## Introduction

The catalytic conversion of CO2 to fuels or valuable chemical products provides a carbon-neutral cycle that can mitigate the rapid consumption of fossil resources and increasing CO2 emissions1,2. Recently, in accordance with global CO2 reduction regulations, carbon capture utilization and storage (CCUS) technology for large-scale greenhouse gas reduction and conversion to high-value-added products has been intensively researched. In particular, the electrocatalytic reduction of CO2 has attracted interest owing to recent developments in electricity production from renewable energy sources such as solar and wind3,4. As a widely used raw material in the pharmaceutical, tanning, and textile industries that can also act as a hydrogen carrier for fuel cells5, formate is a very attractive product of the CO2 reduction reaction (CO2RR). Furthermore, given the required energy input and the market price of formate, the electrochemical reduction of CO2 to formate is an economically valuable process6. Various studies on the use of heteroatom-doped/alloy catalysts or catalyst structures with large active areas for the electrocatalytic reduction of CO2 to formate have been reported7,8,9,10,11. However, a highly efficient catalyst that can meet commercial requirements for activity, selectivity, and stability has not yet been achieved.

$${{{{{{\rm{CO}}}}}}}_{2}+2{{{{{{\rm{H}}}}}}}^{+}+2{{{{{{\rm{e}}}}}}}^{-}\to {{{{{\rm{HCOOH}}}}}}\,{{{{{\rm{E}}}}}}^\circ =-0.12\,({{{{{\rm{V}}}}}}\,{{{{{\rm{vs}}}}}}\,{{{{{\rm{RHE}}}}}})$$
(1)
$$2{{{{{{\rm{H}}}}}}}^{+}+2{{{{{{\rm{e}}}}}}}^{-}\to {{{{{{\rm{H}}}}}}}_{2}\,{{{{{\rm{E}}}}}}^\circ =0\,({{{{{\rm{V}}}}}}\,{{{{{\rm{vs}}}}}}\,{{{{{\rm{RHE}}}}}})$$
(2)

Until now, metal-based catalysts have generally been used for CO2 electroreduction because of their high activity and stability12,13,14,15,16,17,18,19,20,21. As the CO2RR to formate (Eq. 1) competes with the hydrogen evolution reaction (HER; Eq. 2), inhibiting the HER is essential for obtaining a high selectivity for formate. According to Trassati’s volcano plot22, metals such as Pb23, Bi24, In25, Hg24, and Sn23, which are generally located on the left branch of the volcano plot, exhibit high CO2RR selectivity for formate. The weak metal–hydrogen bonds of these metals result in good CO2RR activity for formate production.

Among these materials, Sn is the most reasonable material because the toxicity (Pb, Hg) or relatively scarcity in the earth’s crust (Bi) of other typical catalyst materials limit their commercialization26. Considerable efforts have been focused on developing Sn-based catalysts for formate production, with recent advances including the use of gaseous CO2 flow cells27,28,29,30,31,32,33,34,35,36,37,38,39,40. Although improvements in the catalytic activity for formate production have been achieved, the long-term durability of Sn-based catalysts under reduction conditions remains a critical issue (Supplementary Table 1). Previous research has indicated that a Sn species with a high oxidation state is key for achieving high catalytic activity for the CO2RR to formate41,42. Bocarsly et al. observed intermediates on the Sn electrode by in situ infrared spectroscopy and suggested that the oxidized Sn surface is a catalytically active species for the CO2RR43. However, Sn electrocatalysts with high oxidation states are reduced at high reduction overpotentials during the CO2RR, resulting in the formation of a metallic phase and the loss of catalytic activity41. This phenomenon can be suppressed under strongly alkaline conditions, but alkaline electrolytes, such as potassium hydroxide (KOH), can be neutralized during CO2RR due to the purging of CO2. Therefore, it is necessary to develop alternative Sn-based electrocatalysts to ensure stability at high current densities.

