Stabilizing indium sulfide for CO₂ electroreduction to formate at high rate by zinc incorporation

Abstract

Recently developed solid-state catalysts can mediate carbon dioxide (CO₂) electroreduction to valuable products at high rates and selectivities. However, under commercially relevant current densities of > 200 milliamperes per square centimeter (mA cm⁻²), catalysts often undergo particle agglomeration, active-phase change, and/or element dissolution, making the long-term operational stability a considerable challenge. Here we report an indium sulfide catalyst that is stabilized by adding zinc in the structure and shows dramatically improved stability. The obtained ZnIn₂S₄ catalyst can reduce CO₂ to formate with 99.3% Faradaic efficiency at 300 mA cm⁻² over 60 h of continuous operation without decay. By contrast, similarly synthesized indium sulfide without zinc participation deteriorates quickly under the same conditions. Combining experimental and theoretical studies, we unveil that the introduction of zinc largely enhances the covalency of In-S bonds, which “locks” sulfur—a catalytic site that can activate H₂O to react with CO₂, yielding HCOO* intermediates—from being dissolved during high-rate electrolysis.

Introduction

Electrosynthesis of value-added fuels using carbon dioxide (CO₂) as a feedstock provides an appealing route to reduce climate-changing CO₂ emission and a solid stepping-stone towards carbon neutrality^1,2. Over the past decade, the development of catalysts that are active and selective for CO₂ reduction reaction (CO₂RR), and meanwhile, suppress the competing hydrogen evolution, has been the subject of intensive study. This has resulted in a variety of carbon-based products to be synthesized from CO₂, such as carbon monoxide (CO)³, formate (HCOO⁻)⁴, methane⁵, and higher hydrocarbons and oxygenates (e.g., ethylene⁶, ethanol⁷, and n-propanol⁸). Despite remarkable advances, recent techno-economic analyses showed that CO and formate are probably the only products that can achieve the industrialization trend of CO₂RR in the near future^9,10. Regarding the formate, its profitable production requires current densities of ≥ 200 mA cm⁻², Faradic efficiency (FE) of > 90%, and power conversion efficiencies of > 50% (refs. ^9,10). Further, adopting solid-electrolyte electrolyzers permits the continuous production of formic acid without the separation process, making it even more economically viable⁴.

Early research on CO₂RR from Hori and co-workers revealed that a number of metals, such as lead, mercury, indium (In), bismuth (Bi), cadmium, and tin, could convert CO₂ to formate, but many of these metals suffer from unsatisfactory selectivity or toxic issue¹¹. Improvements in efficiency and selectivity have been achieved on nontoxic metallic catalysts via controlling of catalyst morphologies and dimensionalities^12,13, creation of vacancies¹⁴, and introduction of other elements (e.g., O, S, and P) to form new phases^15,16,17,18. When immobilized on the gas diffusion electrodes that surmount CO₂ mass transport limitation, commercially relevant rates (> 200 mA cm⁻²) and FE (> 90%) were observed to be reached on Bi nanosheets¹⁹, Bi₂O₃ nanotubes²⁰, Bi metal-organic layers²¹, and InP quantum dots¹⁷. However, the prospect on the potential of these catalysts for long-term operational stability is elusive. At high current densities, catalyst stability perhaps becomes a very important challenge. Often, CO₂RR activity deteriorates rapidly during high-rate electrolysis, owing to reasons like catalysis agglomeration²², active-phase change^12,23, and element dissolution^15,24. Unfortunately, previous research effort on catalyst stability, especially working at commercially relevant current densities, has remained rather rare. To make renewably powered formate electrosynthesis from CO₂ to be practical, it is critically necessary to develop catalysts that are not only active but also stable, and to gain insights on mechanisms of mediating the intrinsic stability.

Here, we report that incorporation of zinc (Zn) into indium sulfide (In₂S₃) synthesis enables tuning over its phase and structure, which dramatically improves the long-term stability of the resultant catalyst (ZnIn₂S₄) although the catalyst morphology remains almost unchanged. Comprehensive experiments coupled with computational studies reveal an enhanced covalency of In−S bonds mediated by Zn, which overstabilizes sulfur—a catalytic site that can activate H₂O to react with CO₂, leading to the formation of HCOO* intermediates—in the catalyst structure. Consequently, we achieved nearly 100% CO₂-to-formate conversion at a current density of 300 mA cm⁻² over 60 h without degradation, corresponding to a high production rate of 8,894 μmol cm⁻² h⁻¹.

