In-situ structural evolution of Bi2O3 nanoparticle catalysts for CO2 electroreduction

Under the complex external reaction conditions, uncovering the true structural evolution of the catalyst is of profound significance for the establishment of relevant structure–activity relationships and the rational design of electrocatalysts. Here, the surface reconstruction of the catalyst was characterized by ex-situ methods and in-situ Raman spectroscopy in CO2 electroreduction. The final results showed that the Bi2O3 nanoparticles were transformed into Bi/Bi2O3 two-dimensional thin-layer nanosheets (NSs). It is considered to be the active phase in the electrocatalytic process. The Bi/Bi2O3 NSs showed good catalytic performance with a Faraday efficiency (FE) of 94.8% for formate and a current density of 26 mA cm−2 at −1.01 V. While the catalyst maintained a 90% FE in a wide potential range (−0.91 V to −1.21 V) and long-term stability (24 h). Theoretical calculations support the theory that the excellent performance originates from the enhanced bonding state of surface Bi-Bi, which stabilized the adsorption of the key intermediate OCHO* and thus promoted the production of formate.


Introduction
In recent years, the conversion and utilization of renewable energy have become particularly significant for making progress in environmental and energy sustainable development [1]. However, the use of wind and solar energy with intermittent characteristics will lead to high electricity storage costs * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. [2]. Electrocatalytic CO 2 reduction reaction (CO 2 RR) technology to convert the greenhouse gas CO 2 into value-added products for carbon neutrality and energy storage has received considerable attention [3][4][5]. Among the various products of CO 2 RR, the liquid product formic acid has been noticeable. Formic acid is not only an important raw material in chemical production but also can be used as an excellent hydrogen storage material, and it also has wide application potential in fuel cells [6,7]. This makes formic acid/formate one of the most competitive high-value products among the many products of CO 2 RR. However, due to the chemical inertness of the CO 2 molecules and the strong competition with hydrogen evolution reaction (HER), both the product selectivity and the activity of existing catalysts are not satisfactory [8].
Although noble metal-based catalysts have good catalytic performance in formic acid production, their high toxicity or rarity hinders the development of commercialization. Recent studies have suggested that cheap metals represented by Sn and Bi were more widely used in the conversion of CO 2 to formic acid [9]. However, the poor formic acid selectivity of Sn-based catalysts has impeded its further development. In contrast, bismuth-based catalysts were more commonly used in CO 2 RR, and huge progress has been made because of their large reserves, green pollution-free characteristics [10][11][12][13].
With the further progress of the CO 2 RR research, researchers gradually realized that such metal compounds underwent dynamic structural evolution in the reduction process, which could lead to the uncertainty of active species. The structural evolution of the catalyst in the reaction was affected by external potential, electrolytes, and other factors [14][15][16]. Previous studies have demonstrated that the surface of InN nanosheets (NSs) was reconstructed in the CO 2 RR, which hugely affected the catalytic activity of the catalyst [17]. For Bi-based oxides, the dynamic evolution of catalysts at cathodic potential still remains unclear. The CO 2 RR performance of the reported Bi NSs with the in-situ topological transformation of BiOI NSs was excellent [18]. Compared to the zerovalent Bi metal, the presence of oxygen species in the real reaction environment and the potential role of possible metal/metal oxide evolution configuration in the CO 2 reduction process were rarely taken into account [19][20][21][22][23]. More importantly, the micromorphology and structure of the catalyst were greatly restricted by external environmental conditions [24,25]. There is an urgent need to employ the in-situ spectroscopic characterizations to track its real evolution process [26][27][28] and assist in establishing a real active structure model. The above is of great significance for the study of the structure-activity relationship and determining the reaction active sites in the CO 2  Density functional theory (DFT) results indicated that the electrochemical reconstruction process can lead to the charge redistribution, especially when the presence of the Bi 2 O 3 substrate provides more electrons that can be transported to the vicinity of the surface Bi layer. This enhances the bonding strength and helps to form a stable electron-rich Bi metal layer. It optimizes the activation behavior of CO 2 molecules and reduces the energy barrier of subsequent the OCHO * intermediates, which ultimately greatly promotes the production of formate. The above is of great significance to guide the rational design of catalysts with more stable and excellent performance.

