Protonation pathway for CO2 reduction mediated by coordinated H2O on active sites Authors: Huilin Qing, Evan Cline, Zheng Meng, Baiheng Li, Tai-De Li, Katherine A Mirica, Weiyang Li Published: 2025 ABSTRACT AbstractElectrochemical CO2 reduction to fuels necessitates multiple steps of electron and proton transfer. As one of the primary proton donors for CO2 reduction, H2O is essential to regulating the reaction pathway and product selectivity. However, the coordinating nature of H2O and its influence are poorly understood due to challenges in identifying the precise structure of catalytic sites. Here, we employ a well-defined bismuth-based metal-organic framework material with molecularly precise active sites and ordered microporosity for in-depth mechanistic studies. The coordinated H2O on active sites promotes the adsorption of CO2 and alleviates the sluggish proton supply via a protonated carbonic acid pathway involving a surface hydride transfer, which significantly promotes the adsorption and activation of CO2. This results in a 99% selectivity for CO2 reduction to formic acid with a high turnover frequency of 21.1 s−1. Leveraging the advantages of crystalline coordination frameworks in electrocatalysis, this work fills up a lacuna in our understanding of CO2 hydrogenation/reduction and reinforces the importance of H2O to the advanced design of catalytic systems.Subject terms: Electrocatalysis, Metal-organic frameworks, Electrocatalysis Introduction Electrochemical reduction of CO2 (eCO2RR) into value-added fuels and chemical feedstocks under ambient conditions is a promising approach for reducing atmospheric CO2 to achieve carbon neutrality and alleviate environmental issues1–3. CO2RR necessitates proton sources in the system, such as acids, HCO3–, and H2O. It is known as a complicated process undergoing multiple steps of electron and proton transfer3,4. Unfortunately, the presence of protons unavoidably leads to the co-existence of kinetically favored hydrogen evolution reaction (HER) over eCO2RR on the cathode, causing energy loss and the formation of H2 byproduct3,5. Thus, it is crucial to develop new materials and methods that enable the modulation of the proton sources and elucidation of the protonation pathways. Owing to the expense and polarity merits of H2O to tolerate ionic salts, an aqueous electrolyte has been most commonly used for CO2 reduction, where H2O is one of the primary proton donors and the predominant composition of the solid/liquid interface1,6. The transport of H2O to the catalytic interface and its subsequent auto-dissociation have considerable effects on the electrochemical performance of eCO2RR and HER7–9, which can be categorized into two aspects: effects on interfacial electrical diffusion layer and effects on active sites of solid electrocatalysts. For the electrical diffusion layer, tuning the ordered structure, transport rate, and activity of H2O enables the control of electron transfer, local pH value in the microenvironment, and proton supply6,10–13. In terms of solid electrocatalysts, H2O and species dissociated from water (i.e., H+ and OH–) can affect the reaction intermediates and pathways by adsorbing on the surface of electrocatalysts, coordinating with surficial active sites, or inducing surface reconstructions under applied potential bias8,14,15. For example, metal sites bound with H+ or OH– originated from water to form metal hydrides (M–H)16–19 or metal–hydroxides (M–OH)15,20–23 were proposed as essential intermediates or modified active sites for eCO2RR. Accordingly, extensive research interests have been placed on elucidating the dependence of the microenvironments and structures of active sites on H2O. Although endeavors have been made to reveal the essential roles of surface H2O-derived species on electrocatalysts in directing the activity and selectivity of eCO2RR, most reports concentrate strictly on metals21–24, oxides20, and chalcogenides19. For typical heterogeneous electrocatalysts, the aforementioned progress remains limited and restricted by the imprecise tailoring of catalytic structure and a lack of insightful studies on reaction mechanisms5,25–27. In addition, due to the dynamic nature of interfacial H2O, the coverage and distribution of species on the surface of electrocatalysts are not uniform, largely complicating the electrocatalytic process and constraining the utilization and exposure of active sites. Hence, to further reveal the specific influence of H2O on electrocatalysis and provide an in-depth understanding of the protonation pathway for catalyst design, it is critical to study materials with well-defined structures and identified active sites1,27–30. Highly ordered crystalline coordination frameworks, including metal-organic frameworks (MOFs) and coordination polymers (CPs), have shown great promise as heterogenous molecular catalysts for comprehensive catalytic mechanism studies, as well as precise catalyst design and regulation at the atomic level due to the presence of abundant definite active sites and tailorable structures31–34. Their long-range ordered microporosity provides a well-defined environment that accommodates and adsorbs small molecules, such as H2O and CO2, allowing for clear determination of adsorption sites by crystallographic analysis and enabling the trackability of the coordinated/adsorbed species (Supplementary Fig. 1)35,36. However, due to the electrically insulating features of most existing MOFs and CPs34,37, there are limited studies on their direct applications as electrocatalysts with a lack of understanding of the roles of H2O-derived species in their intrinsic activities and variation of protonation pathways38,39. The 2,3,6,7,10,11-hexahydroxytriphenylene (HHTP) has established precedent as an organic ligand for generating conductive MOFs since the large π-conjugated system and planar geometry enable efficient charge delocalization40. Among various structures, two-dimensional (2D) structures with high surface-to-edge ratios and preferential faceting take advantage of well-defined and uniformly distributed interfaces for studying the implications of surface species (e.g., H2O and other solvated ions) on electrochemical reactions1,41–43. Using HHTP to build conductive MOFs and CPs with well-defined 2D morphology can be an ideal platform for in-depth mechanistic studies of electrocatalysis and provide insights into exploring the applicability of crystalline coordination frameworks in electrochemical systems. This paper describes the effort to elucidate the roles of H2O species in a highly ordered bismuth (Bi)-based coordination framework Bi(2,3,6,7,10,11-hexahydroxytriphenylene) (denoted as Bi(HHTP)) in modulating the intrinsic catalytic activity of the active site and protonation pathways for CO2RR. Despite the defined crystal structure reported by our prior study40 and its promise for CO2 reduction44, the underlying electrocatalytic mechanisms remain elusive and there is a lack of understanding of catalytically active interfaces. Herein, we achieved morphological tuning to construct 2D crystalline sheets for controlling the electrocatalytic performance through preferential exposure of well-defined interfaces. We found that the 2D nanosheets of Bi(HHTP) exhibited high selectivity and activity of CO2 electroreduction to formic acid at a wide potential window. Our in-depth mechanistic studies present the direct, structurally resolved evidence underscoring the critical influence of coordinated H2O on CO2 reduction. The coordinated H2O on active