Light-driven restructuring generates nanoisland NiIr alloy for efficient methane dry reforming

Abstract

The simultaneous utilization of methane and carbon dioxide via dry reforming holds promise for sustainable syngas production, yet conventional thermocatalytic processes suffer from energy-intensive operation and catalyst deactivation. Here, we report a light-driven methane dry reforming strategy utilizing sinter-resistant nano-island alloys catalyst, which are dynamically evolved from partially oxidized NiIr nanoclusters anchored on TiO₂ under photoexcitation. In situ characterization reveals interfacial charge oscillations on the catalyst induce a support-Ni-Ir electron transfer pathway, stabilizing oxidized Ni linkages while electronically modulating surface Ir sites. This dual functionality promotes CH_xO* intermediate formation, suppressing coking during 100-h operation under intermittent illumination. By decoupling photoelectric and photothermal contributions, we demonstrate that localized photogenerated electrons dominate balanced syngas production, whereas photothermal effects enhance molecular vibrations. The optimized catalyst achieves a syngas rate of 10841 mmol g_cat^-1 h^-1 with 25.0% light-to-fuel efficiency, establishing a design paradigm for solar-driven alloy catalysts in greenhouse gas valorization.

Introduction

Mitigating greenhouse gas emissions while securing sustainable routes for chemical energy conversion remains a central challenge in the transition toward a low-carbon future. Dry reforming of methane (DRM), which simultaneously converts methane (CH₄) and carbon dioxide (CO₂) into syngas (CO and H₂) as valuable chemical feedstocks, offers a promising pathway for greenhouse gas utilization^1,2. Yet, the highly endothermic nature of DRM (CH₄ (g) + CO₂ (g) → 2CO (g) + 2H₂ (g), ΔH^o_298K = +247.3 kJ mol^-1) necessitates harsh operating conditions (typically 700–1000 °C) in conventional thermocatalysis. Such high temperatures consume substantial fossil fuels, undermining environmental benefits of CO₂ utilization^3,4,5. In addition, the reverse water-gas shift (RWGS) reaction (CO₂ (g) + H₂ (g) → CO (g) + H₂O (g), ΔH^o_298K = +41.2 kJ mol^-1), being thermodynamically favored under the high-temperature conditions, compromises syngas selectivity by lowering the H₂/CO ratio⁶. Solar light-driven catalysis has emerged as an energy-efficient alternative by synergistically combining photoelectric effects and photothermal heating to promote the dissociation of CH₄ and CO₂ without external heating^7,8,9.

In materials terms, noble metals as active sites display high intrinsic activity toward CH₄ and CO₂ activation but suffer from low atom efficiency and over-strong intermediate binding, often leading to side reactions¹⁰. Nickel-based catalysts are more economically viable and exhibit notable CH₄ activation capacity, which are prone to deactivation via carbon deposition¹¹. Alloying noble metals with Ni modulates the electronic structure to optimize the balance between reactant activation and intermediate adsorption, thereby suppressing side reactions and improving noble metal utilization efficiency. Moreover, ultrafine nanoclusters (<2 nm) could further enhance catalytic efficiency through maximized surface exposure and metal utilization^12,13. However, stabilizing such clusters against aggregation under reaction conditions remains challenging. Recent studies^14,15,16 have demonstrated that integrating oxide nano-islands at metal-support interfaces can effectively suppress sintering through strong charge-mediated interactions. For instance, anchoring Pt¹⁴ or Pd¹⁶ on CeO_x nano-islands has been shown to spatially confine metal mobility. However, extending this nano-island strategy to ultrafine alloy architectures requiring both interfacial stabilization and electronic tuning has yet to be reported.

In addition, catalyst structural stability at the atomic scale under intense light irradiation and elevated temperatures, which are conditions inherent to photothermal catalysis, remains challenging for elucidating catalytic mechanisms. Whether such conditions preserve the active phase or drive transformations from metastable intermediates to thermodynamically favored structures is still poorly understood. Beyond thermal vibrations that commonly induce surface reconstruction at high temperatures, photoexcitation can trigger substantial charge redistribution between the support and active site^17,18,19,20. This photoinduced charge transfer may alter the electronic structure of surface chemical bonds, reducing bond energies and promoting dynamic surface reconstruction^21,22. Despite growing interest, elucidating the nature and formation of the active phase under photothermal conditions as well as disentangling the respective roles of photoelectric and photothermal effects in light-driven DRM, remains a critical yet unresolved challenge, limiting progress toward rational catalyst design.

In this work, we demonstrate that photoinduced interfacial charge oscillations in NiIr/TiO₂ trigger dynamic structural reconstruction into TiO₂-supported NiIr nano-island alloys stabilized by oxidized Ni linkages (NiIr/TiO₂-L_Ni). This structural evolution arises from accelerated electron transfer across the support-Ni-Ir interface under light excitation. The spatially and electronically synergistic configuration promotes CH_xO* intermediate formation, effectively mitigates carbon deposition, and maintains stable performance for 100 h under intermittent illumination. Mechanistic investigations reveal that photoelectric effects dominate the reaction, contributing 57.5% of the syngas production rate and 91.7% of the H₂/CO ratio by facilitating C-H/C = O scission and H-H coupling, while photothermal effects enhance molecular vibrations and reactant diffusion. Under full feed gas flow, NiIr/TiO₂-L_Ni achieves a syngas production rate of 10841 mmol g_cat^-1 h^-1, with a light-to-fuel efficiency (LTFE) of 25.0% and a turnover frequency (TOF) of 23 s^-1.

Results

Synthesis and atomic-scale structural analysis of partially oxidized NiIr nanoclusters on TiO₂

To construct bimetallic NiIr nanoclusters on TiO₂ with precise atomic pairing, we adopted a directional adsorption strategy²³ (Fig. 1a; see Methods for synthesis details). This approach leverages sequential electrostatic adsorption of cationic Ni complexes and anionic Ir precursors onto alkalized TiO₂ nanosheets, ensuring preferential Ir anchoring onto pre-adsorbed Ni sites. For the Ir precursor, we used a species complexed with quaternary ammonium cations to enable its phase transfer into an aprotic solvent. This controlled process prevents Ir ripening and leaching. Consequently, by conducting adsorption in this manner, we minimized competitive adsorption of Ni and Ir, effectively overcoming a key limitation of conventional co-impregnation methods. Subsequent H₂ calcination eliminated organic ligands while strengthening Ni-Ir metallic interactions, yielding ultrafine bimetallic clusters with uniform dispersion on the TiO₂ support (Fig. 1b). Inductively coupled plasma-atomic emission spectrometry (ICP-AES) confirms controlled metal loadings across NiIr_x/TiO₂ samples (Supplementary Fig. 1 and Table S1), where x denotes the Ir/Ni theoretical atomic ratio. The sublinear increase in Ir/Ni ratios with x suggests saturation of Ir adsorption onto pre-adsorbed Ni sites, excluding independent Ir deposition. For comparison, monometallic Ni/TiO₂ and Ir/TiO₂ were prepared with a theoretical loading of 1 wt% for each metal. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) analysis revealed that Ni species or Ir species on TiO₂ exhibited sub-1.2 nm clusters (Supplementary Figs. 2–3), whereas the NiIr_x nanoclusters on TiO₂ showed an increase in size, progressively reaching up to 1.61 nm with higher Ir content (Supplementary Figs. 4). This trend supports the co-stabilization of Ni and Ir atoms within the bimetallic clusters. This size evolution directly supports the formation of bimetallic NiIr nanoclusters, as the incorporation of Ir stabilizes the clusters and promotes their growth. Atomic-scale characterization of NiIr_0.5/TiO₂ provided further insights into the spatial distribution of Ni and Ir atoms. Filtered and color-coded HAADF-STEM images and the corresponding intensity profiles (Fig. 1c, d, Supplementary Fig. 5b, c) revealed intertwined Ni-Ir distributions within individual clusters. The 3D surface plot of a selected region (Fig. 1f) highlighted Ir atoms (whitish-yellow peaks within blue dashed marks) surrounding Ni atoms (darker yellow bumps), demonstrating cohesive bimetallic nanocluster formation. Conversely, monometallic samples exhibited only metal-TiO₂ contrast (Supplementary Figs. 2g-h, 3e-f). Moreover, EDS line scans (Fig.1g) and elemental mapping (Fig. 1h and Supplementary Fig. 5d-i) confirmed the intimate contact between Ni and Ir atoms. Owing to the higher atomic ratio of Ni to Ir, trace amounts of isolated Ni species inevitably persist. Crucially, all Ir atoms preferentially occupy Ni-associated sites, with no detectable isolated Ir single atoms observed on the TiO₂ support (Fig. 1b).