Results and discussion
As shown in figure 1(a), Bi 2 O 3 NPs were synthesized by a simple hydrothermal method. Both scanning electron microscope (SEM) and transmission electron microscope (TEM) found that the particle size of the synthesized Bi 2 O 3 NPs was in the range of 50-100 nm (figure S1). The high-resolution TEM (HRTEM) ( figure 1(b)) gave the information that the 0. With the aim of reflecting the dynamic evolution of the catalyst surface during the CO 2 RR process, Bi 2 O 3 NPs, Bi NPs, and Bi 2 O 3 NSs were tested by the in-situ Raman detection (figure S6). A low-energy 633 nm laser (light intensity: strength 5%) excited by He-Ne laser was used to eliminate laser oxidation caused by thermal effect [29]. In the open circuit potential, the Bi 2 O 3 NPs had two low-frequency stretching vibrations Raman peaks at 65 and 167 cm −1 [30,31]. After applying the bias voltage, the peak at 65 cm −1 was gradually replaced by the new peaks at 71 and 98 cm −1 , which were consistent with the typical two first-order optical bands pattern of metal Bi [32]. The peak at 167 cm −1 , although attenuated, still remained exist and then increased slightly after removing the potential. Significant change of the Roman Peak was not found with the extension of reaction time (figure 2(a)). The continuous maintenance of Bi-O vibrations was also verified by Raman spectroscopy after standard electrochemical testing time (2 h) (figure S7). Then, the test was repeated by applying different potentials within the potential window used immediately, resulting in close peak positions (figure 2(b)). Furthermore, the in-situ Raman results of Bi NPs exposed that the intensity of Bi Raman peak remained stable regardless of the change of applied potential or the extension of cathode time (figure S8), and the particles slightly aggregated under cathode current (figure S9). A minor damage was found on the morphology of the NSs after the reaction of Bi 2 O 3 NSs (figure S10). However, the vibration of Bi-O disappeared after applying a bias voltage. With the increase in the reaction time, the 98 cm −1 characteristic peak remained weak (figure S11). Therefore, it can be demonstrated that the quality of Bi crystal was poor, which may have an adverse impact on CO 2 RR. In conclusion, the in-situ Raman test more clearly showcased the dynamic evolution process of Bi 2 O 3 NPs. Also, the transformation from Bi 2 O 3 NPs to Bi/Bi 2 O 3 NSs was determined in combination with the ex-situ test. X-ray photoelectron spectroscopy (XPS) further explained the changes in the elemental composition and valence state of the catalyst surface before and after the electrochemical process. The highresolution spectrum of Bi 4f (figure 2(c)) showed that the peaks at 164.2 eV and 158.9 eV were assigned to the Bi 3+ of the Bi 2 O 3 NPs component. In the deconvoluted Bi 4f spectrum of Bi/Bi 2 O 3 after electroreduction, the peaks position at 163.6 eV and 158.3 eV with lower binding energy can be attributed to the generation of metal Bi and oxygen vacancies [33]. In addition, the peaks at 529.5 eV and 531.3 eV in the O1s spectrum of Bi 2 O 3 NPs are attributed to the lattice oxygen in Bi-O and the adsorbed oxygen in Bi-OH respectively (figure S12). It is obvious that the ratio of lattice oxygen decreased with the increase of ratio of absorbed oxygen in Bi/Bi 2 O 3 . Bridging hydroxyl groups were usually associated with surface oxygen vacancies [34,35], while the Bi-O peak position shifted to the lower binding energy by about 0.1 eV, which also confirms this view [36]. In the meantime, the Mott-Schottky curve (figure 2(d)) also showed a negative shift in the flat band potential (V fb ) of the reconstructed Bi/Bi 2 O 3 , which may be related to the migration of electrons to the Bi surface. The smaller slope of the Bi/Bi 2 O 3 curve represented an increase in carrier concentration [2], which indicated that the reconstructed Bi/Bi 2 O 3 will exhibits excellent electron transfer capabilities during the CO 2 electroreduction process.