Results

Synthesis and characterizations of catalysts

We had an interest in indium sulfide as a catalyst because S-doped In was shown by Wang and co-workers to be effective for catalyzing CO₂RR to formate. The presence of S enables facile activation of H₂O to form adsorbed H*, which consequently reacts with absorbed CO₂ to yield HCOO* intermediates¹⁶. However, the stability of S-doped In was only assessed under ~60 mA cm⁻² during a 10 h period; the prospect of such catalyst for durable high-rate CO₂-to-formate conversion is unclear. Very recently, Xia et al. reported that exfoliated ultrathin ZnIn₂S₄ nanosheets with rich Zn vacancies show improved CO₂RR ability to formate¹⁴. Although interesting, its long-term stability at current densities relevant to commercial operation (> 200 mA cm⁻²) was not evaluated. These results motivated us to examine the ability of indium sulfide instead of S-doped In for mediating CO₂ to formate. We synthesized indium sulfide hydrothermally by the reaction of InCl₃·4H₂O and C₂H₅NS in deionized water (DIW) at 160 °C (Supplementary Fig. 1). Cubic In₂S₃ (JCPDS 65-0459; Fig. 1i) was produced after 6 h, exhibiting flower-like morphology composed of hierarchically organized nanosheets (Supplementary Fig. 2). Indeed, we observed good formate selectivity on In₂S₃, but the performance degraded quickly at high current densities owing to the dissolution of S²⁻ ions (discussion later).

Fig. 1: Physical characterization of ZnIn₂S₄.

a, b SEM images of the ZnIn₂S₄ catalyst. The right panel in b shows the crystal structure of ZnIn₂S₄. Scale bars, 5 μm (a) and 1 μm (b). c STEM-EDX elemental mapping of ZnIn₂S₄, exhibiting a uniform spatial distribution of Zn (red), In (green), and S (yellow), respectively. Scale bar, 1 μm. d, e Atomic-resolution Z-contrast images of ZnIn₂S₄ along [001] zone axis. Scale bars, 1 nm (d) and 0.5 nm (e). f The corresponding FFT pattern of (d). g The line intensity profile acquired along the yellow arrow in (d). h Atomic model of ZnIn₂S₄ along [001] zone axis. i–k XRD patterns (i), UPS spectra (j), and BET surface area analysis (k) of ZnIn₂S₄ and In₂S₃, respectively.

Previous experimental studies revealed that adding Zn in some transition metal chalcogenides (e.g., Co₃S₄)²⁵ can enhance the structure robustness. Thus we sought to improve the stability of high-rate CO₂RR by incorporating Zn into indium sulfide. We used the same hydrothermal method for preparing the desired product except the addition of ZnCl₂ during the synthesis (Supplementary Fig. 1). Intriguingly, we obtained hexagonal-structured ZnIn₂S₄ (JCPDS 65-2023; Fig. 1i) microflowers that consist of hierarchically organized nanosheets (Fig. 1a, b), which closely resemble In₂S₃ described above. The thicknesses of nanosheets were determined to be ~8.69 nm for ZnIn₂S₄ and 9.32 nm for In₂S₃ through atomic force microscopy (AFM) measurements (Supplementary Fig. 3). We note that the synthesis of flower-like ZnIn₂S₄ was previously reported^26,27,28, whereas the analogous morphologies of ZnIn₂S₄ and In₂S₃ that synthesized by the same protocol here will underpin a fair performance comparison. Energy-dispersive X-ray (EDX) spectrum elemental mapping exhibits a uniform spatial distribution of Zn, In, and S (Fig. 1c). This simple synthetic strategy enables the production of high-yield ZnIn₂S₄ material with good fidelity for potential large-scale adoption (Supplementary Fig. 4).

