In situ electrochemical reconstruction of Sr2Fe1.45Ir0.05Mo0.5O6-δ perovskite cathode for CO2 electrolysis in solid oxide electrolysis cells

ABSTRACT Solid oxide electrolysis cells provide a practical solution for the direct conversion of CO2 to other chemicals (i.e. CO), however, an in-depth mechanistic understanding of the dynamic reconstruction of active sites for perovskite cathodes during CO2 electrolysis remains a great challenge. Herein, we identify that iridium-doped Sr2Fe1.45Ir0.05Mo0.5O6-δ (SFIrM) perovskite displays a dynamic electrochemical reconstruction feature during CO2 electrolysis with abundant exsolution of highly dispersed IrFe alloy nanoparticles on the SFIrM surface. The in situ reconstructed IrFe@SFIrM interfaces deliver a current density of 1.46 A cm−2 while maintaining over 99% CO Faradaic efficiency, representing a 25.8% improvement compared with the Sr2Fe1.5Mo0.5O6-δ counterpart. In situ electrochemical spectroscopy measurements and density functional theory calculations suggest that the improved CO2 electrolysis activity originates from the facilitated formation of carbonate intermediates at the IrFe@SFIrM interfaces. Our work may open the possibility of using an in situ electrochemical poling method for CO2 electrolysis in practice.


INTRODUCTION
Converting CO 2 into valuable chemicals plays a key role in carbon capture, utilization and storage technology, which is expected to achieve a carbonneutral sustainable energy economy [1,2]. CO 2 electrolysis via solid oxide electrolysis cells (SOECs) is a promising way to store renewable electricity into chemical energy and is attracting widespread interest [3][4][5]. Owing to their high electronic/ionic conductivity and good redox stability, perovskite oxides (ABO 3 ) are considered as good candidates for electrode catalysts in SOECs. The construction of active metal/oxide interfaces is effective in improving the electrocatalytic activity [6][7][8][9], however, metal nanoparticles (NPs) deposited by infiltration are inevitably deactivated due to sintering, causing irreparable performance degradation at elevated temperatures [10,11]. In addition, the complex fabrication process is another disadvantage that limits its application in practice [12].
Surface self-reconstruction under a reducing atmosphere is an alternative feasible strategy to manipulate active metal NPs via in situ exsolution from the perovskite oxide lattice to the surface [13][14][15][16]. Benefiting from the robust interaction between exsolved metal NPs and the parent support, the exsolved metal NPs exhibit superior coking and sintering resistance during long-term operation at elevated temperatures [17,18]. Numerous studies have shown improved CO 2 electrolysis performance over the metal/oxide interface [3,19,20]. However, it is still relatively lengthy (taking several hours) in a chemically reducing atmosphere owing to the relatively slow diffusion speed of metal cations across the bulk and surface [21]. In situ electrochemical reconstruction via voltage-driven exsolution has been proven to be an efficient method for producing abundant nanostructures with much faster reaction kinetics [21][22][23]. Irvine et al. reported a fast surface reconstruction process via electrochemical poling of solid oxide fuel cells at 2 V [22]. However, the application of electrochemical reconstruction in SOEC cathodes for CO 2 electrolysis has rarely been reported so far. There is a poor understanding of the dominant active sites formed during CO 2 electroreduction and their roles in the formation of adsorbed intermediates. Thus, it is highly desirable to investigate the dynamic structure evolution, and figure out the coherent structure-activity correlation with the assistance of in situ electrochemical spectroscopy techniques during the CO 2 electrolysis process.
