Roadmap on exsolution for energy applications

Over the last decade, exsolution has emerged as a powerful new method for decorating oxide supports with uniformly dispersed nanoparticles for energy and catalytic applications. Due to their exceptional anchorage, resilience to various degradation mechanisms, as well as numerous ways in which they can be produced, transformed and applied, exsolved nanoparticles have set new standards for nanoparticles in terms of activity, durability and functionality. In conjunction with multifunctional supports such as perovskite oxides, exsolution becomes a powerful platform for the design of advanced energy materials. In the following sections, we review the current status of the exsolution approach, seeking to facilitate transfer of ideas between different fields of application. We also explore future directions of research, particularly noting the multi-scale development required to take the concept forward, from fundamentals through operando studies to pilot scale demonstrations.

the design of advanced energy materials. In the following sections, we review the current status of the exsolution approach, seeking to facilitate transfer of ideas between different fields of application. We also explore future directions of research, particularly noting the multi-scale development required to take the concept forward, from fundamentals through operando studies to pilot scale demonstrations.  [26]. CC BY 4.0, and (c) at point defects on the surface of atomic-flat La0.6Sr0.4FeO3 (LSF) thin films. Note that the point-defect-mediated nucleation results in much higher particle density. (d) Monte Carlo simulation showing the correlation between the surface density of oxygen vacancy pairs and that of the exsolved nanoparticles on the LSF surface. These results suggest surface oxygen vacancy pairs are the likely nucleation center for the exsolved nanoparticles. Adapted with permission from [12]. Copyright (2021) American Chemical Society. 10 9 -10 10 cm −2 (figure 2(b)) [25][26][27]. Intriguingly, recent studies found that exsolution on single-crystalline thin films can generate nanoparticles with densities larger than 10 11 cm −2 (figure 2(c)) [12,28], which is more desirable for (electro)catalysis. Therefore, there is a huge motivation for researchers to understand the nucleation mechanism on single-crystalline materials that can produce a higher particle density, and come up with rational design principles to boost exsolution on polycrystalline materials.

Introduction
In a recent effort, Wang and Yildiz et al investigated the Fe 0 nanoparticle nucleation process on an atomically smooth, (001)-oriented epitaxial La 0.6 Sr 0.4 FeO 3 (LSF) thin film and found that the high density of exsolved nanoparticles on single-crystalline thin films is due to point-defect-mediated nucleation. As shown in figure 2(c), the exsolved Fe 0 nanoparticles on the LSF surface are uniformly distributed on the terraces, with no preference for the step edges. This characteristic suggests that point defects at the surface are more favorable for nanoparticle nucleation and growth than step edges [16,17]. To identify the critical surface defect structure that governs the nanoparticle nucleation, Monte Carlo (MC) simulations were used to model the surface concentrations of isolated oxygen vacancies, oxygen vacancy pairs, and large oxygen vacancy clusters on the (001)-oriented LSF surface. As shown in figure 2(d), the surface Fe 0 nanoparticle density agrees well within the range of the surface oxygen vacancy pairs. On the other hand, the calculated density of both the isolated oxygen vacancy and large vacancy clusters are around three orders of magnitude off compared to the particle density. This comparison delivers strong evidence that the surface oxygen vacancy pairs are the likely nucleation center for Fe 0 nanoparticle exsolution. In support of this mechanism, density functional theory (DFT) calculation further confirms that surface oxygen vacancy pairs are the more favorable sites for Fe adatoms and Fe dimers on the LSF surface, compared to the isolated oxygen vacancy sites [12]. These findings thus suggest that promoting point-defect-mediated nucleation can be the key to realizing high particle density in the exsolution process, which calls for approaches to effectively tailor the defect chemistry in host oxides. We will discuss this topic in the next section.
Tuning defect chemistry with external stimuli to control exsolution One recent advance in exsolution is to use external stimuli (such as lattice strain [29], oxygen chemical potential [30], electrical polarization [31], and photo-illumination [32]) to tailor the concentration and stability of point defects in the host oxide, and thus control the morphology of the exsolved nanoparticles. The major advantage of using external stimuli is that it provides another degree of freedom in tailoring the defect chemistry in materials, beyond the nominal host oxide composition. As such, one can tailor the defect chemistry after the materials/devices have already been synthesized, and introduce dynamic modulation under operational conditions [33,34]. Moreover, as discussed earlier, such dopant-free method is also ideal for fundamental studies in elucidating the role of point defects in modulating exsolution.
Elastic lattice strain has been widely explored to tailor exsolution [12,28,35]. For example, Wang et al [12] and Kim et al [35] reported tensile strain to promote metal exsolution in LSF and SrTi 0.75 Co 0.25 O 3 (STC), respectively. On the contrary, Han et al [28] found that compressive strain is favorable for Ni exsolution in La 0.2 Sr 0.7 Ti 0.9 Ni 0.1 O 3 (LSTN). These seemingly contradictory observations can be reconciled in terms of strain-mediated point defect formation. That is, the favored strain direction to promote nanoparticle exsolution is the one that favors the formation of point defects in the host oxides [12]. Take LSF as an example (figure 3(c)), first principles calculation shows that biaxial tensile strain facilitates the formation of both oxygen vacancy and Schottky defects (i.e. the two primary point defects associated with exsolution) in LSF [12]. In accordance with the strain-dependent defect formation, it was also found that the tensile-strained LSF demonstrated enhanced exsolution phenomena, with an increased amount of total exsolved Fe 0 , smaller particle size, and higher particle density [12]. On the other hand, the exsolution on the compressive-strained LSF surface is suppressed ( figure 3(c)).
In addition to strain, researchers have also explored other types of external stimuli to promote and control exsolution via defect engineering, such as electrochemical polarization [36], radiation [37], and thermal shock [38] (figure 3(c)). For electrochemical methods, Myung et al have shown that by using electrochemical reduction, the voltage-exsolved nanoparticles can have higher particle densities, smaller particle sizes, and hence higher catalytic activities than the conventional thermally exsolved nanoparticles [36]. Opitz et al have furthered used electrochemical biasing to realize reversible control of exsolution, which has implications on controllably tailoring materials' H 2 O splitting kinetics [33]. Regarding irradiation, Kyriakou et al employed N 2 plasma treatment to facilitate exsolution, and found that the plasma-assisted exsolution can generate Ni nanoparticles that have up to ten times higher particle density compared to thermal chemical exsolution [37]. Recently, Khalid et al have leveraged argon plasma treatment to promote Ni exsolution out of La 0.43 Ca 0.37 Ti 0.94 Ni 0.06 O 2.955 at room temperature [39]. Finally, in terms of thermal control of exsolution, Sun et al have utilized ultrafast thermal annealing (i.e. thermal shock) to promote exsolution [38]. As a result, they found that the exsolved nanoparticles produced by thermal shock have significantly higher particle density and reduced particle size, as compared to the conventional furnace-based exsolution. The aforementioned exciting phenomena clearly demonstrate that modulating defect chemistry from external stimuli can play a critical role in enabling novel exsolution modulation.

