Electrochemical CO₂ reduction (eCO₂RR) has emerged as a promising carbon-neutral strategy that synergistically integrates renewable energy conversion with environmental remediation [1-3]. This technology's dual capability of mitigating greenhouse gas emissions while producing value-added chemicals positions it at the forefront of sustainable energy research [4-19]. Among various reduction products, formic acid (HCOOH) presents distinctive advantages as an ideal energy carrier, including high volumetric hydrogen storage capacity, compatibility with direct formic acid fuel cells, and superior transport stability compared to gaseous alternatives [20-22]. These merits have driven intensive research efforts toward developing high-performance electrocatalysts for selective CO₂-to-HCOOH conversion [23-30].

Bismuth-based catalysts have garnered particular attention due to their inherent selectivity toward HCOOH production, with recent advances highlighting the potential of bismuth-oxo clusters as molecularly precise analogs of bismuth oxide nanomaterials [26]. These well-defined clusters offer unique opportunities to establish structure-activity relationships through their atomically precise metal cores and tunable ligand environments. However, a critical challenge persists: most reported bismuth-oxo clusters exhibit limited electrochemical stability under operational conditions, particularly in neutral/alkaline electrolytes, due to ligand dissociation and structural reorganization [31-33].

Current stabilization strategies focus on macrocyclic ligand encapsulation, where rigid organic frameworks (e.g., calix[n]arenes, thiacalix^[4]arenes) form protective shells around cluster cores [34-38]. While this approach enhances structural integrity of the nanoclusters, practical implementation faces two key limitations: (1) Geometric mismatch between rigid macrocycles and cluster surfaces often creates unprotected interfacial regions, and (2) complete encapsulation tends to bury catalytically active sites, compromising their accessibility [39-47]. Recent attempts to address these issues through auxiliary ligand filling have achieved partial success, but introduce new trade-offs-excessive ligand loading may either obscure active sites or induce steric strain that destabilizes the cluster architecture [48,49].

To resolve these conflicting requirements, we propose a ligand engineering strategy that achieves both structural stabilization and catalytic site optimization. Through rational design of hybrid ligand systems combining macrocyclic TC4A with modulatory co-ligands, we developed two novel octanuclear bismuth-oxo clusters (Bi₈-DMF and Bi₈-Fc) with tailored surface architectures. The Bi₈-DMF variant features strategically exposed bismuth sites coordinated by labile solvent molecules, while Bi₈-Fc demonstrates complete ligand encapsulation through ferrocene carboxylate integration. Systematic electrochemical evaluation reveals that controlled ligand coverage rather than maximal protection dictates catalytic performance. Catalytic tests revealed that the Bi₈-DMF cluster, protected by TC4A alone, efficiently catalyzes the electroreduction of CO₂ to formic acid, exhibiting a Faradaic efficiency (FE) for HCOOH of over 90% across a broad potential window from −0.8 V to −1.6 V, with excellent durability for over 24 h. This performance significantly outperforms that of the Bi₈-Fc, which, despite being protected by dual ligands, achieves a maximum FE of only 60% for formic acid at a specific potential. Density functional theory (DFT) calculations suggest that the surface-exposed bismuth sites in the Bi₈-DMF cluster are more effective at stabilizing the *OCHO intermediate, thereby facilitating the reduction of CO₂ to formic acid.