In this work, we demonstrate that the Sr 2 Fe 1.45 Ir 0.05 Mo 0.5 O 6-δ (SFIrM) perovskite cathode displays a dynamic electrochemical reconstruction feature during CO 2 electrolysis by exsolving abundant IrFe alloy NPs anchored on the SFIrM surface. The dynamic electrochemical reconstruction feature is well investigated using in situ electrochemical X-ray diffraction (XRD), nearambient pressure X-ray photoelectron spectroscopy (NAP-XPS) and X-ray absorption spectroscopy (XAS). The in situ reconstructed IrFe@SFIrM interfaces deliver a current density of 1.46 A cm −2 while maintaining over 99% CO Faradaic efficiency, representing a 25.8% improvement compared with the Sr 2 Fe 1.5 Mo 0.5 O 6-δ (SFM) counterpart. In situ NAP-XPS studies and density functional theory (DFT) calculations suggest that the activity improvement originates from the facilitated formation of carbonate intermediates at the IrFe@SFIrM interfaces. Furthermore, self-regeneration of IrFe alloy NPs via redox manipulations could constrict particle agglomeration, improving the activity and operation stability of CO 2 electrolysis.

RESULTS AND DISCUSSION
The in situ electrochemical reconstruction process of the SFIrM catalyst occurred rapidly at a low applied voltage during CO 2 electrolysis at 800 • C. As shown in Fig. S1a, the as-prepared SFIrM possessed a smooth morphology and a uniform element distribution. During the electrochemical reconstruction and activation process under constant voltage mode at 1.0 V for ∼4000 s, the current density gradually increased and finally approached a steady state. Scanning transmission electron microscopy (STEM) images demonstrate that the exsolved metal NPs display high dispersion with an average size of ∼1.0 nm and a density above 80 000 μm −2 on the surface (Fig. S1b-d), which is far in advance of other reduction-and polarization-treated catalysts as listed in Table S1 [20][21][22][23][24][25][26][27][28][29]. Moreover, the activation time is reduced rapidly with increasing applied voltage (250 s at 1.2 V, 40 s at 1.4 V, and 20 s at 1.6 V), with similar particle size and density of the exsolved metal NPs during CO 2 electrolysis (Fig. 1a-c; Fig. S1). The higher applied voltage shows a stronger reduction ability for the exsolution of IrFe alloy NPs, and the final equilibrium is controlled by thermodynamics and the amount of the doped Ir, reflecting a similar distribution of IrFe alloy NPs. These results demonstrate that the electrochemical reconstruction of SFIrM is highly efficient in exsolving abundant IrFe alloy NPs under CO 2 electrolysis compared with other voltage shock-triggered exsolutions, which usually operate in harsh conditions (H 2 atmosphere and high applied voltage, Table S1) [21,23].
In situ XRD measurements were performed to clarify the in situ electrochemical reconstruction feature ( Fig. S2a and b). The as-prepared SFIrM catalyst showed a well-preserved double perovskite phase (Fig. S2c, PDF#04-019-7501). When the dynamic structure evolution occurred under the applied voltage, the characteristic peaks gradually shifted to lower degrees ( Fig. 1d; Fig. S2d and e), demonstrating the reduction of SFIrM via oxygen release during CO 2 electrolysis. Meanwhile, two additional peaks at 30.6 • and 43.8 • could be observed ( Fig. 1d and e), which corresponded to the newly formed Ruddlesden-Popper perovskite (RP-SFIrM, PDF#01-075-3655) and the IrFe alloy phase, respectively. Moreover, no further changes could be observed when increasing the applied voltage, which was consistent with the STEM results. These results demonstrate that the voltage-driven surface activation combined with the reconstruction process is fast and that the metal/perovskite interfaces are stable during CO 2 electrolysis.