Fig. 1: Synthesis and structural characterization of TiO₂-supported partially oxidized NiIr nanoclusters.

a Schematic of the directional adsorption strategy for synthesizing NiIr nanoclusters supported on TiO₂. b Aberration-corrected HAADF-STEM image of NiIr_0.5/TiO₂ and cluster size distribution (inset). c Color-coded atomic map of region 1 (marked by the white dashed box in b after image filtering. d Intensity profiles along the white line in c. e Aberration-corrected HAADF-STEM image of NiIr nanoclusters on TiO₂. f 3D surface plot with atomic overlapping region 3 in e. Blue dashed circles indicate the Ir atoms in NiIr nanocluster. Inset: Filtered image of region 3. g EDS line-scan profiles of Ni and Ir across a cluster along the yellow arrow in e. h HAADF-STEM image and corresponding EDS elemental mapping of NiIr_0.5/TiO₂. i k²-weighted Fourier transform of Ir L₃-edge EXAFS spectra (without phase correction) for Ir foil, IrO₂, NiIr_x/TiO₂, and Ir/TiO₂. j In situ CO-probe DRIFTS spectra of NiIr_x/TiO₂, Ni/TiO₂, and Ir/TiO₂ at the 15^th min of Ar-flow during desorption (25 °C); schematic models of surface structures are included in j.

Next, we elucidate the chemical structure and coordination environment of the NiIr nanoclusters. X-ray diffraction (XRD) patterns confirm exclusive anatase-phase TiO₂ (Supplementary Fig. 6), with no detectable reflections from metallic Ni or Ir phases, consistent with HAADF-STEM evidence of dispersed sub-2 nm clusters. Normalized Ir L₃-edge and Ni K-edge X-ray absorption near-edge spectroscopy (XANES) analysis reveal partially oxidized states of Ni and Ir (Supplementary Fig. 7a, b). Linear combination fitting (Supplementary Fig. 7c-d) shows higher Ir oxidation in Ir/TiO₂ (+3.6) than NiIr_0.5/TiO₂ (+2.7), and more reduced Ni in Ni/TiO₂ (+1.6) versus NiIr_0.5/TiO₂ (+1.8), indicating electron transfer from Ni to Ir, later corroborated by in situ analyses. Extended X-ray absorption fine structure (EXAFS) at the Ir L₃-edge provides atomic-scale coordination details (Fig. 1i). All samples exhibit a prominent peak at ~1.60 Å and a relatively weaker peak at ~2.56 Å, corresponding to Ir-O scattering and Ir-Ir scattering, respectively. Notably, NiIr_x/TiO₂ samples exhibit an additional peak at 2.18 Å, which can be attributed to Ni-Ir scattering. These results demonstrate that NiIr nanoclusters exhibit a partially oxidized state, with electronic modulation between Ni and Ir atoms. The partial oxidation probably arises from the interface reconstruction between nanocluster and the surface lattice O atoms of TiO₂ during the calcination process and the exposure to air during storage. This electronic structure, coupled with their ultrafine size, positions these partially oxidized NiIr nanoclusters (O_x-NiIr) as promising candidates for the adsorption of CH₄ and CO₂ due to their inherent thermal instability¹⁹.

In situ CO probe molecule diffuse reflectance infrared fourier transform spectroscopy (DRIFTS) analyses were conducted to unravel the surface structure and adsorption behavior of NiIr nanoclusters (Supplementary Fig. 8 and Fig. 1j). Ni/TiO₂ exhibited negligible *CO chemisorption signals (<2100 cm^-1), which vanished entirely upon heating to 100 °C. This confirms the inadequacy of individual Ni sites for CO adsorption. In contrast, Ir-containing samples (NiIr_x/TiO₂, Ir/TiO₂) display intense *CO chemisorption bands (2100–1990 cm^-1), with adsorption strength scaling with Ir content. Heating-up desorption revealed a volcanic trend: *CO intensity peaked at 200 °C and declined continuously to 400 °C. It demonstrates reversible CO binding, which is a critical feature for sustaining DRM activity. Detailed comparison of the desorption spectra at 25 °C (Fig. 1j) highlights distinct Ir coordination environments. For Ir/TiO₂, three CO species were resolved: (i) multi-carbonyl Ir-(CO)₂ (2068 cm^-1, low-coordination sites)²⁴, (ii) linear CO on metallic Ir (Ir-CO, 2033 cm^-1)²⁵, and (iii) bridged CO at Ir-TiO₂ interfaces (Ir_if-CO, 1994 cm^-1)²⁶. For Ni/TiO₂, only weak peaks of Ni-CO and Ni_if-CO can be observed. In NiIr_x/TiO₂, the dominant Ir-(CO)₂ peak progressively shifts downward (2063 → 2044 cm^-1) with increasing Ir content, signaling enhanced π-backdonation from Ir to CO. The observed redshift arises from Ni-to-Ir electron transfer, which elevates Ir’s d-orbital occupancy, strengthens Ir-C bonds, and weakens C-O bonds. These results confirm that in partially oxidized NiIr nanoclusters, Ir serves as the main center for CO adsorption, while Ni modulates Ir’s electronic properties to optimize the balance between adsorption strength and desorption kinetics.

Photoelectric-photothermal synergy in light-driven dry reforming of methane over NiIr/TiO₂

Light-driven DRM over TiO₂-supported NiIr nanoclusters was investigated under full-spectrum irradiation (Supplementary Fig. 9). Ni and Ir modification significantly enhanced DRM activity compared to pure TiO₂ (Fig. 2a, Supplementary Fig. 10). NiIr_0.5/TiO₂ achieved optimal performance (Fig. 2a, Supplementary Fig. 11 and Note 1) with syngas generation rates of 2512 mmol g_cat^-1 h^-1 and feed gas conversions near 60% at the condition of CH₄: CO₂: Ar =1: 1: 8. Increasing the Ir/Ni ratio (x) from 0.05 to 1.0 elevated H₂/CO from 0.68 to 1.03, demonstrating Ir-Ni synergy in balancing CH₄-CO₂ activation (Fig. 2a). NiIr_0.15/TiO₂ (equivalent Ir loading to 0.5% Ir/TiO₂) exhibited 2.13-fold higher activity (CH₄ conversion: 17% → 42%), underscoring the critical role of bimetallic interactions. Highly diluted conditions (CH₄: CO₂: Ar =1: 1: 98) achieved 90% conversion (Fig. 2b). Strikingly, at undiluted feed gas, the system achieved syngas generation rates of 10841 mmol g_cat^-1 h^-1, with a LTFE of 25.0% and a TOF of 23 s^-1. However, for gas conservation and safety considerations, subsequent tests were conducted using a feed gas mixture of CH₄: CO₂: Ar =1: 1: 8. Light irradiation dramatically reduced the apparent activation energies for CO and H₂ production to 35.33 and 44.65 kJ mol^-1, respectively, which is less than half those of thermal catalysis (68.76 and 99.19 kJ mol^-1), effectively breaking the limitation of thermal catalysis (Fig. 2c, Supplementary Fig. 12).

Fig. 2: Light-driven dry reforming of methane performance of NiIr/TiO₂.

a Syngas generation rates and H₂/CO ratios of TiO₂, Ni/TiO₂, Ir/TiO₂, and NiIr_x/TiO₂ under 3.89 W cm^-2 light irradiation (feed gas: CH₄: CO₂: Ar =1: 1: 8). Note: the same * or ^# marks indicate that the actual Ir loading amounts of the two samples are similar. b Feed-gas composition dependence of syngas generation rates, H₂/CO ratios, and conversion rates for NiIr_0.5/TiO₂ (light intensity: 3.89 W cm^-2). c Syngas generation rates of NiIr_0.5/TiO₂ under light irradiation or external heating. (CH₄:CO₂:Ar =1:1:8, the heating temperature is near to the detected light irradiation temperature at catalyst center.) d DRM activity per-unit light intensity, detected temperatures, and conversion rates of NiIr_0.5/TiO₂ under varied light intensities (CH₄:CO₂:Ar =1:1:8). Note that the T_AVG and T_Center indicate the average and the center temperature of catalyst surface measured by a thermocouple, which agree with the results from the IR camera (Supplementary Fig. 13). e Syngas generation rates and H₂/CO ratio of NiIr_0.5/TiO₂ with (blue lines) or without (yellow lines) condensation equipment. (CH₄: CO₂: Ar =1: 1: 8, condensate temperatures noted in blue font) Note that the PT and PE represent photothermal contribution and photoelectric contribution, respectively. f UV-Vis-NIR spectrum and wavelength-dependent DRM activity (per W cm^-2) of NiIr_0.5/TiO₂ under monochromatic light (with 500 °C external heating, CH₄: CO₂: Ar =1: 1: 8). g 100-hour stability of NiIr_0.5/TiO₂ under light on/off cycles (Light on: 3.89 W cm^-2, CH₄: CO₂: Ar =1: 1: 8; Light off: exposed to air). Reaction conditions for a-g: 5 mg catalyst, flow rate = 20 ml min^-1; The error bars indicate that there were at least three independent tests.

