Nanoscale greenhouse effect for promoting solar-driven CO₂ reduction with water to CH₄

Abstract

Improvement of solar-to-chemical energy conversion in photocatalytic CO₂ reduction remains fundamentally constrained by insufficient utilization of solar energy, particularly low-energy photons. Here we report a nanoscale greenhouse structure (Bi@Fe₂O₃) that enables cascaded utilization of full solar spectrum. The Bi nanocore primarily absorbs low-energy photons, generating localized nanoheating via non-radiative heating through localized surface plasmon resonance effects and energetic hot electrons. Meanwhile, the oxygen-vacancy-rich loose Fe₂O₃ shell absorbs high-energy photons and serves as the catalytic bed, where injected hot electrons and confined heat synergistically promote CO₂ activation and deep hydrogenation. Benefiting from the interplay between photochemical and photothermal effects, the system achieves a CH₄ production rate of 273.81 μmol g^–1 h^–1 with 98.60% selectivity and an apparent quantum efficiency of 0.64% at 850 nm illumination without any external heating or sacrificial agents. This work paves a way for the efficient utilization of the entire solar spectrum.

Photocatalytic CO₂ reduction is fundamentally limited by inefficient utilization of solar energy, particularly low-energy photons. Here, the authors report a nanoscale greenhouse structure (Bi@Fe₂O₃) that enables cascaded full-spectrum solar utilization and enhances solar-driven CO₂ reduction with water to CH₄.

Introduction

Photocatalytic CO₂ reduction (PCO₂R), a pivotal technology that directly converts CO₂, H₂O, and sunlight into hydrocarbon compounds to store intermittent solar energy, represents a critical pathway toward achieving carbon neutrality and enabling a renewable energy economy^1–3. Consequently, a vast array of semiconductors and metallic nanocatalysts have been studied, demonstrating the feasibility of solar-driven CO₂ reduction^4–7. Despite recent progress, the advancement of PCO₂R technologies toward technoeconomic applications is still restricted by two key obstacles. One is the low utilization efficiency of solar energy, especially in the low-energy photon region⁸, and the other is the sluggish kinetics of transferring photogenerated energy into reactant molecules⁹. Furthermore, the selective synthesis of high-value hydrocarbons remains a formidable challenge, due to complex multi-electron and multi-proton transfer steps during PCO₂R¹⁰.

As is well known, the solar spectrum is composed of approximately 6.8% ultraviolet (UV), 38.9% visible (vis), and 54.3% near-infrared (NIR) light. However, most photocatalysts commonly developed for PCO₂R are only capable of harvesting the solar spectrum in the UV and visible ranges for a long time¹¹. Therefore, developing NIR-responsive photocatalysts is considered essential for achieving broad-spectrum-driven PCO₂R. To date, conventional NIR-responsive photocatalysts have mainly relied on some narrow-bandgap semiconductors such as lead¹² and mercury chalcogenides¹³. Their practical application in PCO₂R is severely constrained, which is ascribed to high toxicity, poor stability, and the diminished redox capability associated with bandgap narrowing. Recent research has demonstrated that noble metal plasmonic nanoparticles can interact strongly with incident light through localized surface plasmon resonance (LSPR)^14–16. Adjusting the geometric characteristics of these nanostructures, such as their symmetry and size, allows precise modulation of the LSPR absorption peak of these materials across a wide spectral range from UV to NIR^17–19. This spectral versatility is the cornerstone of their application as broadband light-harvesting antennas, offering immense potential for advanced solar-to-chemical energy conversion systems^20–24. Moreover, some studies have exhibited that with the assistance of organic additives, specific proton nanostructures can convert CO₂ under laser irradiation, indicating their potential in PCO₂R^25,26. However, in the rapid development of such protonic materials, it has been found that the mismatch in timescales between fast LSPR relaxation (up to approximately 100 ps) and slow chemical reduction kinetics (milliseconds or seconds) severely hinders solar-to-chemical energy conversion^27–29. To overcome this bottleneck, researchers have hybridized or coupled plasmonic materials with semiconductor nanostructures (referred to as “antenna reactors”^30–32) to capture hot carriers and accelerate chemical reduction kinetics. Nevertheless, such structures may cause plasmonic nanoparticles to block reaction sites on the semiconductor, resulting in sluggish chemical reduction kinetics and inadequate utilization of solar energy, thereby limiting their enhancement capabilities. Additionally, the surface atoms of plasmonic metals (primarily gold (Au), silver (Ag), and palladium (Pd), among others) exhibit high mobility under incident light, leading to the rapid degradation of such nanostructures^33,34.

Inspired by the Earth’s greenhouse effect, we herein realize a nanoscale greenhouse effect for promoting highly selective CO₂ reduction to CH₄ based on a noble-metal-free Bi@Fe₂O₃ core@shell structure (Fig. 1). In this design, the Bi nanocore functions as a nanoscale transducer, converting solar energy into two synergistic driving forces: energetic hot electrons and intense localized heat. The loose Fe₂O₃ shell plays a dual, critical role by acting as a thermal insulator to confine the localized heat while simultaneously accepting the injected hot electrons from the Bi nanocore to serve as a catalytic platform. Thus, effectively activating CO₂ and water molecules, enhancing the local electric field, and dramatically accelerating the chemical reaction kinetics. This unique coupling of photochemical and photothermal effects enables remarkable performance for solar-to-chemical energy conversion without any external energy (e.g., thermal, electric) or sacrificial agents,  delivering CH₄ production rate of 273.81 μmol g^–1 h^–1 with 98.60% selectivity, demonstrating a record-high apparent quantum efficiency (AQE) of 0.64% at 850 nm NIR light illumination compared with the current reports. Our work provides an insightful pathway for efficiently utilizing the full solar spectrum to drive CO₂ reduction with water into hydrocarbons.

Fig. 1. Contrast of this work with previous works.

Diagram illustrating the integration of intermittent sunlight and residual heat for on-demand CO₂ conversion with water.

Results

Synthesis and microstructural analyses

The 5-Bi@Fe₂O₃ core@shell photocatalyst was obtained via a two-step solvothermal strategy followed by a controlled calcination process under an inert atmosphere. (Supplementary Fig. 1), yielding an intimately coupled metal/oxide heterostructure. As illustrated in Fig. 2a, a metallic Bi nanocrystal (NC) core is conformally encapsulated by a porous Fe₂O₃ shell, which maximizes interfacial contact while maintaining open diffusion pathways. Schematic illustration of the interfacial electronic contact is exhibited in Fig. 2b. Fermi-level equilibration upon contact establishes band bending and an internal electric field between the Bi core and the Fe₂O₃ shell, providing a driving force for directional transfer of plasmon-excited electrons from the Bi core to the Fe₂O₃ shell. According to the results in Fig. 2c, Powder X-ray diffraction (XRD) further confirms the coexistence of Fe₂O₃ (JCPDS No. 87-1164) and metallic Bi (JCPDS No. 85-1329) without discernible impurities. The scanning electron microscope (SEM) and transmission electron microscope (TEM) directly evidence the morphology of the obtained spherical core@shell structure. The TEM image shows a clear contrast between the dense Bi core and the surrounding Fe₂O₃ shell (Fig. 2d), in which a sphere Bi NC was surrounded by a loose Fe₂O₃ shell, with an average size of 101 ± 31 nm (Supplementary Fig. 2). It can also be seen from the SEM images that 5-Bi@Fe₂O₃ possesses a spherical core@shell nanostructure (Supplementary Fig. 3). The high-resolution TEM (HRTEM, Fig. 2e and Supplementary Fig. 4) images indicated that the main exposed lattice planes of Bi NC in 5-Bi@Fe₂O₃ were (012) and (104) facets, corresponding to measured fringes of 0.33 nm and 0.24 nm, respectively. The lattice fringe of 0.25 nm was also exhibited, indexed to (110) facets of Fe₂O₃ in 5-Bi@Fe₂O₃. Selected area electron diffraction (SAED) (inset in Fig. 2e) showed two sets of diffraction patterns, with the (012) or (104) and (110) spots still indexed to Bi and Fe₂O₃. Moreover, HAADF-STEM with EDX elemental mapping measurements (Fig. 2f) and SEM with EDX elemental mapping measurements of 5-Bi@Fe₂O₃ (Supplementary Fig. 5) further verify the spatially separated distribution of Bi (core) and Fe/O (shell), corroborating the uniform encapsulation of Bi NCs by the Fe₂O₃ shell and the formation of a well-defined core@shell junction. Finally, it is worth noting that this sphere core@shell structure exhibited a larger specific surface area (175.5 m²/g) with a uniform mesoporous structure (5.8 nm) (Supplementary Fig. 6 and 7 and Supplementary Table 1), which has the advantage of adsorbing more reactant molecules, exposing a greater number of active sites, and providing convenient channels for mass transport³⁵.