We studied the detailed atomic structure of the ZnIn₂S₄ by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). The atomic-resolution Z-contrast images in Fig. 1d, e clearly reveal hexagonal lattice, where In atoms exhibit higher image intensity than the overlapped Zn and S atoms (Supplementary Fig. 5). The fast Fourier transform (FFT) result exhibits the (100) and (010) reflections (Fig. 1f). Using image contrast, the In and Zn(S) atoms can be further identified by the line intensity profile (Fig. 1g) acquired along the yellow arrow in Fig. 1d. The corresponding atomic model depicts that all the overlapped Zn and S atoms are located at the centers of honeycomb (Fig. 1h). Without the addition of Zn leads to the crystallization of cubic In₂S₃ by the same synthetic protocol (Fig. 1i and Supplementary Fig. 2). Structurally, ZnIn₂S₄ belongs to (ZnS)_mIn₂S₃ (m = 1−3) system²⁹, which bears an orderly alternation of S and Zn(In) (Fig. 1b, right). The sequence of atoms along the [001] direction is S−Zn_T−S−In_O−S−In_T−S, where Zn_T and In_T occupy the tetrahedral coordination and In_O occupies the octahedral site, respectively^29,30. By comparison, one-third of the tetrahedral sites in In₂S₃ is unoccupied³¹. The incorporation of Zn alters the coordination environment of indium sulfide and thus might tailor favorably the electronic structure and catalytic properties.

To probe the electronic structures of ZnIn₂S₄ and In₂S₃, we measured the work function by ultraviolet photoelectron spectroscopy (UPS) (Fig. 1j). Our results show a lower work function of ZnIn₂S₄ (5.12 eV) compared to In₂S₃ (5.29 eV), revealing a superior electronic property by the incorporation of Zn element, consistent with electrochemical impedance spectroscopy (EIS) results (Supplementary Fig. 6). We speculate, on the basis of the above results, that CO₂RR may be highly favored on such ternary In-based sulfide owing to the modulated coordination environment and electronic structure. Moreover, we determined the Brunauer−Emmett−Teller (BET) surface areas of ZnIn₂S₄ and In₂S₃ to be 71.3 and 70.0 m³ g⁻¹ (Fig. 1k), respectively.

CO₂RR performances in a flow cell

We examined CO₂RR properties of ZnIn₂S₄ and In₂S₃ catalysts in a flow cell (Supplementary Fig. 7) using recirculated 1 M KHCO₃ (pH 8.4) as electrolyte. CO₂ gas was fed at the cathode with a flow rate of 24 mL min⁻¹; the outlet gas flow rate was also measured for accurate product analysis (see “Methods”; Supplementary Fig. 8). We quantified the solution-phase and gas-phase products by using nuclear magnetic resonance (NMR) spectroscopy and on-line gas chromatography (Supplementary Fig. 9), respectively. The linear sweep voltammetry curves in Fig. 2a show sharp reduction peaks for ZnIn₂S₄ and In₂S₃ catalysts in a CO₂ environment. In a N₂ environment, however, the two catalysts exhibit a slight current−voltage response. In comparison with In₂S₃, the onset potential for CO₂RR on ZnIn₂S₄ catalyst shifted to a more positive value, implying enhanced CO₂RR kinetics (Fig. 2a). Figure 2b shows that the Faradaic efficiency (FE) for formate on ZnIn₂S₄ catalyst was always greater than on In₂S₃ at all potentials examined (Supplementary Figs. 10, 11). Notably, the ZnIn₂S₄ catalyst yields peak FE of 99.3% for formate at −1.18 V versus a reversible hydrogen electrode (RHE), while the competing hydrogen evolution reaction (HER) on this catalyst was substantially suppressed (Fig. 2b and Supplementary Fig. 12). With this FE, we achieved a CO₂RR to formate partial current density of ~298 mA cm⁻² (Fig. 2c), representing the highest value reported to date under KHCO₃ environments (Fig. 2e). We also performed reference measurements of hexagonal ZnS (JCPDS 39-1363) that synthesized by the identical route for comparison, which, however, overwhelmingly produces H₂ (Supplementary Fig. 13). Additionally, our series of control experiments disclosed that the optimum CO₂RR performance was gained on ZnIn₂S₄ catalyst that hydrothermally synthesized at 160 °C for 6 h with a ZnCl₂:InCl₃·4H₂O ratio of 1:2 (Supplementary Figs. 14–19).

