Advancements and prospects of perovskite-based fuel electrodes in solid oxide cells for CO2 electrolysis to CO

Carbon dioxide (CO2) electrolysis to carbon monoxide (CO) is a very promising strategy for economically converting CO2, with high-temperature solid oxide electrolysis cells (SOECs) being regarded as the most suitable technology due to their high electrode reaction kinetics and nearly 100% faradaic efficiency, while their practical application is highly dependent on the performance of their fuel electrode (cathode), which significantly determines the cell activity, selectivity, and durability. In this review, we provide a timely overview of the recent progress in the understanding and development of fuel electrodes, predominantly based on perovskite oxides, for CO2 electrochemical reduction to CO (CO2RR) in SOECs. Initially, the current understanding of the reaction mechanisms over the perovskite electrocatalyst for CO synthesis from CO2 electrolysis in SOECs is provided. Subsequently, the recent experimental advances in fuel electrodes are summarized, with importance placed on perovskite oxides and their modification, including bulk doping with multiple elements to introduce high entropy effects, various methods for realizing surface nanoparticles or even single atom catalyst modification, and nanocompositing. Additionally, the recent progress in numerical modeling-assisted fast screening of perovskite electrocatalysts for high-temperature CO2RR is summarized, and the advanced characterization techniques for an in-depth understanding of the related fundamentals for the CO2RR over perovskite oxides are also reviewed. The recent pro-industrial application trials of the CO2RR in SOECs are also briefly discussed. Finally, the future prospects and challenges of SOEC cathodes for the CO2RR are suggested.


Introduction
Our current world is over-dependent on fossil fuels to satisfy global energy demands, resulting in substantial CO 2 emissions.With the increasing concerns about healthy energy supply and sustainable ecological and environmental systems, the world is looking for a more sustainable energy system.Many countries have set targets to achieve net-zero carbon dioxide emissions by the mid-21st century, emphasizing the need to establish a clean, low-carbon, secure, and efficient energy infrastructure. 1here are mainly three ways to realize a net-zero CO 2 emissions target, i.e., the use of renewable energy/hydrogen energy, CO 2 sequestration, and CO 2 re-utilization.By developing the latter two technologies, fossil fuels can still be used to achieve the net-zero CO 2 emissions target, while renewable energy can serve as an energy input for CO 2 sequestration and reutilization.This approach is highly attractive since the current infrastructure for fossil fuel production, storage and utilization is highly matured and can be continuously used, while the netzero CO 2 emissions target is reached due to the storage or recycling of the CO 2 produced.Consequently, the development of efficient CO 2 storage and conversion technologies has received considerable attention recently.
Over the past few decades, technologies like carbon capture, utilization and storage (CCUS) and electrochemical conversion have emerged as the most promising pathways to achieving carbon balance.Unlike CCUS, electrochemical conversion methods including solid oxide electrolysis cells (SOECs) (Fig. 1), alkaline electrolysis technology, and proton exchange membrane electrolyzers can efficiently convert CO 2 into valueadded carbon-containing chemicals, likely hydrocarbons, and CO, which are in principle more attractive.Such conversion processes can actually use renewable energy as an energy input for energy storage; thus, CO 2 can be considered just an intermediate energy carrier in these processes.Among the various available electrochemical technologies, high-temperature CO 2 electrolysis via SOECs offers remarkable advantages, such as short circuiting caused by electron transmission. 11For a SOEC with oxygen ion conducting electrolyte, CO 2 is reduced at the cathode to form CO and O 2− by acquiring two electrons as shown in eqn (1); subsequently, O 2− passes through the electrolyte to the anode side, where it forms O 2 with the release of electrons, as expressed in eqn (2).CO 2 (g) + 2e − / CO (g) + O 2− (1) The C]O in the molecular structure of CO 2 possesses a high bond energy (approximately 750 kJ mol −1 ), thereby presenting a signicant challenge to the catalytic activity of the SOEC cathodes for CO 2 activation. 12In SOECs, the CO 2 RR entails three fundamental processes: CO 2 adsorption, CO 2 dissociation, and CO desorption, which are widely acknowledged.
Currently, there are two distinct CO 2 electroreduction mechanisms based on two different types of cathode materials.In pure perovskite oxides, surface oxygen vacancies ðV O Þ are believed to be the hosts for CO 2 chemisorption at high temperatures. 13Kozokaro et al. employed DFT calculations to conduct an extensive investigation into the CO 2 electrolysis mechanism on La 0.3 Sr 0.7 Fe 0.7 Cr 0.3 O 3 (LSFCr) perovskite.Their calculations revealed that CO 2 is predominantly associated with oxygen ions on the BO 2 (B = ]Cr and Fe) plane of LSFCr to form carbonate (CO 3 2− ) species.Furthermore, the high concentration of oxygen vacancies aids in mitigating the strong adsorption of carbonate, thereby facilitating the activation of CO 2 molecules and the subsequent desorption of CO.  (5) Ricardez-Sandoval et al. found that the basicity of surface O 2− ions of the (0 0 1) facet of La 0.5 Sr 0.5 FeO 3−d perovskite makes it more likely to attract and bind with the C atom of CO 2 (Lewis acid center) thus forming carbonates. 13 2 þ * ¼ CO * 2 (7) Metal-oxide interfaces created through inltration or exsolution demonstrated superior CO 2 electrolysis performance compared to pure perovskite oxide interfaces.This is attributed to the enhanced suitability of the active sites at the metal-oxide interface for CO 2 adsorption and activation. 15,16 2 þ * ¼ CO * 2ðadÞ (10) 3. Advances in perovskite-based CO 2 RR electrocatalysts for SOECs from experimental investigations Typical perovskite-based oxides consist predominantly of alkaline earth or rare earth metal ions occupying their A lattice site, alongside transition metal ions occupying the B lattice site.This category encompasses simple perovskite structures (ABO 3−d ), Ruddlesden-Popper phases (A 2 BO 4+d ), and double perovskites (AA 0 B 2 O 5+d ).As mixed ionic-electronic conductivity (MIEC) electrodes, the perovskite may enlarge the active sites towards the entire perovskite surface, compared with the cermet electrode.However, the relatively poor catalytic activity and stability are the main challenges for most perovskite-related oxides, as illustrated in Fig. 2. Strategies towards improving the activity and durability of perovskite CO 2 RR electrocatalysts in SOECs include the introduction of high entropy into the oxide lattice, surface modication with nanoparticles through various methods, or even the introduction of single atom catalysts.By integrating these optimizations with targeted electrode structure design, challenges such as low catalytic activity, poor stability, and carbon deposition in perovskite-based electrode materials are effectively mitigated.

