Unassisted photoelectrochemical CO₂-to-liquid fuel splitting over 12% solar conversion efficiency

Abstract

The increasing need to control anthropogenic CO₂ emissions and conversion to fuels features the necessity for innovative solutions, one of which is photoelectrochemical system. This approach, capable of yielding gaseous production progressively, is facing challenges for liquid fuels generation due to optical, electrical, and catalytic properties. This study employs a standalone photoelectrochemical setup, in which InGaP/GaAs/Ge photoanode is integrated with tin-modified bismuth oxide cathode to convert CO₂ into liquid formic acid. In unassisted two-electrode assembly, setup exemplifies its operational durability for 100 h, during which it maintains an average Faradaic efficiency of 88% with 17.3 mmol L^–1 h^–1 of yield, thereby excelling in average solar-to-fuel conversion efficiency at 12% with 60% of electrical energy efficiency under one sun illumination. This significant performance is further associated with metal-semiconductor interface formation between tin and bismuth oxide, which bridges electronic structures and generates an electric field at their interfaces. This study outperforms conventional solar-driven systems in operational durability and liquid fuel production.

Introduction

Growing atmospheric CO₂ concentrations, the primary contributor to global warming, pose significant challenges^1,2. Projections indicate that, without intervention, anthropogenic CO₂ levels could approach ~590 ppm, potentially triggering an average temperature rise of +1.9 °C by this century³. Numerous renewable strategies have been implemented to mitigate surplus CO₂ emissions, including recycling, capturing, and conversion; however, their practical integration faces challenges to control excessive CO₂ re-emission and implement complicated systems^4,5,6,7. Among those, photoelectrochemical CO₂ reduction reaction (PEC-CO₂RR) systems, known as artificial photosynthesis, present a viable strategy to convert solar energy into value-added by-products^8,9,10. Among those, high-energy density liquid by-products, such as formic acid (HCOOH), are advantageous owing to ease of storage, transportation, separation, and displacement at peak volumetric densities^5,11. Nevertheless, the investigation into solar-assisted CO₂-converted liquid productions remains inadequate, primarily due to selectivity, sub-optimal yields, limited conversion efficiency and operational durability^5,12.

Among reported solar-driven systems, PEC assemblies, synergistically coupled with catalyst layers, can efficiently catalyze CO₂ into valuable carbon-based high-energy density fuels^9,10. Considering PEC as a significant configuration, several studies on dual-photoelectrode configurations have been reported to initiate the CO₂RR^5,13,14. Conversely, a single-photoelectrode system that mimics artificial photosynthesis and is distinguished as a viable and economical alternative for efficient fuel synthesis is rarely investigated in PEC-CO₂RR due to the limited power output^15,16. This single-photoelectrode PEC assembly facilitates CO₂ conversion beyond conventional structures by effectively mitigating charge carrier transport losses and thermodynamic inefficiencies, such as heat transfer between electrodes¹⁷. However, this single-photoelectrode setup either necessitates external bias or drives a half-reaction without applied bias due to the low power output generated by photoelectrode^17,18,19. Therefore, a lack of evidence reveals the ability of a single-photoelectrode design to catalyze the bias-free PEC-CO₂RR, thereby limiting the potential stride towards unbiased operation²⁰.

In addition to the above considerations, selecting a single-photoelectrode with a tandem configuration capable of producing sufficient power output is significant for driving PEC-CO₂RR without external energy input²¹. The tandem structures reported to date are mainly pertinent to efficient carbon monoxide (CO) or syngas productions, and yet, catalyzing liquid fuels with high η_STF, selectivity, and long-run yields remains a prominent concern^{13,14,20,22,23}. Despite the efficient power output generation, the instability of photoelectrodes within aqueous solutions remains a significant challenge that impedes long-run operation⁸. This need arises to incorporate efficient protective layers against photocorrosion and chemical instability under aqueous environmental conditions, thereby prolonging operational durability²⁴. Subsequently, highly active and low-onset potential electrocatalysts are prerequisites to maintain selectivity and yields for liquid fuels during long-run operation. Bismuth(Bi)-based electrocatalysts have gained prominence to convert CO₂ into HCOOH, primarily due to their exceptional catalytic performance and advantageous properties in stabilizing CO₂^•− intermediate^25,26. However, low current density, less electrical conductivity, limited CO₂ adsorption capability, and inadequate selectivity at a more negative potential necessitate an interfacial engineering strategy for heterostructure design to modulate the electronic structure for performance improvement^27,28. Hence, using an interfacial engineering approach to create metal-semiconductor interfaces, where differences in work functions between the metal and semiconductor enable charge transfer and adjust semiconductor band positions, is significant²⁹. In metal-semiconductor interaction, metal components play a significant role in fine-tuning the composite’s electronic structure²⁸. Such interactions lead to orbital rehybridization and charge transport across the metal-support interface, inducing new chemical bond formation and molecular energy level adjustment^28,30. This process adjusts the d-band structure of metals, enhancing the adsorption of reaction intermediates, thus lowering the energy barriers and facilitating the rate-limiting steps^30,31. Additionally, metal-semiconductor interactions contribute to the stability of metal atoms on the support material through the formation of thermodynamically favorable metal-support bonds and promote charge redistribution via electron transfer^28,30. Considering HCOOH yield as a major product, the synergistic interaction between semiconducting bismuth oxide (Bi₂O₃)-based catalysts with appropriate metal-based catalysts is a favorable choice for designing binary components to promote efficient and long-run CO₂ conversion^32,33,34. To address these challenges, the fabrication of tandem-structured photoelectrode with optimal power output and low-onset potential catalyst is a viable choice for converting CO₂ into high-energy density liquid fuels.

This study implemented an unassisted PEC-CO₂RR setup using 3-J photoanode and Sn-modified Bi₂O₃ nanocomposites (NCs) to convert CO₂ to liquid HCOOH. During three-electrode configuration, the Sn-Bi₂O₃||3-J|Ni/RuO_x PEC device recorded an average FE of 90% with 19.7 mmol L^–1 h^–1 yield, resulting in significant η_STF of 15% and electrical energy efficiency (η_EE) of 57% for 100 h. The engineered Sn-Bi₂O₃ catalyst outperforms in CO₂ conversion rate due to electronic structure modulation at their interfaces. Theoretical density of states (DOS) findings further indicate that incorporating Sn with Bi₂O₃ upshifts the Fermi level. These upshifts in the Fermi level result in a stronger binding affinity of the catalyst towards the OCHO*, which enhances HCOOH formation. These observed effects are associated with enhanced CO₂ and OCHO* adsorption capabilities at the catalyst interface. Compared to the Bi₂O₃, the Gibbs free energy further confirms a reduction in formation energy on the Sn-Bi₂O₃ surface, confirming changes in the electronic structure after Sn modification. These insights facilitate PEC assembly and catalyst designs to synthesize high-energy density liquid fuels for scalable operations.

Results

Photoelectrode design, photovoltaic characteristics and devised assembly

In solar-driven reactions, the efficiency of the photoelectrode emerges as a critical determinant of overall performance. However, the most reported photoelectrodes incite severe charge recombination owing to their inherent photovoltage limitations and inadequate passivation layers, ultimately decreasing the efficacy^35,36. The 3-J photoanode with a sufficient power output that can catalyze the CO₂RR without external bias is utilized in this study. At first, the rear side of the 3-J photoanode is integrated with Nickel (Ni)-foil, which serves dual purposes—stabilizing the device against photocorrosion and functioning as an efficient heat dissipation layer to perform PEC-CO₂RR under aqueous conditions³⁷. This layer ensures the 3-J photoanode bottom and Ni interface an ohmic contact, unlike the Schottky barrier or anolyte-induced depletion layer over the Ni surface³⁸. To fulfill the photovoltage prerequisites of PEC-CO₂RR, low-onset potential ruthenium oxide (RuO_x) is further deposited over the Ni-foil to catalyze the water oxidation at a low applied potential. This 3-J|Ni/RuO_x design of the PEC cell presents the distinct advantage of effectively decoupling the optical absorption and electrocatalytic interfaces, extending its longevity and operational range (Fig. 1a).

