Direct time-resolved observation of surface-bound carbon dioxide radical anions on metallic nanocatalysts

Abstract

Time-resolved identification of surface-bound intermediates on metallic nanocatalysts is imperative to develop an accurate understanding of the elementary steps of CO₂ reduction. Direct observation on initial electron transfer to CO₂ to form surface-bound CO₂^•− radicals is lacking due to the technical challenges. Here, we use picosecond pulse radiolysis to generate CO₂^•− via aqueous electron attachment and observe the stabilization processes toward well-defined nanoscale metallic sites. The time-resolved method combined with molecular simulations identifies surface-bound intermediates with characteristic transient absorption bands and distinct kinetics from nanosecond to the second timescale for three typical metallic nanocatalysts: Cu, Au, and Ni. The interfacial interactions are further investigated by varying the important factors, such as catalyst size and the presence of cation in the electrolyte. This work highlights fundamental ultrafast spectroscopy to clarify the critical initial step in the CO₂ catalytic reduction mechanism.

Introduction

The conversion of planet-warming carbon dioxide (CO₂) into valuable chemicals or fuels is considered one of the appealing methods for storing renewable energy and facilitating sustainable carbon cycles^1,2. However, the high activation energy associated with the change in orbital hybridization and geometry of CO₂ upon the first electron attachment, the generation of CO₂ radical anion (CO₂^•−), makes its reduction a significant challenge. As a result, the successful operation depends heavily on the viable catalysts that can accelerate the reaction. In most electrolytic, photolytic, and radiolytic processes, metal-based catalysts primarily contribute to CO₂ transformations^3,4,5,6. Active metals can be categorized into several groups based on their product distribution. For example, Ag and Au predominately yield CO, while Ni and Fe exclusively produce H₂ with low activity, and Cu has an unusual capacity to synthesize hydrocarbon/multi-carbon products^7,8,9. These findings raise questions about the importance of reaction intermediates binding at the interface on the efficiency. Although great progress has been made in understanding the underlying surface chemistry, the reaction sequence remains incomplete, partly due to the lack of direct experimental evidence for unraveling the fast dynamics of CO₂^•− during the initial stages of complex multi-electron processes. On the other hand, advances in synthetic methodology have made well-controlled metallic nanoparticles and nanoclusters available, and these nanostructured materials exhibit better performance in CO₂ conversion and lower cost. Strategies such as surface engineering, morphology control, and composition manipulation have been developed to enhance the efficiency of CO₂ activation^10,11,12. In this regard, the characterization of CO₂^•− coordination complex with metal centers could provide a basic descriptor to establish the structure-property correlation, where any transient state changes with variables of interest at nanoscale spatial resolution could be detected. However, detecting these active sites directly from experiments has been impossible thus far^13,14,15,16.

The existence of CO₂^•− on metal electrodes, such as Au, Cu, and Ag, has been proposed based on the experimental Tafel slope analysis^{17,18,19,20,21}. Besides, the critical role of CO₂^•− formed by hydrated electrons (e_aq⁻) attachment has recently been recognized by plasma or radiation-driven catalytic and nanodiamond-assisted photolytic CO₂ reduction^22,23,24. Different catalytic systems have several rate-determining steps (RDS), such as initial electron transfer to CO₂ to form surface-bound CO₂^•− radicals, the first proton-coupled electron transfer, or the reduction of other bound intermediates^{17,18,19,20,21}. Due to the large structural reorganization of the bound radical anion, the radical anion CO₂^•− formed by the first electron reduction occurs at very negative potentials. Even though CO₂^•− formation lacked experimental evidence, many studies suggested that the process is likely to be one of the critical RDS for CO₂ reduction^7,9,15. If the CO₂^•− radicals stabilization on catalytic surfaces is kinetically non-favorable, the overall activity for CO₂ transformation will be low. This hypothesis is further supported by theoretical studies that developed a contemporary model for various *CO₂^•− coordination structures, including *COOH or *OCOH, uncovering the selective electrocatalytic pathway toward either CO or formate^{25,26,27,28,29,30}. In situ surface-enhanced Raman scattering has reported the presence of this surface species on an operating Cu surface, but the dynamic resolution remains unresolved^14,26,31,32. For nanocatalysts, operando X-ray absorption and photoelectron spectroscopy are sensitive to surface oxidation evolution and to the accurate active site of metals such as Cu, Pd, Zn, and Sn during CO₂ reduction^25,27,33,34. Few studies have been dedicated to the initial step of CO₂ reduction at a nanoscale catalytic surface, owing to the challenges of forming well-defined nanocatalysts and CO₂^•− intermediate. Additionally, unlike carbon monoxide, which is frequently used as a mechanistic probe, CO₂^•− radicals decay rapidly in aqueous solutions, with a short lifetime of microseconds. The existing operando techniques are limited to the resolution of seconds or sub-seconds range, which cannot access the first elementary reactions occurring at nanosecond timescale.