Activity surged above a threshold light intensity of 2 W cm^-2, marking the onset of photoelectric-photothermal synergy (Fig. 2d and Supplementary Fig. 13). Specifically, at lower intensities (<1.95 W cm^-2, T_Center < 336 °C), the observed activity originated predominantly from photoelectric effects. This assignment is confirmed by the sustained syngas production under illumination, which stands in clear contrast to the negligible activity observed in dark thermal catalysis experiments conducted below the 350 °C threshold (as shown in Fig. 2c and Supplementary Fig. 12a). The interplay between photothermal and photoelectric contributions peaked at 3.89 W cm^-2 (T_Center = 506 °C; T_AVG = 403 °C). Beyond 4.27 W cm^-2, the system approached a state of photon absorption saturation. At this point, the catalyst could no longer utilize the additional photons for charge carrier generation, and consequently, the excess energy was redirected into thermal pathways. This mechanistic shift is evidenced by a distinct decline in the per-unit-light efficiency (syngas generation rates/light intensity), which occurred despite the continued rise in the catalyst temperature (T_Center exceeding 600 °C). This nonlinear response highlights the delicate balance between photocarriers generation and thermal dissipation in light-driven systems.

Controlled condensation experiments (Fig. 2e) decoupled photothermal and photoelectric effects by selectively suppressing thermal contributions. The blue trend line maintained constant light intensity (3.89 W cm^-2) to preserve the photoelectric effect and used condensation cooling to suppress photothermal contributions, while the yellow trend line varied light intensity (from 3.89 to 1.30 cm^-2) without condensation and diminished both effects simultaneously. By comparing the syngas production rates and H₂/CO ratios from these two paths at matched catalyst surface temperatures, it was determined that the photoelectric-dominated system (with condensation) contributed 57.5% to syngas rates and 91.7% to H₂/CO balance, compared to the photothermal-assisted system (without condensation). This contrast confirms that photogenerated carriers preferentially drive H-H coupling and carbon intermediate oxidation, while thermal energy primarily enhances vibration and mobility of reactants. Comparative tests on NiIr_0.5/TiO₂ and NiIr_0.5/SiO₂ (Supplementary Fig. 14 and Note 2) further confirmed that TiO₂-supported NiIr nanoclusters uniquely harness coupled photothermal-photoelectric effects to achieve efficient and stable DRM catalysis.

The UV-Vis-NIR spectra (Supplementary Fig. 15) revealed that the introduction of Ni and Ir significantly enhanced broadband light absorption across the entire spectral range compared to TiO₂. Crucially, the absorption profile across the full spectral range of NiIr_0.5/TiO₂ exhibited a strong correlation with the wavelength-dependent DRM activity (Fig. 2f), confirming that the reaction is primarily driven by photoexcitation rather than thermal effects and evidencing its light-driven nature. Long-term stability tests (Supplementary Fig. 16 and Note 3) demonstrated considerable durability as NiIr_0.5/TiO₂ retained >97% activity over 720 min under light, with near-unity carbon balance. In contrast, thermal catalysis at 500 °C suffered a 20% activity drop and significant carbon deposition, emphasizing light’s role in suppressing deactivation. Moreover, NiIr/TiO₂ sustained 100-hour stability under 20% reactant flow during five light/dark cycles (Fig. 2g), retaining >80% syngas generation rates with H₂/CO > 0.9 throughout the test. Monometallic controls (Supplementary Fig. 17) further validated Ni-Ir synergy that Ni/TiO₂ rapidly deactivated due to CH_x accumulation, while Ir/TiO₂ maintained 70% stability over 720 min but exhibited suppressed H₂/CO ratios, probably due to RWGS side reaction.

These results reveal that UV-Vis- induced hot electrons at the NiIr-TiO₂ interface, consistent with ultrafast (<10 fs) UV photoinduced electrons in TiO₂²⁷, collaboratively cleave C-H/C = O bonds and mediate H-H coupling. Meanwhile, infrared absorption enhances local temperatures. This photoelectric-photothermal synergy suppresses carbon deposition and stabilizes active sites, achieving DRM efficiency in photothermal and even certain thermal systems (Supplementary Table 3)^{7,8,9,28,29,30,31,32,33,34,35,36,37,38,39,40}.

Light-induced dynamic restructuring to form NiIr nano-island alloy on TiO₂

Alloy structural evolution under operando light conditions is critical to unraveling true catalytic mechanisms, which was therefore systematically investigated. In situ X-ray absorption spectroscopy (XAS) and near-ambient pressure XPS (NAP-XPS) analyses revealed the light-induced structural reorganization of NiIr nanoclusters. Under light irradiation, Ir L₃-edge EXAFS spectra of NiIr_0.5/TiO₂ (Fig. 3a) show a dominant Ir-Ni scattering peak at 2.18 Å, replacing the initial Ir-O coordination observed in the dark. Concurrently, Ni K-edge EXAFS of NiIr_0.5/TiO₂ (Fig. 3b) retains Ni-O (1.60 Å) and Ni-Ni (2.09 Å) scattering alongside emerging Ni-Ir (2.18 Å) coordination, indicating alloy formation without complete Ni reduction. In contrast, monometallic Ni/TiO₂ under identical conditions transitions fully to metallic Ni-Ni bonds, while Ir/TiO₂ retains residual Ir-O coordination (Supplementary Fig. 18). This indicates that during the dynamic alloy formation under light, electrons are transferred from Ni to Ir. Quantitative EXAFS fitting analysis (Fig. 3c, Supplementary Fig. 19–20, Supplementary Tables 4–5) quantifies this structural evolution. Under light irradiation, NiIr_0.5/TiO₂ exhibits nearly 100% Ir-Ni coordination at the Ir L₃-edge and 32% Ir-Ni coordination at the Ni K-edge, alongside 37% Ni-O interfacial bonds. This contrasts sharply with monometallic Ni/TiO₂. The nearly 100% Ni-Ni coordination in Ni/TiO₂ likely drives its rapid deactivation (Supplementary Fig. 18), as non-oxidized Ni sites exhibit diminished ability to activate CH₄ and oxidize carbonaceous intermediates, leading to CH_x accumulation and irreversible poisoning. The absence of Ir-O coordination in NiIr_0.5/TiO₂ under light confirms that Ir in the alloy phase does not directly interact with TiO₂, while Ni maintains interfacial Ni-O bonds, forming a nano-island architecture that isolates the NiIr alloy from the support (Supplementary Fig. 21). This structure formation is critical for efficient and stable DRM catalysis.

Fig. 3: In situ XAS and NAP-XPS analysis of light-induced structural dynamics in NiIr/TiO₂.

a, b In situ k²-weighted Fourier transform of Ir L₃-edge a and Ni K-edge b EXAFS spectra for NiIr_0.5/TiO₂ during light-driven DRM conditions (CH₄: CO₂: Ar =1: 1: 8, 3.89 W cm^-2; data are not phase-corrected). c Normalized coordination numbers (CN) of different samples from EXAFS fitting. d In situ NAP-XPS spectra of Ir 4 f for NiIr_0.5/TiO₂ under different irradiation conditions in a gas flow of CH₄: CO₂: Ar =1: 1: 8 (Unfocused light: 1.58 W cm^-2; focused light: 3.89 W cm^-2). e Ir⁰/Ir^δ+ ratio evolution of NiIr_0.5/TiO₂ during in situ NAP-XPS conditions and post-air exposure, derived from deconvolution of the XPS spectra. f Schematics of light induced oscillatory of NiIr/TiO₂-L_Ni formation. g DFT-optimized structures of NiIr/TiO₂-L_Ni and NiIr/TiO₂-L_Ir. The key distances between Ni/Ir linkages and TiO₂ are annotated in black font. Ir, Ni, Ti, and O atoms are shown in pink, yellow, grey, and blue colors.