High entropy perovskite oxides
The eld of high-entropy materials (HEM) has been gaining signicant attention since Yeh and Cantor reported their groundbreaking research on metallic alloys. 17,18The homogeneous blending of various elements elevates the system's entropy and yields a structure that is stabilized both thermodynamically (DG = DH − T*DS) and kinetically (slow diffusion), enabling it to endure demanding conditions like high temperatures, corrosion, and high electrochemical potential. 19These properties have also been conrmed in other materials with high entropy.Lately, many classes of non-metallic oxides have been added to the category of HEM. 20Therefore, the highentropy doping strategy is expected to solve the problem of electrochemical activity and high-temperature stability that are difficult to realize simultaneously in conventional perovskites.
As depicted in Fig. 3a, the presence of multiple elements (four, ve, or more) occupying the same crystallographic site could result in considerable congurational disorder in HEM.The denition of high-entropy perovskite asymptotically follows the concept of high-entropy alloys.Nonetheless, there isn't a single, widely accepted denition of what constitutes a highentropy substance.There are two commonly accepted denitions: (1) congurational entropy larger than 1.6R (Fig. 3b), and (2) single phase and equimolar. 224][25] The molar congurational entropy of the perovskite oxide can be calculated according to eqn (13): where x a , x b and x j are the mole fractions of elements found in the cation and anion sites, respectively, with R being the universal gas constant.The Goldschmidt tolerance factor (t) is employed for evaluating the perovskite structure.
where r A and r B are the ionic radii of the cations on the A and B sites, respectively, and r O is the radius of the oxygen ion.For an ideal cubic perovskite structure, t is equal to 1. Research indicates that the t value for a stable perovskite structure is expected to be in the range of 0.78 < t < 1.05.
where R A is the radius of the ith cation on the A site, and c i is the mole fraction of the ith cation.The observed d(R A ) values in the HEPOs varied from 9.9 to 13%, which is in line with earlier research and within the boundaries set by the Hume-Rothery solid-solution rule. 24,26 Despite numerous high-entropy perovskites being documented, the elements chosen for the A-site or B-site are oen similar.
The A-site high-entropy perovskite comprises at least ve near-equimolar metal cations with similar ionic radii (La 3+ , Pr 3+ , Nd 3+ , Pm 3+ , Sm 3+ , Gd 3+ , Ba 2+ , Sr 2+ , Ca 2+ ), thereby minimizing lattice mismatch.Meanwhile, a "cocktail" effect takes place as a result of the crystal lattice distortion and component mixing, wherein the characteristics of the nal material differ dramatically from the properties of the individual components.A-site cations can affect the overall catalytic activity of the perovskite by inuencing the oxygen vacancy concentration of the transition metal at the B-site, the metal-oxygen bond strength, oxidation, and spin states. 50Up to now, only a few Asite HEM have been utilized as cathodes in SOECs to investigate their performance in CO 2 electrolysis.For example, it was reported that the lattice distortion caused by the highly dispersed A-site ions in La 0.2 Pr 0.2 Nd 0.2 Sm 0.2 Sr 0.2 MnO 3−d (HE-LSM) effectively inhibited Sr segregation at high temperatures, 51 and the distinctive high-entropy structure enabled Ni-YSZjYSZjGDCjHE-LSM single cells to exhibit robust stability during long-term operation.This fully demonstrates that the concept of high entropy in the A-site provides a novel perspective for designing a stable cathode of SOFCs.In addition to a variety of alkaline-earth and rare-earth metal doping, alkali metal multiple doping is worth exploring. 52,53he B-site high entropy exhibits signicant potential in the design and optimization of the perovskite by exploiting the distinct effects of each element on electrochemical properties.High-entropy alloys (HEAs) reside in a state of near nonequilibrium thermodynamics due to signicant lattice distortion resulting from atomic size mismatches.As a result, during the catalytic process, HEAs have a lower energy barrier due to their larger potential energy. 54Particularly, when integrated with an exsolution strategy, the incorporation of multiple elements into the B site of perovskite oxides can enable the exsolution of multimetallic or alloy NPs driven by hydrogen reduction or an applied electric potential.He et al. introduced a novel medium-entropy perovskite, Sr 2 (Fe 1.0 Ti 0.25 Cr 0.25 Mn 0.25 -Mo 0.25 )O 6−d (SFTCMM), aiming to augment CO 2 activation by bolstering metal 3d-O 2p hybridization and diminishing the O 2p bond center (Fig. 4a). 27The SFTCMM-based symmetrical SOECs delivered 1.50 A cm −2 at 800 °C and 1.5 V.  30 Through the reduction of PSCFMMN, an in situ exsolution of the CuFe alloy@FeO x (CFA@FeO) core-shell structure was achieved (Fig. 4b).Unlike simple in situ exsolved NPs, the abundance of oxygen vacancies in the FeO x shell permitted the triple phase boundary (TPB) to extend from the CFA@FeO/HE-PSCFMMN interface to the whole FeO x shell.In situ exsolved Fe-Co-Ni-Cu quaternary alloy nanocatalysts from a Pr 0.5 Ba 0.5 Mn 0.2 Fe 0.2 Co 0.2 Ni 0.2 Cu 0.2 O 3−d (HE-PBM) cathode exhibited excellent electrochemical performance and stability (Fig. 4c). 29Recently, a study has reported an Fe 0.1 Co 0.35 Ni 0.35 Cu 0.1 Mo 0.1 (FCNCM) quinary HEA as a cathode in SOECs (Fig. 4d). 55Due to larger particle sizes, predominantly having a magnitude of several tens of micrometers, the high-entropy fuel electrode exhibited a reduced specic surface area compared to the perovskite electrode.Moreover, aer the stability test, slight oxidation of the alloy occurred.Oxidation would lower the electronic conductivity and potentially trigger the mechanical strength to degrade.
High-entropy perovskite oxides (HEOs) have emerged as promising materials for fuel electrodes in the electrochemical reduction of CO 2 .The incorporation of multiple elements into HEOs serves to stabilize the lattice via congurational entropy, thereby preserving the original perovskite structure while concurrently enhancing catalytic activity.Moreover, the integration of HEAs with perovskite-based materials can be achieved through various techniques such as mixing and inltration to develop exceptional cathodes.The HEA formed by the combination of the B-site high entropy and in situ exsolution strategy is expected to exhibit an efficient and robust cathode of SOECs.Despite this, the precise impact of entropy alone on the catalytic activity enhancement of B-site high-entropy perovskites remains uncertain.Traditional high-entropy alloys have not received the same level of scrutiny as high-entropy perovskites.It is imperative to conduct comprehensive studies to enhance our understanding and achieve signicant short-range disorder in HEAs, thereby unlocking their full potential.Moreover, the inherent complexity of multi-elemental synthesis presents a synthesis challenge, compounded by the ease of phase separation and elemental segregation among multiple  distinct elements. 19In conclusion, additional research is necessary to elucidate the distinctive band structure, identify catalytic sites, and explore the in situ formation of advantageous phases in HEAs.

Surface modication with single atom catalysts
As early as the 1960s  59 At high temperature, MoO 3 was employed to trap Sr atoms on the surface of (La 0.6 Sr 0.4 ) 0.95 Co 0.2 Fe 0.8 O 3−d (LSCF), leading to the creation of additional Sr/O vacancies and higher covalency of (Co/Fe)-O bonds, as well as some harmless SrMoO 4 phases.Compared with the pristine LSCF cathode, the oxygen reduction activity of Sr vac /LSCF was signicantly enhanced.Within the temperature range of 873-973 K, the cell performance using Sr vac /LSCF-1% as the cathode exhibited an improvement of over 70%.Unlike the conventional method of controlling oxygen vacancy concentration in perovskites through A/B site doping, reverse atomic trapping achieved the activation of lattice oxygen in perovskites with atomic precision.Li et al. proposed a custom method to selectively anchor Pt atoms to the B-site in Pr 4 Ni 3 -O 10+d using a simple and universally applicable mechanical mixing approach.This single-atom catalyst not only withstood treatment at 700 °C for 800 hours in air but also saw its electrochemical performance improve by nearly 100%. 60Shin et al. reported a single-atom Pt/ceria nanocatalyst loaded in the oxide fuel electrode of a SOC via an in situ synthetic process, which can securely anchor isolated Pt atoms on the surface of ceria NPs and suppress their high-temperature migration. 61Wang et al. recently synthesized SACs (SDCRu-1100) where a solitary ruthenium (Ru) atom was anchored on the surface of an oxygen ion conductor, Ce 0.8 Sm 0.2 O 2−d (SDC), exploiting the phenomenon that robust covalent metal-support interactions under certain conditions can securely anchor a single atom (Fig. 5d  and e). 62This catalyst found application in the direct electrolysis of CO 2 in SOECs.Due to the distinctive coordination environment and electronic structure of SDCRu-1100, the existence of single atom Ru enhanced the oxygen hole concentration and conductivity of SDC.Compared to the pristine SDC, DFT calculations revealed that SDCRu-1100 exhibited robust CO 2 adsorption and CO desorption capabilities.These ndings collectively signify that SDCRu-1100 served as a highly efficient catalyst for the electrolysis of CO 2 in SOECs.SDC and SDCRu-1100 were mixed with La 0.6 Sr 0.4 Co 0.2 Fe 0.8 O 3−d (LSCF) to fabricate composite SDC-LSCF and Ru 1 /SDC-LSCF cathodes for electrochemical performance evaluation.In contrast to SDC-LSCF, the current density of the Ru 1 /SDC-LSCF cathode reached 1.68 A cm −2 at 1.6 V, representing a notable enhancement of 54.13%.The integration of SACs into perovskite materials has demonstrated potential for enhancing the performance of SOECs.However, extensive studies have demonstrated that NPs on electrode surfaces tend to agglomerate during hightemperature operation, leading to performance degradation.The interactions between perovskite carriers and single atoms, as well as the stability of single-atom active sites within electrodes under operational conditions, remain underexplored.In contrast, the eld of low-temperature catalysis has made signicant progress in understanding the stabilization mechanisms of single atoms and their coordination environments with supports.For instance, studies on catalysts using metalorganic frameworks (MOFs) as single-atom carriers have shown that various amine sites within MOFs can form strong metalnitrogen bonds with metal atoms, thereby stabilizing these atoms during high-temperature pyrolysis preparation. 63,64Yao et al. conrmed the capture and anchoring effects of defects in graphene on Ni atoms through advanced characterization techniques such as high-angle annular dark-eld scanning transmission electron microscopy (HAADF-STEM). 65However, current research on SACs in high-temperature applications remains supercial, lacking necessary studies on the coordination of single atoms with perovskite surfaces.Therefore, future research on SACs for SOECs should focus on the state of single atoms within the catalyst.This will provide a deeper understanding of their stability during long-term operation, facilitating widespread commercialization of SOECs.

