, J. Shao, N. Ta, Q. Zhao, Q. Fu, J. Xiao, G. Wang and X. Bao, Nat. Chem., 2025, 17, 695–702.
18. L. C. Buelens, V. V. Galvita, H. Poelman, C. Detavernier and G. B. Marin, Science, 2016, 354, 449–452.
19. L. Zeng, Z. Cheng, J. A. Fan, L.-S. Fan and J. Gong, Nat. Rev. Chem., 2018, 2, 349–364.
20. X. Zhang, Y. Xu, Y. Liu, L. Niu, Y. Diao, Z. Gao, B. Chen, J. Xie, M. Bi, M. Wang, D. Xiao, D. Ma and C. Shi, Chem, 2023, 9, 102–116.
21. S. Zhang, Z. Q. Huang, Y. Ma, W. Gao, J. Li, F. Cao, L. Li, C. R. Chang and Y. Qu, Nat. Commun., 2017, 8, 15266.
22. T. Montini, M. Melchionna, M. Monai and P. Fornasiero, Chem. Rev., 2016, 116, 5987–6041.
23. S. Zhang, Z. Tian, Y. Ma and Y. Qu, ACS Catal., 2023, 13, 4629–4645.
24. J. Graciani, K. Mudiyanselage, F. Xu, A. E. Baber, J. Evans, S. D. Senanayake, D. J. Stacchiola, P. Liu, J. Hrbek, J. F. Sanz and J. A. Rodriguez, Science, 2014, 345, 546–550.
25. Z. Liang, T. Li, M. Kim, A. Asthagiri and J. F. Weaver, Science, 2017, 356, 299–303.
26. Q. Cheng, X. Yao, L. Ou, Z. Hu, L. Zheng, G. Li, N. Morlanes, J. L. Cerrillo, P. Castano, X. Li, J. Gascon and Y. Han, J. Am. Chem. Soc., 2023, 145, 25109–25119.
27. Z. Rao, K. Wang, Y. Cao, Y. Feng, Z. Huang, Y. Chen, S. Wei, L. Liu, Z. Gong, Y. Cui, L. Li, X. Tu, D. Ma and Y. Zhou, J. Am. Chem. Soc., 2023, 145, 24625–24635.
28. M. Filez, V. De Coster, H. Poelman, V. Briois, A. Beauvois, J. Dendooven, M. B. J. Roeffaers, V. Galvita and C. Detavernier, Nat. Catal., 2025, 8, 187–195.
29. S. Kattel, P. Liu and J. G. Chen, J. Am. Chem. Soc., 2017, 139, 9739–9754.
30. S. Navarro-Jaén, M. Virginie, J. Bonin, M. Robert, R. Wojcieszak and A. Y. Khodakov, Nat. Rev. Chem., 2021, 5, 546–579.
31. R. Ye, J. Ding, T. R. Reina, M. S. Duyar, H. Li, W. Luo, R. Zhang, M. Fan, G. Feng, J. Sun and J. Liu, Nat. Synth., 2025, 4, 288–302.
32. S. Zhang, Y. Liu, M. Zhang, Y. Ma, J. Hu and Y. Qu, Nat. Commun., 2022, 13, 5527.
33. S. Zhang, Z. Xia, Y. Zou, F. Cao, Y. Liu, Y. Ma and Y. Qu, J. Am. Chem. Soc., 2019, 141, 11353–11357.
34. J. Ren, Y. Liu, Y. Wang, W. Guo, K. Ma, Y. Qu and S. Zhang, ACS Nano, 2026, 20, 4470–4478.
35. E. Garand, T. Wende, D. J. Goebbert, R. Bergmann, G. Meijer, D. M. Neumark and K. R. Asmis, J. Am. Chem. Soc., 2010, 132, 849–856.
36. S. M. Jin, C. Kwon, A. Bugaev, B. Karakurt, Y. C. Lin, L. Savereide, L. P. Zhong, V. Boureau, O. Safonova, S. Kim and J. S. Luterbacher, Nat. Catal., 2024, 1119–1212.
37. C. Vogt, E. Groeneveld, G. Kamsma, M. Nachtegaal, L. Lu, C. J. Kiely, P. H. Berben, F. Meirer and B. M. Weckhuysen, Nat. Catal., 2018, 1, 127–134.
38. W. Li, J. Gan, Y. Liu, Y. Zou, S. Zhang and Y. Qu, Angew. Chem., Int. Ed., 2023, 62, e202305661.
39. L. F. Bobadilla, J. L. Santos, S. Ivanova, J. A. Odriozola and A. Urakawa, ACS Catal., 2018, 8, 7455–7467.
40. H. Shen, Y. Dong, S. Yang, Y. He, Q. Wang, Y. Cao, W. Wang, T. Wang, Q. Zhang and H. Zhang, Nano Res., 2022, 15, 5831–5841.
41. W. Li, B. Liu, Q. Guo, W. Guo, S. Zhang and Y. Qu, Nat. Commun., 2025, 16, 7335.
42. F. Xue, C. Zhang, C. Cheng, X. Yan, F. Liu, X. Liu, B. Jiang, Q. Zhang, L. Sun, H. Peng, W. H. Huang, C. W. Pao, Z. Hu, M. Chen, D. Su, M. Liu, X. Huang and Y. Xu, Nat. Commun., 2024, 15, 10451.
43. J. Ma, C. Zhu, K. Mao, W. Jiang, J. Low, D. Duan, H. Ju, D. Liu, K. Wang, Y. Zang, S. Chen, H. Zhang, Z. Qi, R. Long, Z. Liu, L. Song and Y. Xiong, Nat. Commun., 2023, 14, 1410.
44. J. F. Weaver, C. Hakanoglu, A. Antony and A. Asthagiri, Chem. Sev. Rev., 2014, 43, 7536–7547.
45. Q. Fu, H. Saltsburg and M. Flytzani-Stephanopoulos, Science, 2003, 301, 935–938.
46. K. Ding, A. Gulec, A. M. Johnson, N. M. Schweitzer, G. D. Stucky, L. D. Marks and P. C. Stair, Science, 2015, 350, 189–192.

---