In situ NAP-XPS analyses corroborate the XAS findings. To decouple photochemical and photothermal contributions, we first monitored structural evolution under unfocused light. The Ir 4 f spectra (Fig. 3d) reveal a pronounced shift toward lower binding energy, with the Ir⁰ fraction increasing from 36.96% (dark) to 60.65%. Upon switching to focused light, the Ir⁰ fraction further rises to 62.31%. Figure 3e quantifies this trend: the Ir⁰/Ir^δ⁺ ratio increases by 0.95 under unfocused light, with only an additional 0.11 gain under focused irradiation, demonstrating that low-intensity light suffices to initiate alloying. Monometallic Ir/TiO₂ exhibits a smaller Ir⁰/Ir^δ⁺ increase of 0.61 under unfocused light (Supplementary Fig. 22), aligning with its higher oxidation state observed in in situ XAS. Moreover, air exposure restores the initial oxidized state (Fig. 3e, Supplementary Figs. 22–23), confirming the reversibility of the NiIr-L_Ni phase. Parallel in situ Ni 2p NAP-XPS (Supplementary Fig. 24a-c) shows analogous trends, with Ni⁰/Ni²⁺ ratios tracking Ir’s redox dynamics. While Ti 2p and O 1 s exhibit the positive shift in the binding energy of Ti and O_Lattice under light irradiation (Supplementary Fig. 24d-e), this demonstrates that electrons migrate from the TiO₂ support to Ni-Ir under illumination. Post-reaction characterizations (Supplementary Figs. 25–26) reveal that prolonged light-driven DRM preserves Ni-Ir intimacy. Re-oxidation in post-EXAFS (Supplementary Fig. 27 and Note 4) and reversible NAP-XPS trends confirm that NiIr/TiO₂-L_Ni exists transiently only under light irradiation. Thermal catalysis induces limited Ni or Ir reduction (Ir⁰/Ir^δ⁺=0.61 at 500 °C vs. 1.54 under unfocused light) and fails to form obvious Ni-Ir bonds (Supplementary Figs. 28–29), resulting in poor DRM performance. Conversely, light irradiation in pure Ar still forms dominant Ni-Ir scattering (Supplementary Figs. 30–31 and Note 5), proving that reactant gases are non-essential for NiIr alloying. Moreover, light cessation triggers rapid re-oxidation, with full structural recovery after air exposure.

Figure 3f illustrates the light-induced structural dynamics in NiIr/TiO₂. Initial O_x-NiIr/TiO₂ adsorbs CH₄/CO₂ but remains inefficient and prone to coking. Under light irradiation, interfacial oscillations drive electrons from Ni to Ir, and Ir atoms migrate away from the interface, forming NiIr island alloy anchored via oxidized Ni linkages on TiO₂. The Ir-Ni binary phase diagram (Supplementary Fig. 32) and positive segregation energy (0.23 eV)⁴¹ confirm alloy feasibility under photoexcitation. We further constructed DFT structures (Supplementary Fig. 33) for a series of samples based on the above analysis and actual Ni/Ir atomic ratios. Figure 3g confirms NiIr/TiO₂-L_Ni’s stability with a lower formation energy than the hypothetical Ir-linked model (NiIr/TiO₂-L_Ir), where Ir-O bonds elongate (black arrow indication, >3.0 Å), destabilizing the interface. These results demonstrate that light excitation dynamically restructures O_x-NiIr/TiO₂ into NiIr/TiO₂-L_Ni. The reversibility of NiIr/TiO₂-L_Ni under ambient conditions highlights its transient, light-governed property, serving as a suitable model for further investigation of photoelectric-dominated reaction mechanisms.

It is worth elucidating that the light-induced surface restructuring observed in this study is supported by prior findings regarding bimetallic systems. It is established that electron transfer in bimetallic clusters can drive surface segregation of one of the metals^42,43. A pertinent mechanism involves atomic size, in which the introduction of a larger-radius atom (e.g., Sn in PtCu) can promote surface extraction and alloy reorganization^21,44. In our system, Ir has a larger inherent atomic radius than Ni. Critically, the light-induced electron transfer from Ni to Ir reduces Ir’s oxidation state, further exacerbating this radius disparity (Ir⁰ > Ir^δ⁺ > Ni^δ⁺) and thereby strongly promoting the surface exposure of Ir. This process is further enabled by the photo-weakening of metal-support bonds, a phenomenon known to induce structural dynamics under illumination^22,44,45. To confirm that the improved catalytic activity is induced by light-driven structural reorganization, DRM reaction experiment with external heating (Supplementary Fig. 34 and Note 6) were conducted. Under light irradiation with additional heating, a gradual decline in activity is observed over 30 min after the light is turned off. This is attributed to the immediate loss of the photoelectric–photothermal synergy, while the structural reversion of the alloy occurs more slowly, underscoring that light-induced structural reorganization together with the photoelectric-photothermal effect jointly contribute to the high DRM performance of the NiIr_0.5/TiO₂ catalyst.

Intermediates tracking and photoexcited charge transfer investigation

In situ synchrotron radiation-diffuse reflectance infrared Fourier transform spectroscopy (SR-DRIFTS) analysis during CO₂-CH₄ co-adsorption revealed distinct molecular activation patterns on NiIr_0.5/TiO₂ under dark conditions (Fig. 4a and Supplementary Fig. 35a). Characteristic vibrational modes emerged at 3014/1304 cm^-1 (CH₄ deformation), 2346 cm^-1 (C = O stretching), 1354 cm^-1 (OCO stretching vibration, ν(OCO)), and 1536 cm^-1 (monodentate carbonate, m-CO₃^2-), confirming robust adsorption of both reactants on partially oxidized NiIr nanoclusters. Temperature-programmed desorption (TPD) (Fig. 4b, Supplementary Fig. 35b) further demonstrated that NiIr_0.5/TiO₂ exhibited high CH₄ chemisorption with a prominent desorption peak at 417 °C attributed to Ni-Ir dual sites, while maintaining stable CO₂ adsorption through ensemble-dependent Ni or Ir adsorption at 546 °C.

Fig. 4: Time-resolved in situ tracking of reaction intermediates and photoexcited charge dynamics in NiIr/TiO₂.

a In-situ SR-DRIFTS of NiIr_0.5/TiO₂ during CO₂-CH₄ co-adsorption in the dark (25 °C, CH₄: CO₂: Ar =1: 1: 8). b CH₄-TPD profiles of TiO₂, Ni/TiO₂, NiIr_0.5/TiO₂, and Ir/TiO₂. c, d In-situ SR-DRIFTS (4000–1200 cm^-1) of NiIr_0.5/TiO₂ under 3.89 W cm^-2 focused light (c) vs. 500 °C dark heating (d) in a gas flow of CH₄: CO₂: Ar =1: 1: 8. e, f Time-resolved evolution of dominant sub-1200 cm^-1 SR-DRIFTS peaks under focused light e and dark heating f. g 2D fs-TA spectroscopy mapping of NiIr_0.5/TiO₂. h Normalized fs-TA decay kinetics of NiIr_0.5/TiO₂ at 560 nm. i TD-DFT calculated electron-hole pairs distribution on NiIr/TiO₂-L_Ni under S₀ → S₁/S₂ excitations, and the corresponding electron/hole (e^-/h⁺) ratio statistics of total Ir and Ni sites. The green and purple group represent the distribution of electrons and holes, respectively. The isosurface level is 0.005 e Å^-3. Ir, Ni, Ti, and O atoms are shown in pink, yellow, grey, and blue colors, respectively.

Light irradiation or external heating induced surface intermediates over NiIr/TiO₂-L_Ni during DRM reaction as monitored by time-resolved SR-DRIFTS (Fig. 4c-f). Under focused light, the complete conversion of m-CO₃^2- to intensified ν(OCO) indicates light-triggered formation of NiIr/TiO₂-L_Ni, optimizing CO₂ adsorption configuration on NiIr alloy sites. Continuously detected gaseous CO (2024 cm^-1), characteristic vibrations of HCO* (1786 cm^-1) and *CH₂ (1490 cm^-1)^8,46, confirmed efficient light-driven DRM activity (Fig. 4c). In contrast, dark thermal activation at 500 °C resulted in gradual emergence of attenuated CO signals after 12 min (Fig. 4d), aligning with its poor catalytic performance. Notably, transient intermediate signatures between 1120-920 cm^-1 (Fig. 4e, Supplementary Fig. 36a) demonstrated ultrafast generation and decomposition of CH_xO* species^37,47 under illumination. Conversely, thermal catalysis produced prominent C-C stretching vibrations (974 cm^-1)⁴⁸ (Fig. 4f, Supplementary Fig. 36b), consistent with coke accumulation observed in Supplementary Fig. 16. Isotopic mass spectrometry (Supplementary Fig. 37) revealed that photoactivated *O species primarily originated from CO₂ dissociation; these *O species combined with *CH_x fragments to form metastable CH_xO* intermediates, establishing a dehydrogenation-dominated pathway that suppressed carbon deposition. Comparative studies of monometallic systems (Supplementary Fig. 38) highlighted the necessity of Ni-Ir synergy. Ni/TiO₂ exhibited negligible CO production with C-C coupling signals (873/893 cm^-¹), while Ir/TiO₂ promoted *COOH formation (1678 cm^-1) via RWGS pathways. These findings underscore the necessity of Ni-Ir interaction in maintaining both C-H activation kinetics and intermediate stabilization.