In situ NAP-XPS measurements were performed to monitor surface metal valence states during CO 2 electrolysis ( Fig. 1f and g; Fig. S3a and b), and the oxidation states of Ir and Fe cations were detected in the vacuum test chamber. Upon an applied voltage of 1.2 V across the electrolysis cell, an additional broader peak at ∼60 eV assigned to the metallic Ir species could be observed ( Fig. 1f and Fig. S3c). Simultaneously, a shoulder peak at ∼707.1 eV assigned to the metallic Fe species was also observed ( Fig. 1g; Fig. S3d and e). When the applied voltage was switched off, the signals of Fe 0 and Ir 0 gradually weakened as the exsolved IrFe alloy NPs were re-oxidized by CO 2 ( Fig. 1f and g, Fig. S3f, and Table S2). When the applied voltage was switched back on and the CO 2 electrolysis process proceeded, metallic signals of Ir 0 and Fe 0 were generated again. In situ NAP-XPS results further confirmed the dynamic electrochemical reconstruction on the SFIrM surface via in situ exsolution of IrFe alloy NPs, which behaved as the catalytically active sites (IrFe@SFIrM interfaces) for CO 2 electrolysis.
To shed light on the coordination environment change during CO 2 electrolysis, in situ XAS measurements at the Ir L 3 -edge were performed (Fig. S4). The absorption energy of the pristine SFIrM cathode in a CO 2 atmosphere was located between Ir black and IrO 2 , indicating the partial oxidation of the Ir cations (Fig. 1h, red line). When CO 2 electrolysis was performed at 1.4 V, the white line peak exhibited a decrease in intensity, and a slight edge shift could be observed toward lower energy (blue line), indicative of a reduction in Ir cations under electrochemical polarization. (See online supplementary material for a colour version of this figure). Meanwhile, the Fourier-transformed extended X-ray absorption fine-structure spectra (EXAFS, Fig. 1i) demonstrated the formation of the Ir-Ir scattering path in the electrochemically polarized SFIrM cathode. The distinguished second shell structure at 2.7Å from those of IrO 2 implied the coordination environment change of the Ir center, which could be attributed to the scattering path between Ir and Fe [30][31][32]. Both newly added scattering paths demonstrated the formation of IrFe alloy phase during CO 2 electrolysis, in line with the STEM, in situ XRD and NAP-XPS results.
The reconstructed catalyst was confirmed with RP-SFIrM and exsolved IrFe alloy phases in STEM images ( Fig. 2a and b). The lattice spacing of 0.208 nm could be indexed to the exsolved IrFe alloy phase, and the reconstructed RP-SFIrM perovskite support showed a tetragonal structure with a lattice spacing of 0.283 nm that is attributed to its (105) plane. STEM-energy dispersive spectroscopy (EDS) elemental analysis confirmed that the exsolved NPs were IrFe alloys with an atomic ratio of ∼1 : 1 ( Fig. 2c and Fig. S5). The exsolved IrFe alloy NPs are partially submerged into the bulk substrate, which may endow a high stability and catalytic activity at high temperatures (Fig. 2a-c) [25,33,34]. To understand the exsolution mechanism of Ir and Fe on the SFIrM surface, DFT calculations were performed to simulate the surface segregation of Ir and Fe. As shown in Fig. 2d and Fig. S6, the segregation energies of Ir and Fe are −1.70 eV and −0.95 eV, respectively, suggesting that the segregation of Fe is less energetically favored than that of Ir. Meanwhile, oxygen vacancy (V O ) formation plays a vital role in the segregation of Ir and Fe from the bulk phase to the surface [13]. The segregation of Ir leads to easier formation of V O in bulk by reducing the formation energy from 0.25 eV to −0.47 eV (Fig. S6b). More V O facilitates the exsolution of Fe by decreasing the segregation energy to −1.38 eV, demonstrating a sequential reduction of Ir and Fe in the exsolution of IrFe alloy NPs.
The SFIrM cathode was then measured in an electrolyte-supported electrolysis cell for activity evaluation. Figure 3a displays the linear sweep voltammetry (LSV) curves for CO 2 electrolysis. Then, successive reduction treatments in H 2 and polarization at 1.6 V for CO 2 electrolysis were operated. The three overlapping LSV curves (Fig. S7a) indicated that the electrochemical reconstruction via in situ exsolution of IrFe@SFIrM interfaces had been completed after the first LSV measurement ( Fig. S5c and d). For comparison, the CO 2 electrolysis performance of the SFM cathode was also measured under the same conditions. The current density of CO 2 electrolysis was increased from 1.16 A cm −2 for SFM to 1.46 A cm −2 for SFIrM at 800 • C and 1.6 V. Considering the use of the same anode and electrolyte membrane, it can be reasonably speculated that the electrochemically reconstructed IrFe@SFIrM interfaces on SFIrM catalyst are responsible for the improved electrocatalytic performance.