sites not only facilitates CO2 adsorption and activation but also serves as a local proton supplier, promoting kinetics by forming a protonated carbonic acid intermediate. The adjunct electronegative atom in the vicinity of the metal center is key to stabilizing the adsorbed intermediates and enabling a hydride (H−) attack on electrophilic carbon to create C–H bonding for CO2 reduction. This work brings a fundamental understanding of electrocatalytic CO2 conversion for the rational design of catalysts and extends the applicability of MOFs and CPs to electrocatalysis. Moreover, it is anticipated that due to the ability of H2O to coordinate with diverse metals, the proposed protonated carbonic acid pathway may be extended to a variety of catalytic systems (e.g., metal oxides and single-atom catalysts) where the coordinated H2O can be an integral part of the active site, providing an avenue to promote specific reaction kinetics and control the electrocatalytic activity and selectivity. Results Materials design and structural characterization Direct use of pristine Bi-based MOFs and CPs for electrocatalysis is severely limited by the poor structural stability and electrical conductivity. Instead, this general class of materials has been primarily used as “pre-catalysts” to construct heterogeneous metallic Bi/bismuth oxides/bismuth carbonates upon electroreduction45–47. Preceding research revealed that ligand substitution between carboxylate-based organic ligands in the frameworks and bicarbonate ions in electrolytes induced structural transformation of coordination frameworks46,47. In contrast, Bi-MOF with catechol-containing ligands showed structural stability in neutral electrolytes for CO2 electroreduction45,48. Prompted by those findings, we anticipated that Bi(HHTP) could be an appropriate electrocatalyst to support mechanism studies for eCO2RR. Besides the advantage of building electrical conductivity through the large π-conjugation, the catechol-based coordination sites, a planar geometry, and the robust nature of HHTP are likely to prevent structural changes during the electrochemical process. Bi(HHTP) was synthesized via a bottom-up hydrothermal reaction between bismuth acetate and HHTP ligands (Supplementary Figs. 2–5) in an aqueous solution based on our previous work40. Structural elucidation by microcrystal electron diffraction (MicroED) showed that Bi(HHTP) exhibits two different structural forms upon dehydration and hydration, Bi(HHTP)-α and Bi(HHTP)-β (Fig. 1a)40. Compared to the α-structure, the pores in Bi(HHTP)-β are occupied by water molecules. Considering the large amount of water in aqueous electrolyte, the hydrated form may be the dominant species in aqueous environments and could facilitate rapid transfer of proton source during CO2 reduction. To construct Bi(HHTP) with well-defined 2D geometry, we applied chemical interface confinement using Pluronic F127 surfactants as structuring agents to control the growth of Bi(HHTP) and regulate the morphologies (Fig. 1b). The as-prepared materials were denoted as Bi(HHTP)-x wt% F127 where “x wt%” is the weight percentage of F127. Notably, F127 structuring agents were thoroughly washed out after synthesis as evidenced by the elemental analysis (Supplementary Table 1), and the as-purified materials were used for further characterizations and electrochemical measurements. Powder X-ray diffraction (pXRD) patterns of the as-synthesized samples (Fig. 2a and Supplementary Fig. 6) showed the characteristic diffraction peaks of Bi(HHTP) at 8.44, 10.40, 12.46, 14.10, and 27.08° for (002), (200), (202¯), (202), and (321¯), respectively. The considerable increase of peak intensity ratio of (002) facet with the addition of F127 indicated a preferential exposure of the family of {001} facets in the Bi(HHTP) frameworks (Fig. 2a and Supplementary Fig. 7). Crystal structure and Synthesis of Bi(HHTP). a Crystal structures of Bi(HHTP)-α and Bi(HHTP)-β within one unit cell. b Schematic depiction of chemical interface confinement growth of Bi(HHTP) nanosheets. The 3D visualization was generated using the Mercury 2025 software based on the reported crystallographic data40. The schematic was drawn using 3ds Max 2025 software. Materials characterizations of Bi(HHTP) synthesized at different concentrations of F127. a PXRD patterns and b SEM images of the as-synthesized Bi(HHTP) at different concentrations of F127: without (0 wt%) F127, 0.06 wt% F127, 0.6 wt% F127, and 1.8 wt% F127 (from bottom to top). To carefully assess their phase structures, the peak of (002) facet was selected as a reference to normalize the intensities. All the pXRD analysis and SEM imaging were conducted on Bi(HHTP) frameworks after the removal of F127. c AFM image (inset: height profile of white line), d TEM image, e high-resolution TEM image of a lattice pattern (inset: schematic drawing of the relation between crystal orientations and observed lattice in TEM images (top) and corresponding FFT pattern (bottom)), and f HAADF-STEM image and corresponding EDS mapping in the area. Source data are provided as a Source Data file. The morphologies and microstructures of Bi(HHTP) materials were investigated by scanning electron microscopy (SEM). Without F127, Bi(HHTP) crystals featured 3D cuboid-shaped bulk materials, while the morphological dimensionality of Bi(HHTP) was reduced to 2D nanosheets at a high concentration of F127 (Fig. 2b and Supplementary Fig. 8). The oriented crystal growth of Bi(HHTP) was ascribed to the competitive coordination between hydroxyl groups at the end of HHTP ligands and Pluronic polymers to Bi3+, influencing the nucleation and crystal growth rates (Supplementary Fig. 9). This morphological tuning principle was supported by the observation of interactions between bismuth acetate and F127 from attenuated total reflection-infrared spectra (ATR-IR) (Supplementary Figs. 10–12). The more significant surface density of Bi on {001} facets over other facets (Supplementary Fig. 13) favored the adsorption of Pluronic polymer on the specific {001} surfaces, impeding epitaxial assembly in the direction of the c-axis and resulting in the preferential faceting. The atomic force microscope (AFM) images in Fig. 2c and Supplementary Fig. 14 demonstrated the morphology of a rectangular nanosheet for Bi(HHTP)-1.8 wt% F127 with a thickness of about 10–20 nm. Transmission electron microscopy (TEM) images in Fig. 2d–e and Supplementary Fig. 15 also presented rectangular morphology of as-prepared Bi(HHTP)-1.8 wt% F127 nanosheets. The well-defined lattice fringes with an interplanar spacing of 1.68 nm and fast Fourier transformation (FFT) pattern in Fig. 2e could be assigned to the {100} facets of Bi(HHTP), the basal planes along the [001] zone axis, which further confirmed the good crystallinity and enhanced exposure of {001} facets in nanosheets. High-angle annular dark-field imaging (HAADF) and energy-dispersive X-ray spectroscopy (EDS) spectra in Supplementary Fig. 16 and Fig. 2f confirmed the presence of Bi, C, and O elements and their uniform distribution in materials. N2 sorption isotherms at 77 K suggested that 2D nanosheets showed larger surface areas than 3D bulk materials, where the Brunauer-Emmett-Teller (BET) surface areas were calculated as ~25, 33, 58 and 80 m2/g for Bi(HHTP)-0, -0.06, -0.6 and -1.8 wt% F127, respectively (Supplementary Fig. 17). CO2 sorption isotherms at 273 K in Supplementary Fig. 18 also supported the 2D nanosheets with larger CO2 uptake capacity. X-ray photoelectron spectra (XPS) (Supplementary Figs. 