Here, we studied a fluorine-doped tin oxide (FTO) nanocatalyst that not only showed high CO2RR activity over a wide range of current densities but also maintained its performance for more than a week. The electrochemical CO2RR performance is significantly affected by the device design and the type of purged CO244,45. Therefore, a homemade gaseous CO2 fed flow cell was used to achieve a high current density. In-situ/operando analysis was conducted under similar conditions. We found that replacing Sn–O bonds with Sn–F bonds promotes the interactions of catalyst surface and HCOO, and modified electronic structure of CO2 to facilitate electron transfer. The F dopant was also revealed to play a significant role in maintaining the oxidation state of Sn at high reduction overpotentials. This work provides an advanced strategy for synthesizing cost-effective CO2RR electrocatalysts with high activity and selectivity.

## Results

### Structure and physical properties of stannic oxide electrocatalysts

SnO2, fluorine-doped-SnO2 (FTO), antimony-doped tin oxide (ATO), and indium-doped tin oxide (ITO) nanoparticles supported on carbon black were synthesized using a sol-gel method with hydrothermal treatment. The overall synthesis scheme is illustrated in Fig. 1a and includes the following steps:46 (1) formation of a metal-surfactant complex, (2) hydrolysis and condensation, (3) formation of a micelle-like surfactant template with a SnO2 phase, and (4) hydrothermal treatment for crystallization. The mechanism is described in more detail in Supplementary Note 1. Figure 1b, c and Supplementary Fig. 1 shows high-resolution transmission electron microscopy (HR-TEM) images of the synthesized SnO2 and doped-SnO2 catalysts with corresponding particle size distributions (insets, Fig. 1b, c). The SnO2 and doped-SnO2 samples consist of very small oxide clusters (<5 nm) with uniform distributions. All the catalysts had similar average particle sizes [2.567 nm (SnO2), 2.121 nm (FTO), 2.391 nm (ITO), and 2.154 nm (ATO)], and the dopants were uniformly distributed in the doped-SnO2 particles (Supplementary Figs. 24). On the contrary, while the SnO2 catalyst without tetradecylamine (TDA) surfactant has a similar particle size compared to SnO2 with TDA surfactant (2.603 nm, Supplementary Fig. 5), it aggregated to show a disordered mesoporous structure.

The SnO2 and doped-SnO2 nanoparticles all exhibited rutile tetragonal crystal structures, as identified by analyzing the zone axis of the images47,48. Furthermore, the crystal structures and average particle sizes of the synthesized nanoparticles were analyzed using X-ray diffraction (XRD) (Fig. 1d). In the XRD patterns, the (110), (101), and (211) reflections of tetragonal SnO2 were observed at 2θ values of 26.3°, 33.6°, and 51.9°, respectively46,49. The quality of the synthesized SnO2 and doped-SnO2 catalysts was analyzed by thermogravimetric analysis (TGA) (Supplementary Fig. 6). All the catalysts showed a weight loss of ~20% from 30 to 500 °C owing to the removal of adsorbed water molecules and stable oxygen functional groups in the carbon support50,51. Under flowing O2, the catalyst weight slightly increased and then decreased rapidly, which was attributed to complete oxidation of the carbon support to CO2 gas after the formation of oxygen functional groups. The SnO2 and doped-SnO2 catalysts were both found to have oxide contents of approximately 40 wt%. The similarities in the morphologies, crystal structures, and oxide contents of the SnO2 and doped-SnO2 nanoparticles allowed comparisons of their catalytic activities and efficiencies for CO2 reduction to formate under the same conditions.

### CO2-to-formate conversion performance

The electrochemical CO2RR activities of the SnO2-based catalysts were evaluated in a homemade flow cell using gaseous CO2 to accelerate the CO2RR while minimizing the mass transfer resistance. A detailed schematic of the flow cell is shown in Fig. 2a and Supplementary Fig. 7. The SnO2-based catalysts were loaded onto a gas diffusion layer (GDL) and gaseous CO2 was supplied to the cathode. An electrolyte of 1 M KOH or 1 M KHCO3 was used in both the cathode and anode flow channels, which were physically separated using an anion exchange membrane (AEM). Supplementary Figure 8a shows the linear sweep voltammetry (LSV) curves for SnO2/C and SnO2/C without tetradecylamine (TDA) using 1 M KOH as the electrolyte. SnO2/C exhibited significantly lower overpotentials than SnO2/C without TDA. At current densities below 300 mA cm−2, SnO2 showed a faradaic efficiency for formate (FEformate) of more than 70% (Supplementary Fig. 8b). In contrast, for SnO2/C without TDA, FEformate was reduced to 28.6% at 200 mA cm−2. The maximum formate production rate of SnO2/C (4.11 mmol h−1 cm−2) was almost three times higher than that of SnO2/C without TDA (1.44 mmol h−1 cm−2). The effect of TDA on the SnO2 particle size before and after CO2RR is not substantial. However, without TDA, the SnO2 particles are agglomerated (Supplementary Fig. 9: HR-TEM images of the SnO2 catalyst with and without TDA). These results show that the uniformity of the SnO2 particles significantly affects their CO2RR activity.