Fig. 2. Morphological and structural characterizations.

a Schematic illustration of the Bi@Fe₂O₃ core@shell nanospheres. b Schematic illustration of the heterojunction in Bi@Fe₂O₃. W_m and W_s represent the work functions of the Bi NC and Fe₂O₃, respectively. E_F,m and E_F,s represent Fermi-energy levels of the Bi NC and Fe₂O₃, respectively; E_CBM is the bottom of the conduction band; E_VBM is the top of the valence band; E_vac is the vacuum-energy level; E represents the built-in electric field. c XRD patterns of 5-Bi@Fe₂O₃ catalyst. d TEM images of 5-Bi@Fe₂O₃ catalyst. e HRTEM image and SAED pattern of the 5-Bi@Fe₂O₃ catalyst. f HAADF-STEM image and corresponding EDS elemental mapping of 5-Bi@Fe₂O₃ catalyst. g, h Normalized X-ray absorption near-edge spectra (XANES) at the Bi L₃-edge (g) k₂-weighted Fourier transform extended X-ray absorption fine-structure spectra (EXAFS) in r-space (h) for 5-Bi@Fe₂O₃ catalyst. i Wavelet-transformed k2-weighted EXAFS spectra of the Bi foil and 5-Bi@Fe₂O₃ catalyst.

To gain deeper insight into the atomic and electronic structure of 5-Bi@Fe₂O₃ core@shell, synchrotron radiation XAS was further applied to test the Bi L₃-edge and Fe K-edge. The Bi L₃-edge X-ray absorption near-edge structure (XANES) spectrum of 5-Bi@Fe₂O₃ closely resembles that of Bi foil, indicating Bi in 5-Bi@Fe₂O₃ exists predominantly in the metallic Bi⁰ state (Fig. 2g), which is consistent with X-ray photoelectron spectroscopy (XPS) results (Supplementary Fig. 8a). The results of the Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) further corroborate this metallic nature by showing two obvious Bi–Bi coordination features (Fig. 2h), and quantitative EXAFS fitting resolves two Bi–Bi shells at 3.07 ± 0.02 Å and 3.50 ± 0.03 Å (Supplementary Table 2). In addition to the dominant Bi–Bi coordination, an extra Bi–O contribution is resolved at 2.16 ± 0.04 Å (Supplementary Table 2), indicating interfacial Bi–O bonding between the metallic Bi core and the Fe₂O₃ shell. The coexistence of Bi–Bi and Bi–O scattering paths is further supported by wavelet-transform analysis and EXAFS fitting (Fig. 2i and Supplementary Fig. 8b–d). This directly confirmed that 5-Bi@Fe₂O₃ is a chemically coupled core@shell interface that can shorten the transfer distance and facilitate efficient transport of plasmon-excited carriers from Bi to the Fe₂O₃ shell.

Furthermore, the Fe K-edge XAS spectra were analyzed to confirm the nature of the shell. The Fe K-edge EXAFS fitting reveals an Fe–O first-shell distance of 2.01 ± 0.03 Å and a higher-shell Fe–Fe contribution at 3.33 ± 0.02 Å (Supplementary Table 3 and Supplementary Fig. 9). This is consistent with a defect-rich Fe₂O₃ shell framework³⁶, while indicating a locally distorted coordination environment associated with defect formation in the shell. The Fe XANES spectrum and the Fe XPS result of 5-Bi@Fe₂O₃ also exhibit the same results about the Fe state (Supplementary Fig. 10), which suggests the possible formation of oxygen vacancies (Vo) within the Fe₂O₃ shell. The presence of oxygen vacancies is further verified by the EPR signal centered at g ≈ 2.003 and O 1 s XPS (Supplementary Fig. 11). Collectively, these characterizations fully demonstrate the core@shell structure of 5-Bi@Fe₂O₃, featuring a metallic Bi⁰ core, a V_O-enriched Fe₂O₃ shell, and an interfacial Bi–O linkage, which together provide a robust structure for efficient interfacial charge transfer and defect-mediated CO₂ reduction. As a comparison, a series of x-Bi@Fe₂O₃ was prepared by adjusting the amount of Bi precursor (Supplementary Fig. 12). On the other hand, Bi NCs, pure Fe₂O₃, and Bi-Fe₂O₃ were also prepared and characterized in Supplementary Fig. 13.

Light and thermal effects over the unique core@shell structure

After confirming the unique structure, we further explore the fundamental optical properties and solar-to-heat energy conversion ability. Three representative catalysts are shown in Supplementary Fig. 14. As displayed in Fig. 3a, the Fe₂O₃ catalyst exhibits a strong, broad absorption band below ~600 nm. Bi NCs display strong, broadband absorption covering the UV-vis-NIR region, which is characteristic of bismuth plasmonic and interband transitions. Compared to pure Fe₂O₃, the Bi-Fe₂O₃ catalyst and Bi@Fe₂O₃ catalyst maintain this excellent broadband light-harvesting ability of Bi NCs (300–2000 nm). Moreover, increasing the Bi content effectively enhances the light absorption of x-Bi@Fe₂O₃, particularly in the NIR region (Supplementary Fig. 15). This enhanced NIR absorption is crucial for the efficient cascaded utilization of the full solar spectrum.

Fig. 3. Full-spectrum light harvesting and nanoscale heat-localization behavior of 5-Bi@Fe₂O₃.

a UV-vis-NIR DRS spectra of Bi NCs, Fe₂O₃, Bi-Fe₂O₃, and 5-Bi@Fe₂O₃ catalysts. b Dependence of steady-state surface temperature of catalysts on light intensity. The error bars represent the standard deviation of five independent measurements of the same catalyst. c Monitoring of temperatures over different Fe₂O₃-based catalysts under continuous illumination, using a full-Arc 300 W Xe lamp with an optical density of 455 mW cm^–2. The error bars represent the standard deviation of three independent measurements of the same catalyst. d The infrared thermal images of 5-Bi@Fe₂O₃ catalyst under UV-vis, NIR, vis-NIR, and UV-vis-NIR light under a full-Arc 300 W Xe lamp with an optical density of 455 mW cm^–2. e Schematics of heat dissipation for Bi-Fe₂O₃ catalyst and the nanoscale greenhouse effect for Bi@Fe₂O₃ catalyst. f Electric field distribution and spatial distribution of temperature rise under different excitation wavelengths that were excited at an incident power density of 2.5 mW μm^–2 over Bi@Fe₂O₃. Scale bars, 50 nm.

The heat generated by the non-radiative relaxation of hot electrons can cause an increase in temperature in the catalyst bed (without any external heating), thereby accelerating the reaction³⁷, especially in gas-solid phase reaction systems without temperature control. Under the reaction conditions of the following photocatalytic performance tests, the equilibrium temperature at the center of the illuminated surface of the catalyst was measured using an embedded contact thermocouple attached to the surface of the photocatalyst (Supplementary Fig. 16a). As the light intensity increased, the surface equilibrium temperature of Fe₂O₃, Bi-Fe₂O₃ and 5-Bi@Fe₂O₃ continued to rise (Fig. 3b). Numerical simulation results also revealed that light power density significantly affects the average temperature of 5-Bi@Fe₂O₃ (Supplementary Fig. 17). Specifically, after 20 min of illumination, the temperature of the 5-Bi@Fe₂O₃ reached above 170 °C, higher than that of the blank comparsion (56 °C) (Fig. 3c). Compared to Fe₂O₃ and Bi-Fe₂O₃, the temperatures of 5-Bi@Fe₂O₃ are obviously higher, indicating that the core@shell structure exhibits the best photothermal effect^38–40. The average temperature of specific regions, which is in the infrared images measured using SmartView software (Fig. 3d and Supplementary Fig. 16b), is close to the temperatures measured by thermocouples in Fig. 3c. It is worth noting that the values measured by the thermocouple and infrared thermal imager represent the macroscopic surface temperatures of different catalysts, which are lower than the local temperatures of Bi NCs.

Importantly, compared to Bi-Fe₂O₃, the 5-Bi@Fe₂O₃ catalyst effectively suppresses heat dissipation from the Bi core to the surrounding environment after illumination (Fig. 3e). COMSOL Multiphysics simulations further reveal obvious enhancement of local electric fields and surface thermal energy in 5-Bi@Fe₂O₃ under NIR excitation, accompanied by a temperature increase of ~135 K (Fig. 3f and Supplementary Figs. 18–20). Additionally, to decouple thermal confinement from electronic effects, a Bi@SiO₂ core@shell structure with an optically transparent and electronically insulating shell was synthesized as a control (Supplementary Fig. 21). Although Bi@SiO₂ exhibits a certain photothermal temperature rise due to suppressed heat dissipation (Supplementary Fig. 22), its heating efficiency remains lower than that of Bi@Fe₂O₃ under identical illumination conditions (as shown in Supplementary Fig. 22b and Fig. 3c). These results indicate that the elevated temperature of 5-Bi@Fe₂O₃ originates not only from thermal insulation by the shell, but also from enhanced broadband light absorption and interfacial energy conversion enabled by the Fe₂O₃ shell. Based on simulation and experimental results, the observed temperature increase is more likely attributed to the enhanced broadband absorption of the Bi core after core@shell structure construction, as well as the intrinsically low thermal conductivity of the Fe₂O₃ shell layer. The resulting localized high-temperature environment promotes molecular collision frequency and reaction kinetics, thereby enhancing the solar-to-chemical energy conversion efficiency^41,42.