Fig. 2: CO₂RR performances.

a, b The linear sweep voltammetry curves (a) and potential-dependent Faradaic efficiencies for products (b) on ZnIn₂S₄ and In₂S₃. c, d Partial current density (c) and half-cell PCE (d) for CO₂-to-formate conversion on ZnIn₂S₄ and In₂S₃. e, f Comparison of formate partial current densities and FEs (e), and formate production rates (f) for various catalysts reported under KHCO₃ environments (see Supplementary Table 1 for details). g Stability test of the ZnIn₂S₄ and In₂S₃ at 300 mA cm⁻². The electrolyte was occasionally replaced by new 1 M KHCO₃ solution (red arrows) to recover the ionic concentration and conductivity of the anolyte. The error bars represent the standard deviation of three independent measurements.

Figure 2d presents the half-cell power conversion efficiency (PCE) for CO₂-to-formate conversion under various applied potentials. At −1.18 V versus RHE, our full-cell (CO₂ + H₂O → formate + O₂) device shows a half-cell formate PCE exceeding 50% on ZnIn₂S₄ catalyst. A comprehensive review of recent literature revealed that our ZnIn₂S₄ catalyst exhibits superb selectivity and partial current density (Fig. 2e), which result in a formate production rate of up to 8,894 μmol cm⁻² h⁻¹, outperforming all previous results^{4,14,16,19,20,21,24,32,33,34,35,36,37,38,39,40,41,42,43} that have been reported under KHCO₃ environments (Fig. 2f).

We used density functional theory (DFT) to obtain insights into the CO₂RR properties of the studied catalyst (see “Methods” for details). We compared the Gibbs free energies (△G) for the formation of formate intermediate (HCOO*) on the surfaces of ZnIn₂S₄, In₂S₃, and In models (Supplementary Figs. 20–22). The computed barrier of HCOO* formation is 117 meV for ZnIn₂S₄ and 120 meV for In₂S₃, smaller than that of 270 meV for In, implying that S sites favorably mediate the HCOO* formation. Our calculations further reveal the lowest barrier of HCOOH* formation on ZnIn₂S₄, leading to its superior CO₂-to-formate ability. These results indicate that formate preferentially generates on ZnIn₂S₄ catalyst. By contrast, *COOH (intermediate of CO product) formation is highly endergonic on the three In-based catalysts (Supplementary Figs. 20–22), causing the production of CO to be virtually prohibited. Although HER process is largely hampered, our DFT results reveal that, as compared to In₂S₃ and In, the S sites of ZnIn₂S₄ enable much smaller hydrogen adsorption free energy of 370 meV. Early works^15,16 have reported that S acts as a promotor to enhance CO₂-to-formate conversion, we thus reasonably surmise that S sites on ZnIn₂S₄ surface permit easier H₂O dissociation to adsorbed H* species, which then react with CO₂ to yield HCOO* intermediates.

Comprehensive stability study

Aside from activity, long-term stability—especially operating at high current densities (> 200 mA cm⁻²)—is another critical metric for CO₂ electrolysis technique to be practical¹⁰. Figure 2g shows the key finding that we wish to report in this work: that is, the CO₂RR stability of indium sulfide can be remarkably improved by the incorporation of Zn. We tested the stability of ZnIn₂S₄ catalyst at a profitable current density as large as 300 mA cm⁻², during this process portions of the electrolyte were frequently taken out for NMR analysis. The formate FE could be held at > 97% over 60 h of continuous electrolysis without the need of additional overpotentials (Fig. 2g). By contrast, In₂S₃ reference exhibited a rapid drop in formate selectivity, whereas the FE toward H₂ climbed up to ~90% within 8 h. We hypothesize that such severe performance drop might be caused by the structure degradation during high-rate electrolysis. Notably, the exceptional stability of ZnIn₂S₄ catalyst enables us to produce ~327 mmol formate after 60 h (Fig. 2g).