Concluding remarks
In summary, point defects play a crucial role in determining the exsolution products both thermodynamically and kinetically. As a result, tuning point defects in the host oxide via external stimuli (such as strain, electric field, oxygen chemical potential, and radiation) can be an effective approach to optimize the size and density of the exsolved nano-catalysts. To enable rational defect engineering, further studies are required to illustrate the detailed exsolution mechanism. Therefore, a combination of in-situ/operando characterization (such as near ambient pressure x-ray photoelectron spectroscopy (XPS)  [12]. Copyright (2021) American Chemical Society. [34], environmental transmission electron microscopy (TEM) [40], and grazing incident x-ray scattering [33]) and multiscale modeling (such as first-principles calculation [41], MC simulation [12], and phase-field modeling [42]) are suggested for the future studies to establish the precise defect structure-property relationship in exsolution.

Status
When exsolution of reducible transition metals from perovskite-type oxides was discovered as a novel way for preparing oxide supported metal particles, the obtained catalysts were typically studied by ex-situ techniques. With the growing interest for a basic understanding of the exsolution processes and of materials changes during operation, this highly promising catalyst type was more and more investigated also under in-situ conditions. This is particularly important because the particles obtained via an exsolution process are commonly very small (some nm) and in case of non-noble metals also highly prone to oxidation. Transfer of an exsolution catalyst from the gas phase and temperature where formation has taken place to any ex-situ analytical instrument can thus irreversibly alter the metal particles. In-situ characterization allows avoiding this problem and, moreover, can reveal time dependent changes of the catalyst. However, the term 'in-situ' refers to different characterization approaches that should be distinguished here.
On the one hand, the catalytic performance increase upon exsolution can be monitored in-situ, which means that the rate of a reaction proceeding on the surface of the respective perovskite-type catalyst is tracked while the exsolution process itself occurs. This can be done either by measuring the amount of an expected product in a gas stream [43][44][45][46], or-if the exsolution catalyst was employed as an electrode in a solid oxide electrochemical cell-by recording the current passing through this cell upon applying a bias voltage (see figure 4) [8,33,34,36,[47][48][49][50].
On the other hand, the chemical, crystallographic, compositional, and morphological state of the exsolution particles (as well as of the parent matrix) can be tracked during particle formation and during catalyst operation. Here, different tools come into play as also illustrated in figure 4. First, the appearance of metal particles and their changes can be studied in-situ with x-ray diffraction (XRD) during the abovementioned chemical or electrochemical reactions [33,40,43,44,51,52]. Second, spectroscopic approaches such as ambient pressure x-ray photoelectron spectroscopy (AP-XPS) and x-ray absorption spectroscopy (XAS) are used to study the exsolution of metal particles from their perovskite-type host oxide [10,12,24,34,53,54]. With both methods not only the formation of metal particles but also their reversible switching between reduced and oxidized state can be investigated [33,34]. The third and latest development visualizes the exsolution process from the very first moment of nucleation by in-situ TEM [11,40]. This helped gaining deep insights into the very first seconds of an exsolution particle's life.

Current and future challenges
Some challenges for in-situ characterization of exsolved particles arise from an extension of the scope in this field. Initially, exsolution was used as a method to prepare catalyst particles with superior properties on oxide surfaces. More recently, the possibility of preparing bulk exsolutions (also called endo-particles) came into focus. This allows obtaining self-assembled nano-composites with unprecedented structure and properties [3,54,55]. In-situ studying the evolution of these buried particles and the associated surrounding nanostructures, however, is still rather difficult.
For surface exsolution catalysts, the current research activities go into several directions. Reduction of particle size is desired and low temperature exsolution is a possible strategy to achieve this goal. In addition, alloy and dual phase exsolution is in the focus of interest. In those cases it is very important to know the achievable particle size, composition, oxidation state, and crystallographic phase under the operating conditions of the catalyzed reaction. Especially for catalysis on bimetallic particles, such a knowledge is crucial, as the catalytic activity or selectivity for a particular product can depend sensitively on the composition of the exsolved alloy particles, and the same applies to changes in the crystallographic phase.
Another important topic to be understood is the mechanism of the electrochemical switching process of exsolution particles, which is particularly interesting, since it allows catalytic properties to be switched on and off by applying a voltage. The complexity of this question arises from the fact that two independent driving forces affect the oxidation state of the exsolved particles-the oxygen chemical potential µ O in the gas phase and µ O in the supporting oxide. In conventional exsolution catalysts these two chemical potentials are the same. However, in polarized oxides used as electrodes of electrochemical cells, the electrode overpotential can strongly shift µ O from its (gas) equilibrium value [56], as oxide ions and electrons from the mixed conducting parent oxide are electrochemically pumped into or out of the particle. It is a mostly unknown interplay of kinetic steps (transport, reactions at surfaces and interfaces) that finally determine whether the gas or the oxide 'wins' that game and how fast the corresponding changes take place. Achieving a fundamental understanding of this switching kinetics would enable tailored catalytic properties and could expand the application field of exsolution catalysts in electrochemical cells also to non-faradaic reactions. There, the reaction rate is no longer coupled to the current, and a voltage would only be needed to keep the particle in its desired (active or inactive) state.

Advances in science and technology to meet challenges
For at least three of the abovementioned challenges-endo-particle formation, alloy and dual-phase exsolution, as well as understanding the mechanism of electrochemical activity switching-in-situ TEM is expected to provide decisive insights for their deeper understanding, similar to heterogeneous catalysis, where this technique has been successfully contributing to a deeper understanding for some time [57][58][59][60]. In endo-particle, alloy, and dual-phase exsolution, in-situ TEM is essentially the logical continuation of previous efforts, as it is capable of virtually simultaneously providing information on morphology, chemical composition, and crystallographic phase. The combination of several other in-situ techniques may be an alternative or at least a very good supplement to in-situ TEM studies.
The goal of studying the mechanism of electrochemical activity switching raises the bar even higher. Here, not only a parent oxide with its exsolved particles needs to be studied, but an entire solid oxide electrochemical cell is required for being able to pump oxygen in and out of the surface particles. It is probably most promising to use thin film cells in current state-of-the-art (SoA) in-situ TEM sample holders, though many challenges remain, such as performing the TEM experiments under defined atmosphere, elevated temperature and with applied electrochemical bias voltage. Expanding the limitations of other in-situ methods is also important. For example, in-situ XRD and XPS measurements with optimized time resolution can provide highly valuable additional information on the kinetics of electrochemical activity switching and on the origin of its hysteresis-like behavior [33].
Substantial improvements could also be expected from adding other measurement techniques to the toolbox of in-situ studies, such as in-situ gravimetric measurements [61] under polarization or mapping phase changes by chemical capacitance changes during impedance studies. In addition to such sophisticated in-situ experiments, a better understanding of electrochemical switching also requires progress in modeling these processes. So far even the concepts for describing the kinetics of the electrochemical activity switching are not fully developed yet.

