Lanthanide single-atom catalysts for efficient CO2-to-CO electroreduction

DFT Calculations

To elucidate the improved adsorption/desorption on Ln single-atom catalysts (SACs) for CO₂RR, we conducted a systematic investigation using Density Functional Theory (DFT) calculations, comparing them with traditional SACs. For clarity without loss of generality, erbium (Er) was chosen as the representative Ln, while calcium (Ca) and iron (Fe) were selected as reference elements. Both Ca and Fe are employed as typical SACs for CO₂RR, with Fe known for its low energy barriers for CO₂ activation and Ca favoured for its facile CO desorption^9,10,29.

Schematically, the intermediate *COOH exhibits a bridge adsorption state on Er SAC, in contrast to the linear adsorption on Fe and Ca SACs (Fig. 1a, b and Supplementary Fig. 1). However, the intermediate *CO shares a linear adsorption on Er SAC, consistent with those on Fe and Ca SACs. Free energy diagram indicates that Er SAC (0.61 eV) and Fe SAC (0.63 eV) exhibit lower energy barriers for CO₂ activation than that of Ca SAC (1.46 eV), proving the bridge adsorption benefits *COOH formation (Fig. 1c). In addition, the CO desorption on Er SAC (0.34 eV) and Ca SAC (−0.07 eV) are easier than that on Fe SAC (1.06 eV). Thus, the bridge adsorption does not transition onto subsequent *CO linear adsorption on Er SAC, thereby circumventing the scaling relationship in terms of *COOH and *CO observed in typical SAC.

Structure and adsorption configurations of key intermediates on Fe SAC (a) and Er SAC (b). c Free energy diagram for CO₂ reduction to CO. d Charge density differences for *COOH and *CO adsorbed on SAC. Yellow, charge density accumulation; green, charge density depletion. Bader charge transfer analysis for COOH adsorbed on SAC (e) and CO adsorbed on Metal (M) SAC (f). Atom color-coding: Red, oxygen; white, hydrogen; grey, carbon; blue, nitrogen; orange, Fe; pink, Er.

To further explore the interaction between metal sites and intermediates, we examined the catalyst-intermediate adduct (M-COOH, M-CO) (Fig. 1d and Supplementary Fig. 2). Charge density differences and Bader charge analysis are conducted to analyze the intensity of chemical bonds between intermediates and metal sites. As shown in Fig. 1d–f, show that charge transfer from Er and Fe SAC to *COOH is 0.500 and 0.584 e respectively, larger than that of Ca SAC (0.363 e), indicating strong *COOH adsorption on Er sites by virtue of bridge adsorption. Furthermore, Er SAC (0.012 e) and Ca SAC (0.005 e) provide negligible charge to *CO, compared to Fe SAC (0.136 e), demonstrating the easier CO desorption on Er and Ca site than that on Fe site. Thus, Er SAC facilitates the CO₂RR to CO pathway through the bridge adsorption of *COOH as well as the linear adsorption of CO.

Catalyst synthesis and characterization

To utilize the unique catalytical properties predicted by DFT calculations, Er SAC was prepared on carbon nanotubes (CNTs) through a facile calcination method (Scheme in Supplementary Fig. 3). Only the XRD diffraction peaks of CNT substrate was found, suggesting the absence of metallic phases and a homogeneous dispersion of the Er atoms (Supplementary Fig. 4)^30,31. No metal contamination was found in these precursors of Er SAC (Supplementary Figs. 5, 6). The metal contents in Er SAC were estimated to be ~2.17 wt% (Supplementary Table 1), according to inductively coupled plasma optical emission spectrometer (ICP-OES). The morphology and the atomic Er dispersion were investigated through different types of microscopies. The results from scanning electron microscope (SEM), high-resolution transmission electron microscope (HRTEM) and aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) confirmed the single-atom nature of the Er sites on CNT (Fig. 2a and Supplementary Fig. 8). In addition to Er, we developed a systematic procedure to SACs supported on CNTs utilizing all Ln expect radioactive Promethium (Pm), encompassing Lanthanum (La), Cerium (Ce), Praseodymium (Pr), Neodymium (Nd), Samarium (Sm), Europium (Eu), Gadolinium (Gd), Terbium (Tb), Dysprosium (Dy), Holmium (Ho), Thulium (Tm), Ytterbium (Yb), and Lutetium (Lu). Significantly, the size of Ln atom is much larger than that of Ca, and Fe atoms. Further investigations show that 16 SACs including Ln, Ca, and Fe SACs have similar components and morphologies as those of Er SAC (Supplementary Figs. 7–25), demonstrating the universality of preparation.

a HRTEM image for Er SAC. b AC HAADF-STEM images of Ln SACs. Scale bar, 2 nm. c XANES spectra of C K-edge. d XANES spectra of N K-edge. e Er L₃-edge XANES spectra of Er foil, Er₂O₃ and Er SAC. f k³ weighted Fourier transform spectra of Er foil, Er₂O₃ and Er SAC. g the corresponding EXAFS R space fitting curves for Er SAC.