Surface modication with nanoparticles
Since the perovskite oxide in SOECs is well sintered with limited surface area, and the surface properties of the perovskite are sometimes difficult to tailor, the deposition of the perovskite surface with some NPs was found to be an effective way to increase the surface active surfaces and to tailor the surface adsorption capability and other properties, to adjust the catalytic performance of the perovskite electrocatalyst for the CO 2 RR in SOECs.There are several ways to introduce the NPs onto the perovskite surface, including inltration and exsolution, and each method has its advantages and disadvantages.
3.3.1 Inltration.Inltration is a simple method to introduce NPs on the surface of the perovskite scaffold, which can effectively increase CO 2 RR active sites in SOEC cathodes.By inltrating catalytically active species onto the perovskite support, it is possible to enhance the chemical adsorption of CO 2 and foster the formation of active carbonate intermediates.Given that only low-temperature calcination (<800 °C) is necessary during the inltration process, excessive growth and agglomeration of the inltrated particles are avoided. 66Directly inltrating metal catalysts (Co, Fe, Ni, and Ru) as active species on the surface of La 0.6 Sr 0.4 Fe 0.9 Mn 0.1 O 3 (LSFM) can signicantly promote the CO 2 RR activity of the cathodes. 67Research comparisons indicated that under a 50% CO/CO 2 atmosphere, the Fe particle-inltrated LSFM cathode demonstrated the most optimal electrolytic performance, achieving an electrolysis performance of 2.20 A cm −2 at 1.5 V and 850 °C.CeO 2 is an outstanding mixed ion/electron conductor.The abundance of the reversible redox pair Ce 3+ /Ce 4+ endowed it with remarkable anti-carbon deposition properties during the catalytic conversion of carbon-containing molecules. 68La 0.3 Sr 0.7 Ti 0.3 Fe 0.7 O 3−d -CeO 2 (LSTF0.7-CeO 2 ) was prepared by using the inltration method as a fuel electrode for the 50% CO/CO 2 conversion reaction in a reversible SOC (LSTF0.7-CeO 2 jScSZjLSM-ScSZ). 66In SOEC mode, the current densities reached 2.36 A cm −2 at 2.0 V and 800 °C.The composite cathode SFM-SDC is a promising material for SOECs, enhancing CO 2 RR efficiency through increased active sites and TPBs compared to the SFM cathode.The inltrated electrode can further increase the electrode TPB area compared to the composite electrode, offering an effective method for constructing highly active electrode materials.Lv et al. inltrated Gd 0.2 Ce 0.8 O 1.9 (GDC) NPs onto the Sr 2 Fe 1.5 -Mo 0.5 O 6−d (SFM) surface enhancing the oxide ionic conductivity of the SFM electrode, leading to improved CO 2 RR efficiency. 69he current density reaches 0.446 A cm −2 at 1.6 V and 800 °C.Improving the CO 2 adsorption capacity on the perovskite oxide surface enhances the generation of active carbon intermediates, thereby signicantly enhancing electrochemical performance.Zhang et al. show that different alkaline-earth metal compounds were inltrated into SFM perovskite oxides as cathodes enhancing the chemisorption and activation capacity of CO 2 . 70Moreover, the co-inltration of Pd and GDC NPs onto the LSCM cathode signicantly enhanced the electrocatalytic activity and anti-coking properties. 71The extensively distributed Pd-GDC NPs improved the electrochemical performance of LSCM by increasing the number of active reaction sites and forming many metal/oxide interfaces.
Loading NPs on the surface of perovskites to enhance the electrochemical performance of electrodes is a well-established and effective methodology.Through a simple immersion process that inltrates precursor solutions into the electrode scaffold, nanostructured catalyst composite electrodes can be fabricated.This approach leads to the enhancement of the triple phase boundary (TPB), thereby effectively improving the electrochemical performance.Due to the weak NP-substrate interaction and high surface energy of the NPs of the electrode prepared by the inltration method, at high operating temperatures, agglomeration of NPs remains a major problem leading to performance degradation of the cathode. 72Improving the substrate-NP interaction is key to further enhancing the operational stability of such NP modied perovskite electrodes in high-temperature CO 2 RR.

3.3.2
In situ exsolution.An alternative method to inltration for introducing NPs onto the perovskite surface is the in situ exsolution method.Through the incorporation of reducible metals such as Fe, Co, Ni, and Cu into the perovskite oxide, coupled with high-temperature reduction treatment, these elements can undergo migration from the bulk to the surface of the perovskite, ultimately coalescing to form metal NPs.Unlike precipitating NPs onto the perovskite surface, exsolved analogs become tightly embedded within the host perovskite structure, thereby effectively mitigating the tendency for NP agglomeration. 73Following reduction at 800 °C, the in situ exsolved Ni NPs derived from (La 0.75 Sr 0.25 ) 0.9 (Cr 0.5 Mn 0.5 ) 0.9 Ni 0.1 O 3−d markedly elevated the electrochemical performance, with current density reaching 0.38 A cm −2 at 2.0 V, and substantially augmented the faradaic efficiency from 50% to 80% during CO 2 electrolysis. 74umerous studies have demonstrated that prolonged annealing in a hydrogen atmosphere induced structural evolution in perovskites.In situ exsolved Fe NPs incorporated onto the La 1.2 Sr 0.8 Mn 0.4 Fe 0.6 O 4−a surface originated from the phase transition of La 0.6 Sr 0.4 Mn 0.2 Fe 0.8 O 3−d induced by treating in a H 2 atmosphere at 800 °C for 10 h.This material displayed noteworthy current densities of 2.04, 1.43, and 0.88 A cm −2 at 850, 800, and 750 °C, respectively, at a voltage of 1.5 V. 75 To further augment the catalytic activity of perovskite electrodes, the design of bimetals (Fe-Cu 76 ), alloys (NiFe, 77 RuFe, 78 CoFe, 79 IrFe, 80 etc.) and core-shell (NiFe@FeO x ) 81 NPs on the perovskite surface is becoming increasingly prevalent.For example, Wang et al. systematically investigated the CoFe alloy exsolved from the La 0.4 Sr 0.6 Co 0.2 Fe 0.7 Mo 0.1 O 3−d (LSCFM) using in situ STEM and DFT calculations at the atomic scale.Compared with the LSCFM cathode, CoFe@LSCFM demonstrated an approximately 1.21 times increase in CO 2 electrolysis performance.Through DFT calculations, this notable enhancement primarily was found to have stemmed from the robust CO 2 adsorption and dissociation capabilities, facilitated by the highly active metaloxide interface between CoFe and LSCFM. 15In addition to the exploration of diverse exsolved metals, the size and distribution of the exsolved particles are also critical factors inuencing catalytic activity.To optimize the metal/oxide interfaces, Wang et al. proposed a strategy of repeated redox manipulations to enhance the exsolution of RuFe alloy NPs on the surface of Sr 2 Fe 1.4 Ru 0.1 Mo 0.5 O 6−d (SFRuM) perovskite. 78This strategy aims to enrich the Ru beneath the perovskite surface, facilitating the exsolution process.Aer repeated redox manipulations, a substantial number of RuFe particles (∼21 000 mm −2 ) exsolved on the surface of SFRuM, marking a 3.6 times increase compared to the exsolution observed during the rst reduction manipulations.The abundant metal/oxide interfaces formed aer the repeated redox processes signicantly enhanced the CO 2 RR, thereby bolstering the electrochemical performance of the SOECs employing SFRuM-GDC as the cathode.In comparison to the cathode before the reduction, the electrochemical performance of the SOECs aer six repeated redox manipulations exhibited an 86.4% increase at 800 °C and 1.2 V.It is noteworthy that repeated redox manipulations can facilitate exsolution while preserving the Fe/Ru ratio and size of the exsolved metal NPs.Neagu et al. meticulously investigated the Ni NPs exsolved from La 0.5 Ca 0.4 Ni 0.2 Ti 0.8 O 3−g . 82Through detailed analyses encompassing XRD and SEM images, they discerned that neither the reducing gas nor the reduction time was a decisive factor for exsolution control.Instead, the reduction temperature emerged as the primary determinant, with the particle population decreasing and the particle size increasing as the reduction temperature increased.Upon employing the Ni-exsolved La 0.5 Ca 0.4 N i0.2 Ti 0.8 O 3−g in methane conversion, it became evident that the particle characteristics, encompassing size and population, played a pivotal role in the methane activation temperature, selectivity, and anti-coking properties of the NPs.Moreover, the continual enlargement of NPs in a reducing atmosphere led to particle agglomeration, representing one of the principal causes of the reduction in catalytic activity.As shown in Fig. 6a 79 They also investigated the connection between the electrochemical performance of SOECs and reduction time, density, and particle size (Fig. 6b-d).Their investigation conrmed that the size of NPs exsolved aer prolonged reduction treatment (4 hours) was larger, but the CoFe alloy particle density experienced a sharp decrease, and the electrochemical performance was only threequarters that achieved with a reduction time of 2 hours.
To address the issue of aggregation of exsolved particles arising from oversized particles in a hydrogen atmosphere, we introduced, for the rst time in our previous work, a strategy to selectively regulate the exsolution energy of active metals at the B-site of perovskites, aiming for simple, efficient, and precise control of NPs (Fig. 6e). 83By incorporating an appropriate amount of Zn element into Sr 2 Fe 1.5 Mo 0.5 O 6−d (SFM), the exsolution energy of Fe from the initial −1.66 eV increased to −1.56 eV, thereby controlling the passivation exsolution of Fe metal and effectively preventing overgrowth and agglomeration of exsolved particles on porous SFM electrodes (Fig. 6f).In comparison to the SFM electrode, the electrolytic cell utilizing Sr 2 Fe 1.4 Zn 0.1 Mo 0.5 O 6−d as the fuel electrode exhibited a current density of 2.74 A cm −2 at 850 °C and 1.6 V, demonstrating an improvement of approximately 93% in performance.
In addition to hydrogen-driven exsolution, the electrochemical technique of voltage-driven exsolution was employed to construct a more robust metal/oxide interface. 84Owing to the sluggish ion diffusion rate in perovskites, the hydrogen-driven exsolution process required several hours to complete.In contrast, the voltage-driven exsolution behavior can typically be completed within a few seconds.Wang et al. conducted a comprehensive investigation into the exsolution of IrFe alloy in Sr 2 Fe 1.45 Ir 0.05 Mo 0.5 O 6−d (SFIrM) under voltage-driven conditions.The investigation revealed a close correlation between the activation process (i.e.exsolution) of the electrode and the applied voltage.Continuous application of a 1.0 V voltage for 4000 s resulted in the gradual stabilization of the current density of the electrolytic cell, indicative of electrode activation completion.At this juncture, particles exsolved on the surface of SFIrM measured approximately 1 nm, with a density of 80 000 mm −2 .With increasing applied voltage, the duration of electrode activation is signicantly shortened (250 s at 1.2 V, 40 s at 1.4 V, 20 s at 1.6 V).Intriguingly, variation in applied voltages does not affect the nal density and average particle size of particles on SFIrM.Relative to the SFM-based electrolytic cell, the current density of the IrFe@SFIrM-based electrolytic cell increased by 25.8% at 800 °C and 1.6 V (1.46A cm −2 vs. 1.16A cm −2 ), owing to the heightened activity and density modication of IrFe alloy particles. 80Current research indicates that the primary hindrance to the practical implementation of voltagedriven exsolution is the limited stability of perovskites under high voltages. 85Despite the thermal stability of exsolved NPs being indeed bolstered by certain methods mentioned above, numerous studies have demonstrated that prolonged annealing in a hydrogen atmosphere induced structural evolution in perovskites, thereby affecting the stability of exsolutionfacilitated perovskites. 83Luo et al. discovered in their research that the exsolution of B-site reducible cations unavoidably resulted in numerous B-site vacancies within the perovskite bulk, which caused unfavorable A-site segregation and a reduction in electronic conductivity. 86They proposed a B-site supplement mechanism to mitigate the degradation of perovskite following reduction.Through the precise introduction of foreign Fe ions to supplement the B-site vacancies in reduced Sr 2 Fe 1.3 Ni 0.2 Mo 0.5 O 6−d , they achieved a robust perovskite scaffold and ensured long-term stable performance.
Compared to directly inltrating NPs onto the perovskite surface, the embedded structure of in situ exsolved NPs within the perovskite matrix confers greater stability.Additionally, in situ exsolved NPs exhibit a more uniform distribution on the electrode surface, possess higher particle density, and provide more active sites.Nevertheless, during prolonged operation, these exsolved NPs inevitably coarsen and agglomerate.Consequently, research increasingly focuses on the controllable precipitation of nanoparticles, utilizing methods such as repeated redox manipulations, electrochemical reconstruction, and exsolution energy passivation.These strategies effectively optimize NP size and density, mitigating agglomeration during actual operation.Notably, NP exsolution oen coincides with phase changes in the perovskite matrix, which are positively correlated.Further detailed exploration is required to determine whether phase changes in the perovskite inuence electrode stability during prolonged operation.