Concluding remarks
With the growing interest in understanding the exsolution process on an atomistic level, the demand for analytical methods has strongly increased in recent years. Methods such as in-situ AP-XPS, XRD, and TEM have been successfully applied to reveal highly valuable insights into this topic. Among these methods in-situ TEM with all its analytical capabilities appears to be the most powerful, but also the most challenging characterization technique, to address the currently most important questions in the field of exsolutions. It is still a highly non-trivial task to apply these analytical tools also to the very small feature sizes of exsolved metal particles. Moreover, studying switchable exsolutions on electrodes of solid oxide cells requires implementation of model cells into current in-situ TEM sample holders and their operation during TEM experiments. Together with developing and implementing additional in-situ characterization tools this should lead to a major leap forward in understanding the redox behavior of switchable exsolution particles. properties, validated experiments and easily computable descriptors are needed [69]. These descriptors could be linked to mechanisms such as segregation energetics, crystal stability, and electrokinetic reactivity. For compositional search and descriptor development to take place, extensive materials databases are required [70]. It should be noted that first-principle atomistic methods are poorly suited to searching for optimal materials over large compositional spaces. These methods are typically computationally expensive and can handle only a few hundred atoms. Realistic models (i.e. those involving defects such as grain boundaries and large supercells) may contain thousands of atoms, causing a considerable bottleneck for computational resources. To tackle this challenge, artificial intelligence will likely play a key role as appropriately trained graph neural networks [71,72] could be used to relax structures and predict materials properties (i.e. exsolution energies) with an accuracy comparable to that achieved by DFT [73]. Consequently, the search for materials could be accomplished directly by optimizing a descriptor evaluated using easily computable neural networks. This procedure would allow the probing of large compositional spaces at a fraction of the cost of conventional DFT simulations.
While DFT is only capable of producing ground-state energies (at 0 K), exsolution is a process that typically requires between 400 • C-1000 • C. However, Ab initio molecular dynamics simulations can overcome this discrepancy as these simulations depend on pressure as well as temperature. Although ab initio molecular dynamics simulations are computationally demanding and limited to small atomic systems with short simulation times (below a fraction of a nanosecond) [74], these constraints could be alleviated by artificial intelligence as machine-learning-based force fields. Such can achieve force evaluations with quantum mechanical accuracy at high computational speeds. In turn, larger atomic systems could be used to capture the dynamics of transition metal atom migration and surface clustering more realistically (e.g. at grain boundaries as a function of temperature and pressure, etc) [75].
From the discussion above, it becomes apparent that the next generation of exsolution models will need to be matched with a range of experiments that accurately represent atomic scales (∼1 nm) and microscale-level (∼10 mm) processes [76]. The phase-field crystal (PFC) method, which can bridge atomic and mesoscopic scales, could be used to simulate the nucleation and growth of exsolved nanoparticles. By combining non-equilibrium thermodynamics and phase transformation dynamics, the PFC approach can, in principle, handle the characteristic time and length scales of diffusional, nucleation, and growth processes that underpin exsolution [77]. However, for larger three-dimensional systems, computational scaling remains a significant challenge [78]. These issues can be overcome by using modern scale-bridging multiscale modeling. Multiscale modeling requires a cascade of models tailored to specific length scales (e.g. quantum mechanics of metal atom diffusion and reaction energetics at the <1 nm scale, molecular dynamics to capture metal atom clustering at the 10 nm scale, and continuum models to simulate particle growth and strain above that) [79]. These could then be linked rigorously through a heterogeneous multiscale method [80]. In addition, physics-informed neural networks have emerged as a unique approach for learning structures and defects in materials [81]. Using these advanced models on time-resolved, environmental, and TEM data could help determine physical models for exsolution automatically.
Finally, it is important to stress that many potential applications, including the prediction of electrocatalytic activity degradation and the optimal control of nanoparticle properties during operation and synthesis, require models that are far less computationally demanding than those described above. Therefore, future research should also be focused on reduced-order models [82] that can capture the physics and chemistry of the exsolution process and the degradation of nanoparticles.

Concluding remarks
Exsolution has emerged as a major strategy to produce nanoparticles that are stable and highly active. However, as the details of the mechanism by which nanoparticles exsolve are still not fully understood, opportunities for designing and optimizing new electrocatalysts remain limited. To overcome this challenge, models will play a key role. Due to the complexity of the exsolution mechanism, which involves multiple time/length scales and physical processes, multiscale models are needed. In parallel, optimal compositions should be obtained by optimizing suitable descriptors. In this context, acceleration through artificial networks would allow for the probing of large compositional spaces, ultimately realizing the application of high-performance materials with optimized nanoparticles.

The effect of extended defects during exsolution and implications for activity
Brian A Rosen Department of Materials Science and Engineering, Tel Aviv University, Ramat Aviv, 6997801, Israel

Status
Heterogeneous catalysis is regarded as a surface-specific process; therefore, the majority of research related to defect engineering of catalysts relates to point defects which influence the surface chemistry. With the rise in popularity of defect engineering came an explosion of work on how 0D point crystalline defects (e.g. atomic substitutions, vacancies, and interstitials) impacts surface adsorption, kinetics, and phase stability [83]. By comparison, the impact of extended (1D and 2D) defects such as dislocations, grain boundaries, twin boundaries, and stacking faults on catalysis are far less studied since these types of defects are thought to influence bulk rather than the surface [84].
Recent work has shown that the presence of extended defects in a catalyst or its precursor can influence the exsolution mechanism and resulting catalytic properties. 2D crystalline defects have been shown to act as nucleation points for the exsolution process. For example, in-situ TEM under an H 2 atmosphere at 350 • C shown in figure 6 shows 1 nm Ni particles forming specifically on the Ruddlesden-Popper (RP) stacking faults of the LaNiO 3 backbone, as opposed to the defect-free regions where no Ni nanoparticles are found [85]. In fact, the exsolution process started nearly 150 • C higher when theLaNiO 3 was free of RP defects. Similar results have been shown for exsolution at grain boundaries, another type of 2D defect [7]. Interestingly, the impact of 2D defects for exsolution goes beyond the growth mechanism, and extends to the performance of the resultant catalyst. As an example, Ni particles grown on RP stacking faults in LaNiO 3 showed enhance stability against coking during methane dry reforming (DRM) compared to Ni particles grown from the pristine backbone [86].
Additionally, there is experimental and computational evidence that the presence of sub-surface stacking faults in the oxide support can influence the adsorption energy of atoms and molecules, whereby modulating catalyst performance via Sabatier's principle [87]. These findings underscore the potential importance of multi-dimensional defect engineering, particularly in the design of catalysts prepared by exsolution. Further advances in this field can lead to materials where these two effects are simultaneously exploited. In such a material, 2D defects are inserted in pre-determined positions and/or at predetermined densities, whereby modulating the electronic and adsorption properties of the catalyst. When exsolution begins, metal particle growth would initiate at those same locales, allowing for the active site to be planted near the region of enhanced transport or adsorption.