PCO₂R driven by coupling photo-/photothermal effect

This unique structure design, composed of Bi spheres and a loose Fe₂O₃ shell, aims to offer sufficient reaction driving force and to provide a channel for the mass transfer of reactants and products. To reveal the advantages of the unique structure design, PCO₂R performance was tested under full-Arc Xe lamp irradiation for all photocatalysts (Supplementary Fig. 23). Detailed experimental procedures for PCO₂R are shown in the Supporting Information. By optimizing the volume of H₂O and the partial pressure of CO₂ (Supplementary Fig. 24, with the entire system operating at atmospheric pressure), PCO₂R was performed under full-spectrum illumination at different light intensities without any external heating or any sacrificial agent or photosensitizer. Firstly, to distinguish the individual contributions of the core and the shell of unique structure in photocatalytic CO₂ to CH₄, a series of control experiments were performed under light power 50–455 mW cm^–2 for pure Fe₂O₃ and Bi NCs. At different light powers, nearly no products can be detected for pure Bi NCs, clearly indicating that Bi NCs itself is not active for PCO₂R (Supplementary Fig. 25). While pure Fe₂O₃ produces only a small amount of products and intermediates, which may be associated with defect-derived surface states rather than intrinsic band-edge reduction (Supplementary Figs. 26 and 27). Interestingly, after introducing Bi nanoparticles, CH₄ becomes the dominant product instead of CO (Fig. 4a and Supplementary Fig. 28), which may be attributed to an additional driving force from Bi nanoparticles to break the thermodynamic barrier for CO₂ to hydrocarbons. More importantly, constructing the Bi@Fe₂O₃ core@shell structure further markedly increases the CH₄ selectivity (Fig. 4a and  Fig. 29). Moreover, by investigating the effect of light intensity on the catalytic activity, it is found that the catalytic activity of the Bi-Fe₂O₃ increases with increasing light power, reaching a maximum (20.69 μmol g^–1 h^–1) at a critical value of 455 mW cm^–2 (Fig. 4b). In sharp contrast, the spherical 5-Bi@Fe₂O₃ exhibits much higher CH₄ production rate at each light power compared to Bi-Fe₂O₃, the maximum value of 273.81 μmol g^–1 h^–1 with a selectivity 98.60% at 455 mW cm^–2 (Fig. 4c). Besides these gaseous products (CO, CH₄), only trace amounts of H₂ (0.27 ~ 0.58 μmol g^–1 h^–1) were detected over these photocatalysts, indicating that the competing hydrogen evolution reaction (HER) was effectively suppressed⁴³.

Fig. 4. Photo-/photothermo-synergistic catalysis of PCO₂R.

a The PCO₂R performance over different catalysts at 455 mW cm^–2. Error bars in the figure represent standard deviations obtained from three independent measurements, which also include other figures about the performance measurements in this work. b Dependence of the PCO₂R performance on light intensity over Bi-Fe₂O₃ catalyst. c Dependence of the PCO₂R performance on light intensity over 5-Bi@Fe₂O₃ catalyst. d The long-term stability test over 5-Bi@Fe₂O₃ catalyst at 455 mW cm^–2 without external heating. e The comparison of the PCO₂R performance in this work with the previous works (see Supplementary Table 7), which directly convert CO₂, H₂O, and sunlight into compounds. f Average CH₄ production rates over 5-Bi@Fe₂O₃ catalyst under different conditions at 455 mW cm^–2. g CH₄ production rate of 5-Bi@Fe₂O₃ catalyst as a function of time under irradiation by full-Arc 300 W Xe lamp with different filters at 455 mW cm^–2. h CH₄ production rate of the 5-Bi@Fe₂O₃ catalyst at various temperatures and under 520 nm light irradiation with various intensities.

Verifying the origin of the products is another critical step for PCO₂R. As shown in Supplementary Fig. 30 and Supplementary Table 4, the presence of the catalyst, light, CO₂, and H₂O is indispensable for the PCO₂R over 5-Bi@Fe₂O₃. To further corroborate the carbon source, we performed isotopic labeling experiments using ¹³CO₂ as the feedstock. As shown in Supplementary Fig. 31a, when ¹²CO₂ and ¹³CO₂ were used as the respective feed gases, mass signals at m/z = 28 and 29 were assigned to ¹²CO and ¹³CO, respectively. Similarly, signals at m/z = 16 and 17 were attributed to ¹²CH₄ and ¹³CH₄. These results confirm that the reduced carbon products were only generated by PCO₂R rather than other reactions. Furthermore, previous studies indicated that the reduced carbon products were difficult to generate in the photocatalytic system without adding water⁴⁴, thus the water was indispensable in the PCO₂R. Therefore, an H₂¹⁸O isotope-labeling experiment was performed to investigate whether water participates in the overall reaction (Supplementary Fig. 32a). Although there might be oxygen atom exchange between CO₂ and H₂O, the production of ¹⁸O₂ implied the involvement of water in the whole PCO₂R process. Moreover, quantitative analysis of the evolved gases revealed the concurrent formation of O₂ throughout the CO₂ reduction process (Supplementary Fig. 32b, c), with no detectable by-products such as H₂O₂. This confirms that the photocatalytic conversion between CO₂ and H₂O proceeds through a complete and closed redox cycle. Moreover, deuterium oxide (D₂O) was employed in the reaction to trace the origin of the H element in the produced CH₄. The product peak at m/z = 18 was indexed to CD₄ (Supplementary Fig. 31b) when using D₂O in the reaction, indicating that the proton in CH₄ comes from H₂O. This result further manifests that the formation of CH₄ is a multistep hydrogenation process and that the proton is provided by H₂O^45–47.

To explore the high stability of the core@shell structure with the proposed design, a comprehensive characterization was performed. The thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) results demonstrate that the 5-Bi@Fe₂O₃ catalyst exhibits excellent high-temperature tolerance (Supplementary Fig. 33). More importantly, the 5-Bi@Fe₂O₃ catalyst nearly retained its original morphology rather than collapsing after PCO₂R reaction (Supplementary Fig. 34a, b). The XRD patterns of 5-Bi@Fe₂O₃ after PCO₂R reaction exhibited no discernible changes (Supplementary Fig. 34c), confirming the strong thermal stability of the catalyst structure. Furthermore, according to the XPS and XAS spectra of the 5-Bi@Fe₂O₃ catalyst after PCO₂R reaction (Supplementary Figs. 35 and 36; and Supplementary Tables 5 and 6), the valence state and coordination environment of Bi and Fe in the core@shell catalyst did not clearly change, indicating that the Bi core does not undergo oxidation during the reaction, and the Fe-V_O active sites keep stable in the reaction process (Supplementary Fig. 37). Concurrently, 21-h consecutive testing and another 11-h consecutive testing for 5-Bi@Fe₂O₃ were carried out (Fig. 4d). Between the two tests, the 5-Bi@Fe₂O₃ was kept in Ar gas at room temperature for 45 days. During the first 21-h test, there is almost no significant degradation of PCO₂R activity. Surprisingly, the next 11-h test still exhibits stable selectivity in spite of the slightly fluctuating yield of CH₄. These results clearly confirm that the outer Fe₂O₃ shell effectively protects the encapsulated Bi nanoparticles from thermal sintering at elevated temperatures⁴⁸. The core@shell design not only provides a super-photothermal effect through the thermal insulation and the electronic-structure modulation of the nanoporous shell, but also enhances the thermal stability of the Bi nanoparticles against sintering due to spatial confinement within the Fe₂O₃ sheath. This enhanced photothermal effect and thermal stability are expected to synergistically improve the photothermal catalytic CO₂ activity. We compared the performance in this work to typical reports of PCO₂R without an external heater. Impressively, to the best of our knowledge, this represents a record-high performance for noble-metal-free photocatalysts in the reduction of CO₂ with water to CH₄, substantially surpassing most previously reported systems (Fig. 4e and Supplementary Table 7). And the AQE at 850 nm is promoted to 0.64% (Supplementary Fig. 38). Such a superior photoconversion efficiency under low-energy NIR light illumination, which is even comparable with the efficiencies of the representative works in PCO₂R with high-energy photons, achieves the record-high AQE of PCO₂R under NIR light illumination (Supplementary Table 8).