Perovskite-based nanocomposites from self-assembly
In addition to surface modication, the formation of nanocomposites can introduce rich grain boundaries and tailor the lattice strain to induce an effect on the bulk ion conducting properties and surface physicochemical properties, and materials sintering behavior, thus causing an effect on the catalytic performance and durability in the CO 2 RR for the perovskite oxides in SOECs.
Self-assembly is widely used in the creation of functional perovskite-based nanocomposites for different applications.The self-assembly strategy involves initially uniformly mixing metal precursor liquids in each phase, following a calculated stoichiometric ratio, to form a gel-like precursor.Subsequently, the gel-like precursor is subjected to high-temperature calcination, inducing cationic self-assembly into thermodynamically stable multiphase composites.These phases exhibit uniform distribution, close three-dimensional connectivity, abundant TPBs, and facilitate the establishment of rapid ion/electron transmission pathways during electrochemical reactions. 87For example, He et al. fabricated a Ni@Ni-SFM/Ni-GDC cathode via a one-pot synthesis approach. 88Based on DFT calculations, they demonstrated superior CO 2 adsorption capacity and enhanced catalytic conversion ability of Ni@Ni-SFM and Ni@Ni-GDC compared to pristine SFM.Employing Ni@Ni-SFM/Ni-GDC as the cathode, the SOEC attained a remarkable current density of 1.72 A cm −2 at 800 °C and 1.5 V. Recently, Shao et al. employed a one-pot method for synthesizing Cu x -Sr 2 Fe 1.5 Mo 0.5 O 6−d -Gd 0.1 Ce 0.9 O 1.9 (Cu x /SFM-GDC) and utilized it as a cathode in SOECs for direct electrolysis of CO 2 . 89Experimental evidence has demonstrated that Cu x /SFM-GDC exhibited outstanding capabilities in CO 2 adsorption and activation, signicantly enhancing the O 2− /e − conduction rate during the CO 2 RR process.In comparison to the traditional physical hybrid SFM-GDC cathode, the Cu x /SFM-GDC demonstrated a signicantly higher current density of 2.22 A cm −2 at 850 °C and 1.6 V, marking a notable increase of 65.7%.
Nanostructured composite materials, prepared through selfassembly, offer a promising alternative to multiphase electrodes produced by physical mixing methods such as ball milling.Self-assembly composites can be efficiently achieved via a single heat treatment at relatively low sintering temperatures, resulting in enhanced interfacial areas.Furthermore, the self-assembly technique presents advantages such as heightened catalytic activity, excellent chemical and physical phase compatibility, and robust structural stability. 90,91Despite these benets, the precise design principles governing the selfassembly process remain incompletely elucidated.Further indepth investigations into reaction mechanisms are imperative to facilitate the rational development of superior heterogeneous electrodes.

Morphology-controlled perovskites
In addition to the compositional tailoring and surface modication, or the formation of nanocomposites, morphology control is also an effective way to optimize the catalytic activity of perovskite catalysts for the CO 2 RR in SOECs, since the microstructural characteristics of the electrode can also greatly impact the performance of SOECs.During operation, the sintering and agglomeration of particles within the electrode would result in loss of TPB and exert adverse effects on the overall cell performance.Fiber materials fabricated via electrospinning exhibit a distinctive three-dimensional network structure, which is favorable for efficient gas diffusion within the electrode.Moreover, this distinct structure boasts a higher specic surface area, consequently offering a greater number of active sites for the CO 2 RR and enhancing the catalytic activity of the electrode. 92Their unique structure and morphology can also result in a fast exsolution of metal NPs.Furthermore, the distinctive structure, when coupled with exsolution engineering, enables the swi exsolution of metal NPs.For example, Ahmedjonov et al. fabricated a La 0.6 Sr 0.4 Co 0.15 Fe 0.8 Pd 0.05 O 3−d (H-LSCFP) nanober framework electrode via electrospinning technology, subsequently crushing and integrating LSCFP into the nanober framework, resulting in the formation of hybrid nanober electrodes. 93By using a Focused Ion Beam-Scanning Electron Microscopy (FIB-SEM) dual-beam system and cutting edge 3D reconstruction techniques to digitally twin two nano-ber electrodes, they established a correlation between the microstructural attributes of the fuel electrode and its CO 2 electrolysis efficacy.Crushed nanobers were added to the original nanober matrix, which resulted in a signicant increase in the volume fraction (48.7%) and tortuosity factor (2.4).This improved O 2− transport within the H-LSCFP and improved the CO 2 surface adsorption/desorption kinetics of the H-LSCFP.Besides, a honeycomb structure with straight channels prepared by freeze casting has also been used in anodesupported SOECs. 10n comparison to alternative thin lm preparation technologies, atomic layer deposition (ALD) offers distinct advantages in nanomaterial synthesis, such as uniform and conformal deposition onto the substrate, along with meticulous control over the thickness of the deposited layer at the atomic level. 94ncontrolled chemical reactions are prevented due to their selflimiting behavior. 95Moreover, the layers deposited via ALD exhibit notable consistency and repeatability in surface features.8][89][90] The thickness of the catalyst layer is an essential factor as the increased bulk charge transport resistance in the thicker ALD. 96,97For example, Pandiyan et al. deposited Pt NPs (18 nm) on the surface of La 0.75 Sr 0.25 Cr 0.5 -Mn 0.5 O 3 (LSCrM) perovskite oxides via ALD, boosting the electrocatalytic properties. 98Nevertheless, the Pt NPs deposited on the LSCrM are prone to agglomeration during annealing, which is not conducive to the long-term stable operation of the cell.
While morphology-controlled methods have demonstrated the ability to increase the contact area with CO 2 , achieving precise control over the preparation process remains a signicant challenge.An alternative strategy worth exploring is the development of single-phase oxides with exceptional catalytic activity to advance the industrialization of SOECs for CO 2 electrolysis.