We combined multiple characterization techniques to track the structural evolution of ZnIn₂S₄ and In₂S₃ catalysts during CO₂ electrolysis under various current densities and operating times (Fig. 3). X-ray diffraction (XRD) patterns and scanning electron microscopy (SEM) studies reveal that the phase and morphology of ZnIn₂S₄ catalyst were well retained when progressively increasing the current density even up to 500 mA cm⁻² (Fig. 3a, c). By contrast, In₂S₃ catalyst undergoes a complete phase transition to metallic In (JCPDS 65-9292; Fig. 3b) at a current density of mere 50 mA cm⁻², accompanied by a dramatic morphology change (Fig. 3c, below images) owing to the loss of S that leads to structure collapse. Our Raman spectroscopy measurements on ZnIn₂S₄ show that two characteristic peaks at 248 (LO₁ mode) and 340 cm⁻¹ (LO₂ mode)^44,45 were retained after 60 h of operation at 300 mA cm⁻² (Fig. 3d). However, the characteristic Raman modes (A_1g and E_g)⁴⁶ of In₂S₃ disappeared while Raman signals from metallic In (ref. ⁴⁷) were detected within mere 1 min (Fig. 3e and Supplementary Figs. 23a, 24). The Raman results are consistent well with our post-mortem SEM analyses (Supplementary Fig. 25) and XRD results (Supplementary Fig. 23b).

Fig. 3: Structural stability of ZnIn₂S₄.

a, b XRD patterns of ZnIn₂S₄ (a) and In₂S₃ (b) after CO₂ /Subscript> electrolysis under various current densities for 10 min. c Corresponding SEM images of ZnIn₂S₄ (above) and In₂S₃ (bottom). Scale bars, 1 μm (above) and 500 nm (bottom). d–g Raman spectra of ZnIn₂S₄ (d) and In₂S₃ (e), and S 2p XPS spectra of ZnIn₂S₄ (f) and In₂S₃ (g) after CO₂ electrolysis for various times at 300 mA cm⁻². h STEM-EDX elemental mappings of ZnIn₂S₄ (scale bar: 1 μm) and In₂S₃ (scale bar: 600 nm) after running for 60 h and 8 h at 300 mA cm⁻², respectively. i SEM-EDX measurements of the remained sulfur in catalysts after running for various times at 300 mA cm⁻². The error bars represent the standard deviation of three independent measurements. j TEM (above, scale bars: 50 nm) and SAED patterns (down, scale bars: 5 1/nm) of ZnIn₂S₄ catalyst after CO₂ electrolysis for various times at 300 mA cm⁻².

Of note that the severe loss of S for In₂S₃ catalyst was further verified by X-ray photoelectron spectroscopy (XPS; Fig. 3g), STEM-EDX elemental mapping (Fig. 3h) and SEM-EDX (Fig. 3i). This is starkly contrasted with ZnIn₂S₄ whose chemical state and content of each elements (i.e., Zn, In, and S) were nearly unaltered after 60 h of high-rate CO₂ electrolysis (Fig. 3f, h and Supplementary Figs. 26, 27). We quantified the amount of S remained in ZnIn₂S₄ and In₂S₃ catalysts by using SEM-EDX (Supplementary Figs. 28–30), which permits a quantitative compositional analysis at a relatively large scale. As shown in Fig. 3i, the amount of S in In₂S₃ drops to 2.13 wt% from its original value (23.6 wt%) within the first 1 h, followed by a slow drop to almost zero over the next 2 h. This result is consistent with the significantly increased S amount in an electrolyte that measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES; Supplementary Fig. 31). By contrast, the ZnIn₂S₄ catalyst shows negligible loss of S after 60 h. Moreover, selected-area electron diffraction (SAED) analyses of the used ZnIn₂S₄ catalyst reveal that the single-crystalline hexagonal phase well maintains after the aggressive long-term stability test (Fig. 3j). We further note that ZnS catalyst also performs very stable at high current densities (Supplementary Fig. 32) owing to the strong interaction between Zn and S (refs. ^16,48,49), although it mainly produces H₂ (Supplementary Fig. 13).