Fig. 1: Schematics of the PEC assembly for CO₂-to-liquid fuel conversion and photovoltaic characteristics of the 3-J photoanode.

a Mechanistic demonstration of the CO₂R_cat||3-J|Ni/RuO_x PEC device during CO₂RR; enlarged view of 3-J|Ni/RuO_x photoanode on the right-hand side and HAADF-STEM mapping of CO₂R_cat (Sn-Bi₂O₃) on the left-hand side. b HAADF-STEM of 3-J photoanode with elemental mapping of In, Ga, P, As, and Ge, with 1 µm size, respectively. c Operating current-voltage characteristics, and (d) EQE% of the 3-J photoanode. ARC: Anti Reflection Coating, RE: Reference electrode (Ag/AgCl), BPM: Bi-polar membrane.

The photovoltaic characteristics of the 3-J photoanode under dry conditions were analyzed under AM 1.5 G illumination to validate the photovoltage requirements for PEC-CO₂RR. The performance metrics of the 3-J photoanode, as indicated by the current–voltage (J–V) profile, including open-circuit voltage (V_OC) of 2.6 V, a short-circuit current density (J_SC) of 13.97 mA cm^–2, a fill factor of 83.5%, and a power conversion efficiency standing at 29.8% (Fig. 1c). The J–V outcomes confirm that the photovoltages generated by 3-J photoanode surpass the potential requirements of PEC-CO₂RR (~2.2 eV, see Fig. 2a). To illustrate the light-absorption properties further and describe the integral role of 3-J photoanode in PEC performance, external quantum efficiency (EQE) with the absorption edge of each constituent sub-cell was evaluated (Fig. 1d). The uppermost sub-cell (GaInP, E_g = ~1.85 eV) exhibits efficient photon absorption within the 300–600 nm λ spectrum. Subsequently, photons that traverse the upper layer and reach the middle sub-cell (GaAs, E_g = ~1.40 eV) are absorbed within the 600–900 nm λ spectrum. Finally, the lowermost sub-cell (Ge, Eg = ~0.67 eV) captures high-frequency photons within the λ spectrum within the 900–1800 nm spectrum. The consistency between these EQE calculations and the J_SC values reveals the sufficient light-harvesting and effective charge carrier separation capabilities of the 3-J photoanode. Subsequently, using elemental mapping, the layered structure of the 3-J photoanode was confirmed with high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM). With a structural composition incorporating semiconductors such as InGaP, GaAs, Ge, and layers coated with Ni and RuO_x, the cumulative thickness of the device was estimated to be ~6.2 μm. The illuminated surface of the 3-J photoanode was precisely characterized via TEM further, which presents numerous quantum wells distribution within the mediating region (Fig. S1). In terms of elemental dispersion, a ~500 nm thick layer of InGaP resides at the bottom structure, succeeded by a substantial ~5 μm thick layer of GaAs, interspersed with multiple quantum wells in the intermediary region, and concluded by a ~400 nm thick layer of Ge at the top, as elucidated in Fig. 1b.

Fig. 2: Three-electrode PEC-CO₂RR operation under AM 1.5 G illumination with CO₂R_cat||3-J|Ni/RuO_x PEC device.

a Current–voltage characteristics of 3-J photoanode with electrochemically activated currents: anodic RuO_x with Ag/AgCl Reference Electrode and cathodic Sn-Bi₂O₃ NCs catalyst in the dark. b Calculated FE% (left-hand axis) and obtained HCOOH and H₂ production rates (right-hand axis) with Sn-Bi₂O₃ NCs cathode. The error bars indicate the results from ten individual tests. c Quantified Electrical Energy efficiency (η_EE%, right-hand axis) and operating photovoltages (left-hand axis) during 100 h of PEC-CO₂RR with 3-J photoanode and Sn-Bi₂O₃ cathode. d η_STF% of the cathodic products (right-hand axis) with the operating anodic photocurrent density (left-hand axis) during 100 h of PEC-CO₂RR using 3-J photoanode and Sn-Bi₂O₃ cathode. The setup recorded 2.14 ±  0.013 Ω resistance, and cell voltage readings have not been iR-corrected.

In addition to the above considerations, an efficient, stable, and cost-effective system is a prerequisite for efficient PEC performance. To address this, an artificial photosynthesis-inspired, unassisted, two-electrode PEC-CO₂RR setup employing a single photoelectrode to catalyze both the cathodic reduction and anodic oxidation is devised. Figure 1a elucidates the schematics insights of the unassisted PEC-CO₂RR assembly, comprising 3-J|Ni/RuO_x photoanode with Sn-Bi₂O₃ (CO₂R_cat)-cathode. The photogenerated electrons of the 3-J photoanode drift towards the front collection point, which is directly connected to the Sn-Bi₂O₃-cathode (separated by the bipolar membrane) to initiate CO₂RR, and holes transfer to the rear side of the photoanode on the Ni/RuO_x to facilitate water oxidation. Moreover, the physical representation of this PEC-CO₂RR setup is further depicted in Fig. S2.

Photoelectrochemical CO₂ reduction

The bias-free PEC system, operated by a single-photoelectrode, exemplifies a promising, cost-effective, and efficient approach to upscaling energy conversion processes. During PEC operation, the current-voltage characteristics of the 3-J|Ni/RuO_x photoanode (red) under 1-sun illumination with the electrochemically activated currents: anodic RuO_x (black) and cathodic Sn-Bi₂O₃ (green) under dark in aqueous conditions are measured first as displayed in Fig. 2a. Here, the red-green intersection curve of Sn-Bi₂O₃ signify the operating point for CO₂R at –0.73 V vs RHE and 12.99 mA cm^–2 current density. Similarly, a potential of 1.46 V vs RHE is required to initiate water oxidation (indicated by the red-black intersection of RuO_x). Additionally, the PEC system experiences external voltage losses of ~0.4 V due to the bipolar membrane, as detailed in Supplementary Note 1 and Figs. S4–S5. Considering this, the PEC-CO₂RR necessitates ~2.59 V to derive the complete PEC-CO₂RR.

During PEC-CO₂RR, liquid HCOOH was a major product and gaseous H₂ (Supplementary Note 2) was analyzed as a minor product. Before PEC operation, 3-J|Ni/RuO_x photoanode in 0.5 M aqueous KOH (pH ~14) anolyte with CO₂-saturated CO₂R_cat in a 0.5 M aqueous potassium bicarbonate (KHCO₃) (pH ~7.4) catholyte was immersed separately. Within the three-electrode PEC-CO₂RR setup, the potential at the counter electrode (Sn-Bi₂O₃) was recorded first, where the system stabilized 0.25 V potential for 100 h (Fig. S3a). On hourly three-electrode PEC-CO₂RR operation, the CO₂R_cat||3-J|Ni/RuO_x setup exhibited 94 ± 1.9% FE, corresponding to a significant production amount of 20.3 ± 1.3 mmol L^–1 h^–1 (Fig. S6). Subsequently, this PEC assembly stabilized an average 90 ± 2.1% FE with a consistent yield of 19.7 ± 1.7 mmol L^–1 h^–1 for 100 h (Fig. 2b). The hourly nuclear magnetic resonance (NMR) performance for HCOOH is quantified further to confirm the amount of obtained liquid products (Fig. S7). To stimulate artificial photosynthesis-inspired, unassisted two-electrode PEC setup was executed further to synthesize the storable and economically viable liquid fuel synthesis. Prior to unassisted PEC operation, the J–V characteristics of the 3-J|Ni/RuO_x photoanode in two-electrode configurations under standard AM 1.5 G illumination in aqueous conditions were analyzed. In an unassisted two-electrode setup, this 3-J|Ni/RuO_x photoanode maintained a J_SC of 10.5 mA cm^–2 (Fig. S8). Subsequently, in a two-electrode PEC-CO₂RR setup, the counter electrode (Sn-Bi₂O₃) recorded 0.25 V potential during long-run continuous operation of 100 h (Fig. S3b). In the initial hourly evaluation, the unassisted CO₂R_cat||3-J|Ni/RuO_x PEC device demonstrated 91 ± 2.3% FE with 17.9 ± 1.9 mmol L^–1 h^–1 product generation (Fig. S6). Subsequently, long-run 100 h of operation maintained a consistent FE of 88 ± 3.2% and delivered an average production of 17.3 ± 1.9 mmol L^–1 h^–1 (Fig. 3a). Additionally, the performance metrics of PEC-CO₂RR utilizing individual Ni-foil as an OER catalyst were assessed, and findings are thoroughly detailed in the Supplementary Note 3 and Figs. S9–11. Moreover, the half-anodic oxygen evolution reaction in both the three- and two-electrode configurations, occurring on the rear-side of the 3-J|Ni/RuO_x photoanode under AM 1.5 G illumination, are demonstrated in Supplementary Movie 1 and 2.