To distinguish the respective contributions of light and heat, the PCO₂R over the 5-Bi@Fe₂O₃ catalyst was conducted under four distinct conditions: (i) in the dark at room temperature; (ii) under 300 W Xe lamp illumination with a broadband pass filter (λ < 800 nm) to exclude infrared light; (iii) under full-spectrum irradiation from the 300 W Xe lamp; and (iv) in the dark at 220 °C. As shown in Fig. 4f, PCO₂R cannot occur in the dark at either room temperature or at an elevated temperature of 220 °C, which unambiguously confirms that the PCO₂R over 5-Bi@Fe₂O₃ is exclusively light-driven. Under full-spectrum irradiation from the 300 W Xe lamp, the 5-Bi@Fe₂O₃ catalyst exhibited a remarkable average CH₄ production rate of 273.81 μmol g^–1 h^–1 over a 60-min reaction period, during which the catalyst surface temperature reached approximately 212 °C. This rate is 2.7 times higher than that achieved under purely photocatalytic conditions (i.e., with IR light excluded, λ < 800 nm), where the surface temperature only reached ~73 °C. These results powerfully demonstrate a synergistic enhancement between photocatalysis and photothermal catalysis, leading to a superior solar energy utilization efficiency.

Further investigation into the wavelength-dependent activity revealed the distinct roles of different spectral regions. The relative spectral distribution for this series of tests, labeled I through IV, was shown in Supplementary Fig. 39 for (I) no filter (UV-vis-NIR), (II) l > 420 nm (vis-NIR), (III) l > 800 nm (NIR), and (IV) l < 800 nm (UV-vis). The temperature variations under different spectral regions were visualized using an infrared (IR) camera (Fig. 3d). As exhibited in Fig. 4g, both visible and NIR light can independently drive the reaction, while filtering out either the NIR portion (λ < 800 nm) or the UV-vis portion (λ > 800 nm) leads to a dramatic drop of the activity. The UV-vis photons are primarily responsible for initiating the photochemical pathway via bandgap excitation in the Fe₂O₃ shell. Meanwhile, the NIR photons are selectively harvested by the Bi core to generate energetic hot electrons and a strong photothermal effect. This coupling of a semiconductor component, confirmed as indispensable by control experiments with Bi NCs, and a plasmonic nanoheater is the key to the catalyst’s exceptional performance in utilizing low-energy photons. To further reveal the influence of thermal effect, temperature-dependent photocatalytic tests were conducted by systematically varying the reaction temperature under 520 nm illumination while tuning the light intensity. As shown in Fig. 4h, under the same light intensity, the catalytic performance of 5-Bi@Fe₂O₃ exhibited a clear correlation with the reaction temperature, indicating that light-induced heat can promote the PCO₂R process. Furthermore, a similar positive correlation between performance and reaction temperature was observed across various light intensities, indicating that at a constant reaction temperature, increasing the light intensity enhances the catalytic activity. This finding aligns with the observed temperature varying trends (Fig. 3b) and catalytic performance changes (Fig. 4a, b) as a function of light intensity.

Charge carrier dynamics

To elucidate the charge carrier dynamics over 5-Bi@Fe₂O₃ for the PCO₂R process, femtosecond transient absorption (fs-TA) spectroscopy was performed under dual excitation at 560 nm and 870 nm. Under 560 nm excitation, pristine Fe₂O₃ exhibits a broad photoinduced absorption (PIA) in the 500–800 nm region (Fig. 5a)⁴⁹, which is attributed to the absorption of trapped photogenerated charge carriers. In contrast, 5-Bi@Fe₂O₃ displays a modified transient spectral profile with a weakened PIA at short wavelengths and a notable ground-state bleach (GSB) around ~700 nm (Fig. 5b), consistent with interfacial electronic coupling between the Bi core and Fe₂O₃ shell. Upon 870 nm excitation, a distinct plasmon-induced bleach in the NIR region (800–1000 nm) is observed for 5-Bi@Fe₂O₃ (Fig. 5c), whereas negligible signals are detected for pristine Fe₂O₃, confirming that the NIR response originates exclusively from the LSPR of Bi nanospheres. Kinetic fitting analysis further reveals that the carrier lifetimes of 5-Bi@Fe₂O₃ are markedly longer than pristine Fe₂O₃ (Supplementary Fig. 40), particularly under NIR illumination. This lifetime extension indicates suppressed electron-hole recombination enabled by ultrafast interfacial charge separation at the Bi/Fe₂O₃ interface, which stabilizes plasmon-generated hot electrons and promotes their participation in subsequent surface reactions.

Fig. 5. Charge carrier dynamics.

2D pseudo-color plot for a pristine Fe₂O₃ pumped at 560 nm, b 5-Bi@Fe₂O₃ catalyst pumped at 560 nm, and c 870 nm, respectively. KPFM images of 5-Bi@Fe₂O₃ catalyst d in the dark, e under UV-vis, and f under UV-vis-NIR illumination. g Corresponding to the surface potential analysis of the 5-Bi@Fe₂O₃ catalyst. Schematic illustration of the charge transfer processes of the 5-Bi@Fe₂O₃ catalyst during PCO₂R under h UV-vis and i UV-vis-NIR excitation conditions. k_ET represents the rate constant for hot-electron injection from Bi core to the interfacial accumulation layer; k_HT1 denotes the hole transfer rates of hole-trapping processes between Bi core and Fe₂O₃ shell; k_HT2 represents the rate constant of the corresponding holes from the Fe₂O₃ are consumed mainly for water oxidation; k_(e-h)1 and k_(e-h)2 correspond to the electron-hole recombination rate constant in Bi core and Fe₂O₃ shell, respectively; k_CO2R represents CO₂ reduction reaction rate constant.

To further visualize the interfacial charge redistribution, light-assisted Kelvin probe force microscopy (KPFM) was employed to investigate the surface potential distribution of 5-Bi@Fe₂O₃ (Supplementary Fig. 41). As shown in Fig. 5d–g, the surface potential (CPD) of both the Bi core and Fe₂O₃ shell under UV-vis-NIR irradiation exhibits a significant positive shift compared to that under dark or UV-vis illumination, indicating enhanced electron accumulation under full-spectrum excitation. This enhanced CPD response is attributed to the additional absorption of NIR photons, which activates the LSPR of Bi and generates a larger population of energetic hot electrons. Additionally, the transient photocurrent of 5-Bi@Fe₂O₃ under full-spectrum illumination was markedly higher than that under UV-vis irradiation (Supplementary Fig. 42). These results demonstrate that UV-vis and NIR light synergistically enhance charge separation and carrier transport in 5-Bi@Fe₂O₃, thereby providing an optimal electronic environment for plasmon-assisted CO₂ reduction.

Based on the above complementary spectroscopic analysis results, a simplified PCO₂R model based on the cascaded utilization of full-solar spectrum-driven over the 5-Bi@Fe₂O₃ core@shell catalyst is depicted in Fig. 5h, i. Due to the favorable work function alignment (Supplementary Figs. 43 and 44), photoexcited electrons in the Bi core tend to transfer toward the Bi/Fe₂O₃ junction and accumulate in an interfacial electron-rich region. PCO₂R preferentially occurs on the interfacial electron-accumulation layer associated with Fe-V_O active sites. Specifically, under UV-vis illumination (Fig. 5h), the electrons generated from Bi interband excitation are transferred to the interfacial accumulation layer (rate constant k_ET)⁵⁰, which is corroborated by theoretical computation electron redistribution and electrostatic potential (Supplementary Fig. 45). These energetic hot electrons drive CO₂ reduction toward CO/CH₄ (rate constant k_CO2R), which competes with the electron-hole recombination (rate constant k_(e-h)1) of Bi NC and in the Fe₂O₃ region (rate constant k_(e-h)2). The corresponding holes from the Fe₂O₃ are consumed mainly through water oxidation (rate constant k_HT2) or hole-trapping processes (maintaining charge balance, rate constant k_HT1). Under UV-vis-NIR illumination (Fig. 5i), the Bi LSPR effect triggers more energetic hot electrons (kˊ_ET). These hot electrons are efficiently injected and strongly trapped with reactant molecules in the interfacial accumulation layer, creating a significantly higher electron density than under UV-vis illumination. And the large localized electric field in 5-Bi@Fe₂O₃ promotes charge separation and transfer, thereby enhancing the reduction potential at the reaction interface. Moreover, the prolonged lifetime of photogenerated electrons under NIR excitation enables efficient utilization of the full solar spectrum, while concurrently creating a “self-heating greenhouse bed” effect at the nanoscale (Fig. 3e). This synergistic photothermal enhancement is particularly critical for overcoming the high kinetic barriers associated with multielectron CO₂ activation and multielectron-proton CH₄ production. Therefore, the CH₄ yield under full-spectrum irradiation is significantly higher than that obtained under UV-vis or NIR light alone, which is consistent with the PCO₂R performance (Fig. 4g).