To dissect photogenerated charge dynamics, femtosecond transient absorption (fs-TA) spectroscopy was employed (Fig. 4g, Supplementary Fig. 39a-c). Upon 325 nm excitation, all samples exhibited a ground-state bleach (GSB) at ca. 415 nm, reflecting the excited state relaxation⁴⁹. Accompanied by excited state absorption (SA) between 460–570 nm, arising from electronic transitions from the singlet state (S₁) to higher energy excited states (S_n)⁵⁰. Notably, NiIr_0.5/TiO₂ displayed the pronounced TA intensity, confirming NiIr alloy-enhanced charge separation. Time-decay fitting of the SA signal at 560 nm (Fig. 4h and Supplementary Fig. 39d-f) revealed a 520.3 fs lifetime for NiIr_0.5/TiO₂, which was markedly longer than TiO₂ (110.9 fs) or Ni/TiO₂ (233.1 fs), directly evidencing retarded charge recombination via Ni-Ir interaction. In addition, compared to pure TiO₂, NiIr nanoclusters show enhanced photocurrent density, lower electrochemical impedance, and quenched PL intensity; these respectively indicate stronger charge separation, faster charge transfer, and a greater number of effective photocarriers collected for reactions (Supplementary Fig. 39g-i and Note 7).

Ground-state and excited-state calculations resolved the interfacial charge transfer mechanism. Electron-density distributions (Supplementary Fig. 40a-d and Note 8) indicate Ni-mediated charge polarization enriches electrons at Ir sites. Simultaneously, density-of-states profiles (Supplementary Fig. 40e-h) reveal that NiIr alloy-induced d-band center modulation optimizes the trade-off between adsorption strength and intermediate desorption kinetics, suppressing both carbon deposition and parasitic reactions. Time-dependent DFT (TD-DFT) simulations (Fig. 4i) further identified excitation-dependent charge redistribution on NiIr/TiO₂-L_Ni. Electrons in photoexcited states preferentially accumulate at Ir sites distal to the TiO₂ interface, designating these Ir atoms as primary active centers for CO₂ reduction or H-H coupling during DRM. Electron/hole (e^-/h⁺) ratio statistics reveal preferential electron capture at both Ir and Ni sites under the S₀ → S₁ excited state, evidencing electron migration from the TiO₂ support to NiIr nanoclusters. The S₀ → S₂ excited state exhibits a smaller Sr index (the geometric mean overlap function of hole and electron distributions), representing less overlap between the electron and hole, and further indicating superior charge separation. Mechanistically, S₀ → S₂ promotes hole accumulation at Ni sites alongside electron enrichment at Ir sites, which creates a spatial charge segregation that drives photoexcited electrons toward Ir centers in higher-energy states. Computational analysis of chemical bond strength changes between the ground and excited states (Supplementary Table 6) provides further mechanistic insight. The photoinduced charge transfer weakens the chemical bonds, which synergizes with thermal effects, facilitates atomic rearrangement into a more stable configuration. A detailed comparison reveals a more pronounced bond energy difference between the excited state and the ground state in Ir-O bond (-28.06 kJ mol^-1) compared to that of the Ni-O bond (-9.36 kJ mol^-1). This indicates a preferential destabilization and breaking of the Ir-O bonds, thereby facilitating the migration of Ir atoms. Furthermore, under the excited state, the Ni-Ir bond (92.35 kJ mol^-1) is stronger than the Ir-O bond (69.64 kJ mol^-1). These bond energy differences provide support for the formation of the Ni-Ir alloy under light irradiation.

Synergy between Ni-mediated charge transfer and Ir-tailored adsorption establishes a self-optimizing catalytic cycle under illumination. Transient intermediate tracking demonstrates light-driven suppression of coking pathways through metastable CH_xO* formation, while fs-TA reveals hierarchical charge transfer from TiO₂ → Ni→Ir. This dual functionality, achieved through NiIr nano-island architecture, enables simultaneous enhancement of photogenerated carrier utilization and surface reaction kinetics.

Theoretical calculation of reaction mechanism and natural sunlight validation

DFT calculations (Supplementary data 1) coupled with in situ spectroscopic evidence unraveled the light-induced dynamic restructuring mechanism governing reaction pathway selectivity. Site-resolved adsorption analysis on the NiIr/TiO₂-L_Ni model (Supplementary Fig. 41) revealed that the Ir¹-Ni³ asymmetric sites exhibited optimal CO₂ adsorption (E_ads = -1.22 eV) via ν(OCO) configuration. In contrast, the Ni¹-Ni²-O domain exhibited negligible CO₂ adsorption in the form of m-CO₃^2- (E_ads = −0.03 eV). Charge complementarity between Ni^δ+ and Ir^δ- drives complete CO₂ adsorption mode transition from m-CO₃^2- under dark to ν(OCO) upon illumination, as evidenced by time-resolved SR-DRIFTS. CH₄ preferentially anchors at Ir² sites through H···Ir interaction, where accumulated electrons from Ni-mediated interfacial transfer stabilize the C-H scission transition state. Comparative studies confirmed NiIr/TiO₂-L_Ni’s superior adsorption (Supplementary Fig. 42), with Ni/TiO₂ performing poorest, which is consistent with TPD trends.

For the initial bond cleavage of CO₂ and CH₄ molecules as shown in Fig. 5a-b and Supplementary Figs. 43–44, NiIr/TiO₂-L_Ni demonstrates both thermodynamic and kinetic advantages. Compared to monometallic Ni/TiO₂ and Ir/TiO₂, NiIr/TiO₂-L_Ni exhibits a significantly greater energy reduction after direct C = O scission of CO₂ and the first C-H scission of CH₄, indicating its superior thermodynamic favorability. Meanwhile, NiIr/TiO₂-L_Ni exhibits activation energies (Ea) of the transition states at +0.26 eV for C = O scission and +0.30 eV for C-H scission. These values are lower than the transition state barriers for the first C-H scission of CH₄ and direct CO₂ scission on Ni/TiO₂ and Ir/TiO₂, indicating the kinetic advantage of the nanoisland NiIr alloy. Supplementary Fig. 45 further compares the pathway of direct C = O scission of CO₂ with the indirect activation route via *COOH pathways. The results reveal Ir/TiO₂’s propensity for *COOH formation, aligning with in situ SR-DRIFTS detected RWGS activity and a suboptimal H₂/CO ratio. In contrast, NiIr/TiO₂-L_Ni enables concerted C = O scission, where Ni site stabilizes dissociative oxygen and Ir site coordinates carbon, directly yielding CO and *O species.

Fig. 5: DFT calculations of reaction pathway and natural sunlight experiments.

a Reaction energy profiles (including transition states) for C = O bond scission of CO₂ on NiIr/TiO₂-L_Ni, Ni/TiO₂, and Ir/TiO₂ models. b Reaction energy profiles (including transition states) for C-H bond scission of CH₄ on NiIr/TiO₂-L_Ni, Ni/TiO₂, and Ir/TiO₂ models. c Reaction energy profiles for different CH₄ activation pathways on NiIr/TiO₂-L_Ni model. d Energy evolution of key steps (pink-shaded indicator in Supplementary Fig. 46a) during the first C-H scission of CH₄ and the formation of *CH₃O intermediate. Ir, Ni, Ti, and O atoms are shown in pink, yellow, grey, and blue colors, respectively. e DRM performance over NiIr_0.5/TiO₂ under concentrated natural sunlight, with measured reaction temperature and incident light intensity (CH₄: CO₂: Ar =1: 1: 8, 20 ml min^-1). Inset: experimental setup. f DRM reaction demonstration under concentrated sunlight, with gas outlet directly interfaced to a PdCl₂ indicator solution for CO detection. g Schematic illustration of light-driven DRM mechanism.