For initial CO₂ activation, pulse radiolysis provides a time-resolved prerequisite to insight into the general transient absorption kinetics of CO₂^•−. Radiolysis is often referred to as “an electrolysis process without electrodes”. The high-energy radiation (X-rays/accelerated e⁻) with sufficient energy ejects electrons directly from the water. Unlike electrochemical or photochemical reduction, where electrons are provided by the external electric field or photo-induced carriers (e⁻, h⁺) separation in semiconductor materials, the radiation-driven approach produces electrons in bulk solutions via water ionization, known as the e_aq⁻ with characteristic transient absorption at 715 nm. With the standard potential of −2.87 V_SHE, e_aq⁻ can easily overcome the first CO₂ reduction step barrier³⁵. The picosecond pulse radiolysis coupled with transient absorption spectroscopy (ELYSE platform, Paris-Saclay University) could form e_aq⁻ in CO₂-saturated aqueous conditions within 7 ps, and e_aq⁻ reacts directly and quantitatively with CO₂ forming CO₂^•− radicals with an almost diffusion-controlled rate (8 × 10⁹M⁻¹ s⁻¹) within 10 ns. Until now, the current in situ/operando spectroscopic measurements on CO₂ reduction were performed with limited time resolution in a few studies with subsecond time resolution based on infrared, Raman, MS, and XAS^30,36,37,38. Compared to these spectroscopic techniques based on molecular vibration or photoelectron excitation, transient absorption profiles can disclose the electronic transition of surface-bound intermediates on metallic nanocatalysts with the nanosecond time resolution, which allows for the direct observation of the reaction between CO₂^•− and nanoparticles (NPs) as well as their stability kinetics on the surfaces. The method avoids external potential supported catalysts perturbation and the complexity associated with processes involving photosensitizers or electrodes. Therefore, pulse radiolysis measurements allow us to gain critical insights into the first elementary step of CO₂ reduction, such as the intermediate binding and transformation, as well as charge delocalization on nanoscale metallic surfaces. This methodology would set up a model to investigate the factors in operating conditions, such as cations in electrolytes and the nano-size effect of catalysts, which are important but under intense debate in CO₂ reduction because the existing in situ spectroscopic measurements often deal with combined signals caused by various possible consequences including solution effects, competing reactions, and surface structure evolution or catalysts oxidation.

Herein, by using pulse radiolysis^39,40, we present direct transient spectroscopic evidence of CO₂^•− radical stabilization process on the surface of nanoscale catalysts, leading to the formation of surface-bound CO₂^•−. The characteristic absorption spectrum of surface species and density functional theory (DFT) simulations revealed the coordinated structure and interactions between metal clusters and CO₂^•− radicals. The Tafel analysis and electrochemical catalytic performance with the same NPs used in the pulse radiolysis observations are also performed. As a result, our distinct transient kinetics on various well-defined metal (Au, Cu, Ni) nanoparticles (NP), which are commonly included in electrocatalysis, provided fundamental insights to revisit the activity and selectivity of metal catalysts in CO₂ reduction.

Results

Reaction kinetics of CO₂ ^•− radicals in bulk solution

Supplementary Figure 1a presents the transient absorption profiles of CO₂ radical formation in CO₂-saturated formate aqueous solution in the absence of any catalyst. Immediately following the electron pulse, the typical broad band in the 400–720 nm range corresponds to e_aq⁻. The appearance of a new absorption band below 420 nm on the nanosecond timescale, which is associated with e_aq⁻ decay, indicated the rapid formation of CO₂^•− radicals via e_aq⁻ reduction (e_aq⁻ + CO₂ → CO₂^•−, k₁ = 8.2 × 10⁹M⁻¹ s⁻¹). Meanwhile, 0.1 M formate was used to convert ^•OH and ^•H into CO₂^•− radicals (^•OH + HCOO⁻ → CO₂^•− + H₂O, k₂ = 4.1 × 10⁹M⁻¹ s⁻¹; ^•H + HCOO⁻ → CO₂^•− + H₂, k₃ = 2.1 × 10⁹M⁻¹ s⁻¹), thereby converting all water radiolysis radicals into CO₂^•− radicals. Subsequently, CO₂^•− radicals underwent a recombination reaction (CO₂^•− + CO₂^•− → C₂O₄²⁻, k₄ = 1.9 × 10⁹M⁻¹ s⁻¹) and lived for approximately 10 µs under the experimental conditions (Supplementary Fig. 1b–d). These data depicted a landscape of the formation and decay process of CO₂^•− radicals in our conditions and in the absence of metal catalysts, in line with those reported in the literature^41,42.