Current and future challenges
The primary reason why extended defects are not frequently reported in exsolution studies is that controlling the density of such defects is highly challenging [88]. Zero-dimensional crystalline defects such as vacancies and substitutions can be systematically and confidently studied as controlling such defects only requires changing the metal stoichiometry [13]. By comparison, inducing a known amount of growth mistakes in multiple dimensions is more complex. Typically, the parameters of the synthesis itself must be modulated in order to significantly change the extended defect structure of a material. Perovskite oxides represent an excellent platform for studying this phenomenon, as most of them can be synthesized using a wide variety of techniques including aqueous co-precipitation, non-aqueous co-precipitation, sol-gel, combustion, hydrothermal, flame spray pyrolysis, co-deflagration, solid-state, and gas-phase condensation (e.g. chemical vapour deposition (CVD), atomic layer deposition (ALD)) techniques. Since the formation of a pristine single crystal generally requires a steady and controlled growth process, it may be said that the probability for forming multi-dimensional defects increases by triggering a dramatic environmental change during the crystallization process (e.g. a rapid temperature or pH swing). Indeed, out of all the synthesis techniques mentioned above, the density of stacking faults in La-based perovskite oxides was found to be highest during co-precipitation, where the pH of the solvent is rapidly increased.
These challenges are amplified when trying to explain enhanced catalytic activity through computational modeling (e.g. DFT) of materials with extended defects. Even in the case where the experimentalist has coarse control over the density of stacking faults or grain boundaries in the catalyst, the exact locations of these defects cannot yet be determined ab-initio, meaning there will be significant discrepancies between defect structure of the real catalyst and the simulated one [87]. A possible solution to this problem is the use of the PLD (or similar) technique to precisely decide where the stacking fault will occur [89,90]; however, this strategy is typically is limited to fundamental rather than practical catalyst design.
The final challenge stems for the fact that there is not a standardized way to describe the density of extended defects for the sake of modulating catalytic action. A simple proposal for stacking faults is to count

Advances in science and technology to meet challenges
Solution-based oxide synthesis, while theoretically simple, comes with a myriad of free variables which can impact the growth of oxides destined for nanoparticles growth by exsolution. The development of synthesis-defect-structure relationships for oxides using solution-based methods requires careful experimental control and tuning. Well-planned studies can often yield clear results about how subtle changes such as the stirring rate, pH ramp, temperature ramp, or annealing temperature can influence the formation of extended defects [91]. In this context, a rigorous study which utilizes design of experiments [92] together with a quantitative descriptor of stacking faults in the resultant oxide may yield significant advancements.
The number of techniques available to identify, quantify, and characterize extended defects has also grown beyond ex-situ S/TEM and atomic force microscopy (AFM) [68] to include atom probe tomography [93] and in-situ TEM [94]. The increased availability of in-situ holders allows for more studies on the exsolution process itself, particularly allowing for pinpointing the influence of temperature ramp on the growth mechanism. XRD is an under-utilized but important technique that may show promise for the global quantification of stacking faults in certain samples by line-shape analysis. The peak shape of highly faulted perovskites have been shown to have increased Lorentzian character [95]. This technique requires further study, but if successful can easily be applied under in-situ conditions using an in-situ XRD chamber (e.g. XRK-900 by AntonParr) or a synchrotron, the latter providing time resolution on the order of seconds. Techniques which characterize the electrical properties of microstructural defects and interfaces have also seen advancement in recent years; for example, nanometric four-point probe experiments have been shown to allow for resistivity-structure relationships in the vicinity of a grain boundary or stacking fault [96].
The increased computing resources available to researchers allows for the study of more complex systems. Calculating the kinetic barriers for adsorption of gasses on faulted perovskite surfaces can often have lower numerical stability and present as highly nonconvex problems. The effect of electronic correlation and magnetism in certain backbones used in exsolution (such as LaFeO 3 ) can make barrier calculations prohibitive, and yet, as defect engineering gains more attention and computing power is improved, the computational contributions to this field will rapidly increase. Figure 7 shows an example where the adsorption energies of CO, CO + O and CO + O 2 (co)-adsorption on defective LaFeO 3 were calculated using the nudged elastic band method, however the barrier heights were not able to be calculated.

Concluding remarks
Exploiting the defect structure for exsolution catalysts requires a multi-pronged approach with improved control, identification, classification, and simulation of extended defects during the exsolution process and subsequent reaction. Significant improvements in in-situ microscopy, tomography, and computing power have already allowed for advancement, but clever techniques for ab-initio placement of extended defects in nanomaterials are still in its nascent stages. With these improved capabilities and initial reports, however, there must also come an expansion of research in the field of catalyst defect engineering to include multi-dimensional defects.