Proposed reaction mechanism for enhanced photoactivity

After elucidating the charge carrier dynamics mechanism in this unique core@shell structure, it is necessary to explore deep insight into the origins of the enhanced activity of 5-Bi@Fe₂O₃ in PCO₂R with the support of in situ experimental results and theoretical calculations. As widely reported, oxygen vacancies often serve as the primary active sites for CO₂ adsorption and activation⁵¹. Generally, the generated oxygen vacancies can act as reactive centers that modulate the local charge density distribution, thereby increasing the number of adsorbed and activated CO₂ molecules and lowering the energy barrier for the CO₂ reduction reaction. Firstly, to validate this enhanced CO₂ adsorption capacity, CO₂ adsorption isotherms and temperature-programmed desorption (TPD) experiments were performed. CO₂ adsorption isotherms demonstrated that the adsorption capacity of 5-Bi@Fe₂O₃ is stronger than that of Bi NCs, Fe₂O₃, and Bi-Fe₂O₃ (Supplementary Fig. 46a), which is likely due to the defect-enriched shell chemistry and the porous curved shell microenvironment. This result indicates that 5-Bi@Fe₂O₃ may provide more active sites for adsorbing and converting CO₂ molecules during the PCO₂R process, thus facilitating the occurrence of the reaction. As the shell composition was optimized, the overall CO₂ adsorption capacity increased significantly, consistent with the independent CO₂ adsorption results (Supplementary Fig. 46b). Notably, the similarity in CO₂ adsorption isotherms among the core@shell samples with different core loadings suggests that CO₂ activation on the interfacial sites is sufficiently efficient and does not constitute the rate-determining step in the catalytic reaction. The CO₂ adsorption properties of the various catalysts were further evaluated by CO₂-TPD, as shown in Fig. 6a. The desorption profiles can be broadly divided into three temperature regions, namely 50 °C–300 °C, 300 °C–450 °C, and above 450 °C, which are generally associated with weak, moderate, and strong CO₂ adsorption sites, respectively. Compared with the Fe₂O₃ and Bi-Fe₂O₃ catalysts, the 5-Bi@Fe₂O₃ catalyst showed markedly enhanced CO₂ adsorption after the construction of the spherical core@shell structure and active component incorporation. In particular, the appearance of obvious medium- and high-temperature desorption features indicates the generation of stronger CO₂ binding sites (a broad peak at 300 °C–450 °C and a sharp peak at >450 °C). Strong CO₂ adsorption led to a high CO₂ conversion rate (Fig. 4a). These sites are mainly attributed to the abundant interfacial regions formed at the core@shell junction, which effectively promote CO₂ adsorption and activation. These results powerfully underscore the critical importance of a CO₂-activating support (Fe-V_O active centers) for promoting efficient PCO₂R. CO is a key reaction intermediate during CO₂ conversion. As shown in Fig. 6b, the evolution of CO adsorption on the spherical core@shell catalysts closely followed the trend observed for CO₂ adsorption, indicating that the construction of the core@shell structure effectively increased both the number and strength of CO adsorption sites. These enhanced CO-binding sites, mainly originating from the abundant core@shell interfacial regions (Fe-V_O active centers), stabilized the *CO intermediates and favored their subsequent hydrogenation or dissociation rather than desorption. Consequently, CO selectivity was suppressed, while CH₄ formation was significantly promoted. To gain deeper insight into the protonation behavior of *CO, the projected crystal orbital Hamilton population (COHP) analysis was also performed. As shown in Supplementary Fig. 47, the spherical 5-Bi@Fe₂O₃ catalyst exhibits a markedly more positive integrated COHP (-ICOHP) value than the Fe₂O₃ catalyst (3.92 eV vs 0.07 eV), indicating a stronger interfacial electronic coupling between the core@shell junction and the *CO intermediate. This strengthened bond interaction arises from the unique core@shell interface and effectively stabilizes *CO, thereby facilitating its subsequent protonation and promoting the downstream hydrogenation reaction.

Fig. 6. Mechanism analysis of photoactivity enhancement.

a CO₂-TPD profile of different catalysts. b CO-TPD profiles of different catalysts. c In-situ XPS analysis of the Fe 2p signals in 5-Bi@Fe₂O₃ under full-spectrum light irradiation. d In-situ DRIFTS of Bi-Fe₂O₃ catalyst. e In-situ DRIFTS of 5-Bi@Fe₂O₃ catalyst. f Gibbs Free energy diagram for PCO₂R over the pristine Fe₂O₃ surface and the Bi/Fe₂O₃ interfacial model, both with adsorbed CO₂ and an explicit H₂O solvent layer. (Blue: Bi; Orange: Fe; Red: O; White: H; Gray: C). g Mechanism of PCO₂R driven by the synergy of photochemical and photothermal effects based on the cascaded utilization of full-solar spectrum.

Combined with charge carrier dynamics in the above section, to further confirm the proposed mechanism, in situ near-ambient pressure XPS (NAP-XPS) was conducted to deeply explore the specific reconstruction mechanism of the Fe-V_O active centers at the reaction interface. As shown in Fig. 6c, the Fe 2p_1/2 and 2p_3/2 peaks of the Fe₂O₃-based catalysts shifted toward lower binding energies under light irradiation, accompanied by an obvious increase in the surface Fe²⁺ concentration compared to the dark condition. This observation indicates that hot electrons generated in the Bi core were efficiently transferred to the interface between the Fe₂O₃ shell and the Bi core. Notably, the 5-Bi@Fe₂O₃ sample exhibited a substantially higher Fe²⁺/Fe³⁺ proportion than the Bi-Fe₂O₃ catalyst (Supplementary Table 9), confirming that plasmon-induced hot-electron transfer from Bi nanoparticles promoted the formation of abundant Fe-V_O active centers. These sites provided higher electron density for CO₂ adsorption and activation (Supplementary Figs. 48), thereby accelerating the superior catalytic activity observed (Fig. 4c). The In-situ EPR measurements further supported this conclusion (Supplementary Fig. 49). The V_O-associated signal increased slightly as photoexcited electrons were trapped, and a characteristic peak at g = 2.003 appeared under illumination, demonstrating the presence of excited electrons. In contrast, the EPR spectra of Bi-Fe₂O₃ before and after Xe lamp irradiation were nearly identical, suggesting less efficient photoinduced electron migration. These results highlight the pivotal role of Fe-V_O centers in PCO₂R by simultaneously enhancing the intrinsic reactivity of electron-deficient Fe sites and promoting charge separation and transfer. Moreover, under light irradiation, the Bi 4 f peaks exhibited an evident positive shift (Supplementary Fig. 50). Combined with the Fe XPS results, this demonstrates that hot electrons in the Bi core are injected into the Fe₂O₃ shell during PCO₂R.

Subsequently, in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements were employed to fundamentally study the reaction intermediates involved in PCO₂R. The infrared signals of intermediates for photoreduction CO₂ over 5-Bi@Fe₂O₃ after illumination demonstrated more abundant characteristic fingerprints compared to pure Fe₂O₃ and Bi-Fe₂O₃ (Fig. 6d, e and Supplementary Figs. 27 and 51), which is well in line with its higher performance and selectivity toward photoreduction CO₂ (Fig. 4a). After introducing CO₂, a peak at 1378 cm⁻¹ appeared, which could be attributed to the monodentate carbonate ion (m-CO₃²⁻)⁵². The asymmetric stretching vibration peaks of *HCO₃ were observed at 1318 cm⁻¹. Multiple new peaks appeared after illumination. The peak at 1269 cm⁻¹ can be attributed to the stretching vibration of CO₂⁻. The infrared signals at 1590 cm⁻¹ and 1719 cm⁻¹ can be attributed to the *COOH group, which is a critical intermediate in the conversion of CO₂ to other solar fuels^53,54. The increase in the peak intensity of the absorption band at 2082 cm⁻¹ and 2134 cm⁻¹ confirmed the formation of *CO intermediates in the reaction. Importantly, the absorption bands at 1011, 1473, and 1112 cm⁻¹ correspond to the *CHO, *CH₂O, and *CH₃O species, respectively, which are key intermediates in CH₄ formation. Furthermore, the peak at 3026 cm⁻¹ is attributed to *CH₃ stretching vibrations, clearly indicating CH₄ formation.