Due to the dynamic evolution of Bi 2 O 3 NPs in the electrochemical process (figure S13), the active phase Bi/Bi 2 O 3 and stable Bi NPs were used as the main specimens to evaluate the performance of CO 2 ERR. The performance indicators of stabilized Bi 2 O 3 NSs and commercial Bi powder were also given as a reference (figures S14 and S15). The linear sweep voltammetry curve (figure 3(a)) was measured in 0.5 M NaHCO 3 aqueous solution saturated with Ar/CO 2 . In the Ar-saturated electrolyte, the HER process occurs; while in the CO 2 -saturated electrolyte, the significantly increased current can be attributed to the CO 2 RR process. By comparing the current density values, it can be preliminarily concluded based on the result that Bi/Bi 2 O 3 has stronger CO 2 reduction ability. Furthermore, the constant potential electrolysis behavior of the catalyst in the CO 2 saturated NaHCO 3 solution was also investigated. The gas and liquid  phase products produced in the cathode compartment were detected by gas chromatography and nuclear magnetic resonance spectroscopy ( figure 3(b)). The onset potential of Bi NPs and Bi/Bi 2 O 3 for formate production was −0.71 V vs. RHE, while the onset potential of commercial Bi powder was −0.81 V vs. RHE (figure S15). With the increase of applied bias voltage, the Faraday efficiency (FE) of formate increased rapidly; at −1.01 V vs. RHE, the FE of formate for Bi NPs and Bi/Bi 2 O 3 reached the maximum value at the same time, which was close to 94.8% and 79.8% with current density ∼26 mA cm −2 and ∼14 mA cm −2 . The excellent electrochemical performance of Bi/Bi 2 O 3 exceeded that of most previous Bi 2 O 3 based catalysts (table S1). However, the FE of formate for Bi 2 O 3 NSs and Bi powder was only 74% and 70.1%, respectively ( figure 3(c)). Subsequently, the FE of formate for Bi NPs dropped rapidly to below 60% after crossing the optimal potential. While the FE of formate for Bi/Bi 2 O 3 was still close to 90% at −1.21 V vs. RHE, and remained a high formate selectivity within a large potential window (−0.91 V < E < −1.21 V). The CO 2 reduction ability of the catalysts can be demonstrated more comprehensively by comparing the FE of the main carbonaceous products (HCOO − , CO), which can clearly reflect that Bi/Bi 2 O 3 has a stronger CO 2 RR efficiency in the potential window involved (figure S16). The potential-dependent formate partial current diagram (figure 3(d)) can clearly show the above changes. A detailed description was provided that the formate partial current of Bi NPs hardly increases after exceeding the optimal potential, whereas Bi/Bi 2 O 3 increases slowly. Surprisingly, there was a small amount of acetate in the liquid phase product at a higher potential but further discussion will not be presented in this paper (figure S17). The FE of gas phase product diagram (figures 3(e) and S18) demonstrates the lower CO 2 catalytic ability of Bi NPs due to its prominent HER process. In addition, the reaction kinetics of CO 2 RR can be showcased by the Nyquist diagram (figure 3(f)) and Tafel curve (figure 3(g)). It is relatively apparent that the Bi/Bi 2 O 3 structure showed a faster electron transport capacity and lower Tafel slope (161 mV dec −1 ). This supported the conclusion that Bi NPs (266 mV dec −1 ) need higher overpotential to realize the electron transfer process. At high potential, the higher Tafel slope of catalyst was caused by the low solubility of CO 2 in the electrolyte, which may reach the mass transfer limit in the electrolysis process [37]. The specific BET surface area and electric double-layer capacitance (C dl ) also helped support the theory that the Bi/Bi 2 O 3 configuration was beneficial for exposing more active areas to participate in the catalytic reaction (figures 3(h) and S19). The stability of the Bi/Bi 2 O 3 reaction was further evaluated (figure 3(i)), and the current density remained relatively stable in the 24 h electrolysis test. It is worth noting that there was a slight decrease on the FE of formate, but it still remained above 90%. After the reaction, the catalyst remained relatively stable (figure S20). It is evident that Bi/Bi 2 O 3 structure formed in-situ by Bi 2 O 3 NPs can effectively promote the CO 2 RR, and maintain high FE of formate with reasonable stability over a wide potential window. In order to explore the origin of the structural stability and the activity of Bi/Bi 2 O 3 (figure S21) during the electroreduction, crystal orbital Hamilton population (COHP) analysis (figure 4(a)) was used to study the Bi-Bi bonding state on the catalyst surface by DFT calculation. The region with -COHP greater than 0 represents the bonding state, while the other regions are in the anti-bonding state. The displayed integrated crystal orbital Hamiltonian population (ICOHP) value is integrated up to E f , which is an effective way to measure the bond strength [38]. As shown in the blue shaded area near the E f in the figure, the anti-bonding state in the Bi configuration is transformed into a bonding state filled with a large number of electrons in Bi/Bi 2 O 3 , which is favorable for the bonding between Bi atoms. In contrast, the lower ICOHP (−2.53 eV) delivers a more stable bonding state. It is also consistent with the conclusion of the in-situ Raman detection that a stable Bi structure based on Bi 2 O 3 is formed during the electrochemical process. Similarly, the adsorption capacity between intermediate and catalyst in the reaction process was closely related to the change of the electronic structure of the catalyst surface, which will significantly affect the catalytic performance. The electron localization function (ELF) can clearly show the localization of the charge around the atom ( figure 4(b)). In the Bi configuration, the strong localized charges were spherical around the atoms and have little overlap with the surrounding atoms. In Bi/Bi 2 O 3 configuration, the charges were no longer localized, but instead pulled each other to form a flat ellipsoid. The intercepted 2D picture (figure S22(b)) also demonstrates a similar trend, which represents that charge transfer occurred between the Bi layer and the Bi 2 O 3 carrier. Afterward, the surface differential charge density (figure S22(d)) indicates that the outermost Bi atoms in Bi/Bi 2 O 3 gained more electrons (0.175 |e|, surface average) from the Bi 2 O 3 carrier than the Bi configuration (0.005 |e|), which is consistent with the negative shift in the binding energy of Bi atoms observed in XPS. This suggests that the electron-filled Bi layer in Bi/Bi 2 O 3 can become the activity center for the subsequent chemical reaction steps, thereby better promoting the reaction. It is also found that the charge in the Bi/Bi 2 O 3 configuration accumulates strongly between adjacent Bi atoms to better facilitate the bonding, which is consistent with the COHP conclusion. Interestingly, the Tafel value of Bi/Bi 2 O 3 is closer to 118 mV dec −1 , which means that the initial single-electron transfer step CO 2 •− is a rate determining step (RDS) [39,40] and is closely related to the initial adsorption process of CO 2 molecules on the surface. Consequently, partial density of state (PDOS) (figure 4(c)) of the model was calculated using the hybrid function HSE06 to further explore that the active electron density center of Bi/Bi 2 O 3 is closer to the Fermi level (E f ) than Bi. This suggests that the Bi/Bi 2 O 3 model has more abundant active states near the E f , which will contribute to the interaction between CO 2 molecules and the catalyst surface, thereby promoting the charge transfer between each other [41,42]. In a word, Bi 2 O 3 as a catalytically active carrier can adjust the electronic structure state of the surface Bi nanocrystalline to form a more stable surface electronic structure with abundant active electrons, which can help facilitate the interaction between intermediate and catalyst surface to promote the smooth reaction.
The reaction path of CO 2 RR was analyzed by DFT calculation to understand the high selectivity of formate for Bi/Bi 2 O 3 (figures 5(a) and (b)). The adsorption configurations and Gibbs free energy (G) values of the key intermediates involved were used as a reference (figures 5(c) and S23). For the two-electron process of the HCOOH production pathway, the first step proton coupled electron transfer of OCHO * was the RDS in this reaction pathway. The G OCHO * (0.65 eV) of the Bi/Bi 2 O 3 configuration was significantly lower than the G OCHO * (1.29 eV) of the Bi configuration. The smaller the G was, the higher the product activity was. For further understanding, the differential charge density ( figure 5(d)) of the OCHO * on the catalyst surface indicates that a large number of electrons are consumed between the O atom in OCHO * and the Bi atom on the catalyst surface, which helps facilitate the bonding and adsorption on the catalyst surface. Compared to the Bi configuration, OCHO * (0.699 |e|) on the Bi/Bi 2 O 3 configuration obtained more electrons and the bond length of Bi-O was smaller (2.35 Å), which is related to the abundant active electrons on the surface ( figure S22). For the Bi/Bi 2 O 3 , the blue shading area in the PDOS (figure 5(e)) indicates that there is a stronger harmonic overlap between the p orbital of the O atom in the OCHO * and the p orbital of the Bi atom [2]. All results mentioned indicate that the adsorption of OCHO * on the Bi/Bi 2 O 3 surface is tighter, and thereby reducing the over potential of the reaction pathway. Moreover, other competitive reaction paths of Bi/Bi 2 O 3 were also calculated ( figure 5(b)). It can be clearly demonstrated that OCHO * , COOH * , and H * are regarded as the RDS affecting the HCOOH, CO, and H 2 reaction pathways. It was difficult to produce CO due to the high G COOH * . Similarly, G H * was higher than G OCHO * , which supports the theory that the HER activity is lower. Finally, the product of the electroreduction reaction can be considered as HCOOH, which is in good agreement with the experimental results and therefore explains the source of the high selectivity of HCOO -. Overall, the Bi/Bi 2 O 3 interface optimizes the surface charge to make it easier to contact OCHO * . Therefore, the reaction barrier decreases, and formate becomes the final dominant product.

Conclusion
In summary, the dynamic evolution of the Bi 2 O 3 NPs in the cathode environment is detected by the in-situ Raman spectroscopy. With the 2D trend of the morphology and Bi nanocrystals appearing on the surface, the in-situ reconstructed Bi/Bi 2 O 3 NSs participate in the CO 2 RR process as the real active phase. At −1.01 V vs. RHE, the FE of formate reaches 94.8% and maintains high-efficiency CO 2 RR over a wide potential window. The DFT calculation results show that the reconstruction process leads to the accumulation of surface charges, which not only enhances the bonding strength of the surface Bi layer, but also optimizes the absorption of the key intermediate OCHO * to promote the formation of formate. This work explores the structural evolution of the catalyst under cathode conditions and constructs a model to understand the origin of the activity, which provides a reference for the rational design of the CO 2 reduction electrocatalyst and the future theoretical research of the activity mechanism.