SFM-and SFIrM-based cells were subjected to electrochemical impedance spectroscopy (EIS) analysis ( Fig. 3b and Fig. S7). The ohmic resistance (R o ) represents the total ionic and electronic resis-tance coming from the electrolyte and electrode. The polarization resistance (R p ) values of SFIrM-based cell are comparably smaller than those of SFMbased cell. The low R p values expound fast cathode kinetics for CO 2 activation, which confirms the high catalytic activity via in situ electrochemical reconstruction. The distribution function of relaxation times (DRT) was applied to analyze the elementary kinetic process of EIS data. As shown in Fig. 3c, four peaks labeled P1 to P4 were distinguished. Due to the same electrolyte and anode materials, oxygen ion transportation (P1), oxygen evolution reaction process (P2), and gas diffusion process (P4) were considered to be identical [35,36]. The difference between SFM and SFIrM cells was manifested in the P3 process ( Fig. 3c; Fig. S7c and d, and Table  S3), which was ascribed to the CO 2 adsorption and activation process, including charge transfer and intermediate species migration [3,37]. The electrochemical reconstruction provided more catalytic sites that facilitated CO 2 adsorption and activation, resulting in the accelerated P3 process that represents a higher CO 2 electrolysis performance.
In situ NAP-XPS measurements were employed to monitor their catalytic process during CO 2 electrolysis to reveal the intrinsic reactive mechanism ( Fig. 3d-f). A sharp peak at ∼292.9 eV was identified as gaseous CO 2 under a CO 2 atmosphere without additional polarization. Upon application of electrochemical polarization, an additional broader peak appeared at ∼289.3 eV in the C 1s spectra (Fig. 3d), which could be attributed to carbonate species and was most probably decisive for CO 2 electrolysis [38][39][40][41]. Meanwhile, its counterpart in the O 1s spectra was also observed as a shoulder peak at ∼532.2 eV (Fig. 3e; Fig. S8a and b), which is related to the oxygenated carbon species for the SFIrM cathode [41]. The peak area of carbonate species on the SFM cathode was weaker than that on the SFIrM cathode ( Fig. 3g and Fig. S8c). The enhanced carbonate intermediate species signal confirmed that the in situ reconstructed IrFe@SFIrM interfaces facilitated CO 2 adsorption and activation. IrFe@SFIrM interfaces were thus proposed as the catalytically active sites that were devoted to a higher CO 2 electrolysis performance than SFM (Fig. 3a).
When the applied voltage was switched off, the carbonate signal was weakened due to the reoxidation of IrFe alloy NPs by CO 2 , which is consistent with the results in Fig. 1f-i. IrFe alloy NPs were regenerated when the applied voltage was switched on again, and the appearance of the intermediate carbonate peak was further obtained (Fig. 3d). Since the carbonate species only existed under sufficient electrochemical polarization but immediately vanished upon retracting the bias, an observation of the decisive intermediate carbonate species is clearly only possible by means of in situ NAP-XPS, as employed in this work.