Fig. 3: Two-electrode PEC-CO₂RR operation under AM 1.5 G illumination with CO₂R_cat||3-J|Ni/RuO_x PEC device.

a Obtained FE% (left-hand axis) and production yields (mmol L^–1 h^–1, right-hand axis) of HCOOH and H₂ using CO₂R_cat||3-J|Ni/RuO_x PEC device. The error bars indicate the results from ten separate measurements. b Calculated η_EE% (right-hand axis) and operating photovoltages (left-hand axis) during 100 h of PEC-CO₂RR with 3-J photoanode and Sn-Bi₂O₃ cathode. c η_STF% of the cathodic products (right-hand axis) and operating anodic photocurrent density (left-hand axis) during 100 h of PEC-CO₂RR using 3-J photoanode and Sn-Bi₂O₃ cathode. 3-E: Three-electrode, 2-E: Two-electrode. The setup recorded 2.14 ±  0.013 Ω resistance, and cell voltage readings have not been iR-corrected.

Electrical energy efficiency (η_EE), which reveals the energy difference during CO₂ conversion into targeted products, is a key factor for further scalability estimations is measured. Based on the obtained selectivity and generated photovoltage, CO₂R_cat||3-J|Ni/RuO_x PEC device demonstrated a remarkable average η_EE of 57.2 ± 1.5% for HCOOH during a three-electrode and 60.0 ± 1.9% with an unassisted two-electrode assembly during 100 h of PEC-CO₂RR (Figs. 2c, 3c). During 100 h of PEC-CO₂RR, the generation of photovoltages demonstrated a commendable consistency, thereby substantiating the stability of the PEC device within aqueous electrolytic conditions (Figs. 2c, 3c). These η_EE results confirm that an appropriate potential is crucial for targeting the desired products. In the PEC-CO₂RR, the η_STF mark is state-of-the-art, indicating that the incident light induces the formation of the chemical bonds of various fuels. Based on obtained selectivity and photocurrent, CO₂R_cat||3-J|Ni/RuO_x PEC device in a three-electrode system exhibits remarkable η_STF of 15.5 ± 0.7% during hourly and recorded an average η_STF of 15.1 ± 0.9% for HCOOH throughout 100 h of experimentation (Fig. 2d). Operating under identical conditions, artificial leaf-inspired, unassisted two-electrode CO₂R_cat||3-J|Ni/RuO_x PEC device in hourly investigation recorded 12.5 ± 0.6% of η_STF, which was recorded at an average of 12 ± 0.9% during 100 h operation (Fig. 3d).

The lack of stability inherent to the photoelectrode represents a major challenge, effectively limiting the practical implementation of PEC devices. In the initial investigation, a 3-J photoanode without a protective layer (Ni-foil) was employed, which exhibited limited durability of <1 h within an aqueous electrolyte, thereby confirming its chemical instability and susceptibility to corrosion (Fig. S12). Consequently, the rear-side protected PEC device with Ni-foil demonstrated an enduring photocurrent for 100 h of CO₂R operation (Figs. 1e, 3c). Owing to the instability of the cathodic catalyst beyond 100 h in PEC-CO₂RR, the integrated 3-J|Ni/RuO_x photoanode with Pt-foil as a cathode was further experimented to validate its durability under aqueous circumstances. This individual PEC device exhibited robust operation over 500 h, with a negligible irreversible variation of 0.5% in photocurrent (Fig. S13). A slight deviation in the J–V characteristics under aqueous condition measurements was observed before and after 500 h of operation (Fig. S14). This indicates that the 3-J|Ni/RuO_x photoanode is durable for long-run experimentation. This exceptional performance confirms the Ni foil’s efficient electrolyte permeation, thereby improving the chemical stability under aqueous conditions. Overall, passivation by integrating the Ni-foil as a protection layer efficiently maintains chemical stability, dissipating heat and preventing corrosion. Furthermore, using RuO_x as a low-onset potential catalyst effectively catalyzes water oxidation at a lower potential.

Significance of liquid-fuel and comparison analyses of PEC assembly

Among CO₂ converted liquid fuels, HCOOH is an exceptional choice due to its unique electron-per-mole advantage and high energy density³⁹ Annually, ~800,000 metric tons of HCOOH are produced and employed in diverse applications, subsequently considered a remarkable H₂ storage medium due to rapid catalyst-aided dehydrogenation⁴⁰. Its salient features in the liquid phase at ambient environmental conditions exhibit low toxicity and offer high H₂ storage volumetric density (approximately 53 g per litter)⁴⁰. Inspired by artificial photosynthesis, PEC-CO₂ conversion into HCOOH is a compelling scheme for cost-effective commercial viability, given that the resulting liquid products can be precisely stored, seamlessly transported, and directly utilized^5,16,22. Despite numerous PEC-CO₂RR-based gaseous generations, liquid fuel synthesis remains a significant concern due to high energy demands. Such systems consistently undergo several limitations, including inadequate photovoltages, chemical instability, high onset potential and catalyst durability constraints, thereby limiting these implementations on their road to commercial viability.

Considering PEC as a renewable method and HCOOH as a high-energy density liquid fuel, our research demonstrates a standalone bias-free CO₂R_cat||3-J|Ni/RuO_x device. Under AM 1.5 G illumination, unassisted CO₂R_cat||3-J|Ni/RuO_x PEC device, marked by high selectivity and long-run production yield, leads to significant η_STF over 100 h. Remarkably, the selectivity, operational durability, and recorded η_STF outperform those of previously reported photoanode- and photocathode-promoted PEC-CO₂RR setups (Table S1). The assembly, strategically segregating anodic and cathodic compartments within the PEC assembly, facilitates independent reaction adjustment further, thereby preventing the need for electrolyte flow, concentrated solar irradiance, gas-diffused photoelectrodes and thermoelectric systems^22,41. The significant performance achieved in this work is credited to the multi-layered photoelectrodes design, which generates sufficient photovoltage, and the implementation of an artificial photosynthesis-inspired unassisted PEC assembly to induce bias-free CO₂RR. Progressing further on this achievement, catalyst selection and engineering emerge as crucial factors in long-run operations.

Electrochemical CO₂ reduction

To validate the selectivity, production yields and durability of cathode catalysts, the electrochemical performance was performed. The linear sweep voltammetry (LSV) of the cathode-catalysts (Bi₂O₃ and Sn-Bi₂O₃) revealed higher negative currents under CO₂ saturation than Argon, indicating strong interactions between CO₂ and electrode surfaces (Fig. S15a). Combining Sn (i.e., 1 to 4 wt.%) with Bi₂O₃ maximizes the current density, as LSV curves, selectivity, and production yields are presented in Fig. S16a, b and Table S2. With 3% Sn loading, the Sn-Bi₂O₃ demonstrated a FE of 95 ± 2.3% and 3.1 ± 1.2% with a yield rate of 20.7 ± 1.2 and 0.6 ± 0.8 mmol L^–1 h^–1 for HCOOH and H₂ at –0.73 V vs RHE potential, respectively (Fig. S15b). Under the same operation conditions, the Bi₂O₃ presented FE of 85 ± 1.7% and 5 ± 0.9% with a yield of 13.6 ± 1.1 and 1.4 ± 1.1 mmol L^–1 h^–1 for HCOOH and H₂, respectively (Fig. S15b). During 100 h of operation at –0.73 V vs RHE, the average FE of 90 ± 1.9% and 3.5 ± 1.3% with a production amount of 20.1 ± 1.2 and 0.7 ± 0.5 mmol L^–1 h^–1 corresponding to HCOOH and H₂ is achieved with Sn-Bi₂O₃, respectively. Moreover, the Bi₂O₃ performed an average of 80 ± 1.3% and 7.9 ± 1.1% FE with 11.9 ± 0.9 and 1.8 ± 0.8 mmol L^–1 h^–1 production relative to HCOOH and H₂, during 100 h of operation (Fig. S17). These enhancements in performance at a low applied potential and positive shift in onset potential are attributed to the appropriate loading of Sn onto Bi₂O₃. This considerable improvement primarily originates from the synergistic effects between the catalytically active sites of Bi and Sn. During this metal-semiconductor interface, the Sn facilitates electron transport to Bi₂O₃ at the interface, thereby increasing the number of active sites and stabilizing the reaction intermediates, thus improving the overall catalytic performance^33,42,43.