The density functional theory (DFT) calculations were performed to elucidate the thermodynamic origins of the enhanced activity and selectivity in the PCO₂R to CH₄ over 5-Bi@Fe₂O₃. To explicitly probe the interfacial electronic effect and solvent participation, the simplified slab models with explicit water molecules were constructed (details in Methods; optimized structures of pure Fe₂O₃ and 5-Bi@Fe₂O₃ in Supplementary Figs. 52−54). The calculated Gibbs free-energy for CO₂ reduction with H₂O toward CH₄ reveals a markedly more favorable thermodynamic advantage over 5-Bi@Fe₂O₃ than over Fe₂O₃ (Fig. 6f and Supplementary Figs. 55 and 56). Specifically, CO₂ was adsorbed on the surface of 5-Bi@Fe₂O₃ and Fe₂O₃ to produce *CO₂, and then *CO₂ was hydrogenated to produce *COOH. Owing to the high energy barrier, the first proton-coupled electron transfer for *COOH generation was considered a rate-limiting step. Importantly, the formation energy of *COOH on 5-Bi@Fe₂O₃ is 0.67 eV, distinctly lower than that on pristine Fe₂O₃ (0.85 eV), indicating that the electron-rich Bi/Fe₂O₃ interface significantly decreases the energy barrier for CO₂ activation. After that, *COOH was then protonated and dehydrated to form a stable intermediate *CO, which was associated with the C − O bond breakage through a proton-assisted electron transfer process⁵⁰. After *CO formation, pristine Fe₂O₃ tends to favor CO desorption rather than further hydrogenation (0.20 eV vs 0.98 eV). In contrast, the *CO intermediate is more strongly stabilized for 5-Bi@Fe₂O₃, and the subsequent hydrogenation to *CHO becomes thermodynamically preferred over CO desorption, thereby resulting in the pathway toward deep hydrogenation. The subsequent hydrogenation steps (*CHO → *CH₂O → *CH₃O → *+CH₄ ↑ ) proceed with overall downhill energetics and a small thermodynamic barrier, consistent with the enhanced CH₄ selectivity. Notably, these theoretical trends are in good agreement with the experimental results. Specifically, the increased Fe²⁺/Fe³⁺ proportion under irradiation suggests an electron-enriched, defect-associated surface environment, while time-resolved DRIFTS shows the continuous accumulation of *CHO/*COOH-related bands on 5-Bi@Fe₂O₃. This is consistent with the predicted stabilization of key hydrogenation intermediates and the suppression of premature CO desorption. Therefore, the combined DFT and experiment analyses provide a clear logical framework for highly selective CH₄ formation on the Bi@Fe₂O₃ core@shell catalyst.

Based on these results, we therefore propose a CO₂-to-CH₄ mechanism driven by the synergistic interplay between photochemical effect and photothermal effect over the Bi@Fe₂O₃ catalyst (Fig. 6g). Specifically, the Bi core absorbs vis-NIR light to excite LSPR hot electrons, while the Fe₂O₃ shell absorbs UV-visible light to generate electron-hole pairs. The non-radiative relaxation of these excited electrons generates localized heat, elevating the temperature at the interface between the Bi core and Fe₂O₃ shell. The Fe₂O₃ shell, possessing intrinsically low thermal conductivity, effectively confines this energy to establish a highly localized photothermal environment. Crucially, the hot electrons injected from Bi are continuously trapped by Fe-V_O sites (Fe³⁺ to form transient Fe²⁺ species), which serve as highly active centers for CO₂ activation. These Fe-V_O sites are not only the platform of the initial activation for CO₂ to the *COOH intermediate but also support the subsequent proton-coupled electron transfer deep reduction steps (*COOH → *CO → *CHO → *CH₂O → *CH₃O → *CH₄). Under continuous illumination, plasmon-induced hot-electron injection and localized photothermal heating sustain a dynamic Fe³⁺/Fe²⁺ redox cycle, ensuring the persistent availability of Fe-V_O sites during multi-electron proton-coupled transfer steps toward CH₄ formation. Consequently, the synergistic coupling of plasmonic nanoheating and hot-electron injection based on the Bi@Fe₂O₃ structure not only accelerates charge-carrier dynamics but also promotes the deep reduction of CO₂ and H₂O to CH₄, as confirmed by in situ spectroscopy and control experiments.

Discussion

In this work, we designed a noble-metal-free core@shell structure (Bi@Fe₂O₃), which enables cascaded utilization of the full solar spectrum. The metallic Bi core acts as a plasmonic transducer: interband excitation under UV-vis light and LSPR excitation under vis-NIR light generate energetic electrons together with localized nanoheating. Meanwhile, the Fe₂O₃ shell provides abundant adsorption/activation sites and thermally confines heat to create a “nanoscale greenhouse” microenvironment. This interplay promotes ultrafast hot-electron injection from the Bi core to the Fe₂O₃ shell, leading to long-lived charge separation and electron accumulation at defect-associated interfacial sites (Fe-V_O sites), which serve as highly active centers for CO₂ activation and deep hydrogenation toward CH₄ formation requiring multielectron-proton coupling steps. Supported by in situ spectroscopy and temperature-dependent analyses, these findings validate the proposed nanoscale greenhouse effect mechanism, and reveal how interfacial charge transfer (photochemical effect) and plasmon-induced nanoheating (photothermal effect) synergistically drive the highly selective production of CH₄ from CO₂ and H₂O. This work offers an effective strategy for solar-to-fuel conversion based on full-spectrum utilization, and provides design insights for future artificial photosynthesis systems.

Methods

Materials and synthesis

Chemicals

All chemicals are of analytical grade and used without further purification. Polyvinyl pyrrolidone (PVP, molecular weight of 360,000) was purchased from Aladdin Biochemical Technology Co. Ltd. (Shanghai, China); sodium borohydride (NaBH₄), ethanol, bismuth nitrate pentahydrate (Bi(NO₃)₃·5H₂O) and ferric nitrate nonahydrate (Fe(NO₃)₃·9H₂O) were obtained from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). The deionized water was supplied by a Millipore system (Outlet water resistivity > 18 MΩ cm at 25 °C) in all experiments.

Preparation of Bi@Fe₂O₃ spherical nanostructures

In a typical procedure, 10 mmol of urea was dissolved in ethylene glycol (EG) under continuous Ar purging. Subsequently, 0.5 mmol of Fe(NO₃)₃·9H₂O was added under stirring for 2 h, obtaining solution A. Additionally, 1 mmol of Bi(NO₃)₃·5H₂O was dispersed in 100 mL EG in a three-necked flask equipped with a reflux condenser, thermometer, and gas inlet/outlet. An appropriate amount of PVP was then added. The mixture was ultrasonicated for 30 min and then vigorously stirred for 1 h. Ar was continuously bubbled through the system for an additional 1 h. The reaction mixture was heated to 150 °C and maintained for 4 h under an Ar atmosphere. After cooling to 70 °C, the resulting solution was marked as solution B. Subsequently, solution A was added dropwise into solution B under an Ar atmosphere. After addition, the reaction was maintained at 150 °C for 8 h under continuous stirring and Ar flow. Then, the mixture was cooled to room temperature and allowed to stand for several hours. The obtained mixture was washed sequentially with ethanol and deionized water, followed by vacuum drying at 60 °C for 12 h. The dried powder was then transferred into a tubular furnace and annealed at 250 °C for 4 h under an Ar atmosphere with a ramping rate of 1 °C/min. The final product, Bi@Fe₂O₃, featuring a core@shell architecture, was obtained. By varying the amount of Bi(NO₃)₃·5H₂O precursor, a series of samples with different Bi (1-Bi@Fe₂O₃, 3-Bi@Fe₂O₃, 5-Bi@Fe₂O₃, 6-Bi@Fe₂O₃, and 7-Bi@Fe₂O₃) were synthesized accordingly.

Preparation of Fe₂O₃ catalyst

A pristine Fe₂O₃ sample was synthesized under identical conditions but without adding the Bi precursor. All products were stored in a desiccator before further characterization and catalytic evaluation.

Preparation of Bi-Fe₂O₃ flat nanostructures

Fe(NO₃)₃·9H₂O (2 mmol) and urea (6 mmol) were dissolved in 40 mL of deionized water and subsequently treated hydrothermally at 120 °C for 2 h, obtaining FeOOH precursor. The precursor mixture was washed, dried, and calcined at 350 °C in air for 2 h to obtain crystalline Fe₂O₃. Then, 100 mg Fe₂O₃ was dispersed in 50 mL EG with PVP under ultrasonication. After N₂ purging (30 min), Bi(NO₃)₃·5H₂O (0.5 mmol, dissolved in 5 mL EG) was added, followed by dropwise addition of NaBH₄ (0.1 M, 10 mL) at 80 °C. After being stirred for 2 h, the mixture was washed alternately with ethanol and water, then vacuum-dried at 60 °C. Finally, the dried mixture was annealed at 250 °C for 1 h under flowing Ar with a ramping rate of 2.5 °C/min to stabilize the Bi nanoparticles.

Fundamental characterizations

The powder X-ray diffraction (PXRD) spectra were acquired on a PANalytical X’pert MPD Pro diffractometer with Ni-filtered Cu Kα radiation (λ = 0.1538 nm, 40 kV × 40 mA). The microcrystalline structure and surface characteristics were examined by a JEOL JSM-7800F field emission scanning electron microscope (FE-SEM) and an FEI Tecnai G2 F30 S-Twin transmission microscope (TEM). The energy-dispersive X-ray spectroscopy (EDS) analysis was performed using the installed energy-dispersive X-ray detector, OXFORD MAX-80. SEM and EDS images were recorded on the XL-30 ESEM-FEG SEM. XPS measurements were conducted using a Kratos spectrometer (Axis Ultra DLD) with monochromatic Al Kα radiation (hν = 1486.69 eV). The in situ XPS was recorded using United States-Thermo Fisher SCIENTIFIC ESCALAB 250Xi. The X-ray source was an aluminum target (Al Kα line energy of 1486.69 eV after monochromator filtration), and the detection pressure was 2 × 10⁻⁸ Pa. The UV-vis diffuse reflectance spectrum (DRS) was acquired using a Hitachi double-beam UV4100 UV-vis-NIR spectrophotometer equipped with an integrating sphere in which BaSO₄ acted as the background. Electron paramagnetic resonance (EPR) measurements were performed on a Bruker A300 spectrometer.