The effect of nanoscale metal catalysts (Au/Cu/Ni) on the CO₂ ^•− radical reactivity

To elucidate the CO₂^•− radical reactivity with nanocatalysts, we synthesized three typical (Au/Cu/Ni) metal NP by an established radiolytic reduction method with surfactants to regulate the size^43,44 (details in Methods and Supplementary Fig. 2). In addition to our time-resolved measurements, we performed electrochemical analysis on these three NPs. The obtained Tafel slope for CO production with (Au/Cu/Ni) NPs as-synthesized in our work agreed with previous reports^{17,18,19,20,21}. The distinct activity indicates a rate-determining initial e⁻ transfer from CO₂ to CO₂^•− intermediates (Supplementary Fig. 3), excluding the possible RDS relative to protonation and the coupling reaction of CO₂^•− intermediates. The transient absorption profiles of CO₂^•− radicals in the presence of these nanoscale metal catalysts from nanoseconds to microseconds are shown in Fig. 1. Figure 1a illustrates the transient kinetics of CO₂^•− at 350 nm in various concentrations of Au NP solution within 700 ns. Obviously, Au NP suppresses the decay of CO₂^•− and even induces the growth of a new absorption band extending to nearly 520 nm during 7 µs as the concentration of Au rises to 0.5 mM. The optical spectral observations suggest that the gradual introduction of Au NPs inhibited the recombination of CO₂^•− radicals, promoting the formation of new surface species instead of oxalate (Fig. 1b).

Fig. 1: Time-resolved absorption of CO₂^•− radical stabilization process with different metal NP.

a–d Transient kinetics at 350 nm within 700 ns as a function of Au concentration (a), fitted transient absorption spectra within 7 µs in 0.5 mM Au solution (b), transient kinetics at 420 nm as a function of Au concentration (c), and transient absorption matrix (d) within 80 µs in 0.5 mM Au solution. e–h Transient kinetics at 350 nm within 700 ns as a function of Cu concentration (e), fitted transient absorption spectra within 7 µs in 0.5 mM Cu solution (f), transient kinetics at 450 nm as a function of Cu concentration (g), and transient absorption matrix (h) within 80 µs in 0.5 mM Cu solution. i–l Transient kinetics at 350 nm within 700 ns as a function of Ni concentration (i), fitted transient absorption spectra within 7 µs in 0.5 mM Ni solution (j), transient kinetics at 350 nm as a function of Ni concentration (k), and transient absorption matrix (l) within 80 µs in 0.5 mM Ni solution. Source data are provided as a Source Data file.

To investigate the origin of the absorbing species, we performed control experiments and analyzed the kinetics of the supernatant of Au suspensions (solution prepared for pulse radiolysis measurements but underwent centrifugation to remove Au NP). Our results ruled out the possibility that the new absorption band is caused by the reaction of CO₂^•− with any other components in the solution (Supplementary Fig. 4a–e) than Au NP. To ensure the signals are from the surface-bound CO₂^•− on Au NPs, we performed an additional control experiment of the Au NPs system by replacing CO₂ with Ar. In these conditions, the only reducing species is e_aq⁻ since ^•OH radical is scavenged by tert-butanol. The results clearly show the absence of the absorption band as in the case of CO₂-saturated solutions (Supplementary Fig. 4f, g). Additionally, we found that the absorption intensity of the new species correlates positively with the concentration of Au NP (Supplementary Fig. 5d). These experimental data lead us to conclude that the new species absorbing in the visible and UV range corresponds to CO₂^•− radicals adsorbed on Au surfaces (CO₂^•− + Au → $({{{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-}})_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$), and the broader absorption band compared to that of free CO₂^•− is attributed to the more electron delocalization of $({{{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-}})_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ on Au surfaces. Considering the low concentration of Au NP, such kinetics suggest the effective occurrence of CO₂^•− radical stabilization on Au surfaces, outcompeting the fast recombination reaction.

The transient kinetics at 450 nm in the range of 80 µs offers insights into the stabilization process since it corresponds to almost exclusive absorption for ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ with respect to free CO₂^•− radicals in solutions (Fig. 1c). The formation kinetics of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ is accelerated significantly with Au NPs concentration during the first 20 µs. After 20 µs, free CO₂^•− radicals are nearly eliminated by the recombination reaction, and the stabilization on Au surfaces is complete. However, the absorption at 450 nm continues to grow until 80 µs, combined with the slight broadening of the spectra, suggesting an optimization process of the adsorbed CO₂^•− radicals on Au surfaces (Supplementary Fig. 5a). After a few 100 microseconds, the spectral band shape hardly changes with time. The protonation reaction can also accompany this stabilization process by forming HCO₂^• on the surface of NPs. Nevertheless, as the pKa of CO₂^•− is 2.8, the possibility of the CO₂^•− protonation is limited. These critical features demonstrate that CO₂^•− radicals stabilize on Au surfaces swiftly within 20 μs, after which the coordination state progressively orientates and sustains for more than 80 μs (Fig. 1d).

Next, we extended our studies to nanoscale Cu and Ni. In both cases, we pay attention to avoiding any oxidized sites on the surface NPs by performing all experiments in reducing conditions. As shown in Fig. 1e, compared with Au, the transient kinetics at 350 nm displays no growth during the first 700 ns, even at the highest concentration of Cu (0.5 mM). However, within 10 µs, a distinct species issued from the CO₂^•− radicals in solution is identified (CO₂^•− + Cu → ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$) through a detailed comparison of the transient kinetics at 350 and 420 nm (Supplementary Fig. 6). The transient absorption in the Cu system also evidences the formation of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ intermediates, which shows a characteristic absorption band stretching to around 420 nm (Fig. 1f). By comparing the kinetics for solutions containing the same size of Au (5.3 nm) and Cu (5.4 nm) NPs (Supplementary Fig. 7), within 80 μs, the kinetics and intensity of CO₂^•− radicals are similar due to the almost identical surface areas. However, after 80 μs, surface-bound CO₂^•− radicals exhibited accelerated decay on Cu NPs yet still increased on Au NPs (Supplementary Fig. 5b, e). These diverging kinetics between surface-bound CO₂^•− radicals on Au and Cu NPs at identical sizes suggest distinct stabilization processes, which may correlate with catalytic CO₂ activity and selectivity.