The hourly current densities in the −0.6 to −1.2 V vs RHE range with a decrease of −0.1 V in each run are further investigated (Fig. S18a, b). The Sn-Bi₂O₃ displays over 90 ± 2.5% FE towards HCOOH in the potential range from –0.7 to –1.1 V, which is restricted to ~82 ± 2.9% while utilizing the Bi₂O₃. This ultimate FE at a more negative potential is due to the enriched adsorption capacities of adsorbed CO₂^•− and OCHO*. Moreover, when the potential of −1.2 V is applied to the Sn-Bi₂O₃, the FE of HCOOH drops to 80 ± 3.8%, while that of H₂ improves to 18 ± 2.8%. Conversely, when a Bi₂O₃ NSs was employed, the FE for HCOOH decreased to 60 ± 5.1%, with a substantial rise of H₂ to 24 ± 3.8%. This decrease in the FE of HCOOH at a more negative potential indicates dominance of H₂ over HCOOH, as demonstrated by the product distribution (Fig. S19).

To further validate the enhanced catalytic CO₂R activity and active site number, the double-layer capacitance (C_dl) principle was employed in the non-Faradaic region to measure the electrochemical surface area (ECSA). Linear fitting of the cyclic voltammetry curves at varied scan rates indicated that the C_dl of Sn-Bi₂O₃ at 2.23 mF cm⁻² (Fig. S20a, c), which is notably higher than Bi₂O₃ (0.76 mF cm⁻²) (Fig. S20d, c). The calculated ECSA results indicate that the Sn-Bi₂O₃ demonstrates a three times higher active surface area than Bi₂O₃. This increase in ECSA is associated with the decoration of Sn onto Bi₂O₃, which improves catalytic active sites. To further verify the catalytic performances, the partial current densities (j) at various potentials were analyzed (Fig. S21a). The highest j_HCOOH of 33.6 ± 1.2 mA cm^–2 was observed with Sn-Bi₂O₃, 1.5 times that of Bi₂O₃ (22.4 ± 1.6 mA cm^–2). A comparison analysis of this study with earlier reported Bi-based catalysts towards HCOOH is summarized in Table S3. Moreover, electrochemical impedance spectroscopy (EIS) was investigated to validate the charge transfer process. Detailed insights are provided through Nyquist plots and equivalent electronic circuits (inset), as depicted in Fig. S21b. Notably, the Nyquist plots demonstrated a charge transfer resistance (R_ct) for Bi₂O₃ at 4.13 Ω, which was further reduced to 2.17 Ω for Sn-Bi₂O₃. These EIS outcomes support the accelerated faradic process in EC-CO₂RR, indicating an enhanced capability of hydrated CO₂ molecules to accept electrons, thus readily forming CO₂^•− intermediates. The Sn-Bi₂O₃ potentially strengthens electrical conductivity, facilitating faster charge transfer at the catalyst surface and reducing impedance. This leads to a product distribution towards HCOOH, which is associated with a rise in current density under the same potential conditions. The decreased adsorption at more negative potential is further confirmed through CO₂ adsorption studies (Fig. 22a). Isotherms substantiate the superior adsorption capacity of Sn-Bi₂O₃ (0.192 ± 0.05 mmol g^–1) than Bi₂O₃ (0.118 ± 0.03 mmol g^–1). Moreover, Tafel slope analyses shed light on the reaction kinetics of CO₂RR to validate the selectivity towards HCOOH (Fig. 22b). The low Tafel slope for Sn-Bi₂O₃ NCs at 141 ± 1.7 mV dec^–1 is lower than the Bi₂O₃ (223 ± 3.3 mV dec^–1), implies a faster conversion of CO₂ to HCOOH. The determined slope for Bi₂O₃ indicates that the chemical rate-limiting step entails the primary electron transfer. It is important to note that Tafel slopes increase towards more negative potentials, likely attributed to mass-transport restrictions.

Morphological, crystal and electronic structure analyses

Using micro- and spectroscopic techniques, the morphology of Bi₂O₃ nanosheets (NSs) and Sn-Bi₂O₃ NCs were characterized first. Scanning electron microscopy (SEM) image reveals the flower-like vertical growth of Bi₂O₃ NSs with uniform thickness (Fig. 4a). However, no noticeable morphological alteration was observed by SEM when Sn was incorporated over the Bi₂O₃ NSs (Fig. 4b). Further, the HAADF-STEM image of Bi₂O₃ NSs displays the flower-like morphological structure of the NSs. Subsequently, while characterizing the Sn-Bi₂O₃ NCs, Sn with uniform distribution was observed over Bi₂O₃ NSs (Fig. 4c, d). Further, Energy-Dispersive spectroscopy (EDS) was performed in conjunction with HAADF-STEM to reveal the elemental distribution of the Sn, confirming homogenous distributions of Bi, O, and Sn elements in Sn-Bi₂O₃ (Figs. 4g–j and S23a–c). The Bi₂O₃ NSs exhibit a consistent layered morphology before and after Sn modification, as confirmed via TEM, favouring the appropriate exposure of active sites to perform CO₂RR (Fig. S23d–e). Furthermore, the High-Resolution TEM (HRTEM) image of Bi₂O₃ NSs reveals lattice fringes with d-spacings of 0.380 and 0.320 nm, assigned to Bi(110) and Bi(111) facets (Fig. S24), in agreement with the X-ray diffraction (XRD) pattern. In the HRTEM image of Sn-Bi₂O₃ NCs, the interplanar distances of 0.335 and 0.274 nm relate to the Sn(110) and Sn(200) facets, with spacings of 0.380 and 0.320 nm associated with Bi(110) and Bi(111) planes, respectively (Fig. 4e). Selected-area electron diffraction (SAED) patterns were examined to validate the structural properties of Bi₂O₃ NSs and compositionally fine-tuned Sn-Bi₂O₃ NCs. Crystal lattices with Bi(110) and Bi(111) planes at respective interplanar distances of 0.380 and 0.320 nm (Figs. 4e and S24 insets) and Sn(110) and Sn(200) planes at respective d-spacings of 0.335 and 0.274 nm (Fig. 4f) was confirmed through SAED pattern in Sn-Bi₂O₃ NCs.

Fig. 4: Morphological studies of Bi₂O₃ NSs and Sn-Bi₂O₃ NCs.

a Field emission scanning electron microscopy images (FE-SEM) of pristine Bi₂O₃ NSs and (b) Sn-Bi₂O₃ NCs. c, HAADF-STEM images of Bi₂O₃ NSs and d Sn-Bi₂O₃ NCs. e High-resolution TEM image of Sn-Bi₂O₃ NCs. f The selected area electron diffraction pattern (SAEED) of Sn-Bi₂O₃ for Bi and Sn planes. g–j HAADF-STEM image of Sn-Bi₂O₃ NCs and elemental mapping of Sn, Bi, and O, respectively.

Moreover, the crystal structure of synthesized Bi₂O₃ NSs and Sn-Bi₂O₃ NCs was investigated using XRD. The sharp diffraction peaks of Bi₂O₃ align with the standard cubic pattern of Bi₂O₃ NSs (JCPDS no. 27-0052, Fig. S25)⁴⁴. However, in the Sn-Bi₂O₃ pattern, no Sn-containing crystalline phases were detected, possibly attributable to the low Sn concentration or the superposition of broad peaks, indicating that Sn modification does not disrupt the crystal structure of Bi₂O₃. Moreover, X-ray photoelectron spectroscopy (XPS) was performed to investigate the catalysts’ electronic structure, phase compositions, and valence states. The Bi 4f spectrum of Bi₂O₃ displays two significant peaks relating to Bi 4f_5/2 and Bi 4f_7/2 at 164.11 and 158.88 eV, which confirms negative shifts of 0.16 eV in the binding energy of Sn-Bi₂O₃ NCs, verifying electron transfer, which modulates electronic structures (Fig. 26a). The Bi₂O₃ also displayed an O 1s spectrum with two distinct peaks corresponding to the lattice oxygen in Bi–O bonds and on the adsorbed surface at 530.35 and 531.41 eV; thus, a negative shift of ~0.17 eV in the O 1s peak was observed in the Sn-Bi₂O₃ (Fig. S26b). The commercial metallic Sn displays two Sn⁰ corresponding peaks of 3d_3/2 and 3d_5/2 at 493.63 and 485.05 eV, and compared to this, positive shifts of 0.31 eV in the binding energies of Sn⁰ with a noticeable change in peak shapes are observed in Sn-Bi₂O₃ NCs (Fig. S26c). This positive shift in the Sn⁰ peaks authenticates its partial oxidation, confirming an electron interaction between Sn and Bi₂O₃. Overall, a noticeable downshift in the binding energies of Bi 4f, O 1s and an obvious upshift in Sn 3d confirms electron transfer from Sn to Bi₂O₃ due to the difference in electronegativities, indicating charge redistribution, thereby improving the CO₂R rate. Finally, the survey spectra verify the presence of Sn, Bi, and O in the Sn-Bi₂O₃ and Bi and O in Bi₂O₃ (Fig. S26d).