The stability and self-regeneration feature of IrFe@SFIrM interfaces under operational conditions were investigated (Fig. 4a and Fig. S9). The decay rate of the SFIrM cathode-based cell during the whole 210 h stability test was 0.015% h −1 , which was much smaller than that of the SFM cathodebased cell (0.14% h −1 ) and was also comparable with the optimal values in previously reported literatures (Table S4). Interestingly, CO 2 electrolysis performance could rebound after a self-regeneration process [42,43]. The initial exsolved IrFe alloy NPs were re-dispersed sufficiently into smaller nanoclusters with an average size of ∼0.9 nm via a brief oxidation treatment in air for ∼3 min (Fig. S10), which facilitated richer IrFe alloy NPs when the external voltage was switched on again. The recovered metallic IrFe and carbonate peaks confirm that the cathode has been regenerated ( Fig. 4b and  c; Fig. S11). Therefore, more abundant catalytic IrFe@SFIrM interfaces were electrochemically generated, resulting in a rebounded CO 2 electrolysis performance (Fig. 4a). However, the regenerated ultrafine IrFe alloy NPs may not be very stable during the CO 2 electrolysis process. There is a process of optimizing surface (IrFe alloy) energy and interface (IrFe@SFIrM) energy, which results in the slow growth of NPs and hence in the degradation process of cell performance at the initial stage after oxidative regeneration. Although the slow aggregation of NPs is inevitable at high temperatures, the oxidative re-dispersion strategy could efficiently improve the stability by delaying particle aggregation (Fig. S12a-c).
After the stability test, the average size of the exsolved IrFe alloy NPs was ∼2.0 nm, which were uniformly anchored on the perovskite without much agglomeration (Fig. 4d and Fig. S12d). The almost unchanged polarization resistance  further confirmed that the catalyst was regenerable via sequential redox exsolution manipulations (Fig. S12e). In addition, no obvious delamination could be observed at the electrolyte-cathode interface (Fig. S13), and the Raman spectrum elaborates a high coke resistance of the reconstructed catalyst (Fig. S14). Therefore, the electrochemical reconstruction for the exsolution of IrFe alloy NPs with distinctive self-regeneration features is considered as an alternative method to deliver thermally stable and evenly dispersed metal NPs, which could thereby afford a stable and high catalytic activity for CO 2 electrolysis. The catalytic mechanism was further investigated using DFT calculations (Fig. 5 and Figs S15-20). The CO 2 adsorption and CO formation processes are simulated on SFM and RP-SFIrM, which possess similar O 2− migration capabilities [5]. The free energy profiles are depicted for CO 2 adsorption, CO 2 dissociation to CO, and CO desorption in the catalytic process on IrFe@SFIrM and Fe-SFM. The CO 2 adsorption energy is 0.69 eV on the Fe site in SFM (Fe-SFM) through a carbonate configuration (CO 3 * ). The adsorption of CO 2 is greatly improved as the adsorption energy decreases to 0.03 eV at the IrFe@SFIrM interface. The subsequent CO formation is energetically facile on IrFe@SFIrM with −0.77 eV, which is easier than on Fe-SFM with −0.01 eV. CO desorption is energetically uphill on IrFe@SFIrM with 0.42 eV. We can see that compared to the difficult CO 2 adsorption on Fe-SFM (0.69 eV), the IrFe@SFIrM interface is much more active with a small energy barrier of 0.42 eV in the rate-determining step of CO desorption ( Fig. 5a and b). * CO 2 -bent adsorption with the O atom in CO 2 inserted into the V O on both catalysts is also studied, which shows an inferior performance compared to the carbonate configuration (Figs S16-20). Therefore, CO 2 adsorption with carbonate formation is the most favorable pathway, which is consistent with the strong carbonate signal observed in the NAP-XPS results (Fig. 3g). The charge differential diagram shown in Fig. 5c indicates that charge transfer occurs from Fe and Ir atoms to CO 3 * on IrFe@SFIrM, while only Fe atoms participate in charge transfer in SFM, indicating an active role by both Fe and Ir. The partial density of states (PDOS) in Fig. 5d shows that the Ir atom in IrFe@SFIrM possesses a more pronounced state around the Fermi level than the Fe in SFM and a better overlap between oxygen and carbon in CO 2 , which is consistent with the higher activity for CO 2 electrolysis.