To acquire the structural information of SACs, synchrotron-based X-ray adsorption near edge structure (XANES) spectra were conducted. Er SAC shows a clear increase of C K-edge peak intensity at ∼288.5 eV, suggesting the possible formation of C-N-Er bonds, compared to nitrogen doped carbon nanotubes (NC) (Fig. 2c)^32,33,34. The increase of pyridinic N peak at 400.7 eV indicates Er atoms dominantly coordinate with pyridinic N atoms in Er SAC (Fig. 2d)^35,36. Fourier transformed (FT) extended X-ray adsorption fine structure (EXAFS) manifests the atomic dispersion features of the Er, Fe and Ca atoms, with a coordination number of ~6, ~4 and ~4 based on well-fitting process respectively, consistent with the results from theoretical calculations (Fig. 2e–g and Supplementary Figs. 10, 11, Supplementary Table. 2).

To study the real-time intermediates formed on metal sites during CO₂RR to CO, we carried out operando attenuated total reflection infrared spectra (ATR-IR) and L₃-edge XAS measurements for Er (Supplementary Figs. 26, 27). Peaks located at range from 1910 to 1950 cm⁻¹ and around 1640 cm⁻¹ can be attributed to *CO (linear-bonded CO) and H₂O bending respectively (Fig. 3a, b and Supplementary Fig. 28 and Supplementary Table 3)^37,38. Notably, a new peak ranging from 1800 to 1840 cm⁻¹ was identified, assigned to bridge adsorption of *COOH on Er SAC, distinguishing it from Fe and Ca SAC^9,39,40. Similarly, Er and Fe SACs show stronger CO₂ adsorption signals of CO₂ adsorption based on temperature program desorption, compared with that of Ca SAC, facilitating the subsequent CO₂ activation on Ca and Fe sites (Supplementary Fig. 29). Furthermore, differing with the distinct CO adsorption peaks on Fe SAC, no such peak was observed on Er and Ca SACs, suggesting facile CO desorption from Er and Ca sites, consistent with the results from DFT calculations.

a operando ATR-IR spectra of Er SAC (without iR-correction). b operando ATR-IR spectra of Fe SAC (without iR-correction). c operando XANES spectra of Er SAC measured at different potentials (without iR-correction). d Corresponding operando k³ weighted Fourier transform EXAFS spectra of Er SAC (without iR-correction). Atom color-coding: Red, oxygen; white, hydrogen; grey, carbon; blue, nitrogen; orange, Fe; pink, Er.

We further investigated the chemical structure and coordinating environment of Er SAC under operating conditions using XAS. The intensity of white line heightened slightly due to the increased chemical state during CO₂ electroreduction (Fig. 3c). The Er-N bond position of Er SAC shifted towards the Er-O side during CO₂RR, demonstrating the existence of Er-O from Er-*COOH bridge adsorption^41,42. Importantly, over the applied potential range relevant to CO₂RR, the Er-Er bond was not observed (Fig. 3d), suggesting that Er SAC maintained its original complex status without reduction to Er nanoparticles or nanoclusters under operating conditions^39,43.

Evaluating catalyst performance for CO₂RR

To evaluate the performance of catalysts, electrochemical tests were first conducted in CO₂ saturated 0.5 M KHCO₃ electrolyte (Supplementary Figs. 30–40). Er SAC shows high Faradaic efficiencies of CO (FE_CO) ≥90% over a wide potential range from −0.47 to −0.97 V vs. RHE (Fig. 4a), with a maximal FE_CO reaching ~99% (Supplementary Fig. 38). Furthermore, Er SAC retains a FE_CO above 90% at −0.67 to −0.87 V vs. RHE, even when the CO₂ concentration was reduced to 30% (Supplementary Fig. 36). The flow cell fabricated with Er SAC shows a large current density of 500 mA cm⁻² under high CO Faradaic efficiency ≥90% in 1 M KHCO₃ electrolyte (Fig. 4b and Supplementary Figs. 41, 42).

a FE_CO for Ln SACs at −0.67 V vs. RHE (without iR-correction) in H-cell under CO₂ saturated 0.5 M KHCO₃ solution. b Potentials and J_CO for Er SAC at different current densities with 1 M KHCO₃ solution (upper) and 1 M KCl (pH = 1, down) in flow cell (95% iR-correction, Resistance = 2.4 ± 0.3 Ω and 1.8 ± 0.2 Ω for neutral and acidic electrolyte, respectively). c FE_CO and current densities for Er SAC at different CO₂ flow rates at 1 M KCl (pH = 1). d FE_CO and SPCE of CO₂ to CO on at different CO₂ flow rates (pH=1, applied current density of −200 mA cm⁻²). e Stability of Er SAC at a current density of −100 mA cm⁻² with (pH = 1, potential with 95% iR-correction). f Comparison of CO₂RR-to-CO performance with the state-of-art electrocatalysts in acidic media^{68,69,70,71,72,73}. The error bars correspond to the standard deviations of measurements over three separately prepared samples under the same testing conditions.