The Bi₈-DMF cluster was synthesized through a one-step solvothermal approach. The reaction system comprised Bi(NO₃)₃, TC4A, and 2, 4-dihydroxybenzoic acid in a mixed CH₃OH/DMF solvent (4 mL total, v/v = 2:2), followed by thermal treatment at 80 ℃ for three days. Single-crystal X-ray diffraction (SCXRD) analysis revealed the molecular formula {H₄Bi₈O₄(TC4A)₅(DMF)₃}·2DMF (Figs. 1A and B), crystallizing in the trigonal system with space group P-1. The asymmetric unit contains eight distinct Bi atoms, two μ₄O atoms, two μ₃O atoms, five TC4A ligands, and three coordinated DMF molecules, forming an elongated rod-like architecture. The molecular structure features two distinct subunits: (1) A sandwich-type {Bi₃O₂@(TC4A)₂} subunit (Fig. 1C), where three Bi atoms form a triangular plane bridged by a μ₃O atom, asymmetrically encapsulated between two TC4A ligands; (2) A {Bi₅O₂@(TC4A)₃} subunit (Fig. 1D) containing five Bi atoms interconnected through both μ₃O and μ₄O bridges, creating a Bi₅O₂ core stabilized by three triangularly arranged TC4A ligands. These subunits are interconnected via an additional μ₄O bridge to form the final {Bi₈O₄@(TC4A)₅} framework. As illustrated in Fig. 1E, the Bi centers exhibit diverse coordination geometries categorized into five groups with coordination numbers ranging from 4 to 7. Notably, the terminal Bi atoms coordinate with one and two labile DMF molecules, respectively. Given the dynamic nature of these solvent interactions, these sites are proposed as potential catalytic centers for subsequent reactivity studies.

Structural analysis reveals the Bi₈-DMF cluster can be conceptually divided into two fundamental building units: {Bi₃O₂@(TC4A)₂} and {Bi₅O₂@(TC4A)₃}. Intriguingly, this structural division was corroborated through electrospray ionization mass spectrometry (ESI-MS) analysis. Positive-mode ESI-MS of Bi₈-DMF crystals dissolved in CH₂Cl₂ (Fig. 2) displayed characteristic fragmentation patterns. The spectrum prominently featured a + 2 charged species at m/z = 2643.98 (1a), corresponding to the intact {Bi₈O₂(TC4A)₅}²⁺ cluster (calcd. m/z = 2643.94). A weaker +1 ion at m/z = 5304.79 (1b) matched the protonated {HBi₈O₃(TC4A)₅}¹⁺ species (calcd. m/z = 5304.89). Crucially, two intense signals corresponding to the proposed subunits were observed: {Bi₅O₂(TC4A)₃}¹⁺ at m/z = 3227.72 (1c, calcd. 3227.53) and {HBi₃O(TC4A)₂}¹⁺ at m/z = 2076.34 (1d, calcd. 2076.36). This provides direct experimental evidence for the dual module assembly mechanism.

Notably, an additional +1 peak at m/z = 2318.36 suggested the formation of a {Bi₄O@(TC4A)₂}¹⁺ intermediate, likely generated through structural rearrangement of the {Bi₃O(TC4A)₂} unit. This hypothesis was experimentally validated through ligand substitution experiments. Recrystallization of Bi₈-DMF in DMF/CH₃OH with excess bromosalicylic acid (SalH₂) yielded Bi₄-Sal crystals (∞¹{Bi₄O(TC4A)₂(Sal)}), forming a one-dimensional polymeric structure (Fig. 3A). Four Bi sites are bridged by a μ₄O atom, forming a quadrilateral arrangement (Fig. 3B), and are situated between two TC4A ligands, creating a sandwich-type {Bi₄O(TC4A)₂} unit. This Bi₄ unit is further bridged by bromosalicylic acid ligands, forming a one-dimensional infinite polymeric structure.