CONCLUSION
The SFIrM cathode displays a dynamic electrochemical reconstruction feature during CO 2 electrolysis with in situ exsolution of highly dispersed IrFe alloy NPs. The construction of the active IrFe@SFIrM interfaces and the dynamic structure-activity correlation is well established by means of in situ XRD, NAP-XPS and XAS measurements. The in situ exsolved IrFe@SFIrM interfaces as catalytically active sites favor the formation of key carbonate intermediates and thus show improved catalytic activity toward CO 2 electrolysis, which is further verified by DFT calculations. Self-regeneration of the electrochemical reconstruction process further improves the operation stability for CO 2 electrolysis in SOECs. This work provides a surface reconstruction process and reactive mechanism investigation of the cathode, which may provide an in-depth understanding of CO 2 electrolysis in SOECs.

Catalyst preparation and SOEC fabrication
A modified sol-gel method was adopted for the preparation of the SFIrM and SFM perovskite powders. Certain amounts of analytically pure H 2 IrCl 6 ·9H 2 O, Sr(NO 3 ) 2 , (NH 4 ) 6 Mo 7 O 24 ·4H 2 O, and Fe(NO 3 ) 3 ·9H 2 O were dissolved in 400 mL deionized water with citric acid monohydrate and polyvinyl alcohol as chelating materials. Then, the solution was evaporated and self-combusted on the heating plate. After a subsequent calcination at 1100 • C, pure perovskite powder was obtained. Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O 3-δ (BSCF) powder was synthesized via the same method.

Physicochemical characterizations
In situ and ex situ XRD measurements were carried out in reflection mode on a PANalytical Empyrean diffractometer. Detailed structural information was analyzed by GSAS software. For in situ XRD measurements, the SOEC was exposed to 95% CO 2 /5% N 2 at 800 • C, which is the same as the real cathode operation atmosphere. In situ NAP-XPS measurements were performed on an EnviroESCA SPECS spectrometer. The lattice oxygen located at 529.5 eV was used to calibrate the binding energy positions. The electrochemical cell was mounted onto the sample holder by mechanically pressing an Fe clamp onto the SFIrM cathode for electrical contact and mechanical fixation. An IR laser was applied to heat the cell. The temperature was controlled by adjusting the power of the infrared laser. In situ XAS measurements were performed at the BL14W1 beamline station in the Shanghai Synchrotron Radiation Facility (SSRF, China, Fig. S4d and e). The Raman spectrum was recorded at room temperature on a LabRAM HR 800 Raman spectrometer. SEM images of electrochemically reconstructed samples were obtained on a JSM 7900F. STEM and EDS results were obtained on a JEOL JEM F200. In order to obtain the exsolved IrFe alloy NPs with good crystallinity for the clear lattice fringes and elemental distribution, a high voltage at 1.6 V was applied to the cell in a CO 2 atmosphere for 24 h and the edge position of the reconstructed SFIrM perovskite in Fig. 2 was carefully selected.

Electrochemical measurements
The cathode was exposed to CO 2 in which 5% N 2 was filled as an internal standard gas for the Faradaic efficiency determination. The opposite anode side was directly exposed to an air atmosphere. Electrochemical measurements were performed with a Metrohm Autolab potentiostat/galvanostat (PGSTAT 302 N) instrument. The frequency range of EIS was 10 6 -10 −1 Hz with a signal amplitude of 40 mV. DRT and the corresponding established equivalent circuit model were fitted by RelaxIS 3 software supported by rhd instruments using the theory suggested by Francesco Ciucci [44]. The products of outlet gases were determined by online gas chromatography (Agilent GC490).

SUPPLEMENTARY DATA
Supplementary data are available at NSR online.

FUNDING
This work was supported by the National Key R&D Program of China (2017YFA0700102), the National Natural Sci-ence Foundation of China (22125205, 22102175, 92015302 and 22072146), the Dalian National Laboratory for Clean Energy (DNL201923), the China Postdoctoral Science Foundation (2021M693124 and 2022T150636) and the Photon Science Center for Carbon Neutrality. G.X. Wang thanks the financial support from CAS Youth Innovation Promotion (Y201938).