In the Ni case, CO₂^•− radicals show almost the identical transient kinetics as in nanocatalyst-free solution, with nearly negligible concentration dependency (Fig. 1i–l). At the 100 µs timescale, almost no absorption is detected with Ni (Supplementary Fig. 5c, f), underlying no surface CO₂^•− radicals on Ni NP are formed. According to previous literature^8,45, Ni is considered to exhibit a relatively low CO₂RR catalytic activity due to hydrogen evolution reaction (HER) activity and strong binding capabilities with CO intermediates. Nevertheless, our results on CO₂^•− radical kinetics suggest that the activity of metal sites can be predetermined, well before the binding of intermediates from CO₂ dissociation.

To probe the structure evolution of surface-bound (CO₂^•−)^ad radicals, we extended the observation of transient kinetics to the sub-millisecond range. Notably, (CO₂^•−)^ad radicals on both Au and Cu surfaces maintain their specific spectral features even after 800 µs, with intensity dependent on the concentration (Fig. 2a and Supplementary Fig. 8). We observe that the transient kinetics of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ tend to plateau after 200 µs, whereas ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ shows a moderate decay (Fig. 2b–f). However, in the presence of Au NP with the same size as Cu, the decay of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{\rm{Au}}}}}}}^{{{{{{\rm{ad}}}}}}}$ radicals occur as well (Supplementary Fig. 9), which reveals that the presence of larger NP results in a higher concentration of CO₂^•− radicals per particle, and the reactivity of surface-bound ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{\,}^{{{{{{\rm{ad}}}}}}}$ may be altered to occur more readily. However, no new species are observed in the 300–720 nm range; only the moderate decay represents the possible next slow elementary transformation of (CO₂^•−)^ad radicals.

Fig. 2: Stability of (CO₂^•−)^ad radicals on metal surfaces.

a Transient absorption spectra of CO₂^•−, ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$, ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$, ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Ni}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ radicals in 0.5 mM Metal NPs solution at 800 µs. b–d Transient kinetics at 350 nm (b), 450 nm (c), and 520 nm (d) with ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ as a function of Au concentrations. e–g Transient kinetics at 350 nm (e), 420 nm (f), and 500 nm (g) with ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ as a function of Cu concentrations. Source data are provided as a Source Data file.

In addition, the increased bleaching of the surface plasmon absorption band with time delay is found (Fig. 2d, g). Typically, such bleaching occurs in the femtoseconds or picoseconds range via electron thermalization and electron-phonon relaxation, which seems incompatible with our results^46,47,48. We suggested that the rise in the bleaching signal with Au content (Fig. 2d), coupled with the concerted kinetics of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$, implies that ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ radicals affect surface electrons relaxation and may promote a charge transfer from ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ radicals to Au surfaces, resulting in the bleaching of the surface plasmon absorption band. This constitutes another piece of evidence to corroborate our observation of CO₂^•− radical stabilization at the nanocatalyst surfaces as the first elementary step in CO₂RR.

More importantly, we followed the next step of the elementary reaction of surface-bound CO₂^•− on the subsecond timescale. In the case of Cu NPs, we observe the formation of another intermediate following the decay of the first species at 0.05 s, which has an extended absorption band from 300 to 600 nm. This new surface species can be observed up to 900 ms, which is the limitation of observation (Fig. 3a). Moreover, the absorption spectra continue changing within 900 ms, which may indicate the multistep reaction of surface-bound species or the evolution of surface coordination structure. Due to the characteristic absorption band between 300 and 600 nm of the newly observed species, we exclude any stable steady-state product, such as oxalate, formate, or CO. It could be attributed to a new transient species with conjugated structure involving two CO₂^•− radicals as the precursor for CO and formate, already suggested in nanocatalyst-free solutions^10,15,49. As for Au NPs with similar and smaller sizes, no further reaction occurs; only the decay of the absorption band of the first species is observed, indicating that surface-bound CO₂^•− decays slowly within 1 s (Fig. 3b and Supplementary Fig. 10).