19, 20) depicted the oxidation state of +3 of Bi in the frameworks. Thermogravimetric analysis (TGA) demonstrated good thermal stability of Bi(HHTP) frameworks and a weight content of ~39% Bi in material (Supplementary Figs. 21–22). Examined by the four-point probe method, electrical conductivities of Bi(HHTP) were within the range of 103−10−8 S cm−1, indicating the successful implantation of semiconducting properties (Supplementary Table 2). Notably, 2D nanosheets exhibited better electrical conductivity than 3D bulk materials. The above characterizations showed morphological tuning did not alter the inherent chemical compositions and crystal structures of the Bi(HHTP) framework but led to enhancements of surface area and electrical conductivity by reducing dimensionality. Such a well-defined structure and suitable electronic transport properties make this Bi-based coordination framework, especially with 2D-nanosheet morphology, a desirable platform for comprehensive mechanism studies of electrocatalysis. Electrocatalytic CO2 reduction performance To assess the intrinsic catalytic properties of Bi(HHTP)-x wt% F127, we conducted electrochemical CO2 reduction measurements in 0.5 M KHCO3 electrolyte using a H-type cell (Supplementary Fig. 23). Linear sweep voltammetry (LSV) was employed to obtain polarization curves of electrodes in the N2 and CO2-saturated electrolyte with blank carbon fiber paper as a control (Supplementary Fig. 24). Compared to the cathodic polarization curves of samples in N2-saturated electrolyte, the sharp increase of current densities under CO2 saturation suggested the predominant occurrence of CO2 reduction over HER on catalysts (Fig. 3a). Bi(HHTP)-0.6 wt% and 1.8 wt% F127 nanosheets had higher current densities than Bi(HHTP)-0 wt% and 0.06 wt% F127 bulk materials at the same potentials. HCOOH/HCOO– was examined as the only liquid product from the eCO2RR (Supplementary Fig. 25) with a limited number of gaseous products (i.e., CO and H2) and the corresponding Faradaic efficiencies (FEs) were concluded in Fig. 3b and Supplementary Fig. 26. In this work, we referred to the product as “formic acid” to highlight the transient species involved in the reaction mechanism, while acknowledging that the final stable form under the reaction conditions was formate. The Bi(HHTP)-1.8 wt% F127 nanosheets exhibited high FEHCOOH over 80% from –0.82 to –1.42 V vs. RHE and the highest FEHCOOH of 99% at –1.12 V vs. RHE and also showed a better selectivity to HCOOH than the other materials (Fig. 3b, c). Moreover, the minimum contact angle with water of Bi(HHTP)-0 wt% F127 on carbon fiber paper substrate (Supplementary Fig. 27) implied advantageous hydrophilicity over other materials, favoring the competitive HER process during electroreduction. Bi(HHTP)-1.8 wt% F127 exhibited the largest partial current densities toward HCOOH (jHCOOH) over the whole applied potential range, and the highest jHCOOH of 47.8 mA cm−2 (Supplementary Fig. 28) and formic acid production rate of 0.25 μmol s–1 cm–2 (Fig. 3d) were achieved at –1.32 V vs. RHE. Both the selectivity and activity of Bi(HHTP)-1.8 wt% F127 are comparable to the reported performance of CO2RR-to-HCOOH catalysts, including MOF-derivatives, heterogeneous metals, metal compounds, and single-atom catalysts (Supplementary Table 3)30,49–51. Electrochemical performance of Bi(HHTP) for CO2 reduction. a LSV curves of Bi(HHTP) samples in N2 or CO2-saturated electrolyte (scan rate: 10 mV/s) after 85% iRs corrections (Rs = 9.57, 8.98, 9.87, and 10.30 Ω for Bi(HHTP)-0, 0.06, 0.6, 1.8 wt% F127 in N2-saturated electrolyte, respectively; Rs = 9.53, 9.44, 9.32, and 8.48 Ω for Bi(HHTP)-0, 0.06, 0.6, 1.8 wt% F127 in CO2-saturated electrolyte, respectively). b FEs toward HCOOH of Bi(HHTP)-1.8 wt% F127 nanosheets at various potentials in CO2-saturated 0.5 M KHCO3 electrolyte. c FEs of HCOOH and H2 at –1.32 V and –1.42 V versus RHE for Bi(HHTP)-0 wt% F127, Bi(HHTP)-0.06 wt% F127, Bi(HHTP)-0.6 wt% F127, and Bi(HHTP)-1.8 wt% F127. d The production rate of HCOOH at different potentials. e ECSA-corrected partial current densities toward formic acid generation at various potentials. f Cyclic stability of Bi(HHTP)-1.8 wt% F127 during prolonged-time electrolysis at –1.12 V vs. RHE. On-the-fly 85% iRs corrections were applied for potentials during the chronoamperometry measurements (Rs = 9.73 ± 0.80 Ω). The gas flow rate is 30 cc/min. The mass loading is ~0.3 mg/cm2. All electrochemical measurements were conducted on the materials after the successful removal of F127 and performed at room temperature (298 ± 2 K). Error bars were collected from at least three independent experiments. Source data are provided as a Source Data file. We further calculated the turnover frequency (TOF) based on the electrochemical surface area (ECSA) method as a critical parameter for CO2RR efficiency assessment (Supplementary Figs. 29, 30). Because the stacking of 2D nanosheets on substrate blocked the exposure of active sites (Supplementary Fig. 31), Bi(HHTP)-1.8 wt% F127 showed the lowest ECSA. The ECSA-normalized jHCOOH indicated the favorable catalytic activity for CO2 reduction of active sites on Bi(HHTP) {001} facets (Fig. 3e). The maximum TOF value of Bi(HHTP)-1.8 wt% F127 was 21.1 s−1 at −1.32 V vs. RHE, comparable to single-atom catalysts (Supplementary Table 4). Electrochemical impedance spectroscopy (EIS) analysis (Supplementary Fig. 32) demonstrated the lowest charge transfer resistance of Bi(HHTP)-1.8 wt% F127 in good agreement with the trend of electroconductivity from four-point probe measurements. Furthermore, we checked the structural stability of as-prepared Bi(HHTP) materials in 0.5 M KHCO3 solution (Supplementary Figs. 33, 34) and after long-term electrolysis at an applied bias of –1.12 V vs. RHE (Supplementary Figs. 31, 35–40). No obvious structural and morphological changes of Bi(HHTP) materials were observed. Chronoamperometry of Bi(HHTP)-1.8 wt% F127 manifested stable reduction current densities of 34 mA cm–2 and steady product selectivity of CO2RR toward HCOOH over 92% in 8 h, suggesting high stability of Bi(HHTP) in neutral media for eCO2RR (Fig. 4f). H transfer in the Bi(HHTP) framework. a Kinetic effect (KE) of H/D and KHCO3/K2SO4. b Proton conductivity (σ) of Bi(HHTP)-1.8 wt% F127 with increasing temperature at 98% relative humidity. c 1H NMR chemical shifts for liquid products after electrolysis for 2 h at –1.12 V vs. RHE by Bi(HHTP)-D2O and Bi(HHTP)-H2O in 0.5 M KCl-D2O solution. d–f, ToF-SIMS depth profiles and g–i, corresponding schematic illustrations depicting H/D distribution and transfer in Bi(HHTP)-H2O: pristine (d, g), after immersion in 0.5 M KCl-D2O solution for 2 h (e, h), and after 2-hour electrolysis in CO2-saturated 0.5 M KCl-D2O solution at –1.12 V vs. RHE (f, i). ID/(ID + IH) and ID/IC represent the normalized D+ content and the relative ratio of ion counts of D+ to C+, respectively. In schematic illustrations, black frames indicate the boundaries of the solid materials, and cyan blue and pink spheres denote H and D atoms, respectively. Source data are provided as a Source Data file. To overcome the mass diffusion limitation of CO2 resulting from low CO2 solubility in water, a gas diffusion electrode (GDE) employed in a flow cell was used to evaluate the catalytic performance of Bi(HHTP)-1.8 wt% F127 toward CO2RR (Supplementary Fig. 41). The current densities in flow cell were much larger than those in H-type cell. The Bi(HHTP) electrode still exhibited high FEHCOOH of over 96% in a wide potential window from −0.82 to −1.22 V vs. RHE in 0.5 M KHCO3 and 0.5 M KHCO3 + 2.5 M KCl. Remarkably, the specific current density of formic acid in 0.5 M KHCO3 + 2.5 M KCl reached a high density of ~305 mA cm−2 at a polarized potential of −1.22 V, which was 6.9 times larger than in the H-type cell (44.2 mA cm−2). This value exceeds the suggested criterion of 210 mA cm−2 for promising industrialization52. Isotopic kinetic analysis and H transfer studies Tafel analysis was performed to reveal the rate-limiting step and the corresponding reaction kinetics (Supplementary Figs. 42, 43)53. 