To observe the effects of doping on the CO2RR activity, the electrochemical CO2RR activities of FTO/C, ATO/C, and ITO/C were compared with that of SnO2/C (Fig. 2b, c and Supplementary Figs. 8 and 9). All the SnO2-based electrodes exhibited similar LSV curves, but the electrochemical reaction products differed (Supplementary Fig. 10a). In 1 M KOH solution, FTO/C exhibited an excellent FEformate value of 95% at 100 mA cm−2, which is higher than that of ITO/C (85%), ATO/C (80%), and SnO2/C (80%). Moreover, FTO/C maintained a FEformate value of more than 90% up to a current density of 300 mA cm−2. The maximum partial current density and formate production rate of FTO/C were 330 mA cm−2 and 6.31 mmol h−1 cm−2, respectively, which are superior to those of ITO/C (272 mA cm−2, 5.20 mmol h−1 cm−2), ATO/C (242 mA cm−2, 4.62 mmol h−1 cm−2), and SnO2/C (215 mA cm−2, 4.11 mmol h−1 cm−2). These results indicate that F doping of SnO2 promotes the catalytic activity for the CO2RR to formate.

To test the stabilities of FTO/C and SnO2/C, 1 M KHCO3 was used as the electrolyte. In 1 M KOH, the anolyte is neutralized during the electrochemical CO2RR, leading to a high overpotential for OER (Supplementary Fig. 11). Considering the stability of the whole system, the CO2 electrolyzer was optimized for neutral media. At 100 mA cm−2, the FEformate value of FTO/C was ~90%, whereas that of SnO2 was 75% (Supplementary Fig. 12). During the stability tests at a current density of 100 mA cm−2, the FEformate value of SnO2/C decreased significantly after several hours and the cell potential decreased slightly, showing the low stability of SnO2/C. In contrast, the cell potential of FTO/C remained stable for 7 days and a FEformate value of ~90% was maintained (Fig. 2e, f). X-ray photoelectron spectroscopy (XPS), HAADF-STEM, and EDS after the stability tests demonstrated that the structure and chemical state of FTO/C remained the same even after the exposure to long-term cathodic conditions (Supplementary Figs. 13 and 14). These results indicate that F doping of SnO2 provides excellent long-term stability for formate production in a gaseous CO2-fed flow cell. To evaluate the level of developed catalysts, the activity and stability of the FTO/C catalysts were compared with that of other Sn-based literature catalysts (Fig. 2e, f and Supplementary Table 1). The FTO/C catalysts exhibits comparable FEformate and current density to other best-reported catalysts, indicating enhanced intrinsic catalytic properties for CO2 electro-reduction to formate. Remarkable durability of FTO/C compared to other literatures suggests that F doping improves stability of Sn catalysts required for real electrochemical formate production.

### Theoretical investigation of CO2-formate conversion on the SnO2 and FTO surface

The enhanced FEformate of FTO was investigated through density functional theory (DFT) calculations. As the XRD patterns of the as-prepared SnO2 and FTO nanoparticles did not reveal any preferred orientations, the SnO2 and FTO surfaces were modeled using a four-layer slab composed of the tetragonal (110) plane. To build the FTO (110) supercell, 15% of the oxygen atoms in the SnO2 (110) supercell were randomly replaced with fluorine atoms. The elementary steps of the electrochemical conversion reaction of CO2 to HCOOH involving two electron pathways were described in three steps (Method). CO2 adsorption on the catalyst surface was performed to consider the onset potential difference between water and CO2 reduction. The adsorbed CO2 (CO2*) is then converted to the HCOO* intermediate and HCOOH(g) in sequence with two proton-coupled electron transfers. A recent study demonstrated that, in CO2 reduction current densities higher than 35 mA cm−2, the proton can be supplied to CO2* on the electrode surface by the dissociation of water molecules. At these current density regions, a huge amount of unused hydroxide ions is rapidly generated as a by-product, which results to the increase of the local interfacial pH to values above 12. This is regardless of the type of the buffering agent used52. In this regard, the proton for the CO2 reduction reaction is assumed to be predominantly supplied by the local electrolyte.