Fig. 3: Transient absorption profiles at second timescale.

a, b Transient absorption spectra of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{\rm{NP}}}}}}}^{{{{{{\rm{ad}}}}}}}$ radicals at 1 ms, 50 ms, 150 ms, 500 ms, and 900 ms in the presence of 0.5 mM 5.4 nm Cu (a) and 5.3 nm Au (b). c, d The curves of ΔO.D. versus time delay at 350 nm (c) and 540 nm (d). e Schematic diagram of the generation, stabilization, and conversion of CO₂^•− radicals in the presence of Au and Cu NP. CO₂ reacted with e_aq⁻ to form CO₂^•− within a few nanoseconds and the CO₂^•− radicals stabilized on the surface of Cu and Au NPs. After 1 ms, ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{\rm{Cu}}}}}}}^{{{{{{\rm{ad}}}}}}}$ radicals were converted to dimer intermediates, and ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{\rm{Au}}}}}}}^{{{{{{\rm{ad}}}}}}}$ radicals decayed for around 150 ms. For simplicity, cations are not shown. Source data are provided as a Source Data file.

Besides, Au and Cu NPs also showed essential distinctions in the kinetics of surface plasmon absorption bands, implying the different interaction between CO₂^•− radicals and NPs. Within 1 ms, both Au and Cu present slight bleaching of their surface plasmon absorption band due to the charge of surface-bound CO₂^•− radicals, revealing the abovementioned stabilization process. At a longer time (a few hundred ms), once the next reaction happens, a second intermediate species is observed without noticeable bleaching of the absorption band in the case of Cu NPs. Nevertheless, the bleaching of the surface plasmon absorption band in the Au system persisted up to 900 ms, accompanied by the decay of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ radicals, which demonstrated the primary occurrence of charge transfer rather than covalent interaction between Au NPs and CO₂^•− radicals (Fig. 3c, d). The different kinetics observations for the bleaching of the surface plasmon absorption band and characteristic surface-bound CO₂^•− suggested the different character of Cu and Au for CO₂^•− stabilization and conversion, as well as can explain the selective catalysis property of Cu for CO₂ reduction.

In order to further validate our findings of CO₂^•− coverage, we have performed comprehensive electrochemical CO₂RR testing on as-synthesized Au, Cu, and Ni NPs. Our results showed a typical tendency for the model catalysts (Au, Cu, and Ni) in our work (Supplementary Fig. 11). Cu induced the formation of CO, CH₄, and C₂H₄. In contrast, Ni predominately produced hydrogen with very low activity to convert CO₂, and Au mainly produced CO from the conversion of CO₂. The tendency of electrochemical CO₂RR measurements agrees with our pulse radiolysis observation. We found that Ni NPs cannot stabilize CO₂^•− radicals, which correlates with its slight activity of CO₂ reduction; however, Cu and Au NPs can stabilize CO₂^•− radicals. The long-lived characteristic spectra disclosed that the stabilization process of CO₂^•− radicals on the surface of Au and Cu has substantially extended the lifetime of [CO₂^•−]_ad radicals by at least 100 times compared to CO₂^•− radicals in solutions or in the presence of Ni. The extended lifetime lays the groundwork for the subsequent multi-electron transfer reaction. Moreover, the differentiation in the respective [CO₂^•−]_ad radical transient kinetics and characteristic spectra across Au, Cu, and Ni systems suggest diverse stabilization behavior and adsorbed structures of [CO₂^•−]_ad radicals on various metal surfaces. This selective stabilization process determines the subsequent selective reduction pathway of [CO₂^•−]_ad radicals. Notably, we also observed the dimerization pathway at a subsecond range of surface-bound CO₂^•− radicals on Cu rather than Au, which is in accordance with the unique property of Cu to produce C₂H₄.

To better understand the nature of distinct absorption bands from CO₂^•− radicals absorbed on the surface of Au and Cu NP, molecular simulations were performed at the DFT level with solvent effects represented by a dielectric continuum model. Figure 4a, b depicts the optimized structure of the two complexes and highlights a different interaction mode. Specifically, CO₂^•− radical interacts solely via the carbon atom to the Au atom at a distance of 2.17 Å, with geometry comparable to the isolated CO₂^•− (see Supplementary Table 1). The overall charge of CO₂^•− is −0.5 electrons, indicating a partial electron transfer to Au. A Mayer bond order of 0.45, a covalent interaction provided by the DORI plot, and an Au-C ELF valence basin with a population of 1.2 electrons support this conclusion (see Supplementary Information for computational details). In contrast, ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ stabilizes with the carbon atom and one oxygen atom at distances of 2.01 Å for Cu-C and 2.08 Å for Cu-O. The geometry of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ is influenced by lowering the COC angle and increasing the intramolecular C-O distances, suggesting a rather strong interaction between CO₂^•− and Cu atom. The DORI analysis reveals two covalent interaction sites. Moreover, this larger electron sharing between the two entities is reflected by a Mayer bond order and an ELF valence basin population of 0.85 and 2.5 electrons for Cu-C and 0.54 and 2.2 electrons for Cu-O, respectively. The total spin density of the systems at the ground state was computed and represented in Supplementary Fig. 12. In the case of AuNPs, the spin density of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ remains localized on CO₂^•−, so it remains a radical species due to a weakly covalent interaction. For Cu, the result highlights a large delocalization of the spin density as a sign of the stronger covalent interaction in ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$.