2D nanosheets with exposed {001} facets showed broader horizontal region in the Tafel slope plot and had smaller Tafel slopes with values of 179.4, 172.4, 153.6, and 159.7 mV dec–1 for Bi(HHTP)-0, -0.06, -0.6, and -1.8 wt% F127, respectively. Thus, 2D nanosheets possessed faster reaction kinetics than 3D bulks. In the empirical theory of a two–electron transfer electrochemical CO2RR, the rate-limiting step consists of the first electron transfer with Tafel slope about 118 mV/dec, the second electron transfer with Tafel slope about 59 mV/dec, and desorption of products with Tafel slope about 30 mV/dec54. The values of Tafel slopes for Bi(HHTP) frameworks were higher than 118 mV/dec, indicating that the first-step electron transfer was likely to be involved in the rate-limiting step for electrochemical CO2RR. To determine the effect of proton transfer and supply, kinetic effect (KE) studies of isotopic proton/deuterium (H/D) and anions were employed (Fig. 4a and Supplementary Fig. 44). The measured KE values for H2O/D2O and KHCO3/K2SO4 were 1.459 and 1.638 (larger than 1), respectively, indicating that the first proton-coupled electron transfer was involved in the rate-limiting step and that H2O and HCO3− both provided proton supply for CO2RR. With the proton supply involved in the rate-limiting step, hydrophilicity and proton conductivity related to proton transfer were analyzed. The as-synthesized Bi(HHTP) materials showed hydrophilic properties with contact angles on the surfaces less than 90° (Supplementary Fig. 45) as well as good proton conductivity with activation energy (Ea) of 0.14 eV based on Grotthuss mechanism55 (Fig. 4b and Supplementary Figs. 46–48). These results provided substantial support for fast proton transfer for CO2 reduction, facilitating kinetics and contributing to great catalytic activity. Notably, besides H2O and HCO3– from electrolytes, there are coordinated water molecules in Bi(HHTP)-β crystal structures. To track the proton inventory in the crystal structure, we synthesized Bi(HHTP)-D2O materials in D2O solvent, where Bi sites coordinated with D2O instead of H2O (Supplementary Figs. 49, 50). When Bi(HHTP)-D2O was applied for eCO2RR in KHCO3-H2O solution, a peak at m/z ~ 46 appeared in the liquid chromatography-mass spectra (LC-MS), which can be assigned to DCOO− (Supplementary Fig. 51). To avoid the effects of H sources from electrolyte, we substituted the KHCO3-H2O with KCl-D2O solution. The proton nuclear magnetic resonance spectra (1H-NMR) of liquid products of CO2 reduction catalyzed by Bi(HHTP)-D2O indicated the absence of any H-containing products. On the contrary, a non-negligible H peak of HCOO– was observed from the 1H-NMR data (Fig. 4c and Supplementary Fig. 52) for the Bi(HHTP)-H2O catalyzing CO2 reduction in KCl-D2O electrolyte where the coordinated H2O in the Bi(HHTP)-β structure was the only H source in the system. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) was further applied to reveal the H transfer in materials (Fig. 4d–i). Pristine Bi(HHTP)-H2O exhibited a low initial concentration of D across the materials layer. No obvious changes in D concentration and distribution over depth were observed after soaking the materials in KCl-D2O solution for 2 h through solvent exchange, indicating minimal D exchange through simple immersion. However, following eCO2RR, a significantly elevated incorporation of D was detected in the shallow area of the Bi(HHTP)-H2O materials layer, gradually diminishing with depth. The accelerated H/D exchange rate indicated a decomposition-reconstruction process of coordinated H2O could be involved. Taken together, the above observations suggested that during the CO2 reduction process, the H of coordinated H2O in the Bi(HHTP) crystal structure was transferred into the HCOOH product (Supplementary Table 5), and the H transport in the framework was substantially enhanced. In-situ spectroscopic investigations In-situ attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) studies were carried out to monitor the reaction intermediates (Supplementary Fig. 53). In both CO2 or N2-saturated electrolytes, we observed an increase of the distinct vibrational peak at ~1660 cm–1 and broad adsorption peak (3000–3700 cm–1) related to the adsorption and accumulation of interfacial H2O on the surface (Fig. 4a top and Supplementary Fig. 54). This behavior can be attributed to the intrinsic hydrophilicity of catalyst and the formation of hydrogen bond network (HBN), consistent with similar observations for Bi(HHTP)-0 wt% F127 (Supplementary Fig. 55)56. However, a distinct red shift of the broad peak from 3490 to 3300 cm–1 was only observed in a CO2-saturated environment, indicating changes in O−H stretching modes and HBN (Fig. 4b, c). In the presence of CO2, the peak at ~3300 cm−1 increased significantly, corresponding to a strong HBN comprising water molecules with single hydrogen donors57. The new broad peak centered at ~3050 cm−1 ranging from 3600 to 2600 cm−1 appeared upon the electrochemical process, which could be the H-bonded O−H stretching mode in −COOH functional groups58. The switch in the strength of HBN suggested enhanced reorganization of interfacial water to facilitate proton transfer, and the occurrence of a new O−H stretching mode suggested the formation of carboxylic-acid-containing species. The continuous consumption of CO2 and neglectable linear-bound CO were also observed (Supplementary Fig. 56)59. Starting from –0.5 V vs. RHE, the occurrence of two peaks centered at ~1625 and ~1360 cm–1 matched well with characteristic peaks of asymmetric and symmetric C=O stretching modes in HCO3− (Supplementary Fig. 57). The peak at 1498 cm–1 corresponded to the asymmetric C(OH)2 stretching mode in H2CO3, respectively21,60–62. These results indicated that Bi(HHTP) had relatively strong adsorption with H2CO3. During the electrochemical process, the adsorbed H2CO3 was consumed either as reactants for CO2 electroreduction or converted into HCO3– due to the increase of local pH value (CO2 + 2H+ + 2e– → HCOOH) shifting the equilibrium between H2CO3 and HCO3– (H2CO3 ⇌ HCO3– + H+). In addition, the appearance of tiny peak at 1325 cm–1 might be related to trace amounts of other reaction intermediates, such as the symmetric O–C–O stretching of *OCHO. It should be noted that the H2O bending mode initially observed at ~1660 cm−1 could red-shift with the strengthening of HBN, potentially overlapping with the asymmetric C=O stretch in HCO3−. Therefore, to exclude the interference, in-situ ATR-SEIRAS measurement was carried out in a configuration of KHCO3 with D2O (Fig. 4a bottom). The stretching and bending absorbance peaks of D2O shifted to 2565 and 1199 cm−1, respectively. The red shift of the O−D stretching peak to 2435 cm−1 was observed from −1.1 V vs. RHE, consistent with changes in O−D stretching modes and HBN (Supplementary Fig. 58). Peaks of HCO3− and CO32− were retained at the same wavenumbers, but the decreasing