Mass spectrometry analysis combined with structural evolution studies reveal critical stability differences between the two clusters. The Bi₈-DMF assembly, constructed exclusively with TC4A ligands, exhibits limited solution stability and undergoes structural dissociation, as evidenced by its transformation into the Bi₄-Sal derivative. This inherent instability can be effectively mitigated through strategic incorporation of auxiliary ligands - a principle demonstrated in the synthesis of Bi₈-Fc. While maintaining similar synthetic protocols to Bi₈-DMF, the Bi₈-Fc system introduces ferrocene carboxylate co-ligands, yielding the hybrid architecture {H₄Bi₈O₄(TC4A)₄(FcCOO)₂} (Figs. 4A and B). Crystallographic analysis delineates a Bi₈O₄ core comprising eight Bi³⁺ ions interconnected through two μ₄O and two μ₃O bridges (Fig. 4C). This metallo-oxo framework is enveloped by four TC4A ligands exhibiting distinct coordination patterns: two fully deprotonated TC4A units adopt the μ₄-κ(O)²: κ(O)²: κ(O)²: κ(O)² binding mode analogous to {Bi₄@TC4A}, while the remaining two TC4A ligands coordinate via a μ₃-κ(O)¹: κ(O)²: κ(O)¹ configuration with one non-coordinating oxygen per ligand. The structural integrity is further enhanced by two ferrocene carboxylate ligands positioned at apical sites, each employing μ₄-bridging modes through their carboxyl groups to anchor four bismuth centers simultaneously (Fig. 4D). This dual-ligand strategy achieves complete encapsulation of the Bi₈O₄ core within a hybrid organic shell. Notably, Bi₈-Fc represents the first documented bismuth-oxo cluster incorporating metallocene-based ligands, establishing a new paradigm in polynuclear bismuth complex design.

We further examined the behavior of the Bi₈-Fc cluster in solution using ESI-MS (Fig. 5). The spectrum revealed four distinct peaks corresponding to ligand dissociation processes: (1) A prominent +2 charged peak at m/z = 5149.56 aligns with the theoretical m/z = 5149.64 for [H₂Bi₈O₄(TC4A)₄(FcCOO)₂]²⁺, confirming the intact cluster structure; (2) The +1 charged species at m/z = 4702.63 (observed; calcd. 4702.74) corresponds to [H₄NaBi₈O₄(TC4A)₄(DMF)]⁺, indicating partial loss of two FcCOO⁻ ligands; (3) A neutral cluster at m/z = 4464.29 (observed; calcd. 4464.65) assigned to [H₂Bi₈O₄(TC4A)₃(FcCOO)₂(MeOH)] demonstrates removal of one TC4A ligand with methanol coordination; (4) The species at m/z = 4277.32 (observed; calcd. 4277.67), identified as [Na₂Bi₈O₄(TC4A)₃(FcCOO)(DMF)], reveals simultaneous loss of one TC4A and one FcCOO⁻ ligand. These sequential ligand dissociation events contrast sharply with the complete core rearrangement observed in Bi₈-DMF, suggesting that Bi₈-Fc maintains its central Bi₈O₄ framework while undergoing controlled ligand shedding. This partial ligand loss with core preservation highlights the enhanced stability imparted by dual-ligand coordination in Bi₈-Fc compared to the single-ligand-protected Bi₈-DMF system.

The distinct ligand architectures of Bi₈-Fc and Bi₈-DMF clusters create fundamentally different catalytic microenvironments. Bi₈-Fc features a densely packed ligand shell comprising TC4A and ferrocene carboxylic ligands, whereas Bi₈-DMF replaces ferrocene derivatives with additional TC4A moieties and surface-bound DMF molecules, resulting in exposed Bi active sites. To systematically evaluate how these structural variations influence catalytic behavior, we investigated their performance in the electrochemical CO₂ reduction reaction (eCO₂RR) using a three-electrode H-cell configuration.

The electrocatalytic performance of two Bi₈ clusters was systematically investigated using a dual-chamber gas-tight H-type electrochemical cell. Initial characterization through linear sweep voltammetry (LSV) in 0.5 mol/L KHCO₃ revealed distinct current-voltage responses under CO₂-saturated versus N₂-saturated conditions (Fig. 6A). Both catalysts demonstrated enhanced current densities and significantly shifted onset potentials (−0.7 V vs. RHE) in CO₂-rich environments, establishing their preferential selectivity for CO₂RR over the competing H₂ evolution reaction (HER). Remarkably, under CO₂ saturation, the current density surged to 38.0 and 36.0 mA/cm² at −1.6 V vs. RHE for the respective clusters, underscoring their exceptional catalytic activity. Product analysis through gas chromatography (GC) and ion chromatography (IC) following 2-h electrolysis (−0.9~−1.6 V vs. RHE) identified formate as the dominant product, with trace amounts of H₂ and CO as secondary products. Control experiments under N₂ saturation crucially confirmed the exclusive origin of carbon-containing products from CO₂ reduction, as no such products were detected in the absence of CO₂. Additional blank tests without catalysts yielded only H₂ evolution, further validating the intrinsic catalytic function of the Bi₈ clusters.