Fig. 4: The simulation of CO₂^•− radical stabilized on metal nanocatalysts.

a, b DFT optimized structures and DORI plots (repulsive forces in red and covalent interaction in blue) of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ (a) and ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ (b). c, d Electronic transition analysis (hole in blue and electron in green) of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ (c) and ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ (d). e, f Transient absorption spectra (at 800 µs) (e) and TD-DFT electronic spectra (f) of CO₂^•− and CO₂^•− stabilized on nanocatalysts. Source data are provided as a Source Data file.

The TD-DFT electronic spectra and the hole-electron framework reveal the global nature of absorption for electronic transitions⁵⁰. The absorption spectra of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ and ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ are found to exhibit a red-shift toward isolated CO₂^•− in solutions, with a band maximum at 358 nm for Au and 293 nm for Cu, in agreement with experimental observation (Fig. 4e). The hole and electron distributions well describe the electron transitions for two representative excited states for each complex in Fig. 4c, d. In the case of Au, the hole distribution is delocalized over NP. It overlaps with CO₂^•− for the most intense electronic transition at 359 nm, similar at 464 nm with a stronger CO₂^•− contribution. Compared with reported spectra of CO₂ coordinated on Au and Cu surfaces, these surface-bound CO₂^•− spectra all exhibited red-shift^51,52,53. The electron distribution during excitation is mainly delocalized at the surface of NP, reflecting a weak interaction between Au NP and CO₂^•− radical. However, ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ displays a considerable hole distribution of CO₂^•− radical at 299 and 362 nm, whereas the electron distribution remains localized on several Cu atoms at the surface. The greater localized hole-electron distributions of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Cu}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ disclose the stronger covalent interaction on Cu surfaces, which is consistent with previous reports^14,54.

Catalyst size and cation effect on surface-bound CO₂ ^•− radical

In catalytic CO₂RR systems, two essential factors require thorough consideration. First, the intrinsic activity of nanoscale catalysts depends strongly on the size distribution, surface topology, and surface composition. Second, the complex localized aqueous composition, particularly alkali metal cations, profoundly affects the CO₂RR mechanism route through electric double layers and interfacial hydration^55,56. The ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ of different sizes Au NP (1.7, 3.3, and 6.2 nm, Supplementary Fig. 13) exhibited a similar wide-range characteristic absorbance from 300 nm to 500 nm at 75 µs, but their intensity increases by reducing the size of nanocatalysts (Fig. 5a). It is evident that when the concentration of Au precursors for the preparation of the nanocatalyst is constant, the smaller size results in a higher concentration of Au NP; consequently, the larger specific surface area provides more active sites. The competition in transient kinetics of CO₂^•− radicals between dimerization to form oxalate and stabilization on Au surfaces is also size dependent. In the 6.2 nm Au system, the transient kinetics reflect mainly the decay of CO₂^•− radicals due to the recombination process (Fig. 5b, c). However, as the Au NP size reduces to 1.7 nm, the concentration of nanocatalyst increases almost 27 times, the decay presents an inflection around 7 µs, indicating that the ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ stabilization mechanism turns to dominate the competitive processes. As progressed to 750 µs, no apparent changes occur in the spectra shape of ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ radicals relative to 75 µs (Fig. 5d–f).

Fig. 5: The effect of catalyst size and cation in solutions on CO₂^•− radical stabilization process with Au.

a–c Transient absorption spectra of different sizes of Au (0.5 mM) at 75 µs (a) and transient kinetics at 350 nm (b) and 450 nm (c) within 80 µs. d–f Transient absorption spectra of different sizes of Au (0.5 mM) at 750 µs (d) and transient kinetics at 350 nm (e) and 450 nm (f) within 800 µs. g–i Transient kinetics at 350 nm with different concentrations of Li⁺ (g), Na⁺ (h), K⁺ (i). j–l Transient absorption spectra at 750 µs with varying concentrations of Li⁺ (j), Na⁺ (k), K⁺ (l). Source data are provided as a Source Data file.

More importantly, the increase in Au NP size reduces the intensity ratio between two bands (ΔOD_{350 nm}/ΔOD_{450 nm}) due to the greater density of surface-bound CO₂^•− radicals. DFT simulations were performed on the larger size of Au NP with more CO₂^•− radicals (Supplementary Fig. 14) to investigate this phenomenon. The geometry and orientation of CO₂^•− radicals on Au surface atoms are the same at interaction sites (the corner position being the most favorable) and similar to the above results for ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$. However, electronic transition analysis indicated the observation of additional states for ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ at higher wavelengths (especially above 400 nm), where the electrons are more delocalized at the surface of Au NP (Supplementary Fig. 14b). Other simulations on the same size of Au NP with two surface-bound CO₂^•− reflect that ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ tends to undergo electron transfer to Au surfaces instead of localized distribution for covalent interaction with Au NP (Supplementary Fig. 15). The aforementioned data show that the smaller Au NP not only provides more active sites for the stabilization reaction but also facilitates localized electron distribution among ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ radicals and Au surfaces, enhancing the surface-bound interaction.