Regarding the subsequent activation process after the initial bond cleavage of CH₄, *CH₃ may undergo further direct C-H scission until a C atom remains. Alternatively, *CH₃ can be assisted by activation through combination with O species from CO₂ activation (O_CO2) or lattice oxygen (O_L). Figure 5c compares the thermodynamic energy changes before and after the reaction for these three possible activation pathways. The results show that the sequential direct C-H scission pathway (*CH₃ → *CH₂ → *CH → *C) on NiIr/TiO₂-L_Ni requires prohibitively high energy, particularly 1.18 eV for cleaving the second C-H bond. For the *O-assisted C-H scission routes, O_L-assisted pathways proved unfavorable due to the high *CH₃ migration energy of 1.30 eV. Conversely, *O species from CO₂ dissociation adsorbed on Ni sites in NiIr/TiO₂-L_Ni enabled thermodynamically favorable *CH₃O formation with ΔE = -0.61 eV, and subsequent dehydrogenation required only 0.29 eV, consistent with isotopic labeling results. Comparative studies between NiIr/TiO₂-L_Ni and monometallic Ni/TiO₂ and Ir/TiO₂ in Supplementary Fig. 46 further highlight the optimal performance of NiIr/TiO₂-L_Ni. Specifically, in *O_CO2-assisted pathways, while Ni/TiO₂ struggled with CH₄ activation exhibiting a second C-H barrier of 0.75 eV and Ir/TiO₂ suffered from poor *O stabilization, the NiIr alloy achieved balanced kinetics through asymmetric Ni-Ir sites. Figure 5d quantitatively contrasts key steps that Ir/TiO₂ excels in initial C-H scission but falters in *CH₃O stabilization, whereas NiIr/TiO₂-L_Ni synergizes Ir’s C-H activation prowess with Ni’s oxygen affinity, enabling stepwise dehydrogenation without carbon buildup. Product desorption studies (Supplementary Fig. 47) revealed moderate H₂/CO adsorption on NiIr/TiO₂-L_Ni, ensuring sustained catalytic turnover. In addition, O_x-NiIr/TiO₂ exhibited excessive CO₂ adsorption but poor C = O scission capacity (Supplementary Fig. 48), while thermal conditions promoted direct CH₄ cracking into carbon deposits (Supplementary Fig. 49). This starkly contrasts with photoinduced NiIr nano-island alloy, underscoring light’s critical role in structural optimization. We note that our DFT calculations are based on idealized slab models and static configurations and therefore provide qualitative insight into trends rather than a complete description of the real catalyst under operando conditions.

We also conducted an outdoor DRM experiment under actual sunlight using a Fresnel lens for light concentration (Fig. 5e-f). During 2-hour sunlight irradiation, light intensity varied from 3.22 to 4.08 W cm^-2, catalyst temperature ranged 438–505 °C. NiIr_0.5/TiO₂ exhibits a maximum syngas generation rate of 2782 mmol g_cat^-1 h^-1 (65.7% CO₂ conversion) with 20% feed gas flow, indicating the feasibility of this system in the utilization of sunlight. Figure 5f and the Supplementary Movie 1 show the real-time colorimetric response of the PdCl₂ indicator during the DRM reaction under focused natural sunlight. After illumination, the originally yellow PdCl₂ solution gradually turns black, indicating that a large amount of CO generated in the system adsorbs on the Pd²⁺ sites and reduces them to Pd⁰, verifying the high efficiency of the NiIr/TiO₂ system under natural sunlight.

Figure 5g illustrates the overall reaction mechanism. Under light excitation, O_x-NiIr/TiO₂ undergoes interfacial charge oscillations, with electrons sequentially transferring from the TiO₂ to Ni sites and subsequently to Ir atoms. This triggers the upward migration of Ir atoms, leading to the formation of NiIr nano-island alloys. A portion of Ni atoms act as linkages, spatially isolating the NiIr alloy from the support to form the NiIr/TiO₂-L_Ni configuration. During light-driven DRM reaction, photoelectric effects dominate the activation of C-H and C = O bonds at NiIr alloy sites, coupled with efficient H-H coupling via localized photogenerated electrons. Concurrently, the oxygen affinity of the Ni linkages steers the reaction pathway toward the formation of key CH_xO* intermediates, thereby achieving robust coking resistance. Meanwhile, photothermal effects facilitate molecular vibrations, intermediate migration, and product desorption. The synergistic interplay of photoelectric and photothermal effects on the NiIr nano-island alloy enables highly efficient and stable DRM performance.

In summary, this work reveals that light-induced NiIr nano-island alloy formation enables efficient DRM by synergizing photoelectric and photothermal effects. The interfacial charge oscillation mechanism not only stabilizes ultrafine alloy architectures against sintering but also creates dual-active sites. Specifically, Ni linkages with oxygen affinity favor coke suppression, while Ir sites modulated by Ni optimize reactant activation. Our experimental method for decoupling photoelectric/photothermal contributions reveals that localized electron transfer dominates bond cleavage selectivity, whereas thermal effects primarily govern mass transport. The achieved 25.0% LTFE and 100-h stability under practical feed conditions underscore the viability of solar-driven DRM. Importantly, the methodology combining in situ structural analysis with charge dynamics kinetics provides a universal approach to probe light–matter interactions in heterogeneous catalysis.

Methods

Catalyst preparation

O_x-NiIr/TiO₂ catalysts were synthesized via an optimized directional adsorption strategy²³. Typically, 200 mg of ground TiO₂ nanosheets was dispersed in 0.01 M NaOH aqueous solution. After 3 h, the TiO₂ was centrifuged and redispersed in deionized (DI) water through multiple cycles until the supernatant reached pH=9, ensuring a negatively charged TiO₂ surface for cation adsorption. NiCl₂ (Sigma-Aldrich) aqueous solution (containing Ni equivalent to 1 wt% of TiO₂) was added to the TiO₂ suspension and stirred for 12 h prior to centrifugation. Complete metal cation adsorption was confirmed by testing the first supernatant with NaBH₄ (Thermo Fisher), which showed no color change. The cation-loaded TiO₂ was washed with DI water and dried at 60 °C for 12 h. Co²⁺ (Co(NO₃)₂·6H₂O, Macklin), Fe³⁺(FeCl₂·6H₂O, Aladdin), and Pt(NH₃)₄²⁺ ([Pt(NH₃)₄]Cl₂, Sigma-Aldrich) cations were similarly adsorbed using their respective chloride salts.

The dried cation-loaded TiO₂ was subsequently used for complex metal anion adsorption IrCl₆^3-(Na₃IrCl₆, Sigma-Aldrich), PtCl₄^2- (K₂PtCl₄, Sigma-Aldrich), PdCl₄^2- (Na₂PdCl₄, Sigma-Aldrich), AuCl₄^- (HAuCl₄·3H₂O, Sigma-Aldrich), and RuCl₆^2-(NH₄)₂RuCl₆, Sigma-Aldrich) with controlled theoretical anion/cation atomic ratios. Anion solutions were prepared via a phase-transfer method using tetrabutylammonium cations (TBA⁺) in dichloromethane (TBA⁺/anion molar ratio = 3) under vortex mixing. After centrifugation, the dichloromethane phase was retained, and mixed with cation-loaded TiO₂ in dichloromethane. After 12 h of stirring, TiO₂-supported double complex salts were centrifuged, dried at room temperature, and reduced under 5% H₂ flow at 400 °C for 30 min. Monometallic TiO₂-supported catalysts (Ni²⁺ and Ir³⁺) were prepared analogously without anion adsorption steps.

Catalytic reactivity measurements

Light-driven DRM experiments were performed in a custom quartz reactor (Supplementary Fig. 9) at 0.1 MPa without extra heating. The reactor features an outer layer for condensate circulation and an inner chamber containing a quartz sand bed for supporting catalyst and optimize gas distribution. A type-K thermocouple (0–1150 °C range, 0.5 mm diameter, ±0.1 °C accuracy) was embedded in the quartz sand bed, with a 90°-bent tip protruding 1.0 mm above the surface. The thermocouple was pre-calibrated against standard reference sources, and its performance was verified at multiple temperature setpoints to ensure consistency across the experimental range. T_Center (the center temperature of catalyst surface) was detected by the thermocouple probe placed in direct contact with the upper surface of the catalyst at its center, enabling capture of the highest local temperature under irradiation (as shown by the red indicator line representing the thermocouple in Supplementary Fig. 9). This arrangement minimizes the thermal resistance effect by avoiding indirect measurement through the gas phase or reactor materials. Infrared (IR) thermography was used to validate the accuracy of T_Center measurements from the thermocouple probe. T_AVG (average surface temperature) was obtained by moving the thermocouple probe, measuring temperatures at five different points on the catalyst surface (center and edges) and using the formula below:

$${{{T}}}_{{{\rm{AVG}}}}=\frac{{\bar{T}}_{{{\rm{Center}}}}}{3}+\frac{{2\times \bar{T}}_{{{\rm{Edge}}}}}{3}$$

(1)

Catalyst (5 mg) was dispersed in DI water (1.0 mL), spread on a quartz fiber membrane, and vacuum-dried. The membrane was positioned on the quartz sand bed, and depressed by a tweezer align the thermocouple tip flush with the catalyst surface. Real-time temperature monitoring at catalyst center and edge positions was achieved by adjusting the thermocouple position via its stretchable connector. Note that unlike conventional temperature probes positioned below the catalyst, our small-diameter temperature probes are placed directly on the catalyst surface, enabling more accurate and sensitive temperature measurement. Prior to irradiation using a 300 W Xe lamp (PLS-SXE 300 + , Beijing PERFECTLIGHT; spectral range: 300–1000 nm), the reactor was introduced with CH₄: CO₂ (1:1) gas mixture at 20 mL min^-1. A biconvex lens (diameter: 70 mm) was employed for light focusing. Light intensity was controlled by adjusting lamp current and measured using an optical power meter (range: 0–20 W, resolution: 1 mW).