Machine learning
The conventional exploration process of material design based on experimental studies is extremely costly and timeconsuming.Predicting material quality by machine learning (ML) has been shown to be a greatly helpful technique for screening materials, which complements the role of statistics in scientic research. 99ML is an efficient and practical approach for predicting and comprehending the catalytic activity, optimal composition, active sites, and CO 2 RR pathways of materials. 100s depicted in Fig. 7, the process of ML is like emulating the human learning process, wherein the system scrutinizes a vast dataset through algorithms to construct models and render assessments on novel data.
In ML research, a subset of the data is kept aside to evaluate the reliability of the model, with the remaining data being used to train the model (generally following a 1 : 4 ratio).To create the database, Jacobs et al. gathered 749 data points from 313 studies that included 299 distinct perovskite compositions. 101onsidering that biases arising from various experimental conditions during data collection can directly impact ML

Perspective
Chemical Science outcomes, it is imperative to conduct a comprehensive analysis of the underlying statistics for each catalytic property in the database.This analysis aims to mitigate or eliminate these biases, thereby ensuring the reliability of ML outcomes.In contrast to experimentally acquired data, leveraging DFT calculations to furnish training data for ML models offers advantages such as low data acquisition costs and expedited acquisition of substantial datasets.While the Inorganic Crystal Structure Database (ICSD) and Crystallography Open Database (COD) offer abundant material structure information for ML applications, the primary deciency in the realm of SOECs lies in the scarcity of extensive and dependable electrochemical datasets on materials.To effectively investigate protonic ceramic fuel cell (PCFC) cathodes with a minimum dataset, Ni Presently, there are two primary types of learning algorithms for supervised learning, in which we instruct the computer on specic tasks, and unsupervised learning, where the computer learns autonomously.In supervised learning, the selection of electrochemical-related features can profoundly impact the model's performance. 99Therefore, feature engineering is central to the investigation of catalytic materials.By integrating diverse features, ML models can adeptly capture and predict the properties of perovskite electrodes.Furthermore, converting features into descriptors, which should be universal and effective, is necessary.For perovskite oxide electrodes, the crystal structure, and bulk oxygen p-band center, along with fundamental physical and chemical properties, constitute relevant feature parameters that inuence the electrochemical performance of the materials. 103For the CO 2 RR, the adsorption energy of CO 2 on the perovskite surface, the structure features of adsorbates, and the active sites are also important descriptors. 104Hence, the selection of suitable descriptors to elucidate structure-activity relationships is paramount for ML in the realm of SOECs.With the progress of deep learning, descriptors are no longer exclusively provided by researchers but can also be automatically acquired through the amalgamation of deep learning with high-performance computing, facilitating the acquisition of suitable descriptors. 100More specically for SOFCs/SOECs, appropriate descriptors such as ASR values and ionic Lewis acid descriptors can further enhance the performance of ML models. 101,105Choosing a suitable ML algorithm is an essential task.The ideal model should aim to maximize computational accuracy while minimizing resource consumption.Among the various types of models, kernel-based learning methods, decision tree ensemble methods, and articial neural networks are widely used mathematical models in ML.For example, Ni et al. implemented 8 different regression methods to t the prepared dataset to predict the ASR values of unexplored materials.Ultimately, four oxides were successfully predicted and conrmed to have superior electro-activity. 105One has the option to employ a diverse range of algorithms for data training, followed by the selection of the most optimal algorithm based on predictive accuracy.McGuinness et al. used a variety of ML methods including partial least squares (PLS), support vector machine (SVM), principal component regression (PCR), local weighted regression (LWR), multiple linear regression (MLR), articial neural network (ANN), and gradientboosted decision tree (GBDT) to predict the thermal expansion coefficients (TEC) of AA 0 BB 0 O 3 -type perovskites. 106Aer model training, one crucial step is to evaluate for validation and optimization.Regression problems are commonly assessed using metrics like mean absolute error, root mean squared error, and coefficient of determination.Error rate, accuracy, balanced F score, and other metrics are typically used to assess model performance in classication algorithms. 107As an efficient and effective ML algorithm, the extreme learning machine algorithm was designed to predict the performance of H 2 O and/ or CO 2 electrolysis of the SOECs. 108he design of the cathode material in SOECs plays a crucial role in high-temperature electrocatalytic CO 2 reduction technology, signicantly inuencing the efficiency of the CO 2 reduction reaction.The adoption of ML algorithms for the analysis of extensive datasets facilitates the identication of essential catalytic properties, material optimization, and the exploration of potential reaction mechanisms, thereby advancing the development of perovskite-based cathode materials.The availability and quality of raw data in databases are fundamental for ML training purposes.However, accurately capturing every electrochemical step occurring on the catalyst surface and within particles poses challenges under practical operating conditions.In particular, obtaining in situ experimental data at high temperatures is challenging.Therefore, the selection of utilizing DFT calculations to construct a catalyst database for screening the most promising candidate catalysts can be considered as a favorable approach.Based on varying structures, one can calculate properties relevant to catalytic reactions, including band structure, adsorption energy, electronic structure, reaction barriers, and more.By quantifying the relationship between each variable and electrolysis performance, further screening of these descriptors can provide guidance for material design and synthesis.Until now, there has been limited research on the application of ML in SOECs for CO 2 electrolysis.Hence, enhancing existing databases, descriptors, and their renement is crucial for enhancing the decision-making capabilities of cognitive systems.  .Ye et al. employed the DFT method to investigate the CO 2 RR on the surface of SDC surfaces with exsolved Fe clusters (Fe-SDC). 110nalysis of the Density of States (DOS) revealed that Fe doping enhanced the d-band center of Ce in SDC, thus augmenting CO 2 adsorption and activation.Bader charge analysis conrmed that CO 2 adsorption on Fe clusters transpired more readily than on the SDC surface, accompanied by a lower energy barrier for CO desorption from Fe clusters.Furthermore, it was discovered that the highly coupled gradient of 3d-2p-4f in Fe-O-Ce facilitated the generation of Ce 3+ -Ce 4+ redox pairs, consequently enhancing the electronic conductivity of Fe-SDC.Ultimately, these enhancements expedited the CO 2 RR on the surface of Fe-SDC.Furthermore, theoretical calculations have revealed that the CO 2 RR was inuenced by the metal/electrolyte interface.Consequently, the composition, density, and size of NPs emerge as pivotal factors in shaping the CO 2 RR. 111 Moreover, J. Payne et al. conducted a DFT-based thermodynamic analysis, which corroborated the mechanism governing dissolution, NP nucleation, and perovskite structure evolution.The surface energy scans that were obtained suggest that the Ir 3+ ions freely moving on the ideal STO surface will be captured by Sr vacancies, since the Ir ions to move away from the V 00 Sr need to overcome a high energy barrier (z5 eV).Subsequently, these captured ions can pair with other mobile Ir ions forming a stable Ir-O-Ir under the mediation of surrounding O, potentially serving as initial nucleation sites for Ir cluster growth and reduction. 112r surface segregation in perovskites may lead to degradation of the electrode performance.Numerous studies on SOFCs have demonstrated that Sr-containing perovskites undergo segregation during high-temperature operation, resulting in surface enrichment and the formation of various product phases, including SrO and SrCrO 4 . 113During SOEC operation, these segregated A-site metal oxides have demonstrated a signicant enhancement in the CO 2 adsorption for the CO 2 RR on perovskite oxide cathodes, but an optimal doping amount still exists, and the stronger adsorption of carbonate could lead to complete reduction of CO 2 to form carbon during electrolysis. 114ozokaro et al. calculated the energy of Sr segregation with three different amounts of oxygen surface vacancies on the LSFCr surface.The calculations showed that the reduction of CO 2 to CO and the formation of SrCO 3 were competitive reactions on the same LSFCr, and therefore O vacancies were needed not only to improve catalysis but also to block the competing reaction of SrCO 3 adsorption. 14

Advanced characterizations for the fundamental understanding of CO 2 electrolysis in SOECs
The in-depth understanding of the fundamentals in CO 2 electrolytes over the SOEC anode is of critical importance to provide  78 Concurrently, detailed analysis of the peak intensity variations in in situ XRD indicated that repeated redox operations signicantly facilitated the exsolution of the RuFe alloy (Fig. 8a).The dynamic evolution of the structure of Sr 2 Fe 1.45 Rh 0.05 Mo 0.5 O 6−d (SFRhM) during redox activity has also been investigated with in situ high-temperature XRD. 124Aer high-temperature reduction, the double perovskite (SFRhM) gradually evolved into layered perovskite (RP-SFRhM), accompanied by the emergence of RhFe alloy.During the reoxidation process, the RhFe alloy phase gradually disappeared, and the RP-SFRhM recovered to pure SFRhM.The voltage-driven exsolution method for in situ electrochemical reconstruction is an efficient way to produce abundant nanostructures faster than the conventional reduction by hydrogen. 84nvestigating the dynamic structure evolution to understand the correlation between structure and the CO 2 electrolysis process is imperative.By applying a voltage that increases in steps, the Ruddlesden-Popper perovskite (RP-SFIrM) and the IrFe alloy phase could be observed respectively. 80X-ray photoelectron spectroscopy (XPS) is an effective surface analysis technology to monitor surface metal valence states and surface oxygen vacancies by tting peaks. 83However, because of the intense interaction between the photoelectrons ejected and gas phase molecules, conventional XPS needs to be carried out in high vacuum or ultra-high vacuum settings. 125These are far from the real working conditions.In situ near ambient pressure XPS (NAP-XPS) measurements were used to investigate the exsolution of NPs and adsorbed carbonate species, revealing numerous details of the CO 2 electrolysis.Under the simulated electroreduction environment, it was observed by NAP-XPS (Fig. 8b) that Fe 3+ in La 0.66 Ti 0.8 Fe 0.2 O 3−d (LTF2) began to reduce to Fe 2+ at 300 °C, and Fe 0 appeared at 800 °C. 115Under cathodic polarization, inferring possible CO 2 RR mechanisms based on the evolution of carbon species represents a prominent subject of discussion.Opitz et al. employed NAP-XPS to visually unveil the formation of carbonate intermediates during cathodic polarization, subsequently engaging in a comprehensive discussion on the CO 2 electrolysis mechanism at the perovskite surface. 117-ray absorption spectroscopy (XAS), which reects the electronic transition from the inner shell to the valence band, provides valuable information on the oxidation state, site symmetry, bond strength, coordination number, bond distances, and identities of nearest neighbor atoms.X-ray Absorption Fine Structure (XAFS) spectra are commonly separated into the X-ray Absorption Near Edge Structure (XANES) and the Extended X-ray Absorption Fine Structure (EXAFS).126 Typically, the oxidation state of an element and its alterations are primarily ascertained through the displacement of the main absorption edge.XANES spectroscopy discerns the oxidation state of pertinent elements in unfamiliar materials through spectral comparison with similar oxidation states of the same atoms.Kim et al. indicated the increase of Fe-Fe shell intensity from the Fourier-transformed EXAFS spectra, conrming the exsolution of Fe NPs onto the surface aer reduction.127 The development of the single atom synthesis strategy has matured signicantly.On the other hand, the primary difficulty with SACs is the aggregation of metal atoms at high temperatures (600-800 °C).It is a challenge to verify the formation of single atom species using atomic-resolution characterization techniques.Instead of the Pt-O correlation observed in model compounds (such as Pt metal, Pt acetylacetonate, and PtO 2 ), it was conjectured that the predominant peak in the Fourier transform of Pt L3-edge EXAFS spectra of Pt/ceria stemmed from Pt-O-Ce.Additionally, the presence of Pt single atoms was validated through tting and comparative analysis of the EXAFS data from Pt/ceria with the theoretical spectra of Ce 1−x Pt x O 2−d type solid solution and Pt metal.Based on the distinction of the coordination number of 8 for a cubic uorite structure, the presence of Pt single atoms was conrmed.61 Specically, XAS is suitable at low concentrations (0.005-10 wt%) and a broad range of testing conditions (−269-1200 °C) from vacuum to high pressure.In situ XAS measurements provide strong evidence of the chemical state and coordination environment.Shen et al. observed that during the CO 2 electrolysis process at 1.4 V, there was a reduction in the intensity of the white line peak at the Ir L3-edge.The formation of an IrFe alloy was demonstrated in both cases.80 X-ray diffraction (XAS) has emerged as a potent method for obtaining the electronic structure, local coordination environment, and oxidation states, which are critical to investigating electrocatalysts.128 However, the uncertainty in t parameters can be 10-15%, and the insensitivity of minor species in mixtures is also noteworthy.Hence, XAS results and interpretations do not stand alone.Without additional information, analysis of the contributions of different species (different oxidation states, particles of different sizes and composition) in these cases is extremely challenging.129