Stability enhancement mechanism

Our results above conclusively demonstrate that the stability degradation of In₂S₃ can be attributable to S leaching, and show the primacy of Zn as a stabilizer in indium sulfide that prevents S to be leached out. Besides indium sulfide, S leaching was also widely observed in other metal sulfides, while the dissolution mechanism is rather complex^6,18,50,51. We turned to use DFT calculations to study the cause of the enhanced stability of indium sulfide after incorporating Zn. Compared with In₂S₃ having tetrahedral vacancies³¹, the tetrahedral and octahedral sites in ZnIn₂S₄ are fully occupied³⁰ after Zn incorporation. Notably, in ZnIn₂S₄, all Zn atoms bind with S through tetrahedral coordination, which implies the formation of strong Zn_T−S bonds considering that tetrahedral structures commonly give covalent feature⁵². Our computed differential charge density and its projection on the (110) plane map clearly reveal an enhanced electron cloud between Zn and S atoms (Fig. 4a, b), revealing electron donation from Zn to S due to the strong reducibility of Zn atoms. The transfer of electrons from In to S can also be seen more pronounced in ZnIn₂S₄ (Fig. 4a, b) than that in In₂S₃ (Fig. 4d, e), which leads to charge accumulation around In−S bonds and correspondingly higher covalency⁵³.

Fig. 4: Enhanced covalency in ZnIn₂S₄.

a, b Differential charge density (a) and projection on the (110) plane (b). c ELF of ZnIn₂S₄. d, e Differential charge density (d) and projection on the (011) plane (e). f ELF of In₂S₃. The azure and yellow clouds represent electron density depressions and accumulations, respectively. g–i COHPs for In−S bonding (g) and Zn−S bonding (h) of ZnIn₂S₄, as well as In−S bonding (i) of In₂S₃.

The calculated electronic localization function (ELF) of the tetrahedral In_T−S and octahedral In_O−S bonds in ZnIn₂S₄ are 0.84 and 0.79 (Fig. 4c), which compare larger than that of 0.71 and 0.76 in In₂S₃ (Fig. 4f), indicating a greater localization of S−In_O−S−In_T−S (ref. ⁵⁴). Likewise, the ELF of tetrahedral Zn_T−S bond in ZnIn₂S₄ was calculated to be 0.81 (Fig. 4c), pointing to its localized covalent feature. The interatomic bond strengths were further quantitatively analyzed by the projected crystal orbital Hamilton population (pCOHP) method (Fig. 4g−i). We found that the anti-bonding orbitals of In−S and Zn−S for ZnIn₂S₄ are less occupied. Moreover, our calculations yield integrated pCOHP values below the Fermi level of −0.763 and −0.737 for In−S bonds in ZnIn₂S₄ and In₂S₃, respectively, again demonstrating greater bond strengths in ZnIn₂S₄ (ref. ⁵⁵). These results, therefore, indicate that the bond breaking between In(Zn) and S in ZnIn₂S₄ is kinetically cumbersome, which explains the negligible S dissolution and thus exceptional long-term stability of the ZnIn₂S₄ catalyst (Supplementary Fig. 33).

Discussion

In closing, we have shown long-term formate electrosynthesis from CO₂ at a high current density of 300 mA cm⁻² on a cost-effective indium sulfide catalyst modulated by Zn. The extraordinary catalyst stability can be explained by the increase of In−S covalency, which substantially prevents sulfur dissolution during CO₂RR. We achieved selective and fast CO₂-to-formate conversion with a formate FE of 99.3% and a notable formate production rate of 8,894 μmol cm⁻² h⁻¹. These findings will advance the development of efficient and durable catalysts for commercial-scale electrosynthesis of formate.

Methods

Material synthesis

All chemicals were used as received without further purification. Indium chloride tetrahydrate (InCl₃·4H₂O), thioacetamide (C₂H₅NS), and Zinc dichloride (ZnCl₂), were purchased from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). In a typical experiment, 3.0 mmol InCl₃·4H₂O was dissolved in 150 mL deionized water (DIW), and then 6.1 mmol C₂H₅NS was added with vigorous stirring for 20 min. The 20 mL resultant solution was transferred into a 50 mL Teflon-lined autoclave, sealed, and heated at 160 °C for 6 h. After the reaction, the obtained In₂S₃ powders were washed with excess DIW and absolute ethanol for at least three times, and then dried at room temperature in an oven under vacuum for further characterization. The synthesis of ZnIn₂S₄ was the same with the synthesis of In₂S₃, except the addition of 1.5 mmol ZnCl₂ during the first step. For the synthesis of ZnS, it needs to replace 3.0 mmol InCl₃·4H₂O with 1.5 mmol ZnCl₂.