Under CO₂-saturated conditions, Bi₈-DMF demonstrated exceptional Faradaic efficiency (FE) for formate production, maintaining values above 90% across a broad potential window from −0.9 V to −1.6 V vs. RHE. Notably, it achieved a peak FE of 92.79% at −1.0 V (Fig. 6B). In stark contrast, Bi₈-Fc exhibited substantial competitive hydrogen evolution, with formate FE fluctuating between 50%−68% and concomitant H₂ FE ranging from 29% to 40% in the same potential range (Fig. 6C). This pronounced disparity highlights Bi₈-DMF's superior catalytic selectivity for CO₂-to-formate conversion (Fig. 6D). The enhanced performance of Bi₈-DMF was further corroborated by partial current density analysis (Fig. 6E). At −1.6 V, Bi₈-DMF delivered a formate partial current density (j_HCOOH) of 32.67 mA/cm², representing a 1.54-fold enhancement over Bi₈-Fc (21.22 mA/cm²) (Fig. 6F). Comparative turnover frequency (TOF) calculations across all tested potentials consistently revealed the superior intrinsic activity of Bi₈-DMF (Fig. S25 in Supporting information), attributable to its optimized structural architecture and efficient bismuth site utilization.

Electrocatalytic stability assessment through chronoamperometric testing at −1.0 V revealed robust performance: current density remained stable above 20 mA/cm² with sustained formate FE > 90% during the initial 20-h operation. Post-20 h, while FE retention persisted, current density attenuation suggested potential structural degradation. Post-catalysis characterization provided critical insights: MALDI-TOF-MS of recovered Bi₈-DMF showed a dominant +1 charged species at m/z = 5304.79 (Fig. S36 in Supporting information), corresponding to {HBi₈O₃(TC4A)₅}¹⁺ (theoretical m/z = 5304.89). XPS analysis confirmed preservation of Bi³⁺ oxidation states after prolonged operation, indicating chemical stability under reaction conditions (Fig. S34 in Supporting information).

The eCO₂RR mechanism mediated by Bi-based materials unfolds through consecutive stages initiated by chemisorption of CO₂ onto catalytic active sites, followed by proton-coupled electron transfer (PCET) processes that generate key intermediates. The reaction pathway progresses through sequential transformations, beginning with the formation of *OCHO intermediates through initial PCET activation, followed by their conversion into adsorbed HCOOH species via subsequent proton-electron transfer steps, and ultimately culminating in the desorption of final products from the catalyst surface. To unravel this reaction network, we conducted in situ electrochemical attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR) measurements across a potential window from −0.9 V to −1.6 V vs. RHE (Fig. 7). When employing Bi₈-DMF as the electrocatalyst, many characteristic spectral regions revealed voltage-dependent evolution of critical intermediates. As illustrated in Fig. 8, the peaks appear at 1326, 1471 and 1540 cm⁻¹ attributed to the stretching vibration of m-CO₃²⁻, and at 1265, 1290 and 1560 cm⁻¹ belonging to the stretching vibration of b-CO₃²⁻ [44]. The bands at 1407 and 1654 cm⁻¹ in the spectra are attributed to HCO₃⁻ [34]. Most importantly, the *OCHO intermediate signals detected at 1388 and 1588 cm⁻¹ are generally regarded as key intermediates in the electrochemical CO₂/HCOOH conversion and increase gradually with the applied potentials varying between −0.9 V and −1.6 V. Besides, the characteristic peaks of HCOO⁻ (1617, 1685, 1716, 1735, 1735, and 1772 cm⁻¹) start to appear with increasing voltage.