Status
Converting methane, the main component of natural gas and biogas, whose reserves are pretty vast in the world, into value-added chemicals, such as ethylene, has been driven for several decades by the demands in renewable energy along with the long-term decline in petroleum and coal reserves. The development of cost-effective, high conversion, high selectivity and stable catalysts remains a fundamental challenge. The recent several years have seen a revival in the popularity of metal exsolution as an in situ synthesis method to create highly active metal anchored in the surface of an oxide scaffold [97]. Exsolution materials are proposed as quite effective catalysts to enhance catalytic performance for the conversion of small molecules, such as CH 4 → C 2 H 4 [98], C 2 H 6 → C 2 H 4 [99] CH 4 → CO [100], due to the strong interactions between the exsolved metal and oxide interface, which would facilitate electron and ion transport at hetero-interfaces (figure 8) [101].
The oxidative coupling of methane (OCM) reaction (2CH 4 + 2O 2− ⇆ C 2 H 4 + 2 H 2 O + 4e − ) and non-OCM (NOCM) process (2CH 4 ⇆ C 2 H 4 + 4 H + + 4e − ) could be ideally accomplished efficiently in electrochemical processes. Moreover, green/sustainable electrochemical processes could be promoted through synergistically controlling electrochemical redox reactions and advanced catalysis on the porous electrodes. The solid oxide electrolyzers (SOEs) have gained much interest from both academics and industry in the electrochemical conversion of CH 4 to C 2 H 4 at high efficiency using renewable electrical energy, which involves the cracking of the C-H bond in CH 4 and the alternative coupling of the C-C bond to generate C 2 H 4 .
The OCM reaction has been extensively studied, while the problem of methane over-oxidation needs to be addressed. In our previous work [98], we utilized Sr 2 Fe 1.5+x Mo 0.5 O 6−δ (x = 0-0.1) anode with different distribution of exsolved Fe nanoparticles on the perovskite scaffold (xFe-SFMO) to realize OCM reaction in an oxide-ion-conducting SOEs. It shows that the highest C 2 product yields are 16.7% (11.5% C 2 H 4 + 5.2% C 2 H 6 ) with the C 2 selectivity of 82.2%, and that even after 100 h of high temperature (850 • C) operation, it still shows the efficient conversion (figure 9). Exsolution catalyst offers an efficient method to generate C 2 chemicals from methane through the OCM process using a SOEs reactor. The NOCM process using exsolution materials prevents excessive oxidation of methane and ethylene product, however, its thermodynamics is limited in a large range of temperatures and only beneficial at temperatures above 950 • C [102].

Current and future challenges
The exsolution catalysts used in the SOEs reactor for the OCM/NOCM reaction have advanced the impressive performance; however, disruptive advances are still necessary to make such technology commercially available. Several challenges remain for it to become a competitive technological process for obtaining desired products such as ethylene directly from natural gas or biogas.
In order to obtain exsolution catalysts for high activity and excellent durability in the conversion of CH 4 , a set of complex designs is required, resulting in relatively complex oxide formulations, often incorporating multiple elements with various defects chemistries. Such diversity facilitates the output of new compositions tailored for high-temperature applications of methane conversion and more systematic characterization of a reference system, which controls crucial parameters including components, size, distribution or amount, thus consolidating properties of resistance to amalgamation and surface area loss at high operation temperature, and strong interactions and increased triple-phase boundaries between the electron-conducting nanoparticles and the oxide supports. The lack of detailed compositional, structural, thermodynamic and kinetic information for such systems, together with the difficulties of theoretically modeling systems with multi-defects/dopants, have made it difficult to establish a set of commonly used rational design principles that are still largely empirical.  The influence of local interface structure of exsolution catalysts on activity and stability for methane conversion is still not clear. For example, in the oxide-ion conducting SOEs, the electrochemical transport of O 2− ions from the cathode to the anode is indeed a step to generate active species for CH 4 activation with the existence of exsolution catalysts. The mechanism and kinetics of the conversion of O 2ions to different types of active oxygen species, which is considered to be one of the key radicals initiating oxidation of CH 4 and selective coupling to ethylene at the anodic confined active metal-oxide interface, are still required to further in-depth investigate, thus guiding the advance of a novel active exsolved catalyst for the electro-thermal synergistic conversion of CH 4 to C 2 H 4 .
In addition to exsolution catalyst optimization and reaction mechanism investigation, ideal operating conditions for conversion of CH 4 to C 2 in SOEs, such as flow rate, temperature, and concentration of oxygen-containing feed gas at the cathode, CH 4 concentration at the anode, applied current, etc, together with more stable SOEs component materials and microstructural engineering of electrolyzer, need to be further optimized to achieve an optimized SOEs system for the coupling process of methane to C 2 hydrocarbons.

Advances in science and technology to meet challenges
In the face of the challenges and issues discussed above, the significant efforts of researchers in exsolution and SOEs have allowed us to point out some possible approaches to accelerate the industrial processes of CH 4 conversion into C 2 hydrocarbons using such reactor system.
Many parameters to design and control the structure of exsolution materials with the conversion, selectivity, and stability necessary for a commercial application are yet to be explored, including different oxide crystal structures and defect chemistries, less-explored anion sublattice, active elements (e.g. Fe, Ni, Co, FeCo, or NiCu) [100,103]. These could help improve bulk and surface transport of ions, reducing the exsolution temperature and providing new opportunities in methane electrochemical coupling under milder conditions. It is also crucial to diversify the microstructures of oxide perovskites such as thin-film, hierarchical, and porous. In addition to the material itself, more methods to trigger the exsolution of the material should be explored, including but not limited to chemical, electrochemical and light driving forces [102]. For instance, reduction temperature and time determine particle size and amount; the oxygen partial pressure and the nature of reducing atmosphere control the shape and composition of the nanoparticles. But more efforts are needed to exsolve single atoms (one of the ultimate frontiers in catalyst design), such as single iron sites, which show the highest methane conversion (48.1%) and ethylene selectivity (48.4%) at 1090 • C [104].
The development of advanced in situ characterization techniques, such as in situ XRD, in situ XPS, in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), in situ Raman, in situ XAS, in situ TEM, enable probing the elusive, rapidly changing reaction intermediates and morphology of the exsolved interface. Thus, it helps us better understand the mechanism of methane coupling and the relationship between exsolution materials and their catalytic performance. In addition, through computational simulation and modeling, machine learning, high throughput computation, the transport of heat and mass and the degradation mechanisms in electrochemical systems could be understood, providing theoretical guidance for the synthesis of in situ exsolution catalysts and the design of OCM/NOCM systems in SOEs reactors [105,106].

Concluding remarks
In summary, the direct conversion of methane to ethylene through an electro-/thermo-catalyzing process with exsolution materials is an important research topic that would bring significant changes to the traditional petroleum-based ethylene industry, while more efficient and sustainable technologies are still required. With substantial efforts focused on the study of the design of innovative exsolution materials, the mechanism of methane coupling with and an effective combination of advanced in situ characterizations, computational simulation, and machine learning techniques to overcome the challenges in the field, the C 2 production via OCM/NOCM process using SOEs reactors system with exsolved metal-oxide materials is likely to become feasible. Utilizing exsolution materials as a class of highly efficient catalysts for methane conversion to ethylene could be a key technology in future decarbonization of massive petrochemical industrial processes.