As reported in the existing studies^57,58,59, three prominent theories exist regarding the primary function of metal cations during CO₂ reduction: modification of the (local) electric field, buffering of the interfacial pH, and the stabilization of intermediates (such as CO₂^•− radicals) through local field effects. To verify the above assumptions, we carried out experiments in solutions without sodium cations, the samples were centrifugated, and formic acid was used to adjust the concentration of proton. Compared with the above results in the presence of sodium cations, the kinetics of CO₂^•− radicals stabilizing on Au surfaces was retarded, and the absorption intensity of surface-bound CO₂^•− radicals on Au NPs remarkably decreased (Supplementary Fig. 16a–d), indicating that the stabilization process of CO₂^•− radicals on Au surfaces was undermined in the absence of metal cations. Moreover, transient spectro-kinetics data revealed that the lifetime of surface-bound intermediates without metal cations was curtailed regarding the sodium system (Supplementary Fig. 16e–g), suggesting that the primary role of cations is to stabilize the surface-bound CO₂^•− radicals. These observations are in total agreement with recent steady-state electrocatalysis studies^57,58,59. However, excess metal cations disturb the stabilization of intermediates through electric and nonelectric field aspects^60,61,62,63. Figure 5g shows that adding Li⁺ does not affect the transient kinetics of CO₂^•− at 350 nm. The same transient kinetics rule out the possible interference of Na⁺ from HCOONa during radiolytic preparation when replacing HCOONa with HCOOLi for radiolytic preparation (Supplementary Fig. 17a, b). The simulation of the complex ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{Au}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ with Li⁺ also led to a stable structure with an electrostatic interaction between cation and CO₂^•− while keeping a similar orientation with Au. While Fig. 5h, i indicates that elevated concentrations of both Na⁺ and K⁺ suppress the increase of transient absorption within the first 5–10 µs, after this period, the transient kinetics exhibits no dependence on cation concentration until 750 µs (Supplementary Fig. 17c–e). These results demonstrate that metal cations primarily interfere with the diffusion and initial stabilization process of CO₂^•− radicals on Au surfaces. According to previous reports^61,63, we infer that metal cations combine with CO₂^•− radicals through electrostatic interactions to form ion pairs (M^z+ + CO₂^•− ↔ (M^z+∙∙∙CO₂^•−)) that obstruct the stabilization of CO₂^•− radicals on Au surfaces. Also, the trend in transient absorption spectra intensity follows the order Li⁺ < Na⁺ < K⁺ for the same concentration (Fig. 5j–l), which correlates with the cation radius and supports the hypothesis of the pair formation. In contrast to the reported steady-state results, our finding fills in the missing information on the role of metal cations during the stabilization process of CO₂^•− radicals.

In summary, pulse radiolysis allows us to acquire a complete image of the stabilization process of CO₂^•− radicals on catalyst surfaces. Even if the pathway to form surface-bound CO₂^•− radical on the NPs is fundamentally different between electrocatalysis and radiolysis conditions, the final state (CO₂^•− bound on the NPs) is the same as whatever is produced by radiolysis or direct reduction of CO₂ on the surface of the electrode. The surface stabilization process of CO₂^•− radicals on Au and Cu surfaces includes the initial adsorption and the structure optimization process, which extend substantially the lifetime of (CO₂^•−)^ad radicals by at least 100 times compared to CO₂^•− radicals in solutions, thereby laying the groundwork for the subsequent multi-electron transfer reaction. Most notably, the respective (CO₂^•−)^ad radical transient kinetics and characteristic spectra across Au, Cu, and Ni systems suggest diverse stabilization behavior and structures of (CO₂^•−)^ad radicals on various metal surfaces. Extensive research in electrochemistry has resulted in categorizing metal electrodes into four distinct groups based on their primary products. The different selectivity of CO₂ reduction between metals has been explained by their binding energy to key intermediates, including *OCHO, *COOH, and *CO. Therefore, we assume that the stabilized structure and kinetics of the first key intermediate, surface-bound CO₂^•− on metallic sites, influence the catalytic properties and the product selectivity. Au, Ni, and Cu NPs distinctly represent the individual catalytic selectivity and activity of each group. We observed very clearly the difference between the three nanocatalysts. In the same conditions, we found that Ni cannot stabilize the CO₂^•− radicals, which corroborates its slight activity of CO₂ reduction reported in the literature; however, Au and Cu can stabilize the CO₂^•− radicals weakly and strongly, respectively, in agreement with the results obtained by steady-state electrocatalysis. The importance of our work lies not only in explaining the reaction mechanism in electrocatalysis but also in deciphering the function of the metal in CO₂ reduction. This selective stabilization mechanism provides deeper insight into the intrinsic mechanism of CO₂ reduction, extending from steady-state intermediates to transient elementary reactions. Our results are a significant advance in the development of time-resolved techniques for optimizing catalyst design and system performance.

Methods

Chemicals

Cetyltrimethylammonium chloride (CTAC, 98%), sodium formate (99%), copper(II) sulfate (98), nickel(II) chloride (98%), sodium tetrachloroaurate (99%), and cetyltrimethylammonium bromide (98%) were purchased from Sigma-Aldrich. All chemicals were used without further purification. Ultrapure water with a resistivity of 18.25 MΩ cm was obtained from a water purification system. CO₂ (99.999%) and Ar (99.999%) were purchased from Air Liquide Industrial Gases Company.