Material characterizations

XRD was performed on a Japan Rigaku DMax-γA X-ray diffractometer with Cu Kα radiation (λ = 1.54178 Å). The morphology of the samples was investigated by SEM (Zersss Supra 40) and TEM (JEOL 2010F(s)). The STEM and HRTEM images, SAED and EDX elemental mapping were taken on JEMARM 200 F Atomic Resolution Analytical Microscope with an acceleration voltage of 200 kV. SEM-EDX was determined by GeminiSEM 500 with an Oxford Aztec series X-ray energy spectrum. Raman spectra was measured on a Raman microscope (HORIBA) with a 785 nm excitation laser. ICP-AES data were obtained by an Optima 7300 DV instrument. N₂ adsorption/desorption isotherms were recorded on an ASAP 2020 accelerated surface area and a porosimetry instrument (Mictromeritics), equipped with an automated surface area, at 77 K by using Barrett−Emmett−Teller calculations. XPS was taken on an X-ray photoelectron spectrometer (ESCALab MKII) with an X-ray source (Mg Kα hυ = 1253.6 eV).

Preparation of CO₂RR electrodes

The catalyst ink was prepared by ultrasonic dispersion of 10 mg catalyst powders in 1 ml isopropanol, which was mixed with 50 μL of 5 wt% Nafion. The resulted ink was uniformly spread on the gas diffusion layer (GDL, Sigracet 29 BC) of 3 × 3 cm² in the area by using an airbrush, yielding the prepared electrode with a catalyst loading of ~1.0 mg cm⁻².

Electrochemical measurements

All electrochemical measurements were performed in a flow cell with VSP-300 Potentiostat (Bio-Logic, France). For experiments in flow cells, gaseous CO₂ (99.999%) was passed through the gas chamber at the back side of the gas diffusion electrodes. Both catholyte and anolyte (1 M KHCO₃) were continuously circulated through the cathode and anode chambers separated by the cation exchange membrane (Nafion™ 117), which was used to avoid the crossover issues of formate⁵⁶. The cathode is the prepared gas diffusion electrode (GDE, 1 × 1 cm²), and the anode is a piece of nickel foam (1 × 1 cm²). The CO₂ inlet flowrates were kept constant at 24 mL min⁻¹ by a mass flow controller (C100L, Sierra). KHCO₃ electrolyte flowrates were maintained constant at 20 mL min⁻¹ controlled by a peristaltic pump (BT100-2J, Longer Pump). The CO₂ electrolysis lasted for 10 min unless otherwise specified. The linear sweep voltammetry (LSV) curves of ZnIn₂S₄ and In₂S₃ were performed in CO₂-fed and Ar-fed 1 M KHCO₃ solution. All potentials were measured against an Ag/AgCl (saturated KCl) reference electrode and converted to the RHE reference scale with iR correction on account of the equation:

$$E\,({{{{{\mathrm{vs}}}}}}\,{{{{{\rm{RHE}}}}}})=E\,({{{{{\mathrm{vs}}}}}}\,{{{{{\rm{Ag}}}}}}/{{{{{\rm{AgCl}}}}}})+0.205+(0.0591\times {{{{{\rm{pH}}}}}})-i{R}_{s}$$

(1)

Where the solution resistance R_s was determined by EIS over a frequency range from 100 KHz to 10 mHz.