For both the SnO2 and FTO (110) surfaces, the CO2 molecule was gently adsorbed on the Sn atoms. Then, the HCOO* intermediate was formed as the oxygen atoms of CO2 were tightly bound to the Sn atoms. The difference between the CO2 adsorption energies on the SnO2 and FTO (110) surfaces was not substantial. However, approximately, a 1-eV difference was observed on the free energies for the HCOO* intermediate and HCOOH(g) formation steps. The FTO surface could provide thermodynamically favorable conditions for HCOO* formation compared to the SnO2 surface, whereas the conditions on the SnO2 surface favors the HCOOH(g) formation (Fig. 3 and Supplementary Table 2). From the computational hydrogen electrode model, a minimum potential of −2.05 V was required for SnO2 to overcome the activation barrier of HCOO* formation, whereas −1.79 V was needed for FTO to complete HCOOH(g) formation. Identical free energy calculations using strained SnO2 and FTO supercells were performed to simulate the structural changes on the SnO2 and FTO electrodes during CO2 reduction. The in-situ EXAFS data demonstrated that the bonding distance between Sn and Sn (or O) was slightly decreased by applying a potential of −1.0 V. Consequently, the lattice constants of the SnO2 and FTO supercells decreased uniformly by 1.37% and 0.33%, respectively. The observed changes in the lattice constants correspond to the average change of each bonding distance at −1.0 V (Supplementary Fig. 15 and Supplementary Table 3). Although the strained SnO2 (110) surface registered a lower energy for the formation of HCOOH(g) by 0.42 eV, the energies for CO2 adsorption and HCOO* formation were almost identical to those recorded for the SnO2 surface. As such, the free energy for the HCOO* formation, which is the potential-limiting step, was not affected by the compression of the SnO2 (110) crystal. In consequence, a potential of more than 2 V was still required for the strained SnO2. In contrast, the strained FTO (110) surface showed a 0.27-eV lower free energy for the HCOOH(g) formation, while the energies for CO2 adsorption and HCOO* formation slightly increased compared to that on the FTO surface. Since the HCOOH(g) formation is the potential-limiting step for the FTO electrode, the minimum potential to complete the CO2-to-HCOOH conversion reaction decreased from −1.79 V to −1.65 V as the FTO crystal was compressed. The DFT studies suggest that doping fluorine to SnO2 could alter the chemical environment of the oxide surface, making it thermodynamically stable for HCOO* intermediate formation, which is an important step in the CO2-to-HCOOH conversion reaction. In addition, the compression of the crystal through the application of an external bias rendered an effective contribution only for FTO. Therefore, these factors are assumed to facilitate formate production in the FTO electrode.

### Origin of excellent durability with high current density under highly reduction potential

In addition to the high FEformate of FTO/C, the high partial current density for formate at high cathodic overpotentials with excellent stability is another major advantage of this catalyst. To reveal the origin of this behavior, in-situ/operando X-ray absorption near-edge structure (XANES) spectroscopy at the Sn k-edge and Raman spectroscopy were performed for SnO2-based electrodes under gaseous CO2RR conditions in a customized electrochemical cell with a GDL (Supplementary Figs. 16 and 17). XANES is a bulk-sensitive technique that can reveal the oxidation state of materials, whereas Raman spectroscopy is a surface-sensitive technique that can identify the chemical structure of materials. Therefore, the combination of these in-situ/operando spectroscopies can be used to reveal the state of materials during the CO2RR. The ex-situ Sn k-edge XANES spectrum showed that both catalysts were predominantly in the quadrivalent (+4) oxidation state (Figs. 4a, b). The in-situ/operando XANES spectrum of SnO2 exhibited a strong electrolyte-induced energy shift and fitting revealed a large fraction of metallic Sn. In contrast, the spectrum of FTO showed only a small energy shift at the reduction potential, indicating that the change in the oxidation state is extremely small. The linear combination fitting (LCF) results visually represent this trend more clearly (Fig. 4c, d).