Ultrafast spectroscopy characterization

Femtosecond transient absorption spectroscopy was carried out using a PHAROS PH2 Femtosecond laser system. The excitation source was a PHAROS PH2 Femtosecond laser system (LIGHT CONVERSION, Lithuania, 35 fs pulse width, 1 kHz repetition rate, 3 mJ per pulse). The output beam was divided into two paths to feed a pair of optical parametric amplifiers (TOPAS-PRIME) for generating the pump and probe pulses. A white light continuum (350–950 nm) was produced by focusing the 1300 nm signal from one OPA (TOPAS1) onto a sapphire plate, then filtered with a short-pass filter to limit the spectral window. The pump pulses at 560 and 870 nm were generated from the second OPA (TOPAS2), with fluences adjusted to 5.0 mJ cm^–2 and 0.7 mJ cm^–2, respectively. Temporal resolution was achieved by introducing a delay line into the pump beam path, allowing the capture of differential transmission spectra over time. To correct for spectral dispersion in the white light probe, the optical Kerr effect of the substrate was measured following a standard polarization-resolved method. This involved setting the pump polarization at 45° relative to the probe and analyzing the probe polarization changes after transmission through the sample. These polarization shifts reflected pump-induced anisotropy in the substrate’s refractive index. For nanosecond-scale measurements in the NIR, a separate nanosecond TAS system was employed. Here, excitation was provided by 560 and 870 nm pulses from a Nd:YAG laser, while probe light came from a tungsten–halogen lamp. Transient NIR absorption spectra were recorded using a liquid nitrogen-cooled InSb detector, operating in single-channel mode with spectral scanning via a grating. Decay kinetics at each wavelength were captured using a Ni-Scope digital oscilloscope (8-bit, 100 MHz bandwidth, 30,000 data points).

Kelvin probe force microscopy measurements

The atomic force microscope (AFM) and Kelvin probe force microscope (KPFM) signals were detected by an SPM-9700HT (SHIMADZU, Japan) AFM instrument. The powder sample was dispersed in ethanol and then dropped on a Silicon chip. Au was used for the potentiometric calibration. CPD changes under dark and illumination were recorded to explore photoinduced charge dynamics. According to the following Eq. (1), the CPD can be calculated as:

$$V_{CPD}=\frac{{\varphi }_{sample}-{\varphi }_{tip}}{e}$$ 1

where ${\varphi }_{\mathrm{sample}}$ is the work function of the sample, ${\varphi }_{tip}$ is the work function of the AFM tip, and e is the elementary charge.

In-situ DRIFTS characterization

In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurement was performed on a Thermo Fisher Nicolet iS50 FTIR spectrometer equipped with a mercury-cadmium-telluride (MCT) detector cooled by liquid nitrogen and a DRIFTS accessory. Each spectrum was collected over the range of 4000 to 600 cm⁻¹ at a resolution of 0.964 cm⁻¹. The as-prepared samples were compressed at the bottom of an in situ sample cell and then degassed for 1 h under an Ar atmosphere at 120 °C to remove the absorbed species before the photocatalytic experiments. After cooling down to room temperature, humidified CO₂ (CO₂/H₂O = 4/1) was flowed into the specimen chamber to saturate the surface absorption of the catalyst. The DRIFTS spectra were collected after a certain irradiation duration. The simulated light source (UV-vis and UV-vis-NIR) was consistent with that in the photocatalytic experiment above.

Photocurrent and Mott–Schottky measurements

First, the original ITO glass was washed with acetone (30 min), ethanol (30 min), and water (30 min) by ultrasonication. The 5.0 mg catalyst was dispersed in 0.5 mL of ethanol and then treated with ultrasonics for 5 h. The obtained suspension was added to the 1.5 cm × 2 cm ITO glass drop by drop. Before the photoelectrochemical measurement, the ITO glass was treated in a vacuum at 150 °C for 2 h. The CHI 660B electrochemical workstation equipped with three electrodes was used for photoelectrochemical experiments. An aqueous solution of 0.5 M Na₂SO₄ was used as the electrolyte. Using standard three electrodes, the catalyst-coated ITO was used as the working electrode, Pt foil was used as the counter electrode, and the Ag/AgCl electrode was used as the reference electrode. The Mott-Schottky diagram was obtained at the frequencies of 1000 and 1500 Hz in the range of −1 ~ 1 V vs Ag/AgCl.

PCO₂R measurement

The PCO₂R experiments were carried out in a custom-made reactor with a quartz glass top (illuminated area approximately 1 cm²). 5 mg of catalyst was uniformly spread on a glass fiber cloth after plasma cleaning and then placed into the reactor. Except for the temperature comparison experiment, all other experiments were conducted at room temperature (without any external heat source). A PLS-SXE300 Xe lamp (Beijing Perfect-Light Technology Co. Ltd.) with different reflectors and cut filters was used to simulate UV-vis, NIR, and full-spectrum. The light intensities can be adjusted continuously by changing the current of the xenon lamp and the optical path to the reactor. The light intensity on the sample surface was measured by an optical power meter (CEL-NP2000-2A, Beijing China Education Au-light Co., Ltd.), and the measuring range of wavelength was 190 − 11000 nm. Before irradiation, the whole reaction system was purged with CO₂ and H₂O vapor for 1 h to remove the gaseous impurity. High-purity CO₂ was introduced into the reaction chamber through a mass flow controller, while H₂O was injected into the chamber using a syringe pump, which can precisely control the amount of each reactant. The temperature of the catalyst bed was measured by a K-type thermocouple (DT-8891E) and an infrared thermometer (Fisher Scientific). The gas products were qualitatively and quantitatively detected using an Agilent GC-8890 gas chromatograph equipped with a TDX-01 column, a thermal conductivity detector (TCD), and a flame ionization detector (FID) with ultra-high purity argon as the carrier gas. The liquid products were quantified by a liquid chromatography (1260 Infinity II LC System, Agilent Technologies, Inc, Santa Clara, CA, USA) with a refractive index detector (RID) and a column from Bio-Rad.

To detect the difference between photothermal and traditional thermal catalysis, an IKA heating plate equipped with a thermocouple was used as a heat source for both the heating zone and reaction area to test the thermal catalytic activity.

Isotopic labeling experiments

¹³CO₂ gas (isotopic purity 99%) was used as the carbon source instead of pure ¹²CO₂ gas (chemical purity 99.999%), and the same photocatalytic process was used for isotope labeling measurement. The high-purity Ar was purged at a flow rate of 8 sccm for 60 min to remove any existing CO₂ in the reactor. The reactor was then exposed to a 300 W Xe lamp for 1 h while continuously bubbling and stirring at a flow rate of ¹³CO₂ and ¹²CO₂ at 30 mL/min. Photocatalytic species were separated into individual substances via gas chromatography columns. The origin of products was analyzed by gas chromatography-mass spectrometry (GC-MS) (Trace 1310 GC-ISQ quadrupole MS, ThermoFisher, USA).

To confirm the oxygen source of O₂, the labeled H₂¹⁸O experiment was performed as follows: CO₂ was purged continuously into the reactor for 60 min. Then, 1 mL of H₂¹⁸O was introduced into the reactor. The temperature was controlled at room temperature. Then, the photocatalytic reaction experiment was conducted as before by irradiating the mixture for 1 h. The gas samples from the reactor were collected and analyzed by mass spectrometry.

Equations evaluating PCO₂R

The production rate, the selectivity, and the AQE were calculated as follows.

Production rate. The mass-normalized production rate of product $i$ was calculated as⁵⁵:

$${\mathrm{r}}_i\left({\mathrm{\mu }\mathrm{mol\; g}}^{-1}{\mathrm{h}}^{-1}\right)=\frac{N_i}{m_{\mathrm{cat}}t}$$ 2

where $N_i$ is the amount of product $i$ (μmol), $m_{cat}$ is the catalyst mass (g), and $t$ is the irradiation time (h).

Amount of gaseous products. For continuous-flow analysis, the amount of product

$i$ was obtained from the outlet flow rate and the measured volume fraction:

$$N_i\left(\mathrm{mol}\right)=\frac{F_{\mathrm{out}}\left(V_{\mathrm{i}}/100\right)}{V_{\mathrm{m}}}$$ 3

where $F_{\mathrm{out}}$ is the total outlet flow rate (sccm), $V_i$ is the volume percentage of product $i$ (vol%), and $V_m$= 22.414 L mol^–1 is the molar volume of an ideal gas at 273.15 K and 1 atm.