Significance of Sn on charge transfer modulation

For better performance, adjusting the work function by increasing the number of electronic states near the Fermi level is essential for semiconductors. With this objective, metallic Sn was interfaced with semiconducting Bi₂O₃ to tune catalytic conversion efficiency. At first, the Mott-Schottky analyses for pristine Bi₂O₃ and Sn-Bi₂O₃ NCs were investigated, which displayed negative slopes, indicating their p-type semiconducting nature (Fig. S27a). The analysis further revealed that the flat band potential of Sn-Bi₂O₃ NCs was calculated at −0.187 V, notably higher than the −0.279 V value measured for Bi₂O₃. To further understand the influence of electronic structure modulation, contact potential difference (V_CPD) and the work function (Φ), pivotal factors in composites’ interfacial charge transfer dynamics were further investigated using kelvin probe force microscopy (KPFM). The KPFM analysis revealed that metallic Sn demonstrates a lower V_CPD of −0.180 mV than the 330 mV value measured for the Bi₂O₃, resulting in the Φ of 4.93 eV for Bi₂O₃ relatively higher than metallic Sn at 4.42 eV (Fig. S27b, Eq. 12)^45,46. This discrepancy in Φ confirms that Sn possesses a lower electronic Fermi level relative to Bi₂O₃, which can facilitate charge transfer through metal-semiconductor interactions. The KPFM measurements further indicated that metal-semiconductor interfaced Sn-Bi₂O₃ NCs exhibited a V_CPD of 70 mV (Fig. S27b). This deviation in the V_CPD of Sn-Bi₂O₃ indicates alterations in the work function compared to the Bi₂O₃, predominantly attributed to the charge transfer between metal-semiconductor interfaces. Subsequently, using the V_CPD value, the Φ of Sn-Bi₂O₃ was recorded to be 4.67 eV, which is comparatively lower than Bi₂O₃ (Fig. S27c, Eq. 12). This adjustment in Φ generates an electric field at the interfaces that drive charge carrier separation at their contact layers (Fig. S27b). This favorable band alignment promotes efficient electron transfer from Sn to Bi₂O₃, enhancing the electron density at the semiconductor interface, which is advantageous for CO₂RR. Notably, the Sn-Bi₂O₃ NCs modulate charge transfer capabilities relative to pristine Bi₂O₃, facilitating local junction characteristics that align well with Mott-Schottky analysis outcomes (Fig. S27b). The impedance analysis matches well with these findings, which shows that Sn-Bi₂O₃ has reduced charge transfer resistance compared to Bi₂O₃, indicating superior charge transfer kinetics for CO₂RR (Fig. S22b). The measured electronic properties and band alignments for Sn-Bi₂O₃ and Bi₂O₃ are comprehensively summarized in Table S4.

Computational studies

This study further investigated the density functional theory (DFT) calculations to validate the experimental findings. At first, the Gibbs free energy calculation via HCOO* and OCHO* pathways are examined separately to validate how Sn modification affects Bi₂O₃ during HCOOH formation. During CO₂RR, the CO₂ molecule first adsorbed over the catalyst surface, thereby reducing into CO₂^•− and converting into the HCOO*/OCHO*. During the initial protonation, carbon atoms of HCOO*/OCHO* bond with the O-bidentate atom on the catalyst surface by positioning the protonated hydrogen atom upwards, whereas oxygen atoms are bonded with either Bi or Sn surface atoms. During the HCOO* conversion pathway, the required formation energy of Sn-Bi₂O₃ reduced from 2.20 eV to 1.45 eV at the Sn site and 0.91 eV at the Bi sites compared to the Bi₂O₃, presenting notable energy differences (ΔE) of 0.75 eV at the Sn sites and 1.29 eV at the Bi sites (Fig. S28a). Conversely, through the OCHO* pathway, the formation energy of Sn-Bi₂O₃ lowers from 1.48 to 1.15 eV at the Sn sites and 0.61 eV at the Bi sites, demonstrating ΔE of 0.87 eV on the Bi site and 0.34 eV on Sn site (Fig. S29a). This initial protonation reaction validated a significant Gibbs free energy difference (ΔG) between Bi₂O₃ and Sn-Bi₂O₃, validating Bi-site in Sn-Bi₂O₃ favorable for HCOO* or OCHO* formation. Notably, these significant energy barriers in HCOO* and OCHO* formation validate the initial proton-coupled electron transfer as a possible rate-limiting step, consistent with Tafel slope observations (Fig. S22b). Subsequently, in the second protonation of HCOO*/OCHO* on its oxygen atom, these adsorbed species react with an electron-proton pair to form stable HCOOH* with a notable decrease in ΔE from the catalysis surface. Overall, the higher ΔE on the Sn-Bi₂O₃ surface confirms the species’ favorable formation due to alterations in the electronic structure after Sn modification. Consequently, bonds of HCOO* or OCHO* with HCOOH* are cleaved due to intrinsic scaling relation limitations when Bi₂O₃ is modified with Sn⁴⁷. Overall Gibbs energy calculation results confirm CO₂ to HCOOH conversion via the OCHO* pathway more favorable than HCOO*, as the HCOO* path demands a higher potential, leading to a thermodynamically challenging reaction.

To further ensure the conversion of CO₂ to HCOOH, the adsorption energy of HCOO* and OCHO* on the catalyst surface was examined (Figs. S28b and S29b). For the HCOO* intermediate, the Bi-site of Sn-Bi₂O₃ exhibits a higher adsorption energy of −3.83 eV compared to its Sn-site (−4.39 eV) and the Bi-site (−5.59) of Bi₂O₃. Similarly, for the OCHO* intermediate, the Bi-site of Sn-Bi₂O₃ demonstrates more substantial adsorption energy of −3.14 eV than its Sn-site (−3.40 eV) and the Bi-site of Bi₂O₃ (−4.79 eV). Findings confirm that the negative adsorption energy of HCOO* and OCHO* are favorable for HCOOH formation at the Bi and Sn sites of Sn-Bi₂O₃ relative to Bi₂O₃, confirming facile adsorption and desorption during HCOOH formation. To further gain deeper insight into the root cause of favorable adsorption energies on the Sn-Bi₂O₃ surface, p-projected density of states (p-DOS) analyses via HCOO* and OCHO* paths were performed (Fig. S30a, b). Following both the HCOO* and OCHO* adsorption, the p-DOS of Sn-Bi₂O₃ compared to Bi₂O₃ shifts towards the Fermi level (E_f), confirming the stronger binding affinity to HCOO* and OCHO*. However, the p-DOS associated with OCHO* adsorption analysis reveals a higher E_f than HCOO*, thereby validating that OCHO* exhibits superior binding affinity over the HCOO*. Moreover, a significant upward shift in the E_f  of the p-DOS of the Bi and Sn sites in Sn-Bi₂O₃ is observed via OCHO* and HCOO* intermediates. This alteration causes an increase in the antibonding state and decreases the bonding states, resulting in a stronger binding affinity towards intermediates⁴⁸. Hence, an effective HCOOH yield is associated with a shift in adsorption energy and the synergistic effect of Sn and Bi₂O₃. These computed DOS findings agree with the Gibbs free energy simulation outcomes. This is the case even at a high negative potential range; CO₂ and consecutive intermediates possess adequate adsorption capabilities on the catalyst surface.