The in-situ/operando Raman spectra, which show the chemical states of the surface, are shown in Fig. 4e, f. Carbon supports for the catalysts were not used for the in-situ/operando experiments to improve the peak intensity. The Raman peak at 630 cm−153, which is associated with the symmetric stretching of Sn–O bonds (A1g modes), was identical for SnO2 and FTO, confirming the presence of oxide phase (Fig. 4a–d). This peak was observed for both SnO2 and FTO at applied potentials above −0.8 V. However, for SnO2, the Raman peak disappeared at a potential of −1 V, whereas for FTO, the Raman peak was still present at a potential of −1.2 V. These findings demonstrate that the surface of SnO2 is converted to metallic Sn but the oxidized state of the FTO surface is maintained under high cathodic overpotentials, which is consistent with the in-situ/operando XANES results. Despite their similar radii, F ions have a higher electronegativity than O ions. This property would make F–Sn bonds stronger than O–Sn bonds, leading to enhanced stability of FTO. The behavior of SnO2 and FTO catalysts for CO2RR are summarized in Fig. 4e. SnO2 catalysts exhibits good performance for formate production at low overpotential but are reduced to metallic Sn at high cathodic potentials, accelerating HER and lowering CO2RR. On the other hand, FTO possess enhanced catalytic activity for CO2 electro-reduction to formate by improving interaction with HCOO. Furthermore, oxidation state of FTO stabilized by strong F–Sn bonds under high cathodic overpotentials contributes the enhanced current density and durability. Thus, we expect that the enhanced current density and durability of FTO could allow the development of economically feasible CO2RR technology.

## Discussion

In summary, we obtained insights into the performance of doped-SnO2 catalysts for the CO2RR to formate using a combination of flow-type single-cell experiments, in-situ/operando spectroscopy, and DFT calculations. Compared with traditional SnO2 catalysts, the nanoparticles synthesized using TDA were much smaller because this surfactant prevents particle agglomeration during SnO2 growth via micelle formation. The high dispersion of our SnO2 catalysts allowed for a large number of oxide species to act as electrochemically active centers. The FTO/C catalyst exhibits higher performance than other doped catalysts, achieving partial current densities for formates of up to 330 mA cm–2 and high faradaic efficiency (95% at 100 mA cm−2). Notably, using the FTO catalysts, we achieve superior formate selectivity (≈90%) over 7 days of operation at 100 mA cm−2. Based on DFT calculation, fluorine doping not only enhance the interactions between HCOO and FTO surface but also alters electronic structure of CO2 to facilitate electron transfer. Furthermore, in-situ/operando spectroscopy suggests that in FTO catalysts, oxidation state of Sn, which significantly affects the CO2RR activity, did not change significantly under an applied potential. These findings suggest that fluorine dopant played an important role in increasing the selectivity for formate on the FTO catalyst by modulating electronic structure of Sn and enhancing the durability by preventing reduction under reduction potentials. Our study provides insight for designing highly active and durable electrocatalysts for the electrochemical conversion of CO2RR to formate.

## Methods

### Preparation of SnO2 and doped-SnO2 catalysts

To synthesize the SnO2 and doped-SnO2 catalysts, TDA (Sigma-Aldrich) was dissolved ultrasonically in a mixture of deionized water and ethanol. Then, SnCl4 (Sigma-Aldrich) was added, and the mixture was stirred for 1 h. This suspension was ultrasonically blended with carbon black powder, and then ammonium hydroxide solution (NH4OH, Sigma-Aldrich) was added dropwise, followed by stirring for 30 min. Subsequently, the suspension was refluxed at 80 °C for 72 h. The reaction mixture was cooled to room temperature, filtered, and washed several times with an ethanol solution. To remove excess TDA, the as-prepared SnO2 catalyst was transferred to a glass-lined stainless-steel autoclave and hydrothermally treated at 120 °C for 24 h. A detailed description of the formation of the tin ethbutoxide intermediate in the ethanol solution and the formation of SnO2 through the hydrolysis reaction is described in Supplementary Note 1. ATO, FTO, and ITO were synthesized using the same procedure, except that the composition of the metal precursor was varied by adding antimony acetate (C6H9O6Sb, Sigma-Aldrich), ammonium fluoride (NH4F, Sigma-Aldrich), and indium chloride (InCl3, Sigma-Aldrich), respectively. Varying amounts of the dopant salts were added to 712 mg of SnCl4 to synthesize the doped-SnO2 sample. In ATO, 25.0 mg C6H9O6Sb was used, in FTO, 19.48 g NH4F, and in ITO, 109.6 g InCl3. Antimony and indium were doped in the form of mixing Sb2O3 or In2O3 with SnO2 during hydrolysis and condensation, respectively. On the other hand, fluorine was incorporated through oxygen substitution54.