absorbance peak at ~1498 cm−1 shifted to ~1457 cm−1 due to the substitution of H with D in H2CO3. In a 13C-labeling experiment to further confirm the presence of carbon-containing species (Supplementary Fig. 59), peaks at ~1585 and 1330 cm−1 assigned to H13CO3− showed red shifts from ~1625 and 1360 cm−1. For the asymmetric C(OH)2 stretching mode of H2CO3, the isotopic effect on vibrational modes of 12C/13C was less than H/D, so only a slight shift of H213CO3 from ~1498 cm−1 to ~1487 cm−1 was observed. Notably, a residual band at ~1625 cm−1 persisted even under 13C labeling, confirming that the broad 1625 cm−1 feature in CO2-saturated KHCO3 originated from both the HBN-induced red-shifted H2O bending mode and the asymmetric C=O stretch of HCO3−. Our investigation of pH effects on the electrocatalytic performance of Bi(HHTP) also revealed that the CO2 electroreduction rate relied on the concentrations of H2CO3 and HCO3− (Supplementary Fig. 60). These phenomena evidenced the consumption of H2CO3 and variation of the hydrogen bond network during the CO2 electroreduction process. To further elucidate the structural variation of Bi(HHTP) and reaction mechanism, in-situ electrochemical Raman measurements were conducted (Supplementary Figs. 61–62). In the range of Raman shift from 25 to 800 cm−1 (Fig. 4d), five bands centered at ~57, 100, 324, 632, and 683 cm−1 were observed, corresponding to the presence of Bi-specie-involved phonon vibrations63,64. Upon eCO2RR, four of five bands at ~57, 324, 632, and 683 cm−1 were blue-shifted to ~67, 344, 640, and 715 cm−1, respectively, except for a red shift of the band from ~100 cm−1 to ~93 cm−1 in Bi(HHTP) structure during the electrochemical process. The shifts were likely caused by the elasticity of Bi−O bonds and changes in lattice structure64,65. Such behaviors were also observed in the 13C-labeling experiment (Supplementary Fig. 63). Generally, lattice contraction or shortened chemical bonds lead to a blue shift in the Raman peak position, while lattice expansion or elongated bonds result in red shifting behavior65. The shifts in both directions for Bi(HHTP) could be due to the formation of intermediates for CO2RR, which induced the strengthening of most Bi−O bonds but weakened a small portion of them. After the electrochemical process, these bands recovered close to the initial Raman shifts of Bi(HHTP) at OCV, suggesting good structural integrity of Bi(HHTP). Determination of CO2 adsorption sites in crystal The ordered microporosity of MOF crystals enables the well-anchoring of small guest molecules at specific molecular-recognition sites and their orderly arrangement within the crystal, thereby facilitating the clear determination of small molecules and their adsorption sites through crystallographic tools. To gain molecular-level insights into the CO2 adsorption sites in Bi(HHTP) crystal and the effects of coordinated H2O on CO2 electroreduction, MicroED analysis was employed to solve the crystal structure of Bi(HHTP) after CO2 adsorption and probe the atom locations precisely (Fig. 5e-g, Supplementary Figs. 64−65 and Supplementary Table 6). The CO2-loaded Bi(HHTP) retained high crystallinity and produced clear electron diffraction patterns suitable for structural analysis (Fig. 5g). Through continuous rotation and tilting, a 3D reciprocal lattice was reconstructed, allowing structural refinement with a high resolution of up to 0.8 Å (Supplementary Figs. 64−65). The solved crystal structure of CO2-loaded Bi(HHTP) revealed the CO2 molecule was confined by the coordinated H2O rather than the Bi atom. The distance of 2.623 Å between the C of the adsorbed CO2 and the O from coordinated H2O was much smaller than the shortest distance of 4.325 Å between Bi and the C of CO2. It was noted that the CO2 molecule restricted by the coordinated H2O in the channel was slightly bent with an 158.25° angle of O=C=O (Fig. 5f). The geometry of CO2 changing from linear to bent increased the electrophilic nature of C atom of CO2. The observations suggested that the coordinated H2O on Bi sites promoted both the adsorption and activation of CO2 molecules. Experimental spectroscopic investigations of CO2 reduction mechanism catalyzed by Bi(HHTP) and CO2 adsorption site detection. a In-situ ATR-SEIRAS spectra of Bi(HHTP)-1.8 wt% F127 nanosheets in CO2-saturated 0.5 M KHCO3-H2O (top) and KHCO3-D2O (bottom) during the electrochemical process (OCV: open circuit voltage). Gaussian fits of different O–H stretching modes from 4000−2400 cm−1 for in-situ ATR-SEIRAS spectra of Bi(HHTP)-1.8 wt% F127 nanosheets in N2-saturated K2SO4-H2O solution (b) and CO2-saturated KHCO3-H2O solution (c) at different potentials, respectively. d In-situ Raman spectra of Bi(HHTP)-1.8 wt% F127 nanosheets in CO2-saturated 0.5 M KHCO3-H2O. e The single crystal structure of CO2-loaded Bi(HHTP). f Detailed CO2 adsorption site experimentally detected by MicroED analysis. Cyan blue, red, light grey, and white colors indicate Bi, O, C, and H atoms, respectively. g Electron diffraction pattern of CO2-loaded Bi(HHTP) with a maximum resolution of 0.8 Å. Source data are provided as a Source Data file. Crystal structure in Fig. 5e−f is provided in the Supplementary materials. Density Functional Theory (DFT) calculations DFT calculations were performed (Supplementary Tables 7–10) to better understand the CO2 reduction process on Bi(HHTP) surfaces. Based on the crystal structures of Bi(HHTP) and facets revealed by MicroED analysis and pXRD, four different types of active sites were identified and constructed: Bi–O2, H2O–Bi–O2, Bi–O4, and H2O–Bi–O4 sites (Fig. 6a and Supplementary Figs. 66–68). The adsorption energies of CO2 and H on each active site were calculated (Supplementary Fig. 69 and Supplementary Tables 8–10). The examined active sites exhibited stronger adsorption of CO2 over H, but Bi–O4 and H2O–Bi–O4 presented more significant adsorption energy differences than those of Bi–O2 and H2O–Bi–O2 (Fig. 6b). This suggested decrease in coordination number had a more pronounced effect on reducing energy barriers for HER in comparison with CO2RR, which was further evidenced by the projected density of states (PDOS) analysis (Supplementary Figs. 70−72). Highly dense Bi–O4 and H2O–Bi–O4 sites on {001} facets enabled Bi(HHTP) nanosheets with preferential faceting to possess better selectivity for CO2RR than 3D bulk material. More importantly, the coordinated H2O facilitated the physical adsorption of CO2 molecules on the active sites, with more negative values for adsorption energy on H2O–Bi–O2 and H2O–Bi–O4 over their dehydrated forms. Reaction path for CO2 electroreduction and proton transfer. a Structures of the identified active sites of Bi–O2, H2O–Bi–O2, Bi–O4, H2O–Bi–O4 from Bi(HHTP)-α and Bi(HHTP)-β crystals. b Calculated H and CO2 adsorption energy on different active sites. Gibbs free energy of various reaction pathways for CO2 reduction occurring at the Bi–O4 site without H2O molecules (c) and H2O–Bi–O4 site (d), respectively. e Schematic illustration of proposed reaction pathway of CO2 reduction toward HCOOH through protonated carbonic acid on H2O–Bi–O4 site. The dashed line means close contact or hydrogen bond between atoms. Yellow spheres represent H in the coordinated H2O and demonstrate how the H in coordinated water is transferred into the formic acid product. Other H in the ligand and from the electrolyte is indicated by white color. Black arrows indicate the electron transfer. Source data are provided as a Source Data file. The Gibbs free energies of