Post-reaction characterizations

The TEM, XRD, and XPS analyses of the 100 h post-reacted Sn-Bi₂O₃ NCs were conducted to investigate the alterations in surface morphology, structural properties, and chemical state. The post-electrolysis morphological characteristics of Sn-Bi₂O₃ NCs were examined using TEM and HRTEM. Post-reaction TEM analysis shows that the catalyst’s surface maintains its initial NSs morphology significantly. However, a notable alteration in the shape of the post-reacted sample’s NSs is observed, where initial smooth NSs demonstrate a rougher shape. This morphological transformation is likely a result of surface reconstruction processes occurring within the catalyst during the electrolysis period⁴⁹. Moreover, the HRTEM analyses show that Bi₂O₃ partially reduced to metallic Bi while the core retained its original state (Fig. S31a, b). However, Sn undergoes oxidation while partially retaining its metallic state. Observing the consistent selectivity trends in the initial and final hours of CO₂RR, it is estimated that the number of accessible active sites per electrode area primarily influences the partial variation in overall electrode activity. These structural arrangements maintain the CO₂ reactants and by-products through their active sites. Moreover, the XRD analysis of the 100 h post-reacted catalyst is characterized, in which diffraction patterns revealed the conversion of the Bi₂O₃ into the Bi₂O₂CO₃ phase, consistent with JCPDS No. 41-1448 (Fig. S32)⁵⁰. The emergence of the Bi₂O₂CO₃ phase is attributed to aqueous electrolyte-mediated conversion by reaction with HCO₃⁻ of KHCO₃ electrolyte⁵¹. This factor plays a pivotal role in influencing the CO₂ conversion pathway, featuring the critical impact of electrolyte conditions on the reaction mechanism⁵⁰. These XRD results highlight that the Bi₂O₂CO₃ does not compromise the catalyst’s fundamental core structure, affirming its stable crystal structure during long-run CO₂RR. Following 100 h of post-CO₂RR, XPS analysis of the Bi 4f spectrum of the post-reacted Sn-Bi₂O₂CO₃ catalyst presented two distinct peaks corresponding to Bi 4f_5/2 and Bi 4f_7/2, which validate the presence of Bi³⁺ (Fig. S33a). Simultaneously, in the post-reacted catalyst, the rise of two additional energy peaks at 161.2 and 156.5 eV attributed to Bi 4f_5/2 and Bi 4f_7/2 confirms the formation of metallic Bi⁰ (Fig. S33a), consistent with reported literature⁵². The dominant presence of the Bi³⁺ peak in the post-reacted catalyst indicates partial conversion to the metallic Bi⁰. In the subsequent analysis, the O 1s spectra of the post-treated sample demonstrated a higher oxygen concentration than the pre-treated sample, indicating increased oxygen adsorption (Fig. S33b). This rise of the additional oxygen peak is attributed to the CO₃⁻ formed when the Bi₂O₃ reacts with aqueous HCO₃⁻ and is converted into the Bi₂O₂CO₃. Subsequently, in the post-reacted Sn-Bi₂O₃, Sn species exhibited oxidation, with valence states ranging from Sn⁰ to Sn⁺⁴, further authenticating electronic interactions between the Sn and the Bi₂O₃ (Fig. S33c). This observation confirms that the strong interfacial interaction between Sn and Bi₂O₃ causes a significant alteration in the material’s electronic structure. Despite the partial conversion of Bi³⁺ to Bi⁰ and Sn⁰ to Sn⁴⁺, the retention of peaks corresponding to both Bi³⁺ and Sn⁰ at their respective positions confirms the structure’s integrity of the 100-h post-reacted catalyst. Overall, the post-morphology, crystal structure, and chemical states remain intact except for the change of the crystal phase, partial transitions of Bi³⁺ to metallic Bi, and oxidation of Sn⁰. Furthermore, this partial conversion does not hinder the catalyst’s durability and selectivity towards HCOOH, aligning with findings reported during CO₂RR^{50,52,53,54,55}. Post-reaction ECSA was performed further to confirm the electrocatalytic active sites. The cyclic voltammetry curves, linearly fitted at various scan rates, showed that the C_dl value for Bi₂O₃ NSs post-reaction was recorded at 0.68 mF cm⁻², which slightly decreased from its pre-reaction value of 0.76 mF cm⁻² (Fig. S20a–c). In contrast, the Sn-Bi₂O₃ NCs also preserved their ECSA performance after reaction and recorded C_dl at 2.11 mF cm⁻², which is minimally below the pre-reaction value of 2.23 mF cm⁻² (Fig. S20d–f). These post-reaction ECSA results show that catalysts maintained their ECSA performance with a minimal decrease in the active surface area during 100 h of continuous operation. However, this slight decrease in ECSA can be associated with blocking some active sites owing to the strong adsorption of some contaminants onto the catalyst surface. These post-reaction measurements indicate that Bi₂O₃ and Sn-Bi₂O₃ maintain their catalytic durability and are consistent with post-reaction morphological and chemical state results.

Catalytic reconstruction mechanism

Based on the post-reaction findings, a catalytic reconstruction mechanism is proposed (Figs. S34–S35). During CO₂RR, when Bi₂O₃ were immersed into the 0.5 M aqueous KHCO₃ electrolyte, the HCO₃⁻ ions reacted with catalysts, which underwent an electrolyte-mediated transition and reconstructed a metastable Bi₂O₂CO₃ phase as confirmed by XRD analyses (Fig. S32). Further, during CO₂RR, when a negative potential is applied to the Sn-Bi₂O₂, a partial conversion from Bi³⁺ to metallic Bi⁰ in the post-reacted sample is confirmed by XPS analysis (Fig. S33). Subsequently, oxidation in the Sn species of the post-reacted sample, with valence states ranging from Sn⁰ to Sn⁺⁴, was further confirmed. These transitions confirm strong electron coupling between Sn and Bi₂O₃, which promotes electron transfer at their interfaces and changes the electronic structure of active sites, thereby improving the CO₂ conversion rate. During CO₂RR, the reaction environment and negative applied potential are the main driving forces to phase and valance state conversions. Overall, the conversion of Bi₂O₃ into the Bi₂O₂CO₃ phase and partial conversion to metallic Bi with oxidation of Sn⁰ do not affect the post-reacted catalyst’s core structure.

Discussion

This study demonstrates an artificial photosynthesis-inspired unassisted PEC-CO₂RR system design and catalyst engineering approach to convert CO₂ into high-energy density, storable, and economically viable liquid HCOOH. In the devised two-electrode PEC-CO₂RR assembly, a 3-J|Ni/RuO_x-photoanode with Sn-Bi₂O₃-cathode is executed to catalyze the cathodic reduction and anodic oxidation. Under AM 1.5 G irradiation, the unassisted two-electrode CO₂R_cat||3-J|Ni/RuO_x setup exhibited an average FE of 88% with 17.3 mmol L^–1 h^–1 production rate, leading to a remarkable 12% η_STF with 57% of η_EE for 100 h operation. In this significant CO₂R performance, the metal-semiconductor interface formation between Sn and Bi₂O₃ cathode-catalyst played a significant role in electronic structure bridging. Subsequently, due to the difference in work functions, metallic Sn transfers electrons to semiconducting Bi₂O_3, which modulates the electronic properties of Sn-Bi₂O₃ NCs. Subsequently, this metal-semiconductor formation generates an electric field at Sn and Bi₂O₃ interfaces, resulting in upshifting in the Fermi level and facilitating charge separation. This favorable band alignment between Sn and Bi₂O₃ leads to a redistribution of electron density, promoting electron transfer from Sn to Bi₂O₃ and facilitating the CO₂ conversion rate. The XPS analyses further confirm an obvious downshift in the binding energies of Bi 4f, with a noticeable upshift in Sn 3d confirming electron transfer from Sn to Bi₂O₃, indicating charge redistribution. These modulations in the electronic structure and increased catalytic active sites in Bi₂O₃ after Sn-loading confirmed significant CO₂ conversion efficiency. The DOS findings further confirm that Sn loading over Bi₂O₃ upshifts the Fermi level. These upshifts in the Fermi level validate a stronger binding affinity between adsorbed species and the catalyst’s active sites, thereby facilitating efficient HCOOH formation. The Gibbs free energy calculations further confirm that Sn-Bi₂O₃ NCs reduce the activation energy barrier compared to the Bi₂O₃ surface, strengthening the electronic structure after Sn incorporation. Overall, the devised unassisted PEC setup and catalyst interface engineering approach provide significant insights, including rational design, intrinsic activity, selectivity, and long-run durability, paving the way to synthesize high-energy-density liquid fuels for scalable operations.