### Preparation of SnO2-based catalyst electrodes for the CO2RR

A catalyst ink was prepared by ultrasonically mixing 5 wt% of ionomer solution (Dioxide, 15 wt% target of catalyst) and SnO2-based catalyst powder (30 mg) with ethanol (2 mL). The SnO2-based catalyst electrodes were fabricated by spraying the prepared catalyst ink was onto a GDL (Sigracet 39 BB, SGL Carbon) at 70 °C. The electrode area was 2 cm2 and the loading of SnO2 was fixed at 0.5 mg cm−2.

### Preparation of Fe-Ni foam electrodes for the oxygen evolution reaction (OER in alkaline)

Fe-Ni foam electrodes for the OER were fabricated by a simple dip-coating method. The Ni foam was washed with deionized water and dried under nitrogen. After dipping in a 0.125 M FeCl3 solution, the Ni foam was removed and dried in a convection oven at 70 °C. The FeCl3-coated Ni foam was then activated using a three-electrode system in 1 M KOH at a current density of 100 mA cm−2 for 10 min. A graphite rod and Hg/HgO electrode were used as the counter and reference electrodes, respectively.

### Preparation of IrO2/Pt coated Ti-foam electrodes for the oxygen evolution reaction (OER in neutral)

A catalyst ink was prepared by ultrasonically mixing 5 wt% of ionomer solution (Nafion, 10 wt% target of catalyst) and commercial iridium oxide catalyst powder (Alfa Aesar, 30 mg) with ethanol (2 mL). The electrodes were fabricated by spraying the prepared catalyst ink was onto a Pt coated Ti foam at 70 °C. The electrode area was 2 cm2 and the loading of IrO2 was fixed at 1 mg cm−2. The reason of IrO2 catalyst for OER in neutral electrolyte was described in Supplementary Note 2.

### Electrochemical CO2RR flow cell tests

A detailed schematic of the flow cell used for evaluating the electrochemical CO2RR performance is shown in Supplementary Fig. 7. The fabricated SnO2-based catalyst electrodes were used as the cathode. The active area of each electrode was 2 cm2, and CO2 as the reactant gas was fed into the serpentine flow field channel on the cathode side at a flow rate of 50 sccm. The electrolyte solution of 1 M KOH or 1 M KHCO3 was supplied to both the anode and cathode sides using a pump. An AEM (Dioxide Materials, X37-50 Grade RT) was used to separate the anode and cathode flow channels. All electrochemical tests were conducted using a VSP potentiostat (BioLogic, VMP3B-10), which was suitable for measurements up to 10A. The reference electrode (Ag/AgCl, 3.5 M KCl) was inserted in the cathode flow line to measure and control the cathode potential.

The Faradaic efficiency of the catalyst was measured using GC and IC where points were taken at 18 min-intervals. The formate concentrations on the catholyte and anolyte were measured to calculate the total production of formate. The composition of the outlet gas was measured using gas chromatography (GC, Agilent 7890A). The inlet of the GC, fitted with a water trap to prevent water from entering the GC, was connected to the cathode outline. Ultrahigh-purity helium gas (99.9999%) was used as the carrier gas. A flame ionization detector and thermal conductivity detector were used to detect carbon-based gases (CO, CH4, and C2H4) and hydrogen gas, respectively. A methanizer was used to enhance the detection of CO. The faradaic efficiency of each product was calculated using the following equation:

$${{{{{{\rm{FE}}}}}}}_{{{{{{\rm{product}}}}}}}\left( \% \right)=\,\frac{{i}_{{{{{{{\mathrm{product}}}}}}}}}{{i}_{{{{{{{\mathrm{total}}}}}}}}}\,\times 100\,=\,\frac{{V}_{{{{{{{\mathrm{product}}}}}}}}\times Q\,\times \frac{2{Fp}}{{RT}}}{{i}_{{{{{{{\mathrm{total}}}}}}}}}\,\times 100$$
(3)

Where F is the Faraday constant (96485 C mol−1), Q is the flow rate of products, T is room temperature (298 K), and R is the ideal gas constant (8.314 J mol K−1).