two-electron transfer eCO2RR on Bi–O4 (Fig. 6c and Supplementary Table 9) and H2O–Bi–O4 sites (Fig. 6d and Supplementary Table 10) and the Gibbs free energies of competitive HER (Supplementary Fig. 73) were calculated and related conformations of intermediates were shown in Supplementary Figs. 74, 75. In general, two distinct reaction pathways of two-electron CO2RR exist: carboxylic acid pathway (*COOH) and formate pathway (*OCHO). For Bi–O4 sites, the calculated energy barriers for the formation of *OCHO and desorption of HCOOH were lower than the formation of *COOH, favoring CO2RR-to-HCOOH conversion (Fig. 6c). However, for H2O–Bi–O4 sites, when the Bi atom was coordinated with an H2O molecule, the energy barriers for adding the first proton-electron pair to form “*H2O + *COOH” and “*H2O + *OCHO” were larger than that for HER (Fig. 6d), insufficient to support the high CO2RR selectivity and explain the H transfer from coordinated H2O into HCOOH. Notably, in a detailed conformation of CO2 adsorbed on H2O–Bi–O4 (Supplementary Fig. 76), H2O was stabilized by close contact with Bi and hydrogen bonding with adjunct O site. The distance between C atom in CO2 and O atom in H2O was less than that between O atom in CO2 and Bi atom, consistent with the experimentally observed CO2 adsorption site. Adding a proton-electron pair to the O atom of CO2 induced the conformational transition of CO2 from linear to a bending geometry, pushing the C atom to the coordinated H2O. Eventually, an adsorbed H2CO3 was formed with one proton from the coordinated H2O transferred to the adjacent O site (Supplementary Fig. 77), displaying a structural rearrangement. The formed H2CO3 was stabilized by the Bi atom and the newly formed OH site. We denoted this intermediate form as “*H+ *H2CO3” and this pathway as protonated carbonic acid pathway. Upon the adsorption of CO2, there was a decreased covalent character based on the difference in Mulliken charge of the H2O molecule (Supplementary Fig. 78), which might drive the dissociation of H2O and the formation of “*H + *H2CO3”. Subsequently, the adsorbed H2CO3 tends to dissociate, and the H adsorbed on the O site would transfer to the C atom in the form of surface hydride to give out one HCOOH molecule and H2O molecule. The geometrical configuration of the intermolecular six-member ring allows H− to attack the C atom. According to DFT calculations, the energy barrier for “*H + *H2CO3” (0.36 eV) was much lower than those for “*H2O + *COOH” and “*H2O + *OCHO”. The limiting potential difference between the CO2RR and HER switched from –0.19 V for formate pathway and –0.66 V for carboxylic acid pathway to a much more positive value of 0.76 V for protonated carbonic acid pathway (Supplementary Fig. 79), supporting CO2RR surpassing HER on H2O–Bi–O466. This protonation pathway in our proposed mechanism (Fig. 6d, e) also explained the experimental observation of H atom originating from coordinated H2O in HCOOH products. The protonated carbonic acid pathway led to contractions of partial Bi−O distances (Supplementary Figs. 80, 81), providing a possible explanation for the blueshift of Raman peaks during the electrochemical process. Given the critical role H2CO3 played in the Bi(HHTP) catalyzing CO2RR, additional DFT calculations focusing on H2CO3 were conducted. Bi–O4 active sites showed strong adsorption of H2CO3 (Supplementary Figs. 82, 83 and Supplementary Table 11). “*H + *H2CO3” with various conformations of H2CO3 all depicted lower energies than those of “*H2O + *OCHO” and “*H2O + *COOH”, mediating the reaction pathway of eCO2RR. Additionally, the eCO2RR was more favorable than self-reduction in the second step, supporting the structural stability of Bi(HHTP) frameworks (Supplementary Fig. 84). In terms of the Bi–O4 site, the protonated carbonic acid pathway would possibly be stimulated by the co-adsorption of CO2 and H2O or the adsorption of H2CO3 (Supplementary Fig. 85). This proposed mechanism would not be limited to the Bi-based coordination frameworks reported in this study. It may have potential implications on broader catalytic systems (Supplementary Fig. 86). Specifically, this work provides insights into an explanation of enhanced electrocatalytic performance for CO2 reduction on metal oxides and metal hydroxides compared with metallic catalysts67 and it unravels the significance of the coordination effect of H2O for CO2 reduction. Discussion In summary, we synthesized 2D nanosheets of Bi-based coordination framework via a chemical interface confinement strategy, which provided a definitive crystal structure with well-defined morphology for in-depth mechanistic studies of electrochemical CO2 reduction. In an H-type cell, the prepared nanosheet-like Bi(HHTP) material synthesized in the presence of 1.8 wt% F127 polymer exhibited high selectivity of CO2 reduction toward HCOOH in a wide potential window (−0.82 to −1.42 V vs. RHE), and the near unity HCOOH selectivity (FE~99%) was achieved at −1.12 V vs. RHE. The partial current density for producing HCOOH could reach up to 47.8 mA cm–2 at −1.32 V vs. RHE. Furthermore, compared with 3D bulk material analogs, the Bi(HHTP) nanosheets with preferential exposure of {001} facets displayed much-enhanced mass activity of Bi (408.5 A g−1) and large value of TOFs (21.1 s−1). Crystalline organic frameworks with long-range ordered microporosity offered a well-defined platform for gaining fundamental insights into how H2O and CO2 coordinate to active sites and influence reduction pathways. Employing in-depth H/D isotope labeling studies, MicroED analysis, in-situ spectroscopies and DFT calculations, we unraveled the crucial roles of coordinated H2O in CO2 electroreduction: i) the coordinated H2O was beneficial to CO2 adsorption and activation, and ii) the coordination of H2O combined with neighboring electronegative atoms on active site induced a protonated carbonic acid reaction pathway for CO2 electroreduction to formic acid, allowing proton transfer from coordinated H2O into product. This work highlights the applicability of MOFs and CPs for electrocatalysis with the advantages of ordered microporosity, atomically precise sites, and structural tunability. In a broader catalytic context, the coordination effect of H2O and the proposed protonated carbonic acid reaction pathway could be applicable to other catalytic systems, opening a method to optimize catalyst design. Methods Preparation of Bi(HHTP)-x wt% F127 100 mg (0.308 mmol) HHTP was suspended into 100 mL of H2O. Pluronic F127 powder with different masses was added to the suspension, and it was sonicated until all the white Pluronic F127 powder dissolved. 