Exsolution for power generation from ammonia (NH 3 )
Guangming Yang 1 and Zongping Shao 1,2 1 State Key Laboratory of Materials-Oriented Chemical Engineering, Nanjing Tech University, Nanjing, People's Republic of China 2 WA School of Mines: Minerals, Energy and Chemical Engineering, Curtin University, Perth, WA, Australia Status NH 3 is a carbon-free and hydrogen-rich energy source that is widely used in many fields, such as nitrogen fertilizers, refrigerants, and fiber production. In addition to the advantages of green environmental protection, NH 3 energy also has the following unique advantages as a future alternative to hydrogen energy: (1) high energy density; (2) easy storage and transportation; (3) high safety; (4) mature large-scale production process [107]. It can also be directly used in fuel cells, especially solid oxide fuel cells (SOFCs). Compared with other types of fuel cells, SOFCs have the advantages of high electrode reaction rate, high ionic conductivity, high energy efficiency, and a wide range of available fuels. Meanwhile, high temperature can significantly promote the thermal decomposition of NH 3 , thus allowing NH 3 to be used directly as a fuel in SOFCs. According to the type of electrolyte, SOFCs can be divided into two categories: oxygen anion conducting electrolyte-based SOFCs (O-SOFCs) and proton conducting electrolyte-based SOFCs (H-SOFCs). Compared to O-SOFCs, H-SOFCs based on proton conductor electrolytes have unique advantages: firstly, the generation of nitrogen oxides (such as NO) can be effectively avoided; secondly, the reaction-product water will generate at the air electrode, which can avoid the dilution of the fuel and improve the conversion efficiency of the system; in addition, the proton conductor exhibits good ionic conductivity at medium and low temperatures, which can reduce the operating temperature of SOFCs [108]. Figure 10 shows the working principle of NH 3 -fueled O-SOFCs and NH 3 -fueled H-SOFCs.
In direct-NH 3 SOFCs (DA-SOFCs), NH 3 is first thermally decomposed to generate N 2 and H 2 , and then H 2 undergoes electrochemical oxidation to complete the NH 3 electrode reaction. Therefore, NH 3 electrodes have an important impact on the performance of DA-SOFCs. NH 3 electrodes generally can be divided into two categories: one is non-precious metal materials (such as Ni-based electrodes); the other is noble metal materials (Ru, Pd, Pt, etc). Past studies have proved that precious metal catalysts have a high decomposition efficiency of NH 3 , which can still reach more than 90% at 450 • C, but its expensive price limits its large-scale application [109]. Since Ni-based electrodes also have excellent catalytic properties against NH 3 at 550 • C, the most commonly used DA-SOFCs electrode is the traditional composite electrode of Ni and electrolyte ceramic materials. Another method to promote the catalytic activity of electrodes is to introduce nano catalysts into the electrodes. At present, the preparation methods of nanoparticle-modified electrodes that have been more studied in SOFCs are mainly composed of two steps of solution impregnation and low temperature sintering [110]. Although the electrodes prepared by the impregnation method have high catalytic activity, the nanoparticles have some disadvantages such as large particle size, uneven distribution, complex preparation process, easy sintering, and reduced electrode porosity, which are difficult to operate in practical applications. Recently, Myung et al improved the activity of catalysts by enabling in-situ growth of nanoscale phases from non-stoichiometric perovskite surfaces in a reducing environment [36]. Owing to the small size of the exsolved nanoparticles, the uniform distribution on the electrode surface, and the good exsolution-dissolution repeatability of the nanoparticles, they have been widely used in SOFC anodes. In this topic, we summarize the recent reports on a series of nanocatalysts prepared by the exsolution method for the application of DA-SOFCs, and propose the challenges that the exsolution method will face in the application of DA-SOFCs.

Current and future challenges
Among non-precious metal materials, Ni has the highest catalytic activity for NH 3 decomposition and owns the lower cost. Therefore, the current research on NH 3 electrode materials mainly focuses on Ni-based electrodes [108]. However, related studies have shown that high concentrations of NH 3 can cause Ni coarsening and destroy the microstructure of NH 3 electrodes, thereby reducing the performance and stability of DA-SOFCs [111]. Therefore, improving the anti-coarsening ability of Ni-based NH 3 electrodes is the key to enhancing the stability and catalytic activity of NH 3 electrodes. In order to improve the catalytic activity and durability of DA-SOFC anodes, the NH 3 catalytic activity and reaction sites can be effectively improved by using nanoparticle catalysts to modify the anode surface. Recently, Zhang et al reported a Fe-modified Ni-based ceramic anode, which effectively enhanced the activity and stability for NH 3 conversion [112]. Meanwhile, Fe-modified electrode can effectively improve the surface adsorption of NH 3 and reduce the desorption barrier of N 2 , which was also confirmed by first-principles calculations. In addition, the Ni nano-modified anode also showed obvious advantages in NH 3 decomposition reaction. A Ni-doped  [113]. Therefore, the BZCYYbN-based single cell using NH 3 as fuel produces a PPD of 523 mW cm −2 at 650 • C, which is increased by 36.6% compared with un-modified BZCYYb-based single cell. Cavazzani et al also confirmed that the dissolution of Ni also significantly promotes hydrogen oxidation process [114].
Noble metals usually have high electronic conductivity and catalytic activity, among which Ru-based materials have the highest catalytic activity for NH 3 decomposition. Zhu et al reported a high-performance NH 3 catalyst Ru-(BaO) 2 (CaO)(Al 2 O 3 ) (Ru-B 2 CA) and combined it with H-SOFC to obtain a PPD of h of 877 mW cm −2 at 650 • C under NH 3 fuel [115]. The degradation rate of H-SOFCs was significantly decreased by using Ru-B2CA catalyst for the pre-catalysis of NH 3 gas. Ultimately, the DA-SOFCs of this structure operated stably for 1200 h under test conditions. In addition, ruthenium not only provides excellent catalytic activity for the anode, but trace amounts of ruthenium doping can also promote the exsolution of metals in the perovskite in the form of alloys, which can greatly enhance the catalytic activity of DA-SOFC anode. , which led to a significant increase in the catalytic activity of r-PSCFRu for the NH 3 decomposition reaction [116]. The experimental results also show that the DA-SOFC with r-PSCFRu anode has a PPD of 374 mW cm −2 at 800 • C, which is nearly 30% higher than that of the cell with r-PSCF anode, and the operational durability is also improved. This shows the catalytic activity of NH 3 electrodes can be significantly improved by doping Ru in perovskite materials. Besides ruthenium, the noble metal palladium also exhibits excellent catalytic activity in DA-SOFCs. The catalytic activity of palladium for NH 3 [117]. Their BZCY-Pd oxide-metal-based NH 3 DA-SOFCs achieved a stable power output of 0.58 W cm −2 at 600 • C.