Microstructure visualization of electrodes
][132] In situ transmission electron microscopy (in situ TEM) is widely recognized as a powerful characterization technique that enables real-time observation of electrochemical reactions with high spatial resolution (atomic scale) and temporal resolution (subsecond).In situ exsolution is a highly effective strategy for CO 2 electrolysis in SOECs.Comprehending the exsolution process not only enables better adjustment of activity and stability but also offers valuable guidance for developing abundant interfaces with enhanced functionality.Neagu et al. employed in situ TEM to capture pictures at various points in time of the progressive development of a particle socket and nucleation of two more particles from a powder sample of La 0.43 Ca 0.37 Ni 0.06 Ti 0.94 O 3 at 900 °C in an H 2 (20 mbar) environment.Rather than the few hours reduction durations commonly reported in the literatures, the majority of the particle growth occurred in less than 100 s. 133 Neagu et al. further elucidated that the diffusion of Ir originated from SrIr 0.005 Ti 0.995 O 3 bulk and the amalgamation of mobile metal clusters proceeded from the atomic scale to their ultimate nanoscale conguration through in situ high-resolution STEM observations (Fig. 8d). 112he correlation between structure and performance is crucial for optimizing the design and efficiency of CO 2 electrolysis systems.The 3D reconstructions obtained by FIB-SEM tomography can offer valuable insights into the relationship between the microstructural features of the fuel electrode and its CO 2 reactivity. 134Analyzing these reconstructions will help in understanding how the electrode's physical characteristics inuence its functionality in the electrolysis process.Following a duration operation at 800 °C in a CO/CO 2 atmosphere with varying constant current densities (0 and −1 A cm −2 ), FIB-SEM tomography was used to reconstruct the microstructure development of Ni/YSZ-supported planar-type SOECs. 135Akhmadjonov et al. demonstrated the phenomenon of nickel migration at the electrolyte/electrode contact during long-term CO 2 electrolysis, suggesting that the current was a signicant factor in this process.The durability of the hybrid structured nanoberbased fuel electrode was further conrmed by quantied 3D microstructural characteristics based on parameters like volume fraction, tortuosity coefficient, connectivity, surface area, and TPB value analyzed in the total volume of the electrode sample (Fig. 8e). 93n atomic force microscope (AFM) is a powerful tool used to visualize and manipulate materials at the nanoscale level.Similar to lightly gliding your ngertip along the surface of an item, it measures the forces between the tip and the atoms on the sample surface by scanning a sharp tip across it. 136,137urface pitting occurred when concentrated HNO 3 was used to etch La 0.52 Sr 0.28 Ni 0.06 Ti 0.94 O 3 .The number density and size distribution of these pits were comparable to the exsolved particles, suggesting that the exsolved particles were deeply entrenched in the perovskite bulk.This nding was demonstrated in research by Neagu et al. using AFM (Fig. 8f). 73The large coverage of the Sr-rich phase of the surface with the segregated particles can also be detected by AFM. 132,138erspective Chemical Science

Spectroscopy: Raman, IR and Mössbauer
Taking advantage of the ngerprinting of molecules based on their vibrational modes, infrared spectroscopy (IR) is a powerful technique for identifying adsorbed species on electrode surfaces.The majority of studies based on IR on the CO 2 RR have identied CO 3 2− as a common intermediate.The infrared band ranging from 2380 to 2300 cm −1 is attributed to the adsorption of CO 2 molecules on the perovskite surface, whereas the band spanning from 1450 to 1380 cm −1 is associated with CO 3 2− . 6,1392][133] During desorption, the peak intensity of the 2360.79cm −1 absorption peak associated with the C]O vibration of the CO 2 molecule and the 1600-1300 cm −1 absorption vibration frequency double peak related to the tridentate carbonate species gradually decreases (Fig. 8g). 122In addition, in situ infrared can also be used to test the stability of the adsorbed CO 2 on the Sr 2 FeMo 2/3 Mg 1/3 O 6 (SFMM-R) sample at elevated temperatures between 200 and 700 °C. 140aman spectroscopy is a unique chemical ngerprint of specic molecules or materials, which can be used to quickly conrm the type of material or distinguish between different materials. 141Typically, Raman has been successfully used to detect whether there is carbon deposition on the cathode aer stability testing due to the reasonable sensitivity of Raman spectroscopy to carbon species. 142Materials with centrosymmetric crystal structures like cubic metals (BCC or FCC) and perfect perovskites (ABO 3 ) cannot generate Raman scattering. 143his property makes Raman spectroscopy a valuable tool for analyzing changes in lattice symmetry due to factors such as doping, element segregation, contamination, and redox reactions in SOEC electrodes.In situ exsolution of Ni NP-supported (La 4 Sr n−4 ) 0.9 Ti 0.9n Ni 0.1n O 3n+2 (n = 5, 8, 12 (LSTNn)) is a potential CO 2 electrolysis catalyst.Under reducing conditions of 5% H 2 / 95% N 2 from 20 °C to 600 °C, in situ Raman spectroscopy revealed a gradual weakening in the rotational vibration of the TiO 6 octahedron of LSTNn along the c-axis (Fig. 8h). 123This is because oxygen atoms depart from the lattice of the BO 6 octahedron and nickel segregates from the lattice, leading to a signicant contraction in lattice parameters.
Mössbauer spectroscopy has been widely used as a tool for fundamental research in solid-state physics, especially to differentiate between various species of iron. 139,144Iron-based perovskites have been universally used in the development of cathodes for SOECs, and iron is also oen involved in the electrolytic CO 2 reaction as an active site.The Mössbauer method was oen used to monitor the oxidation state of Fe following various long-term operations. 145Mahmoud et al. compared the initial state of the cell with the Mössbauer spectra data acquired for 0, 1774, 6100, and 9000 h of operation to better understand the structural and electrochemical changes that occur aer an operation. 146Aer prolonged operation, the Mössbauer spectrum results revealed the reduction of iron from perhaps pentahedral Fe(IV) to octahedral Fe(III will also be different in the electrolysis process of CO 2 (Fig. 8i). 114Mössbauer spectra of La 0.7 Sr 0.2 FeO 3−d (LSF72) and La 0.9 Sr 0.2 FeO 3−d (LSF92) collected at room temperature show that the LSF72 crystal structure exhibited a broader range of iron species types compared to LFS92.The oxidation state of Fe in the perovskite decreased with an increase in A-site occupancy.
Advanced operando techniques are essential to meet the scientic challenge (Fig. 8k).The new insights they offer have contributed to a mechanism understanding of degradation and real-time monitoring of carbon deposition on the electrode and exsolution of NPs, and are guiding future directions for material optimization and innovation.