Thermal catalytic tests were conducted in a dedicated reactor (PLR-PTSR II, Beijing PERFECTLIGHT) under dark conditions, using identical catalyst supporting and thermocouple configurations to those in the custom quartz reactor. Natural sunlight experiments utilized the custom quartz reactor with Fresnel lens focusing. All other experimental parameters (gas flow, catalyst weight, etc.) remained consistent across light-driven, thermal, and natural sunlight conditions.

The monochromatic light activity tests were carried out in a dedicated reactor (PLR-PTSR II, Beijing PERFECTLIGHT) with external heating at 500 °C. A 300 W Xenon lamp was used as the light source, coupled with optical filters to obtain monochromatic light of specific wavelengths. In each test, an optical power meter was used to accurately measure and calibrate the incident light intensity in front of the reactor window. Dividing the measured syngas generation rates by the light intensity under the corresponding monochromatic light yields the contributions of monochromatic light at different wavelength bands to the activity per unit light intensity. All other test conditions were consistent with those of the light-driven conditions.

The CO products were analyzed by GC with a flame-ionization detector (FID). H₂ was analyzed by GC with a high-sensitivity thermal conductivity detector (TCD). The H₂ and CO generation rates were defined as follows:

$${R}_{{{\mbox{H}}}_{2}}={C}_{{{\mbox{H}}}_{2}}^{{\mbox{out}}}\times {F}_{{\mbox{out}}}\times 60/({V}_{{\mbox{m}}}\times M)$$

(2)

$${R}_{{\mbox{CO}}}={C}_{{\mbox{CO}}}^{{\mbox{out}}}\times {F}_{{\mbox{out}}}\times 60/({V}_{{\mbox{m}}}\times M)$$

(3)

Where ${R}_{{{\mbox{H}}}_{2}}$ (mmol g_cat^-1 h^-1) and ${R}_{{\mbox{CO}}}$ (mmol g_cat^-1 h^-1) are the H₂ and CO generation rates, respectively. The ${C}_{{{\mbox{H}}}_{2}}^{{\mbox{out}}}$ (dimensionless) and ${C}_{{\mbox{CO}}}^{{\mbox{out}}}$ (dimensionless) are the concentrations of H₂ and CO detected by gas chromatography. ${F}_{{\mbox{out}}}$ (ml/min) is the space velocity of the gas exiting the reactor, ${V}_{{\mbox{m}}}$ (L/mol) is the gas molar volume under standard conditions, and $M$ (g) is the mass of catalyst. 60 refers to the exponent needed to convert ml/min into the standard unit ml/h.

The CH₄ and CO₂ conversions were defined as follows:

$${{{\rm{CH}}}}_{4}{{\rm{conversion}}}:{X}_{{{\mbox{CH}}}_{4}}\left(\%\right)=({C}_{{{\mbox{CH}}}_{4}}^{{\mbox{in}}}-{C}_{{{\mbox{CH}}}_{4}}^{{\mbox{out}}})/{C}_{{{\mbox{CH}}}_{4}}^{{\mbox{in}}}\times 100$$

(4)

$${{{\rm{CO}}}}_{2}{conversion}:{X}_{{{\mbox{CO}}}_{2}}\left(\%\right)=({C}_{{{\mbox{CO}}}_{2}}^{{\mbox{in}}}-{C}_{{{\mbox{CO}}}_{2}}^{{\mbox{out}}})/{C}_{{{\mbox{CO}}}_{2}}^{{\mbox{in}}}\times 100$$

(5)

Where ${C}_{{{\mbox{CH}}}_{4}}^{{\mbox{in}}}$ (dimensionless) and ${C}_{{{\mbox{CH}}}_{4}}^{{\mbox{out}}}$ (dimensionless) indicate the inlet and outlet concentration of products detected by gas chromatography, respectively.

Total carbon balance is considered using the following equation:

$${{\rm{Total}}}\; {{\rm{carbon}}}\; {{\rm{balance}}}=\frac{{C}_{{\mbox{CO}}}^{{\mbox{out}}}+{C}_{{{\mbox{CO}}}_{2}}^{{\mbox{out}}}+{C}_{{{\mbox{CH}}}_{4}}^{{\mbox{out}}}}{{C}_{{{\mbox{CO}}}_{2}}^{{\mbox{in}}}+{C}_{{{\mbox{CH}}}_{4}}^{{\mbox{in}}}}\times 100$$

(6)

TOF under kinetics-controlled regime per second can be calculated with the following equation:

$$ {{\rm{TOF}}}({S}^{-1}) \\ =\frac{{0.5\times N}_{{\mbox{Number of produced}}{{\mbox{H}}}_{2}{\mbox{per second}}}+0.5\times {N}_{{\mbox{Number of produced CO per second}}}}{{N}_{{\mbox{active metal}}}\times {{\rm{Dispersion}}}}$$

(7)

The LTFE (η) was determined with the following equation:

$${\eta }=\frac{({R}_{{{\mbox{H}}}_{2}}\times {\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{{\mbox{H}}}_{2}}^{{\mbox{o}}}+{R}_{{\mbox{CO}}}\times {\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{\mbox{CO}}}^{{\mbox{o}}}+{R}_{{\mbox{C}}{{\mbox{H}}}_{4}}\times {\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{\mbox{C}}{{\mbox{H}}}_{4}}^{{\mbox{o}}})\times M}{{P}_{{\mbox{irradiation}}}}$$

(8)

Where ${R}_{{{\mbox{H}}}_{2}}$ (mmol g_cat^-1 h^-1), ${R}_{{\mbox{CO}}}$ (mmol g_cat^-1 h^-1) and ${R}_{{\mbox{C}}{{\mbox{H}}}_{4}}$ (mmol g_cat^-1 h^-1) are the H₂ generation rates, CO generation rates, and CH₄ conversion rates. ${\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{{\mbox{H}}}_{2}}^{{\mbox{o}}}$, ${\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{\mbox{CO}}}^{{\mbox{o}}}$, and ${\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{\mbox{C}}{{\mbox{H}}}_{4}}^{{\mbox{o}}}$ are the standard combustion heats (298.15 K) of H₂, CO, and CH₄ as fuels, respectively. Note that the ${\triangle }_{{\mbox{C}}}{{\mbox{H}}}_{{\mbox{C}}{{\mbox{O}}}_{2}}^{{\mbox{o}}}$ is 0. $M$ (g) is the mass of catalyst. ${P}_{{\mbox{irradiation}}}$ (W) is the irradiation power.

Characterization

HAADF-STEM imaging was performed on an aberration-corrected FEI Themis Z instrument, with a convergence semi-angle set to 11 mrad and annular detection ranges of 59 mrad (inner) and 200 mrad (outer). Elemental mapping and EDS analysis were carried out utilizing the integrated four-quadrant Super-X detector system. Powder XRD measurements were obtained using a Rigaku D/MAX 2550 diffractometer equipped with Cu Kα radiation (λ = 1.5406 Å), operating at 40 kV and 40 mA, with 2θ scanned from 10° to 80°. Raman spectroscopy was performed on a Renishaw in Via system using a 532 nm laser excitation source. UV–Vis-NIR diffuse reflectance spectra were recorded using a Varian Cary 500 spectrophotometer equipped with an integrating sphere, using BaSO₄ as a reference. For sample preparation, the powder sample was pressed into a flat, uniform pellet in a culture dish to ensure data reliability and reproducibility. The absorbance was represented by the Kubelka-Munk function:

$${{\rm{F}}}(R)({{\rm{w}}}{{\rm{h}}}{{\rm{ere\; F}}}({{\rm{R}}})={(1-{{\rm{R}}})}^{2}/2{{\rm{R}}})$$

(9)

which is derived from the measured reflectance data. Inductively coupled plasma atomic emission spectrometry (ICP-AES, Varian 730-ES, Agilent) quantified the metal loadings of catalysts. Thermogravimetric analysis (TGA) of spent catalysts was conducted on a Netzsch STA 449 F5 analyzer under air flow (50 mL min^-1), heating from 100 °C to 800 °C at 10 °C∙min^-1. Temperature-programmed desorption (TPD) experiments were performed on an Auto Chem II 2920 (Micromeritics, USA) in pure He, ramping to 770 °C at 10 °C∙min^-1. H₂-O₂ titration was performed on a temperature programmed chemisorption analyzer (VDSorb-91i) to measure metal dispersion. Photoluminescence (PL) spectra were acquired at room temperature using a Cary Eclipse fluorescence spectrometer (Agilent). Femtosecond transient absorption (fs-TA) spectroscopy was performed on a Helios pump-probe system (Ultrafast Systems, USA) with an 800 nm Ti:sapphire laser (Coherent Legend, 85 fs pulse width, 1 kHz repetition rate). The pump beam (325 nm, 80 μW) was generated via frequency doubling in a β-barium borate (BBO) crystal, while the probe beam (near-infrared continuum) monitored photoexcited states. Isotope labeling experiments employed a high-precision gas isotope mass spectrometer (Nu Instruments Horizon, UK): a 1:1 mixture of CH₄ and C¹⁸O₂ was introduced into a batch reactor in the dark, followed by 15-min irradiation (300 W Xe lamp, 3.89 W cm^-2). Reaction products were cryogenically separated and analyzed for C¹⁶O and C¹⁸O isotopes.