Advances in science and technology to meet challenges
The current research on surface exsolution catalysts for NH 3 fuel cell anodes has numerous hurdles. Although it is well recognized that surface exsolution of some metals can dramatically improve NH 3 catalytic activity, the appropriate temperature, particle size, nanoalloy formation process, and surface electrochemical reaction mechanism of nanoparticle exsolution are still cannot be analyzed exactly. Probing using in-situ TEM (in-situ TEM) and in-situ x-ray absorption fine structure (XAFS) spectroscopy are anticipated to provide decisive insights into these challenges. It is possible to adjust the ideal circumstances for nanoparticle exsolution by combining these two types of in situ characterization, which can simultaneously provide information on the morphology, composition, and diffraction of the crystal phase of exsolved particles with temperature and time. Additionally, the microkinetic modeling and DFT calculations can more clearly explain the reaction mechanism of the NH 3 -catalyzed process on the surface of exsolved nanoparticles. The optimal reaction sites for the surface activity of the desolated nanoparticles were theoretically predicted using the DFT simulations. To calculate the surface kinetic rates of desolation of various elements, the NH 3 adsorption energy and N 2 desorption energy on the surface of various metal particles could be estimated. The DFT calculation can also accurately determine the reaction energy barriers of NH 3 dehydrogenation and its intermediates ( at various active sites, which offers an accurate theoretical foundation for further investigation of the catalytic effect of nanoparticles out of solution on NH 3 decomposition. The stability of in-situ growth nanoparticles also meets great challenge in DA-SOFCs, since nanoparticles usually undergo agglomeration at high temperatures and reducing atmosphere. Especially, traditional nickel-based ceramics can be easily pulverized when suffer from high NH 3 concentration, leading to a serious degradation in long-term operation. Bimetal exsolution, such as Co-Mo, Ni-Fe, Ni-Ru, have been investigated in thermally NH 3 decomposition, which can not only improve the decomposition activity, but also hinder metallic nanoparticles agglomeration, sintering, and poisoning [118,119]. Song et al prepared a high-performance DA-SOFC via the in-situ exsolved NiCo alloy nanoparticles modified La 0.55 Sr 0.30 TiO 3−δ (LST) anode substrate, which is proved to have a fast NH 3 decomposition reaction dynamic rate because it effectively balances the process of adsorption and disconnection of the NH 3 and N 2 [120]. The intense interaction between NiCo alloy and LST perovskite as well as the inherent sintering resistance make DA-SOFC show a good stability at 700 • C. Therefore, bimetallic exsolution strategy serves as the essential candidates applied in DA-SOFCs anodes.
Although the exsolved nano-catalyst has excellent electrochemical performance and stability in DAFC anode, its mechanism is not well understood, especially whether the exsolved nano-particles also play other roles in the reaction? Recently, a new Ba(Zr 0.1 Ce 0.7 Y 0.1 Yb 0.1 ) 0.95 Pd 0.05 O 3−δ (BZCYYbPd) perovskite and used Ni-BZCYYbPd was reported as an anode for H-SOFCs [121]. The catalytic activity and chemical stability of the anode for NH 3 decomposition were successfully enhanced by doping 5 mol% palladium into a conventional proton conductor metal-ceramic material. This is due to the Pd nanoparticles exsolved from BZCYYbPd in a reducing atmosphere can provide additional active sites for NH 3 electrode catalysis. Meanwhile, the strong hydrogen adsorption ability of Pd metal keeps the NH 3 electrode surface with a high hydrogen concentration, which inhibits the coarsening effect of NH 3 on the Ni-based NH 3 electrode, thus enhancing the chemical stability of the NH 3 electrode. Finally, the H-SOFCs with Ni-BZCYYbPd as the NH 3 electrode exhibit superior performance (724 mW cm −2 @ 650 • C) and good stability for 350 h. Therefore, the exsolved Pd nanoparticles not only improve the catalytic activity of NH 3 , but also play a key role in the stability of the electrode.

Conclusions and challenges
The nanoparticle-modified hydrogen electrode material prepared by the exsolution method has excellent catalytic activity for NH 3 splitting and H 2 oxidating performance, which makes the performance of DC-SOFCs comparable to SOFCs fueled by H 2 . However, there are still some challenges in the DA-SOFCs prepared by the exsolution method, such as the long-term stability of the exsolved nanoparticles and the influence on the overall structure of the electrode when the nanoparticles are re-exsolved. At the same time, the mechanism of the exsolved nanoparticles in the operation of DA-SOFCs is not yet clear, and further in-situ characterizations such as Raman spectroscopy, synchrotron radiation photoelectron spectroscopy/absorption spectroscopy, and scanning tunneling microscopy are needed, thus providing a theoretical basis for the design and development of electrodes for the exsolution of highly active nanoparticles.

Yubo Zhang, Jakob Reinke, Travis A Schmauss and Scott A Barnett
Department of Materials Science and Engineering, Northwestern University, Evanston, IL, United States of America

Status
SOFCs have shown great potential to fulfill today's requirement in energy generation and storage due to their high efficiency, fuel flexibility, and low emissions [122]. While conventional cermet SOFC anodes such as Ni-YSZ have shown good electrochemical performance, they suffer from poor redox stability, coking in C-containing fuels, and poor sulfur tolerance [123]. Although extrinsic addition of catalytic nanoparticles via methods like precursor solution infiltration has been reported to alleviate these problems and enhance anode catalytic activity, these nanoparticles usually suffer from coarsening resulting in poor durability [124].
Exsolution  [125], where nano-sized Ru particles were observed on anode surfaces during cell operation and significantly improved the electrochemical performance. Since then exsolution has been extensively studied for SOFC perovskite anodes where metal nanoparticles of Ni, Co, Fe, Ru have been reported as well as bimetallic alloys of these elements [49]. One example is shown in figure 11 [126]. Different from infiltrated nanoparticles, exsolved nanoparticles have been shown to be socketed into the parent perovskite surface, making them more resistant to sintering and thereby offering better long-term stability [2]. Exsolution anodes general provide good redox cycling capability due to the low metal content compared to Ni-YSZ, and it may be possible to regenerate such anodes by redox cycling [127]. Moreover, exsolved anodes are more resistant than conventional cermet anodes to carbon coking [2] and sulfur poisoning [128].