The total current was measured using a VSP potentiostat. The peaks in the GC chromatogram were used to determine the volumes of specific products, which allowed the partial current density of each product to be calculated. Ion chromatography (IC) coupled with inductively coupled plasma optical emission spectroscopy (Thermo Scientific, Dionex ICS-5000+ HPIC) was used to monitor formate, and FEformate was calculated using the following equation:

$${{{{{{\rm{FE}}}}}}}_{{{{{{\rm{formate}}}}}}}\left( \% \right)=\,\frac{{i}_{{{{{{{\mathrm{foramte}}}}}}}}}{{i}_{{{{{{{\mathrm{total}}}}}}}}}\,\times 100\,=\,\frac{{C}_{{{{{{{\mathrm{formate}}}}}}}}\times N\,\times F}{{i}_{{{{{{{\mathrm{total}}}}}}}}}\,\times 100$$
(4)

where N is the number of electrons transferred and F is the Faraday constant. The peak in the IC chromatogram was used to determine the concentration of specific products.

### Physical characterization

The microstructures of the synthesized catalysts were observed using HR-TEM (FEI Co, Titan 300 kV), and the elemental distribution in each catalyst was obtained using energy-dispersive X-ray spectrometry (FEI Co, Talos 200 kV). The nanoparticle size distributions were analyzed using ImageJ software, and the obtained profiles were fitted using a Gaussian function. Ten HR-TEM images were taken for each oxide sample, from which 300 particles were selected for particle size measurements. The surface chemistry of the catalysts was evaluated using X-ray photoelectron spectroscopy (Ulvac Co., PHI 5000 Versaprobe). Wide-angle XRD (Rigaku D-max/2500-PC, Cu-Kα radiation) was used to investigate the crystal structure and identify the nature of the oxides. TGA (TA Instruments, Q600 SDT) of the catalysts was performed from room temperature to 800 °C at a rate of 10 °C min−1. This analysis was performed in a N2 atmosphere up to 500 °C, and O2 gas was injected thereafter.

### In-situ/operando X-ray absorption spectroscopy (XAS)

The Sn k-edge hard-X-ray absorption spectroscopy (XAS) spectra of the synthesized catalysts were recorded at the 10C beamline of the Pohang Acceleration Laboratory (PAL). The setup for the in-situ/operando hard-XAS measurements with a homemade electrochemical single cell is shown in Supplementary Fig. 16. For operando XAS measurements, a 1-cm2 hole was made in the anode and cathode bipolar plate and covered with Kapton film to allow passage of the X-rays. The operational conditions were the same as in the single-cell tests, and 1M KHCO3 was used as the catholyte and anolyte. Before the XAS measurements, the electrode was stabilized for 5 min at each potential. The hard-XAS analysis was carried out in the fluorescence collection mode using a Si (311) monochromator. The hard-XAS spectra were calibrated using Sn foil to ensure a zero shift in the edge energy. The XANES data were fitted using the Athena software (Demeter ver. 0.9.20). To maintain consistency in the analysis, the height of the arctangent function corresponding to the transition to the continuum level was set to one.

For the EXAFS analysis, Artemis (also implemented in Demeter ver. 0.9.20) software was utilized after the processing of data using the Athena software. The background signal was removed to extract the EXAFS signal for Rbkg = 1.0–1.1 Å. The EXAFS data were transformed using the Kaiser–Bessel function. The many-body reduction factor (S02) for Sn was determined to be 0.86 from the EXAFS curve fit of the Sn foil. The statistical quality of the curve fit to the proposed models can be determined from the R-factor and χ2 function available in the refinement.