In situ XAS measurements conditions

In situ XAS measurements at Ir L₃-edge and Ni K-edge were performed with Si (111) crystal monochromators at the BL14W Beam line at the Shanghai Synchrotron Radiation Facility (SSRF) (Shanghai, China). Typically, the powdered sample was pressed into a self-supporting wafer and placed into a reaction microdevice equipped with an extra window for light irradiation. Before the test, the sample surface was purged with Ar gas flow (50 ml min^-1). Afterwards, the gas mixture (CH₄: CO₂: Ar=1: 1: 8) or pure Ar was continuously introduced into the reactor with a flow rate of 20 mL min^-1. The microdevice was irradiated by the 300 W Xe lamp at the light intensity of 3.89 W cm^-2. The XAFS spectra were recorded in situ in the dark or after 10-min light irradiation using a Bruker 5040 4-channel Silicon Drift Detector (SDD) in fluorescence mode. All spectra were collected in ambient conditions. The data reduction and analysis of the XAFS spectra were conducted using the Demeter software package (ATHENA and ARTEMIS, respectively)⁵¹. All fits were performed in the R space with a k-weight of 2.

In situ NAP-XPS conditions

In situ NAP-XPS experiments were conducted on a SPECS Proven-X NAP instrument (Germany) equipped with a monochromatic X-ray source µ-FOCUS 600 NAP. Prior to the test, the sample surface was cleaned using an Ar gas flow (50 ml min^-1). Subsequently, a gas mixture (CH₄:CO₂:Ar=1:1:8) or pure Ar was continuously introduced into the reactor with a flow rate of 20 mL min^-1. The NAP-XPS spectra were recorded in situ in the dark or after 10-min light irradiation. The irradiation started with unfocused light at an intensity of 1.58 W cm^-2 and then switched to focused light at 3.89 W cm^-2. Noted that under light conditions, no additional heating was applied, while under thermal conditions, the heating temperature was set at 500 °C. The NAP - XPS spectra were obtained with charge correction, using the C 1 s binding energy of 284.60 eV as the energy reference.

In situ SR-DRIFTS conditions

In situ synchrotron radiation DRIFTS were performed at the BL01B Beam line at the Shanghai Synchrotron Radiation Facility (SSRF) (Shanghai, China). The spectra were recorded by a Bruker infrared spectrometer (70 V) equipped with a liquid nitrogen-cooled mercury-cadmium-telluride (MCT) detector. Integrated with a high-brightness synchrotron radiation source, the setup achieved a spatial resolution near the theoretical diffraction limit, producing infrared spectra with high intensity and signal-to-noise ratio. The catalysts were first sealed in the chamber and purged with an Ar gas flow (50 ml min^-1). Subsequently, a flowed mixture gas (CH₄: CO₂: Ar=1: 1: 8) or pure Ar was introduced into the chamber with a flow rate of 20 mL min^-1. The chamber was then irradiated with the 300 W Xe lamp at a light intensity of 3.89 W cm^-2.

In situ CO-probe DRIFTS conditions

In situ CO-probe DRIFTS was recorded by a Fourier-transform infrared spectrometer (Bruker VERTEX 80 v, Germany) equipped with a designed reaction cell. Catalysts were initially degassed at 100 °C for 1 h and then cooled to 25 °C under a flow of Ar (50 ml min^-1). Subsequently, catalysts were exposed to a mixture of 20% CO in Ar (30 ml min^-1) for 30 min at 25 °C, which proved sufficient to saturate the catalyst surface. After saturation, the CO flow was replaced with an Ar flow (30 ml min^-1), and CO DRIFTS spectra were collected at 25 °C for 25 min. Then, the temperature was sequentially increased to 100 °C, 200 °C, 300 °C, and 400 °C, with spectra recorded at each temperature for 10 min.

Electrochemical measurements

Electrochemical measurements were conducted at room temperature using a Zahner Zennium potentiostat (Zahner-elektrik GmbH, Germany) with a three-electrode configuration: a fluorine-doped tin oxide (FTO) working electrode, a graphite counter electrode and a saturated calomel reference electrode (SCE). For working electrode preparation, 5 mg of catalyst was dispersed in 0.5 mL ethanol, and 20 μL of the slurry was drop-cast onto FTO (1 cm² active area) and dried under ambient conditions. Transient photocurrent responses were measured in N₂-saturated 0.5 M Na₂SO₄ under 300 W Xe lamp irradiation. The pH of the 0.5 M Na₂SO₄ electrolyte was measured to be 6.8 ± 0.1 (n = 3). Electrochemical impedance spectroscopy (EIS) was performed in the frequency range from 100 kHz to 0.1 Hz using an N₂-saturated potassium ferricyanide mixed electrolyte (25 mM K₃Fe(CN)₆, 25 mM K₄Fe(CN)₆, and 0.1 M KCl), whose pH was 7.0 ± 0.1 (n = 3). All electrochemical measurements were performed without iR correction. All potentials were converted to the reversible hydrogen electrode (RHE) scale according to the formula below.

$${E}_{{RHE}}={E}_{{SCE}}+0.241{{\rm{V}}}+0.059\times {{\rm{pH}}}$$

(10)

where 0.241 V is the standard potential of the saturated calomel electrode versus RHE at 25 °C.

Theoretical calculations

All density functional theory (DFT) calculations were performed using the Materials Studio software and the Vienna ab initio simulation package (VASP)^52,53,54. The exchange-correlation interactions were treated within the generalized gradient approximation (GGA), employing the Perdew–Burke–Ernzerhof (PBE) functional^55,56. The projector augmented-wave (PAW) method was used to describe core-valence electron interactions, with a plane-wave basis set and a kinetic energy cutoff set to 400 eV. The Gaussian smearing scheme with a width of 0.05 eV was applied for partial orbital occupancy. Convergence criteria were set to 10^-5eV for electronic self-consistent iterations and 0.05 eV/Å for ionic relaxation during geometry optimization. A vacuum layer of 15 Å was added perpendicular to the surface model to avoid periodic interactions. For surface calculations, the Brillouin zone was sampled using a 2 × 2 × 1 Monkhorst–Pack k-point mesh. The adsorption energy was calculated according to the following formula:

$${E}_{{ad}}={E}_{A-S}-({E}_{S}+{E}_{A})$$

(11)

where E_A-S is the energy of the slab together with the adsorbate, E_A is the total energy of the free adsorbate and E_S is the total energy of the bare slab.

The d band center (ε_d) was calculated as follow formula:

$${\varepsilon }_{{{\rm{d}}}}={\int }_{-\infty }^{+\infty }{n}_{{{\rm{d}}}}({{\rm{\varepsilon }}}){{\rm{\varepsilon }}}d{{\rm{\varepsilon }}}/{\int }_{-\infty }^{+\infty }{n}_{{{\rm{d}}}}(\varepsilon )d\varepsilon $$

(12)

The transition states corresponding to elementary reaction steps were located using a sequential approach combining the nudged elastic band (NEB) and dimer methods⁵⁷. First, the reaction pathway connecting the reactant and product states was sampled using the NEB method, which represents the path as a discrete set of structural images. From this set, the image possessing the highest energy, indicative of the most probable transition state configuration, was selected. This structure subsequently served as the initial guess for precise transition-state optimization using the dimer method.

Excitation-dependent charge redistribution calculations are performed with ORCA software⁵⁸. The functional employed was PBE0, and the basis set used was def2-SVP. (This basis set is an all-electron basis set for the first four periods and a pseudopotential basis set starting from the fifth period, so no mixed basis set is required.) The electron-hole structures were calculated using Multiwfn⁵⁹, and the visualization data were obtained using the visual molecular dynamics VMD software⁶⁰. The geometric mean overlap function of hole and electron distributions (S_r) was calculated as follows:

$${{\rm{S}}}{{\rm{r}}}(r)=\surd [{\rho }^{h{ole}}\left({{\rm{r}}}\right){\rho }^{{ele}}({{\rm{r}}})]$$

(13)

ρ^hole and ρ^ele represent the electron density of the hole and the excited electron at the spatial position r, respectively.