Synthesis of metal nanoparticles

All metal nanoparticles were synthesized by radiolytic methods: In a typical synthesis for Cu nanoparticles, 20 mmol sodium formate and 0.5 mmol CTAC were added to 10 mL H₂O. Then, 10 mL copper(II) sulfate (1 mmol L⁻¹) solution was added to form homogeneous solutions. The solutions were saturated with Ar and sealed in a homemade cuvette. The cuvette was irradiated at a ⁶⁰Co gamma source (5.35 × 10¹³Bq, located in the Université Paris-Saclay) at a dose rate of 2.2 kGy h⁻¹ at ambient conditions. The dose rate was calibrated using the Fricke dosimeter, and the total absorbed dose reached 6.6 kGy. The resulting Cu nanoparticle solution was directly used for pulse radiolysis experiments and absorption spectroscopy characterization.

For Ni or Au nanoparticles, the precursor was replaced with nickel(II) chloride or sodium tetrachloroaurate. For size modification, it was controlled by varying the concentration of CTAB, with higher concentrations resulting in larger particles. The synthesized 1.7, 5.3, and 6.2 nm Au nanoparticles correspond to 1, 2.5, and 10 mM CTAB concentration systems. The supernatant solution was centrifuged at 164,700×g for 24 h using an ultracentrifuge (Optima XE-90, BECKMAN COULTER) to remove the NPs.

Characterization

Particle morphology and size distribution were examined by high-resolution transmission electron microscopy (HR-TEM, JEOL 2100 PLUS, 150 kV). UV-vis spectra were recorded on a UV-spectrophotometer (HEWLETT PACKARD 8453, HP) in the range from 190 to 800 nm.

Pulse radiolysis

Pulse radiolysis experiments were carried out employing the picosecond laser-triggered electron accelerator, ELYSE, coupled with a time-resolved absorption spectrophotometric detection system³⁹. Laser (260 nm) driven Cs₂Te photocathode allowed the production of short electron pulses with a typical half width of 7 ps, a charge of ≈6 nC, and energy of ≈7.8 MeV at a repetition rate of 10 Hz. During irradiation, metal NP solutions were contained in a homemade cell with a path length of 1 cm. The diameter of the electron beam was 3 mm, and the irradiated volume was less than 0.1 mL.

Absorption spectral measurements were performed using the white light from a homemade Xenon flash lamp. The light was focused on the sample parallel to the electron beam with a smaller diameter and then directed onto a flat field spectrograph (250IS, Chromex), which disperses the light on the entrance optics of a high dynamic range streak-camera (C-7700-01, HAMAMATSU) to obtain an image resolved in wavelength and time. The kinetic data and absorption spectra were extracted from three series of 250 resulting images. In this work, the transient spectra were measured from 290 to 720 nm at 1 µs, 10 µs, 20 µs, 100 µs, and 1 ms⁴⁰.

Computational methods

DFT calculations were done using the ORCA 5.0.3 software⁶⁴. The geometry optimization calculations were performed using the PBE functional, the SDD pseudopotential (with 19 valence-electrons treated explicitly) for the metal atoms, and the def2-TZVP basis set for carbon and oxygen atoms. Dispersion corrections were added to the functional used in the D3 framework proposed by Grimme with the addition of the Becke–Johnson damping (D3BJ) in all cases⁶⁵. ${({{{{{{{\rm{CO}}}}}}}_{2}}^{{{\bullet }}-})}_{{{{{{{\rm{M}}}}}}}_{}}^{{{{{{\rm{ad}}}}}}}$ (M for Au and Cu) has been designed as a model of a metallic NP in interaction with one CO₂^•− radical with solvent effects represented by a dielectric continuum model. The orientation and nature of the interaction between the radical and NP have been probed by several quantum chemical analyses. Numerical frequencies were calculated to ensure the structures corresponded to energy minima. For Au and Cu, the 38-atom nanoparticle is considered in the singlet spin state, and the complex with the radical results in a doublet spin state. The CPCM implicit solvent model was used to represent the water environment. The electronic spectra were computed in the TD-DFT framework using the CAM-B3LYP functional and the implicit solvent. Between 600 and 800 excited states were included in the calculations to obtain the spectra of the different metal systems. The TD-DFT spectrum of the isolated CO₂^•− radical was obtained at the CAM-B3LYP/aug-cc-pVTZ level by calculating 200 excited states. The structure of the 92-atom Au NP with 5 CO₂^•− radicals was optimized at PBE-D3BJ level (with the SVP basis set for C and O atoms) using an implicit solvent. A simplified calculation of the electronic spectrum was computed in vacuum using the simplified sTDA⁶⁶ approach and the wB97X functional. The quantum chemical topological and electronic transition analyses were performed with the Multiwfn code by analyzing the electron density generated from ORCA⁶⁷. The analyses of both covalent and non-covalent interactions were calculated based on the Density Overlap Regions Indicator (DORI)⁶⁸. The electron localization function (ELF) was calculated to determine the valence basins⁶⁹.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.