To avert low carbon utilization limits witnessed in neutral solutions^44,45,46, the performance of Er SAC was tested in acidic media, 1 M KCl (pH adjusted to 1.0 with sulfuric acid). Only tiny degradation of FE_CO was observed when then current densities were increased by 10 times from 50 to 500 mA cm⁻², demonstrating its potential for large-scale production. Such ≥90% Faradaic efficiency at 500 mA cm⁻² both in neutral (top panel, Fig. 4b) and acidic electrolyte (bottom panel, Fig. 4b and Supplementary Fig. 43) endows the SAC a high turnover frequency (TOF) of ~130,000 h⁻¹. Furthermore, Er SAC can achieve a full-cell energy efficiency of 34.7% at 200 mA cm⁻² (Supplementary Figs. 44–47 and Supplementary Tables 4, 5)^47,48,49,50. To decrease resistance of the system, we have tested the performance in the Membrane Electrode Assembly (MEA). As shown in Supplementary Table 5, MEA fabricated Er SAC shows high full-cell energy efficiency of ~32.5% even at high current density of 300 mA cm^-2.

Moreover, we investigated the catalytic performance across various CO₂ flow rates spanning a wide range of current densities. Notably, we achieved a remarkable single-pass carbon efficiency (SPCE) of 70.4% for CO₂RR to CO a large current density of 200 mA cm⁻², indicating that 70.4 out of 100 CO₂ molecules can be successfully transformed into CO at the outlet (Fig. 4c, d and Supplementary Tables 6–11). The flow cell maintained stable operation at 100 mA cm⁻², with FE_CO >90% in acidic electrolyte (100 h, Fig. 4e), and meanwhile showed a long-term stability in neutral electrolyte (Supplementary Fig. 45). No obvious cluster and Er-Er bond are observed on Er SAC after the test according to operando XANES spectra (Fig. 3d) and TEM images, demonstrating the durability of structure (Supplementary Fig. 46).

Figure 4f illustrates a comparative analysis of catalytic performance between our study and a previous flow cell for CO₂ to CO conversion. To provide a comprehensive evaluation, we considered five performance indicators: efficiency (both FE and SPCE), stability (duration for FE_CO exceeding 90%), pH, and the operational current at maximum SPCE. Our Er SAC surpassed previous reported data across all performance metrics, a testament to the efficacy of the unique mechanism elucidated earlier (Supplementary Tables 12–14).

To assess the generic catalytic abilities of Ln, similar tests were conducted for the other thirteen SACs utilizing different elements. All of Ln SACs keep a high FE_CO between 92.0% and 99.6% at −0.67 V vs. RHE (Fig. 4a). The consistently high efficiency observed across SACs using elements from the entire Ln group (except radioactive Pm) unequivocally demonstrates the universal catalytic activity of Ln as predicted by DFT and indicated by operando spectra.

In summary, we proposed and realized a series of Ln SACs with catalytic performance for CO₂RR to CO. The unique electron configuration shared by Ln metals facilitates the bridge adsorption for favorable *COOH formation, which circumvents the scaling relationship between *COOH and *CO. Both DFT calculations and experimental characterizations were employed to prove the bridge adsorption of *COOH for Er SAC.

The adaptability of our strategy was validated through CO₂RR flow cells containing SACs incorporating 14 different Ln metal atoms, all exhibiting FE_CO beyond 90%. Among them, the Er SAC demonstrated a remarkable TOF of ~130,000 h⁻¹ at a high current density of 500 mA cm⁻², maintaining a CO Faradaic efficiency of ≥90% in both neutral and acidic electrolytes. This achievement results in a significant enhancement of full-cell energy efficiency to 34.7% and single-pass CO₂ electrolysis (SPCE) to 70.4% at a large current density of 200 mA cm⁻². These results underscore the promising potential of our approach for practical industrial applications. Furthermore, with the recent pioneering work involving promethium (Pm), the investigation of this only missing lanthanide element in our configurations will be an interesting and valuable direction for future research⁵¹.

While previous studies have primarily focused on identifying specific Ln atoms for targeted chemical transformations such as nitrogen reduction, carbon dioxide reduction, and oxygen reduction reactions^{31,52,53,54,55}, our approach represents a departure from this paradigm. Rather than singling out individual atoms, we present a methodology that harnesses the collective catalytic potential of the entire Ln group. The significance of our work extends beyond the demonstrated conversion of CO₂ to CO; it opens avenues for exploring the applicability of our strategy (using the entire Ln group) to a broader range of reduction reactions and beyond. In addition to its fundamental implications, our approach offers an platform for CO₂ neutralization, with the unique ability to select different Ln atoms without compromising efficiency. This flexibility introduces potential benefits for practical applications, where the choice of Ln metals can be guided by considerations such as availability, cost, and compatibility, in addition to catalytic properties.