### In-situ/operando surface-enhanced Raman spectroscopy

The in-situ/operando surface-enhanced Raman spectroscopy (SERS) measurements for the gas-phase CO2RR were performed using a homemade electrochemical three-electrode cell with a GDL, as shown in Supplementary Fig. 17. The excitation light source was a Nd:YAG laser (532 nm). A platinum wire was used as the counter electrode, Ag/AgCl (3.5 M KCl) was used as the reference electrode, and the catalyst-loaded GDL was used as the working electrode. For the CO2RR, 100 sccm CO2 was supplied to the catalyst-loaded GDL. The homemade electrochemical cell was filled to a thickness of 5 mm with 1 M KHCO3 as the electrolyte. The electrochemical experiments were controlled using an IVIUM CompactStat.h potentiostat/galvanostat.

### DFT calculations

DFT calculations were performed using the Quantum ESPRESSO package55,56. Geometry optimization was performed using the Perdew–Burke–Ernzerhof functional57 with the projector-augmented wave pseudopotentials58,59. The Grimme’s D3 (DFT-D3)60 method was used to account for the van der Waals dispersion correction. A kinetic energy cutoff of 500 eV was used with a plane-wave basis set. Gaussian smearing was applied with a smearing width of 0.1 eV. The geometries were fully relaxed until the residual force on the atoms converged to 0.01 eV/Å. The SnO2 (110) supercell was modeled using a unit cell of the tetragonal space group (mp-856) obtained from the Materials Project61. The four layers of the primitive unit cell that are cleaved to the (110) plane were expanded six times (2 × 3 × 4), yielding a supercell of 24 Sn and 48 O atoms with a lattice constant of 6.83 Å × 9.73 Å × 13.67 Å. To build the FTO (110) supercell, seven O atoms (~15%) were randomly replaced with F atoms. To simulate the SnO2 and FTO crystal conditions under an applied potential of −1.0 V, the lattice constants of the SnO2 and FTO (110) bulk were uniformly reduced by −1.37% and −0.33%, respectively (Supplementary Fig. 15 and Supplementary Table 1). The lattice constants of strained SnO2 and FTO (110) bulk supercell were 6.74 Å × 9.60 Å × 13.48 Å and 6.81 Å × 9.70 Å × 13.62 Å, respectively. The Brillouin zone was sampled with a Monkhorst–Pack k-point mesh of (5 × 4 × 2) for bulk SnO2 and FTO (110) supercells. To build the slab structure, a 20-Å vacuum gap was added along the c-axis of the stabilized supercells. The bottom two layers of the slab were fixed, while the top two layers were allowed to relax during geometry optimization. The Brillouin zone was sampled with a Monkhorst–Pack k-point mesh of (5 × 3 × 1) for SnO2 and FTO (110) slab supercells.

The elementary steps of the electrochemical conversion reaction of CO2 to HCOOH involving 2 electron pathway are described as follows:

$${{{{{{\rm{CO}}}}}}}_{2}+\ast \to {{{{{{\rm{CO}}}}}}}_{2}\ast$$
(5)
$${{{{{{\rm{CO}}}}}}}_{2}\ast +{{{{{{\rm{H}}}}}}}^{+}+{{{{{{\rm{e}}}}}}}^{-}\to {{{{{\rm{HCOO}}}}}}\ast$$
(6)
$${{{{{\rm{HCOO}}}}}}\ast +{{{{{{\rm{H}}}}}}}^{+}+{{{{{{\rm{e}}}}}}}^{-}\to {{{{{\rm{HCOOH}}}}}}+\ast$$
(7)

where * represents the surface sites for molecule adsorption. The change in Gibbs free energy (ΔG) at 298 K and 1 atm was calculated thorough ΔG = ΔE + ΔZPE – TΔS, where ΔE is the total electronic energy obtained from the DFT optimization, ΔZPE is the change in the zero-point energies, T is the temperature, and ΔS is the change in entropy. The computational hydrogen electrode model62 was applied to calculate the chemical potential of proton/electron pairs, which is equal to the half of the chemical potential of H2 gas under standard conditions and electrons with an applied bias of U (−eU). The pH contribution is considered by adding kBT × ln10 × pH to ΔG, where kB is the Boltzmann constant. The ZPE and S of the molecules and adsorbates were determined from the calculated vibrational frequencies and NIST database63,64, where all vibrations were treated in the harmonic oscillator approximation. The ZPE and S data are listed in Supplementary Tables 2 and 5, respectively.