119 mg (0.308 mmol) bismuth acetate (Bi(OAc)3) was then added, and the suspension was sonicated for 20 min. The mixture was heated at 85 °C for 24 h in open Air with stirring. After cooling to room temperature, the dark green solid was collected by centrifugation. The solid was suspended in 100 mL of deionized H2O, heated with stirring at 50 °C for 12 h, and then collected with centrifugation. Repeat the above wash step three times to remove F127 polymer. The suspension was then centrifuged, and the solid was further washed and activated with deionized H2O, ethanol, isopropanol, and ethyl acetate. The dark green or dark blue solid was dried overnight in a vacuum oven at 122 °F. The mass content of F127 varied from 0.06 wt% to 10 wt%. At high concentrations of F127 (larger than 1.8 wt%), the purification to remove F127 became more difficult, and the effects of morphological tunning via F127 were restricted. General procedures for materials characterization Powder X-ray diffraction (pXRD) was performed with a Rigaku sixth-generation MiniFlex X-ray diffractometer with a 600 W (40 kV, 15 mA) Cu-Kα radiation (λ = 1.54 Å) radiation source. NMR spectra were recorded on a Bruker 500 MHz NMR spectrometer. For organic ligands, the chemical shifts (δ) were expressed in ppm with the internal standard tetramethylsilane (TMS) and solvent signals as internal references, and J values were given in Hz. Scanning Electron Microscopy (SEM) and Energy Dispersive X-ray Spectroscopy (EDS) were performed using a Thermo Scientific™ Helios™ 5 CX DualBeam. Transmission electron microscopy (TEM) was carried out at an FEI Talos™ f200i S/TEM instrument under 200 kV beam voltage. Nitrogen adsorption/desorption measurements at 77 K were performed on a Micromeritics 3Flex to obtain pore properties such as the specific surface area, total pore volume, and pore size distribution. The sample was degassed at 100 °C for 48 hrs. Thermal gravimetric analysis (TGA) was performed using a TA Instruments TGA Q150 with a 10 °C/min ramp from room temperature to 600 °C under N2 or Air. Attenuated total reflectance-infrared spectroscopy (ATR-IR) measurements for powder materials were performed on a Nicolet iS50 FT-IR spectrometer with deuterated doped triglycine sulfate (DTGS) detector. Liquid chromatography-mass spectrometry (LC-MS) was performed via the Shimadzu LCMS-8030 Triple Quadrupole system. X-ray photoelectron spectroscopy (XPS) experiments were conducted on Kratos Analytical AXIS Supra X-ray Photoelectron Spectrometer under ultrahigh vacuum (base pressure 1×10−7 Torr). The measurement chamber was equipped with a monochromatic Al (Kα) X-ray source. A beam diameter of 200 μm was used for wide-scan survey and high-resolution spectra. Atomic force microscopy (AFM) imaging was performed using an AIST-NT OmegaScope 1000 Scanning Probe Microscope in non-contact mode. Tips were either uncoated silicon or Al-coated silicon with tip radii <20 nm. Contact angles of a deionized H2O droplet on the materials of pellet and the material-coated carbon fiber paper were measured using an Ossila Contact Angle Goniometer. The bulk electrical conductivity of the pellet was measured using a four-point probe (Pro4 Resistivity Test System equipped with a high resistivity probe head, Signatone) with a uniform 1.5875 mm space between the probes. Details are shown in Supplementary Information. Proton conductivity measurements of Bi(HHTP) materials are shown in the Supplementary Information. Electrochemical measurements The electrochemical tests were first performed on a Bio-Logic model VMP3 potentiostat at ambient temperature ( ~ 298 ± 2 K) using a three-electrode setup in a H-type cell separated by proton exchange membrane (117 µm-thickness Nafion 117). See Supplementary Information for the pre-treatment of the proton exchange membrane. Each compartment contains 40 mL of electrolyte and 10 mL of gas headspace. 0.5 M KHCO3 was used as electrolytes in both cells and the pH of the KHCO3 solution was measured after gently bubbling with N2 (pH = 8.4 ± 0.1) and CO2 (pH = 7.3 ± 0.1) for more than 30 min, respectively. A saturated calomel electrode (SCE, CHI150, CH Instruments, Inc., 44 mm × 9.0 mm + 54 mm × 5.3 mm) was used as the reference electrode. The reference electrodes were calibrated in 0.5 M H2SO4 under H2-saturated conditions using two Pt electrodes. The potentials (E) were measured against SCE with an 85% iRs compensation and were converted to Reversible Hydrogen Electrode potential (RHE) by using the equation: The ohmic resistance (Rs) for LSV measurements was determined through electrochemical impedance spectroscopy (EIS) measurements, which were conducted at open-circuit voltage across a frequency range spanning from 0.1 MHz to 10 mHz. For chronoamperometry measurements, the on-the-fly correction was implemented, where the points were corrected by instruments with built-in iRs compensation (Rs = 9.73 ± 0.80 Ω). Prior to each electrochemical measurement, the electrolyte was purged with N2 or CO2 for 30 min to form N2/CO2-saturated solution, respectively. Active materials drop-cast on carbon fiber paper (CFP) was used as working electrode. 3 mg Bi(HHTP) and 30 μL of a Nafion solution (5 wt%, Sigma-Aldrich) were added into 1 mL ethanol to form a slurry. The slurry was sonicated for 1 hour to form a homogenous ink. 100 μL slurry was pipette onto the microporous layer of 1 cm2 (1 cm × 1 cm) carbon fiber paper (SGL 28 BC with Vulcan-30% FEPD 121 Filler, Fuel Cell Store) and the back of CFP substrates was covered by 1 cm2 hydrophobic Teflon tape to avoid the contact of CFP with aqueous electrolyte. After drying at room temperature in air overnight, the Bi(HHTP)@CFP working electrode was formed. The mass loading of Bi(HHTP) was ~0.3 mg cm−2. Gas products of CO2 electrocatalysis were characterized via gas chromatograph (GC, Shimadzu GC-2014) equipped with molecular sieve 5 Å, Hayesep Q, Hayesep T, and Hayesep N columns. Argon was used as the carrier gas. A flame ionization detector (FID) with methanizer was used to quantify the concentration of carbon monoxide (CO). Hydrogen (H2) product was analyzed with a thermal conductivity detector (TCD). A standard gas with known concentration of know gases was used to calibrate GC through external standard calibration method. The gas flow from the electrolytic cell was injected into GC for 1 min with a constant flow rate. After a run of GC, the peak areas of different gas products were converted to volumetric concentration in gas based on calibration curves. Then the partial current density for a given gas product was calculated as below: where xi is the volumetric concentration of gas product based on the calibration curve of GC with standard gas, f is CO2 flow rate controlled and measured using a mass flow controller (30 cc/min), p0 is the standard pressure (1 atm), ni is the number of electrons transferred for the generation of one product molecule (e.g., formic acid/formate), F is the Faradaic constant (96,485 C mol−1), R is the gas constant (8.3145 C mol−1 K−1), T is the temperature (298 K), and S is the geometric surface area of the electrode. The corresponding Faradaic efficiency (FE) was calculated as: where the jproduct and jtotal are calculated partial current density for a given gas product and current density in total, respectively. After 30 min or 1 h electrochemical process, the electrolyte in which the working electrode was immersed was collected to quantitatively analyze the amounts of liquid products by nuclear magnetic resonance spectroscopy (NMR, Bruker 500 spectrometer). NMR samples were prepared by mixing a discrete mass of dimethyl sulfoxide (DMSO, CAS # 67-68-5, Sigma-Aldrich) as an internal standard into the collected electrolyte and then mixing 450 μL of the solution containing DMSO with 50 μL D2O. The concentration of HCOOH/HCOO− was determined by the integrated ratio of HCOOH/HCOO− to DMSO. The FE of products can be calculated via the following equation: Where n is the number of moles for a given product. The amount of charge (Q) passed to produce products during a period of reaction time in total was obtained through the potentiostat. The calculation of mass activity, production rate, and TOF are shown in the Supplementary Information. Flow cell measurements were performed in a commercial flow cell reactor (Gaoss Union 101017-1.2). For working electrode preparation, the catalyst inks were drop-casted onto a hydrophobically