Selectivity. The electron-based selectivity toward the product $i$ was calculated as:

$$S_i\left(\%\right)=\frac{n_i\times N_i}{{\sum }_j{n}_j\times n_j}\times 100\%$$ 4

where $n_i$ is the number of electrons required to form one molecule of product $i$, and the summation index $j$ refers to all detected reduction products.

AQE, the light was filtered by different monochromatic filters to test the PCO₂R, and the AQE was calculated as:

$$\mathrm{AQE}\left(\%\right)=\frac{n_{\mathrm{e}}{N}_{\mathrm{prod}}}{N_{\mathrm{ph}}}\times 100\%$$ 5

where $n_{\mathrm{e}}$ is the electron number per product molecule (for CH₄, $n_{\mathrm{e}}$= 8); $N_{\mathrm{prod}}$ is the amount of the target product formed (mol).

$$N_{\mathrm{ph}}=\frac{IAt\lambda }{hc}$$ 6

where $N_{\mathrm{ph}}$ is the total number of incident photons during the irradiation time $t$; $I$ is the incident light intensity (W m^–2); $A$ is the illuminated area (m2); $t$ is the irradiation time (s); $\lambda$ is the incident wavelength (m); and $h$ and $c$ are Planck’s constant and the speed of light, respectively.

COMSOL Multiphysics simulations

The finite element modeling (FEM)-based numerical framework in COMSOL Multiphysics 6.3 was used to numerically solve the frequency domain form of Maxwell’s equation:

$$\nabla \times {\mu }_{\mathrm{r}}^{-1}\left(\nabla \times \mathbf{E}\right)-k_0^2\left({\epsilon }_{\mathrm{r}}-\frac{\mathrm{j}\sigma }{\omega {\epsilon }_0}\right)\mathbf{E}=0$$ 7

where $k_0$ represents the wavenumber, $\omega$ is the angular frequency, $\mathbf{E}$ is the scattered field; ${\epsilon }_0$ is the vacuum permittivity, ${\mu }_{\mathrm{r}}$, ${\epsilon }_{\mathrm{r}}$, and $\sigma$ represent the properties of the material, namely, relative permeability, relative permittivity, and electrical conductivity, respectively.

The frequency domain module and thermal conduction module are coupled in one model, in which a single spherical Bi@Fe₂O₃ nanoparticle is illuminated by the linear polarized plane wave, with a power density of 2.5 mW μm^–2. The core radius and shell thickness of the Bi@Fe₂O₃ nanoparticle are 50 and 5 nm, respectively, in line with the TEM and SEM images. A spherical computational domain with a perfectly matched layer (PML) was implemented. To characterize the thermal effects of the plasma itself, the heat-transfer analysis was carried out by solving the steady-state form of the heat equation:

$$\mathbf{U}\nabla T=\frac{k}{\rho C_{\mathrm{p}}}{\nabla }^{2}T+\frac{q}{\rho C_{\mathrm{p}}}$$ 8

where $T$ is the temperature, $\mathbf{U}$ is the velocity vector, $k$, $\rho$, and $C_{\mathrm{p}}$ are the local properties, namely, thermal conductivity, density, and heat capacity, respectively. The heat source term $q\left(\mathbf{r}\right)$ induced by LSPR is given by the equation:

$$q\left(\mathbf{r}\right)=\frac{\omega }{2}{\epsilon }_0\text{Im}\left({\epsilon }_r\right){\mid \mathbf{E}\left(\mathbf{r}\right)\mid }^2$$ 9

where $\omega$ is the angular frequency of the incident light, ${\epsilon }_0$ is the vacuum permittivity of the material, and ${\epsilon }_r$ is the complex relative permittivity, and $\mathbf{E}\left(\mathbf{r}\right)$ is the local electric field. Optical constants and heat-transfer parameters for Bi, Fe₂O₃, and water were taken from the built-in Optical Materials Database. Due to the small length scale of the nanoparticle, the Grashof number is negligible, and the convective contribution to heat transfer can be neglected. Thus, heat transfer was modeled using the conduction equation in a sufficiently large domain that approximates an infinite medium.

Note: (i) The PML is a non-reflective boundary condition that absorbs outgoing electromagnetic waves, minimizing spurious reflections to accurately model wave propagation in unbounded domains.

ii) Due to the high symmetry, only a quarter of the model is simulated to reduce computational costs.

iii) Due to the low air mass fraction, water’s properties are applied to the medium.

DFT calculations

The spin-polarized first-principles calculations in this study were performed using DFT as implemented in the Vienna Ab initio Simulation Package (VASP, version 5.4.4)⁵⁶. The interactions between valence electrons and ionic cores were described using the projector augmented wave (PAW) method^57,58, which ensures accurate treatment of the core-valence interactions. The exchange-correlation energy was treated with the revised Perdew–Burke–Ernzerhof (RPBE) functional within the framework of the generalized gradient approximation (GGA)⁵⁹. A kinetic energy cutoff of 500 eV was employed for the plane-wave basis set to ensure convergence of the total energy.

All structural optimizations were conducted using the conjugate gradient algorithm until the residual force on each atom was less than 0.05 eV/Å, and the electronic self-consistent field (SCF) convergence criterion was set to 10^–6 eV. Brillouin zone integrations were carried out using a Γ-centered 2 × 3 × 1 Monkhorst–Pack k-point grid for the slab models. Gaussian smearing with a width of 0.05 eV was applied for electronic occupations. Where necessary, dipole corrections were applied to eliminate spurious interactions between periodic images. For surface and adsorption calculations, a vacuum layer of at least 15 Å was added along the direction perpendicular to the slab to avoid image interactions. To illustrate the long-range dispersion interactions between the adsorbates and catalysts, we employed the D3 correction method by Grimme et al.⁶⁰. All calculations were carried out under periodic boundary conditions.

The free energy change ($\mathrm{\Delta }G$) of each elementary step was estimated using the following thermodynamic correction:

$$\mathrm{\Delta }G=\mathrm{\Delta }E+\mathrm{\Delta }ZPE+{\int }_0^T\mathrm{\Delta }{C}_{p}dT-T\mathrm{\Delta }S$$ 10

where $\mathrm{\Delta }E$ is the DFT total energy difference between initial and final states, $\mathrm{\Delta }ZPE$ is the zero-point energy correction, ${\int }_0^T\mathrm{\Delta }{C}_{p}dT$ represents the thermal contribution to the enthalpy, and $\mathrm{\Delta }S$ is the entropy change. All thermodynamic corrections were obtained from vibrational frequency analysis and processed using the VASPKIT package⁶¹.

Note: It should be noted that although the experimentally synthesized Bi@Fe₂O₃ catalyst exhibits a core@shell architecture, constructing a full-scale core@shell model is hardly possible in DFT calculations. Therefore, as commonly reported in previous studies^62–65, simplified slab-based models are employed to approximate the essential physicochemical features of core@shell systems.

In a conventional stacked slab model (a layered “vertical” slab model, Supplementary Fig. 57), the Bi/Fe₂O₃ interface is buried and inaccessible to reactant molecules, so that it is extremely difficult to investigate the interfacial electronic effects under CO₂/H₂O coexisting conditions. To better represent the realistic reaction microenvironment at the interface, a new Bi@Fe₂O₃ interfacial model was constructed by loading Bi nanoparticles onto the Fe₂O₃ slab. A layer of 16 H₂O molecules was introduced above the surface to account for solvation and interfacial electric-field effects⁶⁶.

Although this model represents a geometric simplification of the ideal core@shell structure, it captures the intrinsic electronic interaction at the Bi@Fe₂O₃ heterojunction. Importantly, the charge redistribution and activation mechanisms observed in this interfacial model reflect the intrinsic chemical nature of the Bi@Fe₂O₃ heterojunction. The resulting theoretical insights are therefore directly relevant to understanding the experimentally observed high selectivity for CO₂ reduction to CH₄ under CO₂ and H₂O coexisting conditions. Considering all computational constraints and mechanistic requirements, we believe that the adopted modeling approach provides a more physically meaningful and experimentally consistent theoretical description of the catalytic mechanism proposed in this work.

Author contributions

X.K. designed research, carried out the experiments, analyzed the experimental data, and wrote the manuscript. M.J. performed DFT calculations and analyzed data. J.L. performed COMSOL Multiphysics simulations and analyzed data. C.L., X.D., F.W., and S.B. helped analyze data. Y.L. and L.G. helped revise the manuscript. All authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Hong Chen, Lingzhi Wang, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data supporting the findings of this study are available within the article and its Supplementary Information files. Data are available from the corresponding authors upon request. Source data are provided with this paper.

Competing interests

The authors declare no conflict of interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70960-9.