Pre-industrial application trials
The electrolysis of CO 2 using SOEC technology is a promising approach for efficiently converting CO 2 into valuable fuels and chemicals.This process contributes to carbon capture and utilization, offering a potential avenue to mitigate greenhouse gas emissions.The choice of cathode is pivotal in determining the overall performance, efficiency, stability, and economic viability of SOEC systems.When it comes to lower temperature conversions, an SOEC is more efficient at using industrial waste heat than other electrochemical techniques.Since part of the energy is derived directly from heat, this eliminates the intermediate process of converting heat to electricity and improves energy utilization.While there are many potential uses for hightemperature electrolysis technology in the chemical sector, stack performance and degradation may differ signicantly from single cell studies.There are still many key technological challenges that urgently require breakthroughs to achieve largescale commercialization.
The main structures of SOECs adopted at this stage are planar structure, tubular structure, and at tubular structure 140,141,147 (Fig. 9a-c).The structure of a planar SOEC stack includes electrolysis cells, many accessories such as a current collector plate, interconnect, seal layers, and necessary components such as a heat management system, purication system, and automatic control system. 151Since it is easy to achieve the airtight sealing and interconnection of single cells in the manufacturing process of tubular SOECs, compared to planar SOECs, tubular SOECs have certain advantages, but they also face issues such as complex manufacturing processes, low power density, and high manufacturing cost.More recently, micro-tubular and cone-shaped structures have been studied and developed.However, the problems with the planar structure and the tubular structure do not seem to have been completely solved for these designs. 152To address the issue of substrate edge warping in planar structures during long-term operation, researchers have developed a at tubular structured SOEC by integrating the geometric symmetry characteristics of tubular structures.The at tubular structure, which amalgamates the benets of both at and tubular structures, demonstrates commendable performance and mechanical properties.Furthermore, it is more convenient to congure as a stack compared to a traditional tubular structure.Nevertheless, challenges persist, including the high costs associated with stacks and systems, as well as the limited robustness in managing dynamic operations. 153,154With the development of novel electrodes, the performance of SOECs has been greatly enhanced.The CO 2 RR on the cathode is largely inuenced by the number of reaction sites, the size and distribution of porosity for effective gas transport in the high-temperature reduction environment, and the catalytic activity of the cathode surface.Haldor Topsoe's commercial CO 2 electrolysis system based on SOEC technology combines CO 2 electrolysis and gas purication in one unit and is capable of producing 96 N m 3 high-purity carbon monoxide gas per hour. 155Owing to the good electronic conductivity and catalytic activity of Ni metal, cermet composites like Ni-YSZ still stand out as the predominant electrocatalysts utilized in fuel electrodes, having successfully transitioned into widespread commercial usage.Regrettably, the Ni surface operating carbon-containing fuels is susceptible to carbon deposition and agglomeration, thereby inducing performance degradation over prolonged operational durations. 156Zhan et al. used a planar Ni-YSZ electrode-supported SOFC to achieve 1 A cm −2 at 800 °C and 1.3 V with a combination of 25% H 2 and 75% CO 2 inlet gas. 157As Ni-YSZ substitutes, numerous perovskites and uorite oxides were researched.Kaur  The main challenges faced by most perovskite oxide electrodes are the lower electronic ionic conductivity and poor electrochemical catalytic activity.Therefore, the current research focus is on improving the electrocatalytic activity and electronic ionic conductivity of perovskite oxides through methods such as inltration, doping, and in situ exsolution.Employing SOEC high-temperature CO 2 /H 2 O co-electrolysis facilitates the storage of renewable energy in the form of chemical energy, such as syngas or hydrocarbon fuels, thereby boasting potential application scenarios in the realm of energy storage. 159Reytier et al. performed stack-level studies (10-cell stack) in both electrolysis and co-electrolysis modes. 160The conguration of the stack was NiO-8YSZj8YSZjCGOjLSC with an active area of 100 cm 2 .The maximum current density applied per cell at 1.15 V voltage is −0.8A cm −2 with an inlet gas composition of 65 vol% H 2 O + 25 vol% CO 2 + 10 vol% H 2 , the H 2 + CO syngas production of the stack was 0.34 N m 3 h −1 at −80 A and 800 °C.The concurrent occurrence of H 2 O/CO 2 coelectrolysis and methanation processes at the cathode further enhanced the advantage of SOEC co-electrolysis.When the temperature decreased to around 600 °C, methane began to form in the electrolysis products.Methane synthesis properties in tubular SOECs (115 mm in length, 5 mm in inner diameter and 6.59 mm in outer diameter) between 550 and 650 °C were investigated by Luo et al. 161 At 550 °C, when a voltage of 1.5 V was applied, the CH 4 yield reached 12.34%, with an inlet gas composition of 20% H 2 O, 20% CO 2 , 20% H 2 , and 40% Ar.

Perspective Chemical Science
Synthesis gas itself is one of the major chemicals for various chemical and fuel syntheses in modern industry.Gaseous or more complex chemicals and fuels require further downstream processes aer the electrolysis cell.][164] As shown in Fig. 9d, SOECs offer a promising avenue for bridging the energy storage and chemical synthesis industries through electrolysis.Recent studies have demonstrated the feasibility of enhancing the kinetics of the CO 2 electrolysis reaction and reducing power consumption signicantly.Guo et al. introduced methane on the SOEC anode side to produce syngas, leading to accelerated reaction kinetics. 165Chen et al. utilized Sr 2 Ti 0.8 Co 0.6 -Fe 0.6 O 6−d as a dual anode and cathode material to effectively catalyze the conversion of ethane to ethylene on the anode and the reduction of CO 2 to CO on the cathode.It provides an important regulatory knob for achieving high catalytic activity of small molecule conversion. 166In addition to performance improvement, an essential prerequisite for stacks in large systems is the ability of the stack to perform reliably in real-life situations.Long-term operation of SOECs is susceptible to performance deterioration because of their high operating temperature.8][169][170][171] Poisoning is caused either by impurities in the steam or the volatile species evaporated from glass-ceramic sealants, while interconnects reduce reaction sites and hinder charge transfer and oxygen ion transport pathways, leading to electrode degradation and electrolysis performance decline.The purity of the inlet gas is very important for extending the life of the SOECs.Furthermore, a major challenge to CO 2 electrolysis in SOCs is the conict between carbon deposition (CO 2 (g) / C + O 2 (g)) and CO production (2CO 2 (g) / 2CO (g) + O 2 (g)).Königshofer et al. investigated various kinds and intensities of impacts on the stack's performance at various reactant conversion rates, operating temperatures, and gas mixes.Mitigation measures of steam regeneration have been proposed for degradation caused by carbon deposition. 172To illuminate the carbon deposition mechanism, Skae et al. thoroughly studied the relationship between carbon deposition and oxidized carbon intermediates and carbonate coverage. 173Furthermore, the Nernst voltage generated on the SOEC was proportional to the driving voltage provided to the SOEC, making voltage play a crucial role in regulating Ni oxidation and carbon deposition.The electrolysis voltage can simultaneously suppress Ni oxidation and carbon deposition in the range of −0.9 V to −1.4 V. 171 However, the performance degradation caused by nickel coarsening in nickel-based materials remains unavoidable.Potential issues within the Ni cermet fuel electrode encompass Ni particle agglomeration/coarsening and Ni migration, culminating in a notable increase of the series resistance of the cells.The 3D electrode microstructures, aer enduring a 1000 hours durability test under CO 2 electrolysis, were utilized to investigate the mechanism of Ni migration using FIB-SEM tomography.During long-term CO 2 electrolysis, not only will there be coarsening of Ni particles, but Ni will also migrate away from the electrode/electrolyte interface, resulting in irreversible loss of electrochemical performance. 135As a result of stack improvements, degradation rates have dropped to below 0.4% per 1000 h.Meanwhile, the appearance of CO 2 electrolysis mode operating for over 1 year demonstrates that SOECs are one of the most suitable CO 2 RR technologies for commercialization. 174