treated carbon paper (SGL 28 BC with Vulcan-50% FEPD 121 Filler, Fuel Cell Store) with an effective area of 0.5 × 2 cm2. The mass loading of Bi(HHTP) was ~0.3 mg cm−2. An Ag/AgCl electrode (stored in saturated KCl), and a commercial Pt plate were used as the reference electrode and counter electrode, respectively. The potentials were measured against Ag/AgCl with an 85% iRs compensation and were converted to Reversible Hydrogen Electrode potential (RHE) by using the equation: A 280 μm-thickness bipolar membrane (Alfa Chemistry) was used to separate the anolyte compartment and the catholyte compartment. CO2-saturated 0.5 M KHCO3 (Rs = 2.66 ± 1.16 Ω) and a mixture of 0.5 M KHCO3 and 2.5 M KCl (Rs = 0.39 ± 0.22 Ω) were used as catholytes, while 0.5 M KOH and 3 M KOH were used as anolytes accordingly. During the test, the CO2 feed gas was introduced into the gas chamber with a constant flow rate of 30 sccm. The electrolytes in the cathode and anode chambers were circulated by using a peristaltic pump and a gas-liquid mixed flow pump, respectively. ¹H NMR identified the liquid products in the electrolyte. The electrolyte with high salt concentration was diluted prior to ¹H NMR measurements. ToF-SIMS measurements ToF-SIMS measurements were performed using a PHI TRIFT nanoToF system. Giant Cluster Ion Beam (GCIB) with 2000+ Ar+ ions operating at 10 kV and 25 nA ion current was used for sputtering on a 500×500 μm2 area, and a Bi+ primary ion gun was used for analysis. Similar procedures with the working electrode preparation were used to prepare Bi(HHTP)-H2O samples for ToF-SIMS experiments. After immersion and CO2RR in KCl-D2O solution, the samples were rinsed only with ethanol and fully dried. A reference sample was prepared by drop-casting the Bi(HHTP) sample slurry on a Si substrate with a thickness measured by Focused Ion Beam (FIB)-SEM to estimate the sputtering rate. The sputter rate was estimated as 1.6 nm/s. To provide a quantitative comparison among different samples, the normalized D+ content expressed as ion counts of D+ over total ion counts of H+ and D+ (ID/(ID + IH)) and the relative ratio of ion counts of D+ to ion counts of C+ (ID/IC) were used for analysis. In-situ ATR-SEIRAS The attenuated total reflectance-surface enhanced infrared absorption spectroscopy (ATR-SEIRAS) experiments were performed on a Nicolet iS50 FT-IR spectrometer equipped with a mercury cadmium telluride (MCT) detector cooled with liquid nitrogen and PIKE VeeMAX III variable angle ATR sampling accessory. The Au-coated Si prism was used as the conductive substrate for catalysts and the IR reflection element. Plasma sputter coating was used to physically deposit pure Au on a Si prism, and the thickness of Au-coating was controlled at ~30 nm. The catalyst suspensions were dropped on the Au/Si surface as the working electrode. 0.5 M KHCO3 and 0.5 M K2SO4 solutions were used as electrolytes. The electrochemical cell for in situ ATR-IR was connected with a Gamry potentiostat. Ag/AgCl in saturated KCl solution and Pt wire were used as the reference electrode and counter electrode, respectively. The conversion of applied potential to reversible hydrogen electrode (RHE) was based on: Before each measurement, N2 or CO2 gas was bubbled into the electrochemical cell for at least 30 min. Multiple-step chronoamperometry tests with a 200 s step time at each working electrode potential were set, and in-situ ATR-SEIRAS spectra were recorded during the stepping of the working electrode potential. In-situ Raman The in-situ electrochemical Raman experiments were carried out using a Horiba LabRAM HR spectrophotometer. A laser with 633 nm excitation wavelength and a long-working-distance 50× microscope objective with a numerical aperture of 0.55 was used in all measurements. Si peak at 520.5 cm−1 was used to calibrate the Raman spectra. In situ electrochemical Raman experiments were employed in a homemade Raman cell and a Gamry potentiostat was used to control the potential. Plasma Sputter Coating was used to deposit pure Au on CFP, and thick of the Au-coating physically was controlled as 20−30 nm. Bi(HHTP) drop-cast on Au-coated CFP, Ag/AgCl in saturated KCl solution, and Pt wire were used as the working electrode, reference electrode, and counter electrode, respectively. Raman spectra were recorded with a grating with 300 lines mm−1, 10% laser intensity, and 2 accumulations over an acquisition time of 20 s. MicroED analysis The collection of electron diffraction used the JEM-2100 Plus transmission electron microscope produced by Japan Electronics Corporation, which used an accelerating voltage of 200 kV and was equipped with a MerelinEM high-speed direct electron camera. QUANTIFOIL grids (R1.2/1.3) and Fischione 2550 Cryo Transfer Tomography Holder were used. The sample was prepared by placing an appropriate amount of the sample and copper mesh in a centrifuge tube and gently shaking to make the sample stick to the copper mesh. A CO2 gas atmosphere was maintained throughout the operation. The CO2-adsorbed sample on Cu mesh was loaded for MicroED data collection at 77 K. A total of 11 sets of electron diffraction data were collected, which could be used for diffraction data collection and interpretation of single nanocrystals. XDS was used for diffraction data restoration, and to improve completeness, six sets of data were combined to resolve the initial structure in SHELXT, refined with the assistance of SHELXL and OLEX2. DFT calculation DFT calculation methods are shown in the Supplementary Information. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-5 Transparent Peer Review file Source data Source Data Acknowledgements K.M. and W.L. acknowledge support from the National Science Foundation (NSF) EPSCoR Award (#1757371) and the Arthur L. Irving Institute for Energy and Society at Dartmouth. W.L. also acknowledges the partial support under Camille Dreyfus Teacher-Scholar Award (Award TC-22-008) from the Camille and Henry Dreyfus Foundation. We acknowledge the assistance of Paul Defino and Maxime J. Guinel from Dartmouth College (Hanover, NH) for materials characterization and help from John Wilderman from the University of New Hampshire (Durham, NH) for XPS characterization. We thank ReadCrystal Technology Co., Ltd for providing the MicroED analysis. Author contributions H.Q. performed most synthetic, characterization, and electrochemical measurement work. E.C. helped with the synthesis of materials. H.Q. and Z.M. did the computational analysis. B.L. helped with some electrochemical measurements and proton conductivity measurements. T.-D. L. assisted with the ToF-SIMS experiments and analysis. W.L. and K.M. supervised the project and revised the paper with all authors. Peer review Peer review information Nature Communications thanks Hao Bin Wu and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Data availability Full data supporting the findings of this study are available within the article and its Supplementary Information. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Katherine A. Mirica, Email: katherine.a.mirica@dartmouth.edu Weiyang Li, Email: weiyang.li@dartmouth.edu. Supplementary information The online version contains supplementary material available at 10.1038/s41467-025-66145-5. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-5 Transparent Peer Review file Source Data Data Availability Statement Full data supporting the findings of this study are available within the article and its Supplementary Information. Source data are provided with this paper.