The reaction energy pathways for electrochemical CO₂ reduction to HCOOH and CO, along with the competing HER, were systematically investigated through density functional theory (DFT) calculations. Fig. 8A presents the optimized configurations and corresponding energy profiles of key intermediates along the CO₂ reduction reaction (CO₂RR) pathways. To enhance computational efficiency while maintaining structural fidelity, the theoretical models were simplified by substituting tert‑butyl groups on the TC4A ligand with H atoms, thereby preserving the essential coordination environment of the Bi₈-DMF cluster observed in experimental crystal structures. The proposed catalytic mechanisms reveal distinct reaction pathways: The CO production pathway follows the sequence CO₂(g) → *COOH → *CO → CO(g), whereas formate generation proceeds through CO₂(g) → *OCHO → *HCOOH → formate(l). As shown in Fig. 8B, the potential-determining step (PDS) for formate formation in Bi₈-DMF corresponds to the desorption of *HCOOH intermediate, exhibiting a Gibbs free energy change (ΔG) of 0.91 eV. In contrast, the CO pathway demonstrates a significantly higher energy barrier at the initial proton-coupled electron transfer step (*CO₂ + H⁺ + e⁻ → *COOH), with a ΔG value of 1.34 eV. Furthermore, calculated hydrogen adsorption energies (*H) revealed an unfavorable ΔG of 1.69 eV, thermodynamically disfavoring HER. This energy landscape analysis provides a fundamental rationale for the experimentally observed product selectivity, where Bi₈-DMF exhibits superior Faradaic efficiency for HCOOH production compared to CO generation or H₂ evolution. For the Bi₈-Fc catalyst, the complete encapsulation of its bismuth-oxo cluster core by organic ligands results in the shielding of catalytic active sites and insufficient surface exposure. DFT computational analysis reveals that this structural constraint prevents the stable binding of key adsorbed intermediates (e.g., *OCHO or *COOH) to the cluster surface, indicating the inability of bismuth active sites to effectively stabilize these reaction intermediates. This mechanistic insight fundamentally explains the significantly lower CO₂ reduction catalytic activity of the Bi₈-Fc system compared to Bi₈-DMF at the atomic scale.

In this work, we have demonstrated the critical role of ligand engineering in modulating the catalytic performance of polynuclear bismuth-oxo clusters for selective electrochemical CO₂-to-formate conversion. Two structurally distinct octanuclear clusters, Bi₈-DMF and Bi₈-Fc, were synthesized via tailored ligand coordination strategies. The Bi₈-DMF cluster, stabilized solely by TC4A ligands with surface-exposed Bi sites coordinated by labile DMF molecules, exhibits exceptional catalytic activity and selectivity, achieving > 90% FE for formate across a broad potential window with stable operation exceeding 20 h. In contrast, the Bi₈-Fc analogue, featuring a fully encapsulated Bi-O core via dual-ligand coordination (TC4A and ferrocene carboxylate), shows inferior performance (60% maximum FE_HCOOH) due to restricted active site accessibility. DFT calculations reveal that the exposed Bi sites in Bi₈-DMF effectively stabilize the critical *OCHO intermediate through optimized orbital interactions, thereby lowering the energy barrier for CO₂-to-formate conversion. This study establishes a clear structure-activity relationship in polynuclear bismuth catalysts, emphasizing that controlled ligand coverage—rather than maximal encapsulation—is essential for balancing stability and catalytic efficiency. The findings provide a molecular-level blueprint for designing next-generation CO₂​RR catalysts, highlighting the importance of strategically exposed metal sites and dynamic ligand environments in achieving high-performance electrochemical systems.

CRediT authorship contribution statement

Hao-Nan Zhou: Writing – original draft, Investigation, Formal analysis, Data curation. Lan-Yan Li: Theoretical calculations. Hong-Bing Mo: Supervision, Funding acquisition. Yi-Xin Li: Funding acquisition. Jun Yan: Supervision. Chao Liu: Writing – review & editing, Supervision, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare no competing financial interest.

Acknowledgments

This work was supported by the Natural Science Foundation of Hunan Province (No. 2023JJ30650) and the Central South University Innovation-Driven Research Programme (No. 2023CXQD061).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2025.111269.