Summary and outlook
When it comes to storing renewable energy in the form of useful chemicals and fuels, SOECs are seen as one of the best solutions due to their high efficiency with nearly 100% faradaic efficiency and favorable reaction kinetics resulting from elevated operation temperatures.However, the rapid performance degradation, mainly due to the poor durability of state-of-the-art Ni-YSZ based cathodes at high-temperature operating conditions greatly hinders its widespread application.Alternative cathode materials are urgently needed, and the advanced ones should possess certain features to optimize the multiple physical and chemical processes during the CO 2 electrolysis process, such as a large number of active sites, adequate electronic conductivity, and chemical, thermal, and mechanical stability.Compared to the traditional Ni-YSZ cathodes, perovskite oxide cathodes have demonstrated signicant potential when used for direct CO 2 electrolysis in SOECs.In particular, metallic NPs loaded on perovskite oxide surfaces, through inltration, assembly, or codeposition methods, have exhibited excellent catalytic capabilities, but their long-term stability is limited due to weak metalsupport interactions.
In this review article, we have provided a thorough review of the current advancements, challenges, and prospects of perovskite-type cathodes for SOECs in CO 2 electrolysis.It covers a comprehensive summary of recent design concepts for SOECs, aiming to help scholars with more insights into the design of novel cathodes with improved catalytic activity and stability for CO 2 electrolysis.Emphasis was placed on in situ characterization and DFT calculations for an insightful understanding of the mechanism of electrolysis of CO 2 and analysis of the origins of electrode performance degradation.Additionally, the rise of ML has contributed greatly to the rapid screening of superior candidate materials.However, to achieve a high degree of commercialization on an industrial scale, it is essential to develop new cathode materials that signicantly improve durability and stability.Based on the latest research results, here we propose some suggestions for future development directions in perovskite-type cathodes for SOECs towards CO 2 electrolysis.
(1) Exsolution has emerged as a promising method to fabricate NP-decorated oxides, as the NPs are anchored onto the host oxide with strong mutual interactions.This can suppress particle sintering, ultimately improving the stability of the catalysts.Various strategies involving structural modications (defects and phases) and control of operating conditions (gas composition, temperature, potential, and strain) are utilized to tailor the size, composition, and distribution of the exsolved nanostructures.Nevertheless, the mechanisms of exsolution are still not fully understood.Actually, there even exist some conicts in the explanations, like the relationships between the Chemical Science Perspective perovskite surface and the exsolved NPs.With the help of advanced characterization techniques and theoretical calculations, we may get a clearer picture of the electronic and atomic structures of the materials, thus guiding the development of new methods for exsolution to precisely control the density and distribution of exsolved NPs.
(2) Currently, single atoms, high-entropy, and other engineering approaches are being widely explored and exhibit great potential in improving catalytic performance.High-entropy perovskite oxides are outstandingly promising materials for the CO 2 RR via SOECs, which exhibit favorable performance stability due to entropy stabilization.Especially at high temperatures, such stabilization effect is more obvious considering that the effect of increasing entropy is magnied at higher temperatures based on DG = DH − T*DS.On the other hand, the chemical element variation at the A/B sites of perovskite oxides can also help to tailor the catalytic activity of the materials.This leads to a new consideration about the origin of the improved performance, the high-entropy effect or the multielement doping effect.Direct evidence about the real contribution is unfortunately still lacking.The unique maximum atom utilization and remarkable high catalytic performance of SACs will undoubtedly attract the attention of researchers.However, the thermal stability of SACs is crucial for their practical application under reaction conditions (including CO 2 or reducing atmospheres).There is an urgent need to further improve the stability of SACs.Currently, metal loading in SACs is still at low levels; to improve the metal loading and prepare dual atom catalysts, it is necessary to delve deeper into the structure-performance relationship and catalytic mechanisms.This will enable the rational design of superior SACs for hightemperature CO 2 electrolysis.
(3) Fundamental understanding is the basis of material development.An in-depth understanding of the growth and working mechanisms of NPs based on exsolution from a perovskite lattice is still not available.Credible chemical information on the catalysts is provided by advanced in situ characterization technologies, such as XRD, XPS, and XAS.In situ Raman and FT-IR are widely used to investigate the mechanism of the CO 2 RR by identifying the carbonate species.Secondary ion mass spectrometry and isotope exchange investigations will develop into potential characterization tools that provide rich information on structural design.Through the combination of theoretical calculations and advanced in situ characterization, the pathways of CO 2 reduction will be provided to understand the correlation of structure and activity during the CO 2 RR.Most researchers may lack the opportunity for in situ characterization due to the expensiveness and access limitations of synchrotrons, so when ex situ characterization is employed, it should closely approximate the working conditions of the material and provide a thorough explanation of the testing conditions.
(4) ML has great potential in rapid material screening.Predicting catalytic activity, active site identication, and reaction pathways in the CO 2 RR through ML techniques will accelerate the development of high-performance cathodes.In addition, by combining ML with insights obtained from the mechanisms of decayed cell performance, we can understand the primary failure modes and extend that lead to performance degradation.Therefore, establishing a database with a large amount of reliable data and selecting appropriate descriptors are essential to fully leverage the potential of ML.
(5) Scaling SOEC devices to the industrial level is extremely challenging.There has been a large amount of experimental research on single cells.While the current density of single cells (small-area button-type SOECs) can reach 3 A cm −2 (800 °C, 1.5 V) or even higher, the output performance of the SOEC stack experiences a substantial decline with the continuous amplication of single cells.The commercialization challenges of SOEC technology lie in its short lifespan, susceptibility to damage, and inability to operate stably for long periods.Particularly at high current densities, the degradation of the cathode is the main reason.Therefore, while optimizing the design of the electrolysis stack system, it is crucial to develop advanced cathodes.Perovskites could be feasible electrodes to replace traditional Ni-YSZ in the future.At the same time, more theoretical simulations and experimental research should be conducted to further conrm the cost-effectiveness and technical viability of coupling hightemperature electrolysis, renewable energy generation, and chemical synthesis processes.
Overall, while the industrialization of SOECs for CO 2 electrolysis still faces signicant challenges, we rmly believe that combining various in situ characterizations with advanced theoretical calculations can further clarify the reaction mechanisms of high-temperature CO 2 electrolysis in complex reaction environments.This will lay the foundation for researchers to design excellent cathodes, eventually making SOECs a practically applicable technology to help humanity realize a net-zero carbon society.

Chemical Science
Perspective

Fig. 2
Fig. 2 Schematic diagram of the current challenges and their solutions.

Fig. 3
Fig. 3 (a) Typical schematic of a high-entropy perovskite.(b) Ideal configurational entropy as a function of the number of elements, n, in equiatomic proportions.Reproduced from ref. 21.Copyright 2019, The American Chemical Society.

Fig. 4
Fig. 4 (a) Theoretical calculations and I-V curves for CO 2 reduction of SFTCMM and SFM.Reproduced from ref. 27.Copyright 2023, The American Chemical Society.(b) TEM, HRTEM, and the corresponding EDX mapping of PSCFMMN after reduction in H 2 at 800 °C for 2 h.Reproduced from ref. 30.Copyright 2023, Wiley-VCH.(c) HRTEM, HAADF-STEM, and EDS element mappings of reduced HE-PBM with exsolved NPs.Reproduced from ref. 29.Copyright 2023, Elsevier.(d) STEM-EDS image of the five elements in the alloy.Reproduced from ref. 55.Copyright 2024, The American Chemical Society.

Fig. 5
Fig. 5 (a) Schematic image of Pt-loaded onto PNO to form Pt1-PNO. Reprinted with permission from ref. 60.Copyright 2022, Elsevier.(b) Probing atomic Pt: HAADF-STEM images of Pt/ceria.(c) k 2 -Weighted Fourier transform of the Pt L3-edge EXAFS spectra of Pt/ceria and model compounds.Reprinted with permission from ref. 61.Copyright 2022, Elsevier.(d) Atomic-scale HAADF-STEM images and EDS elemental maps of SDCRu-1100 and (e) the corresponding intensity profiles are labeled with the blue boxes in (d).Reproduced from ref. 62.Copyright 2023, Wiley-VCH.

Fig. 6
Fig. 6 (a) In situ STEM images before reduction and after reduction and re-oxidation.(b) XRD patterns under different conditions and (c) the enlarged XRD patterns of (b).(d) The relationship between reduction time, particle size, population density, and current density.Reproduced from ref. 79.Copyright 2020, Wiley-VCH.(e) Design scheme of NPs exsolved from perovskites with different methods.(f) SEM images under different reduction times.Reproduced from ref. 83.Copyright 2023, The Royal Society of Chemistry.
et al. suggested a unique method based on the experimental design paradigm (EDP), 102 which facilitated efficient screening and discovery of novel materials by meticulously selecting elements and design matrices, even in the absence of extensive datasets.Furthermore, the D chem , k chem , area-specic resistance (ASR), and EC response parameters of materials like BaCo 0.667 Ce 0.167 Fe 0.083 Y 0.083 O 3−d and Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O 3−d predicted by EDP are highly consistent with the experimental results, providing further validation of the model's reliability.

Fig. 7
Fig. 7 The flow chart of ML in screening catalytic materials.
Table1lists several typical highentropy perovskite oxides that have been utilized as electrodes for SOFCs/SOECs.HEM primarily consist of high-entropy perovskites and spinels.Present research predominantly centers on the air electrode of SOFCs, and related investigations concerning the utilization of HEM for CO 2 electrolysis in SOECs have emerged.Depending on the doping position, HEM can be categorized into A-site high entropy and B-site high entropy.
149DFT calculationsCombining DFT calculations with experiments has been an increasingly common strategy for perovskite design and discovery.DFT calculations have been extensively employed to corroborate experimental results and elucidate the CO 2 RR mechanism within SOEC cathodes, thereby furnishing a crucial reference for the meticulous design of cathodes.CO 2 adsorption and dissociation reactions are signicant for efficient CO 2 RR.Li et al. calculated the relative energy proles for CO 2 adsorption and decomposition on the surface of Sr 2 Fe 1.5 Mo 0.5 -O 6−d F 0.1 (F-SFM), revealing that F-SFM exhibited a lower CO 2 adsorption and dissociation energy compared to SFM.Based on a series of theoretical calculations, the fundamental steps of the complete electrochemical CO 2 RR cycle within the F-SFM cathode were subsequently proposed.109Kozokaroetal.employed DFT and nudged elastic band methodologies to discern the optimal active sites for CO 2 adsorption on the perovskite surface and demonstrated the signicance of oxygen vacancies in the catalytic process.14Theenergy calculation for CO 2 adsorption revealed that CO 2 adheres to the oxygen atoms on the BO 2 plane via the O surface -C bond, establishing a robust surface connection through chemical adsorption as CO 3

Table 2
[107][108][109]nced characterization techniques applied to the cathode in SOCs XRD) has demonstrated its importance in characterizing physical and chemical properties.Targeting the high temperature of SOC technologies, high-temperature in situ XRD is one of the most attractive methods to reveal structural and chemical information.[107][108][109]Throughhigh-temperature in situ XRD, we can see how the crystal structure of the electrode changes in the temperature range from room temperature to working temperature.For example, Wang et al. utilized in situ XRD technology to visually capture the exsolution and subsequent disappearance of the RuFe alloy within Sr 2 Fe 1.4 Ru 0.1 -Mo 0.5 O 6−d (SFRuM), achieved through cyclic switching of the redox atmosphere at 800 °C.