CO₂RR products analysis

The gas products were analyzed by gas chromatography (GC-2014, Shimadzu) equipped with thermal conductivity detector (TCD) to quantify H₂ concentration and flame ionization detector (FID) to analyze the content of CO. Considering CO₂ consumption, the outlet flow rate was monitored by a mass flowmeter (AST10-HLC, Asert Instruments) before flowing to the on-line GC. The Faradaic efficiency for gas products (FE_x) was calculated by the following formula:

$${{{{{{\rm{FE}}}}}}}_{x}( \% )=\frac{{n}_{x}\times {C}_{x}\times u\times F}{I\times {V}_{M}}\times 100 \%$$

(2)

where F is the Faraday constant (96485 C mol⁻¹), I is the total current density, n_x is electrons transferred for reduction to product x, C_x is volume fraction of the product x detected by GC, u is outlet gas flowrate and V_M is molar volume (22.4 L mol⁻¹).

The formate products were quantified by ¹H NMR spectra measured with a Bruker 400 MHz spectrometer. Typically, 400 µL of the electrolyte after CO₂RR electrolysis was mixed with 200 µL of D₂O containing 50 ppm (m/m) dimethyl sulphoxide (DMSO) as the internal standard. The area ratio of the formate peak to the DMSO peak was compared to the standard curve to quantify the concentration of formate. The molar quantity of formate (n_formate) was calculated via multiplying the concentration of formate with the volume of the catholyte. The Faradaic efficiency of the formate (FE_formate) can be calculated by the following equation:

$${{{{{{\rm{FE}}}}}}}_{{{{{{\rm{formate}}}}}}}( \% )=2\times F\times \frac{{n}_{{{{{{\rm{formate}}}}}}}}{I\,\times \,t}\times 100 \%$$

(3)

where t is the CO₂ electrolysis time.

The half cell (cathodic) power conversion efficiency (PCE, assuming the overpotential of the oxygen evolution reaction is zero) of the formate products was calculated using:

$${{{{{\rm{PCE}}}}}}( \% )=\frac{(1.23-{E}_{{{{{{\rm{formate}}}}}}})\times {{{{{{\rm{FE}}}}}}}_{{{{{{\rm{formate}}}}}}}}{1.23-E}$$

(4)

where E is the applied potential vs RHE, E_formate is thermodynamic potential (−0.02 V vs RHE) of CO₂RR to formate⁵⁷.

The production rate for formate was calculated using the following equation:

$${{{{{\rm{Prodution}}}}}}\,{{{{{\rm{rate}}}}}}=\frac{Q\times {{{{{{\rm{FE}}}}}}}_{{{{{{\rm{formate}}}}}}}}{F\times 2\times t\times S}$$

(5)

where Q is the total charge passed and S is the geometric area of the electrode (1 cm²).

DFT calculations

The DFT calculations were performed by Vienna ab initio simulation package (VASP)⁵⁸ program with projector augmented wave (PAW)⁵⁹ method and the kinetic energy cut off was set to be 500 eV. The convergence criterion for electronic self-consistent iteration was set to be 10⁻⁴ eV. The atomic positions were fully relaxed until the force on each atom is less than 0.02 eV Å⁻¹. The Perdew−Burke−Ernzerhof⁶⁰ generalized gradient approximation exchange-correlation functional was used throughout. The slab model of In (101), In₂S₃ (311), and ZnIn₂S₄ (102) surface were constructed from the optimized In, In₂S₃, and ZnIn₂S₄ crystal structure. At the same time, a vacuum layer of 15 Å is established in the c-axis direction to ensure the separation between slabs. In addition, the surface formate species takes a unit negative charge, and, the present DFT calculation is not so great as to describe this kind of system carrying a neat charge. Thus, HCOOH was considered as the final product to describe this reaction rather than formate, in line with the DFT calculation in many researches^18,61,62. The COHPs were computed using the developed lobster program^63,64,65.

Here, The Gibbs free energies were calculated at 25 °C and 1 atm:

$$\varDelta {{{{{{\rm{G}}}}}}}_{{{{{{\rm{ads}}}}}}}=\varDelta {{{{{{\rm{E}}}}}}}_{{{{{{\rm{ads}}}}}}}+\varDelta {{{{{\rm{ZPE}}}}}}-{{{{{\rm{T}}}}}}\varDelta {{{{{\rm{S}}}}}}+e{{{{{\rm{U}}}}}}$$

(6)

where ΔE_ads, ΔZPE, T, ΔS, U, and e are the binding energy, zero-point energies changes, temperature, entropy changes, applied potential at the electrode, and charge transferred, respectively.