Methods

Catalyst preparation

To initiate the synthesis process of Bi₂O₃ NSs, 1.455 g of Bismuth (III) nitrate pentahydrate (Bi(NO₃)₃·5H₂O) was precisely added into 22.5 mL of ethylene glycol and 51 mL of ethanol. The homogeneous mixture was subjected to ultrasonication and stirring to ensure uniform dispersion of particles and subsequently heated in the autoclave at 160 °C for 5 h. The resultant powder was precisely rinsed and dried overnight in a vacuum to remove residual moisture. Subsequently, the Sn-Bi₂O₃ NCs were synthesized by dispersing the as-prepared Bi₂O₃ NSs into 1 mL of 3 M hydrochloric acid (HCl) and 20 mL of deionized (DI) water. Tin(ii) chloride dihydrate (SnCl₂.2H₂O) was then incorporated into the solution with a precise weight percentage ranging from 1 to 4 wt.% of Sn, i.e., for 1% = 0.0199 g, 2% = 0.038, 3% = 0.057 g, 4% = 0.076 g, respectively. Following sonication to obtain a uniform distribution of Sn, the mixture was incrementally transferred into a 10 mL solution of 10 mmol Sodium borohydride (NaBH₄) and then stirred to ensure a consistent, homogeneous distribution under Argon gas environment. Finally, purification with DI water is carried out to eliminate any unreacted species that might still be present in the solution.

Atomic-layer growth of RuO_x

The Atomic Layer Deposition (ALD) of the RuO_x on Ni-foil was executed using an Oxford Instrument Plasma-ALD System. Bis(ethylcyclopentadienyl)ruthenium (II) (Ru(EtCp)₂, and oxygen (O₂, 99.999%) utilized as precursors. High-purity Argon (Ar, 99.999%), utilized as a purge gas, and oxygen (O₂, 99.999%) as a carrier gas, were supplied throughout the deposition. The Ni-foil affixed to the silicon substrate was subsequently loaded into the reactor chamber. Notably, the chamber temperature was maintained at 270 °C, while the Ru(EtCp)₂ precursor was kept at a steady 110 °C, thereby providing a constant flux of Ru to the reactor. Meanwhile, the gas line temperature was fixed at 150 °C to prevent precursor condensation. The deposition protocol entailed a 1-second Ru(EtCp)₂ pulse and a 5-second oxygen pulse, each followed by a 20-second Argon purge. Modulation of deposition, including density, size, and distribution, was achieved through adjustments to the number of processing cycles.

Fabrication of photoanode

The 3-J|Ni/RuO_x photoanode was precisely fabricated by applying the silver paste to the Ge side of the solar cell (InGaP/GaAs/Ge), serving the dual purpose of creating an ohmic contact and furnishing an adhesive layer. This was successively followed by incorporating a compact RuO_x-deposited Ni-foil, precisely measuring 1 × 1 cm², as elucidated in the Atomic-layer deposition of RuO_x above. This inclusion functions as a protective barrier and an oxidation layer under aqueous environmental conditions. Post to this integration, insulation of the backside was achieved using epoxy, ensuring that the catalyst layer was explicitly exposed to the anolyte, thereby facilitating water oxidation. Furthermore, on the front top finger of the photoelectrode, the ohmic contact was proficiently accomplished by leveraging a Ga-In eutectic alloy. Subsequently, a Copper (Cu) strip, connected via silver paste, was extended to secure the reliable performance measurement throughout the operation. Prior to PEC testing, the epoxy and adhesive layers were subjected to vacuum drying at a temperature of 70 °C for 1 h. The individual 3-J|Ni/RuOx photoanode image is presented in Fig. S36a.

Preparation of CO₂ reduction electrode

The cathode electrode was fabricated by adding 5 mg of the pre-synthesized CO₂R catalyst into a mixture of 0.7 mL of ethanol, 0.3 mL of deionized (DI) water, and 10 µL of Nafion solution. This solution underwent sonication for 1 h to ensure uniform dispersion. The 100 mL resulting catalyst ink was taken and precisely drop-casted onto 1 × 1 cm² carbon paper. The mass loading of the CO₂RR catalyst was measured to 0.51 ± 0.07 mg cm⁻². Following vacuum drying, the treated carbon paper was employed as the cathode electrode during CO₂RR. The photograph of the cathode electrode with catalyst loading over carbon paper is illustrated in Fig. S36b.

Electrochemical and photoelectrochemical analyses

The corresponding EC and PEC experiments were employed with Zahner Zennium Pro potentiostat. All presented cell voltage measurements for EC and PEC have not been corrected by internal resistance (iR). The PEC and electrochemical reactions were performed in a two-compartment setup equipped with a bipolar membrane to inhibit generated products and electrolyte crossover. Prior to use, the bipolar membrane (fumasep® FBM) was soaked in 0.5 M NaCl solution for 24 h and then preserved at room temperature in 0.5 M NaCl aqueous solution. Before using it in EC and PEC setup, it was cut into 3.0 × 3.0 cm² and rinsed with DI water. Moreover, the reactions were conducted with 0.5 M KOH anolyte and 0.5 M KHCO₃ catholyte in the aqueous conditions. The pH value from ten separate measurements was recorded, which is 14.0 ± 0.023 for 0.5 M KOH and 7.4 ± 0.017 for 0.5 M KHCO₃, respectively. Using an electrochemical workstation, the resistance of the electrochemical cell was measured and recorded at 2.14 ±  0.013 Ω. Prior to testing, the cathodic chamber was purged with CO₂ or Ar. gases for 30 min to create pre-gas-saturation, and 10 sccm flow rate was maintained during the experiment.

Initially, an evaluation of the EC performance of the anode and cathode at varying applied potentials was conducted to confirm the selectivity of yielded products and operational potential. The three-electrode setup, which comprised a Bi₂O₃ and Sn-Bi₂O₃ catalysts working, Pt-mesh counter, and Ag/AgCl reference electrodes, facilitated this process. After this, the electrochemical measurements of Ni/RuO_x, acting as an oxidation catalyst, were further scrutinized in a three-electrode configuration to authenticate the operating potential. Furthermore, it is noteworthy that all potentials calculated experimentally during the EC processes were referenced to the reversible hydrogen electrode (RHE) from Ag/AgCl, as per Eq. (1).

$${E}_{{RHE}}={E}_{\left({Ag}/{AgCl}\right)}+0.197+0.059\times {pH}$$

(1)

The EC measurements were performed in the three-electrode setup, using cathode (Bi₂O₃ or Sn-Bi₂O₃)-catalysts—working, Ag/AgCl—reference and Pt-foil anode—counter electrodes, respectively. All EC measurements were conducted within the CO₂-saturated conditions. The electrochemical impedance spectroscopy (EIS) measurements were assessed at −0.8 V within 0.5 M KHCO₃ solution, employing a 5 mV amplitude across a frequency spectrum from 1 Hz to 10000 Hz. A Tafel plot illustrating overpotential versus the log j_HCOOH was generated from the results of controlled potential electrolysis. Moreover, PEC-CO₂RR was executed in three- and two-electrode configurations under 1-sun illumination. In a three-electrode setup, the 3-J|Ni/RuO_x photoanode, Sn-Bi₂O₃ cathode and Ag/AgCl reference electrodes assisted this PEC operation. However, the 3-J|Ni/RuO_x—photoanode and Sn-Bi₂O₃—cathode were the requisite electrodes in the two-electrode setup. All PEC reactions were conducted under 1-sun illumination at ambient conditions, complemented by continuous magnetic stirring. The ensuing photocurrents were subsequently normalized to the photoactive area of the 3-J|Ni/RuO_x photoanode. The SS-X Solar simulator with the solar filter of AM 1.5 G served as a light source. Prior to initiating PEC measurements, the light intensity was precisely regulated at 100 mW cm^–2. For both the two- and three-electrode “Counter Electrode Potential” measurements, Zahner Zennium Pro potentiostat was employed. For performing three-electrode configured “Counter Electrode Potential” measurements, a 3-J|Ni/RuO_x photoanode—working with Ag/AgCl—reference electrode in the anodic chamber, and the Sn-Bi₂O₃ catalyst—counter electrode in the cathodic chamber was executed. Subsequently, during two-electrode “Counter Electrode Potential” quantification, 3-J|Ni/RuO_x photoanode—working without a reference electrode in the anodic chamber, and Sn-Bi₂O₃ catalyst—counter electrode in the cathodic chamber was employed.

In addition, the external quantum efficiency (EQE) was quantified using the QE-R monochromator system, represented by Eq. (2). Here, ‘h’ is the Planck Constant, ‘c’ is the speed of light, ‘J’ is the photocurrent density, ‘e’ is the electronic charge, ‘λ‘ represents the wavelength, and ‘Pλ‘ is the flux of light intensity dependent on wavelength.

$${EQE}\left(\%\right)=\frac{{hcj}}{{e\lambda P}_{\lambda }}$$

(2)

Further, the solar-to-fuel conversion efficiency (η) for the various CO₂R products obtained was calculated using Eq. (3). In this context, ‘J_op’ signifies the operational current, ‘E°‘ is the thermodynamic potential of the chosen product (for HCOOH, –0.17 V vs RHE), and ‘FE’ denotes the Faradaic efficiency.

$${\eta }_{{STF}}\%=\frac{{J}_{{op}}\times \left[1.23V-E^{\circ}\, \right]\times {FE}\,}{{100{mW}.m}^{-2}}$$

(3)

Additionally, solar-to-hydrogen conversion efficiency (η_STH) was quantified via Eq. (4). In the given equation, ‘J_op’ shows the operational current.

$${\eta }_{{STH}}\%=\frac{{J}_{{op}}\times 1.23\times {FE}}{{100{mW}.m}^{-2}}$$

(4)

Moreover, the electrical energy efficiency of the products derived during PEC-CO₂RR was determined using Eq. (5). Here, E_chem is the energy used for the CO₂ conversion, and E_applied is the electrical energy input. Moreover, ‘E°‘ indicates the thermodynamic potential of the cathodic and anodic reactions; for instance, ${E}^{0}={E}_{{OER}}-{E}_{{CO}2{RR}}$. Here, ‘E_OER’ is 1.23 V vs RHE (anodic potential of water oxidation), and ‘E_CO2RR’ is –0.17 V vs RHE (cathodic potential of HCOOH). ‘E_cell’ represents the total supply voltage generated by solar cells (i.e. 2.2 V in three-electrode and 2.05 V in two-electrode configuration).

$${\eta }_{{EE}}=\frac{{E}_{{chem}}}{{E}_{{applied}}}=\frac{{E}^{0}\times {FE}\%}{{E}_{{cell}}}$$

(5)

Product Quantification

Online gas chromatograph quantitatively facilitated the amount of resultant gaseous products (Agilent 7890 B). Subsequently, the liquid products were quantified via a 600 MHz ¹H nuclear magnetic resonance spectroscopy (NMR) to ascertain their constituents. Prior to NMR experimentation, an analytical composition of 500 μl of resultant electrolyte solution, 120 μl D₂O, and 10 μl of a 0.05 mmol L^–1 DMSO internal standard was integrated into an NMR tube for subsequent analysis. Before generating the ¹H spectrum, water signals were attenuated using a pre-saturation methodology, ensuring a more precise assessment of the remaining constituents.

Moreover, the analysis of FE% of HCOOH and hydrogen was determined using Eq. (6), where the number ‘2’ symbolizes the requisite electrons for the formation of either HCOOH or a hydrogen molecule, ‘n’ signifies the number of total moles in the product, ‘F’ stands for Faraday constant (96485), and ‘Q’ corresponds to the cumulative charge supplied by the potentiostat or 3-J photoanode during CO₂RR testing.

$${FE}=\frac{2{nF}}{Q}\times 100\%$$

(6)

The production rate during EC and PEC-CO₂RR processes can be comprehensively determined via Eq. (7). In this formula, ‘ν_HCOOH’ symbolizes the concentration of the obtained product, ‘C_HCOOH’ denotes the concentration of HCOOH, ‘V’ signifies the electrolyte volume used in the reaction, ‘S’ is indicative of the geometric surface area of the engaged electrolyte, and ‘t’ corresponds to the time duration of the operation.

$${{{{\rm{\nu }}}}}_{{HCOOH}}=\frac{{C}_{{HCOOH}}\times V}{S\times t}$$

(7)

Electrochemical surface area measurements

The electrochemical capacitance was investigated further by calculating non-Faradaic capacitive currents related to the C_dl to investigate the effect of the number of active sites. This was achieved by analyzing the scan-rate dependency observed in cyclic voltammograms. In the non-faradaic region, a small current arises from surface adsorption-desorption, generating a capacitance at the electrode-electrolyte interfaces, thereby forming double layers. By linearly fitting the variation of peak current density (J_pc ‒ J_pa) against the scan rate, which correlates to the number of active sites in the electrocatalyst. Generally, the higher C_dl indicates a strong interaction between the electrolyte and the electrodes, reflecting the number of catalytic centers and providing insights into the ECSA. The ECSA of the electrodes was computed from their C_dl values using the given Eq. (8):

$${ECSA}=\frac{{C}_{{dl}}}{{C}_{S}}$$

(8)

Here, C_S represents the specific capacitance of the sample, defined as the capacitance per unit area of an atomically smooth planar surface of the material under the same electrolyte conditions. This ECSA calculation uses a general specific capacitance value of C_S = 0.029 mF⁵⁶.

Computational Details

The computational study deploying Density Functional Theory (DFT) was executed via the Vienna ab initio simulation package⁴³. The interaction between valence electrons and ions was proficiently modeled using projector-augmented wave (PAW) potentials. Additionally, the Perdew–Burke–Ernzerhof (PBE) rendition of the Generalized Gradient Approximation (GGA) was utilized to handle the electron exchange-correlation⁵⁷. Notably, van der Waals (vdW) interaction calculations were implemented through the DFT-D3 methodology. A plane wave cutoff energy of 520 eV, with 3 × 3 × 1 sheet k-point mesh for Bi₂O₃ geometry, was implemented to ensure an accurate density of electronic states. The ionic relaxations were conducted under a conventional energy of 10^–4 eV and convergence criteria force of 0.01 eV/Å. The adjusted lattice parameters for Bi₂O₃ (200) have been determined as a = 26.77 Å, b = 11.77 Å, and c = 19.36 Å, and this structural configuration of Bi₂O₃ is derived from a single-layer model. To further simulate the influence of Sn in the Sn-Bi₂O₃ catalyst system, a Sn-Bi₂O₃ surface was constructed by incorporating Sn atoms onto the pure Bi₂O₃ (200) surface (Fig. S37). Moreover, a Bi₂O₃ slab mimicked the (200) lattice plane, and a 1.5 nm vacuum layer was employed to prevent interactions with adjacent layers.

Additionally, Gibbs free energies of all adsorbed and gaseous species at 298.15 K were calculated using the expression in the corresponding Eq. (9). Where ‘E’ signifies the DFT-determined electronic energy for a given geometric structure, ‘E_ZPE’ represents the zero-point energy estimated at the harmonic approximation by the vibrational frequency of a molecule/adsorbate, ‘T’ stands for temperature, and ‘S’ denotes entropy.

$$\varDelta G={\varDelta E}_{{DFT}}-{\varDelta E}_{{ZPE}}+T\varDelta S$$

(9)

Within this computational framework, each reaction step in the computational hydrogen electrode (CHE) model was regarded as an instantaneous electron-proton pair transfer through the applied potential. Ultimately, the adsorbate’s adsorption energy (E_ads) was calculated using Eq. (10).

$${E}_{{ads}}={E}_{{substrate}+{adsorbate}}-{E}_{{substrate}}{-E}_{{adsorbate}}$$

(10)

Mott-Schottky Analysis

The Admiral Squidstat plus multichannel potentiostats were employed for the Mott-Schottky analysis during flat band potential (V_fb) measurements. Under a three-electrode system, this analysis utilized 0.5 M KHCO₃ as the electrolyte, a Pt wire as the counter and Ag/AgCl as the reference electrodes. Frequency ranged from 1 Hz to 100 kHz within the potential range from 0.2 V to −1.0 vs Ag/AgCl. The measurements of V_fb were based on the following Mott-Schottky Eq. (11).

$${C}^{-2}=\frac{2}{{e}_{0}{{{{\rm{\varepsilon }}}}}_{0}{{{{\rm{\varepsilon }}}}}_{0}{N}_{d}}(V-{V}_{{fb}}-\frac{{kT}}{{e}_{0}})$$

(11)

Where C denotes the electrode surface’s space charge layer capacitance, e₀ the electronic charge, ε₀ vacuum permittivity, ε_r the dielectric constant, N_d the donor site density, V the applied potential, k the Boltzmann constant, and T the absolute temperature, a plot of C⁻² vs V (vs RHE) with a negative slope signifies the material’s p-type semiconducting characteristic.

Kelvin probe force microscopy (KPFM) analysis

KPFM was conducted to assess the contact potential difference and the material’s work function. Utilizing an MFP-3D infinity model atomic force microscopy with a conductive tip in tapping mode, KPFM allowed for probing the work function (Φ in eV), determined from the contact potential difference (V_CPD in mV) between the tip and material through a given Eq. (12). Here, the electronic charge (e) is equal to 1.623 × 10⁻¹⁹.

$${V}_{{CPD}}=\frac{{\varPhi }_{{material}}\left({eV}\right)-{\varPhi }_{{tip}}\left({eV}\right)}{e}$$

(12)