Current and future challenges
Despite the development over the last 15 years, challenges remain for the application of SOFC exsolution anodes. Exsolution is usually induced by high temperature reduction of SOFC anodes, and it has been reported that reduction temperature [129], time [130] and atmosphere (pO 2 ) [11] can all affect the morphology, size and density of the exsolved nanoparticles, leading to different electrochemical properties. Thus, it is still unclear how to tailor the exsolution process to optimize the performance. Moreover, given the fact that exsolution is the result of decomposition, phase transformation of the parent perovskite may occur simultaneously with exsolution. Both positive [14] and negative [131] effects have been reported for the conductivity and electrochemical activity of the exsolved anode due to a phase change, suggesting a complicated synergy between exsolved particles and their parent oxide structure. Furthermore, the optimization and tailoring of the composition of exsolved particles remain challenging, especially for materials with multiple cations on B sites where bi-metallic exsolution happens upon reduction. A model has been proposed to predict the composition of the exsolved bi-metallic nanoparticles [27], but it needs further validation. It is also unclear how the alloy composition affects the electrochemical performance of the SOFC anodes.
In addition to electrochemical performance, operational durability of the exsolved SOFC anodes is also of great importance. Different from SOFCs with conventional cermet anodes, durability tests for SOFCs with exsolved perovskite anodes have only been in the range of hundreds of hours [27,40]; there is no consensus as to the main degradation mechanisms for these anodes. Moreover, there is a lack of long-term stability studies of exsolved SOFC anodes under different real-life fuel conditions (e.g. different steam contents, carbon and sulfur-containing fuels, and different redox cycles).
Finally, due to the operating conditions of SOFC anodes (high temperature, reducing atmosphere), it is challenging to study exsolved nanoparticle formation and oxide phase changes in operando, limiting our understanding of the exsolution process. Measurements made after cell operation and cooling to ambient temperature may introduce significant artifacts that obfuscate the true structure.

Advances in science and technology to meet challenges
To meet the challenges noted above, a better fundamental understanding of the exsolution process during SOFC operation is needed. In situ characterization methods such as in situ XRD [131], thermogravimetric (TG) analysis [27], and environmental TEM [11] have all been used to elucidate the exsolution process (examples shown in figure 12), but the number of studies focusing on the fundamental understanding of exsolution is limited. Developing and utilizing advanced characterization in the SOFC anode environment (high temperature, reducing atmosphere) will help reveal the anode morphological and compositional evolution including exsolved particle size, composition, and density as well as oxide phase changes, improving the mechanistic understanding of exsolution and leading to better control of the exsolution process.
Moreover, theory and modeling will complement the experimental studies to better understand exsolution. Beyond the DFT studies on phase reconstruction and nanoparticle formation during metal exsolution that have been reported [14,21], the scientific community would benefit from more modeling work to investigate the relationship between exsolution conditions, the microstructure and composition of exsolved anodes, and resulting electrochemical properties. Finite modeling analysis could help investigate the change of electrochemical properties of the material during exsolution and phase-field modeling could also provide a powerful tool to simulate the formation and evolution of exsolved particles.
Finally, while some studies have shown that exsolved nanoparticles can enhance the electronic conductivity of the anode [132,133], perovskite-based SOFC anodes generally suffer from low electronic conductivity compared with their cermet counterparts. This could lead to significant resistance losses when exsolved anode materials are scaled up from the button cell to the stack level. Although this could be minimized via optimized interconnect designs, it will likely be important to develop higher conductive exsolution anodes and related oxide current collector materials.

Concluding remarks
Alloy exsolution on SOFC anodes has been a fast-growing research field over the past 15 years. The enhanced electrochemical performance and stability against coking, impurity poisoning, and redox cycling have made exsolved perovskite materials a promising alternative to the incumbent cermet anodes. Despite the progress that has been made, there are great opportunities for further R&D, e.g. to improve the fundamental understanding of the exsolution process, study long-term stability, and develop new compositions with improved electronic conductivity, with the goal of developing highly active, durable exsolved SOFC perovskite anodes. Further development should also include implementation of exsolved materials in large-area SOFCs and stacks.

Status
Solid oxide electrolysis cells (SOECs) are advanced electrochemical energy storage and conversion devices with high energy efficiencies. They offer attractive high-temperature CO 2 and/or H 2 O electrolysis routes that can reduce anthropogenic CO 2 emissions, enabling large-scale energy storage and facilitate the integration of renewable energy into the electric grid [135][136][137]. A SOEC mainly consists of a fuel electrode, an oxygen electrode and a dense ceramic electrolyte. During operation, the electrode materials offer catalytically active sites for CO 2 and/or H 2 O reduction and oxygen evolution reactions respectively, while the electrolyte provides the required pathway for transporting oxide ions from the fuel to the oxygen electrode.
The electrochemical CO 2 and/or H 2 O conversion and utilization was originally presented as a technique for oxygen regeneration in spacecraft habitats in the US manned spaceflight program in the mid-1960s and as an in-situ resource utilization technology for Mars in the 1970s. In view of energy transition in the beginning of the 21st century, the concept has regained momentum as a key technological enabler for renewable energy approach of re-introducing CO 2 in the energy and chemical cycles [135][136][137].
The economic feasibility of transforming CO 2 emissions into value-added products strongly depends on the decarbonization policies (e.g. carbon taxation), as well as electricity prices. Technoeconomic studies have highlighted the potential of combining SOECs CO 2 and/or H 2 O electrolysis with current chemical processes, such as Fischer-Tropsch synthesis for liquid fuels production and methanation. In this context, a recent assessment of power-to-gasoline conversion reported a gasoline price of 2.25 euros liter −1 [138] Moreover, the production of synthetic natural gas through co-electrolysis in combination with thermal integration can achieve overall electric efficiencies of >80% while via low temperature electrolysis the efficiency is limited to 60%-70% [139] However, in such studies the necessity of overcoming material degradations issues is highlighted.
The approaches to deal with degradation issues are focused on altering operating conditions and modifying cell architecture by optimizing or substituting SOEC components, especially electrode or electrolyte materials. The concept of exsolution has been proposed and successfully implemented for increasing electrocatalytic activity and stability of electrode materials in SOECs. Due to its unique feature of probing the electrode interface and bulk, exsolution is considered as an emerging toolbox for solving electrode material degradation issues, thus bringing the concept to the required maturity for commercial applications [3,[135][136][137].

Current and future challenges
The technical maturity of CO 2 and/or H 2 O co-electrolysis is high but still not sufficient to reach the requirements for practical implementation or commercialization. This is due to several major challenges from which degradation is one of the most noteworthy followed by low activation/conversion efficiency [135][136][137]. Under practical operations, the lowest degradation reported at cell level is ∼1.7% per 1000 h at ∼1 A cm −2 , while at short stack level degradation is much higher, i.e. 18% per 1000 h [135]. Therefore, to reduce degradation and prolong the lifetime, SOEC studies has been focused on elucidating the degradation mechanism that can be used as guideline for the development materials [136,140].
The main technological challenges observed for the SOEC co-electrolysis cell with commonly used components include [135][136][137]141]: There are promising alternative materials that have the potential to address several of the aforementioned challenges, however, their implementation in SOECs will necessitate optimization of alternative fabrication methods. Current, focus has been given in the development of two classes of SOECs i.e. Ni-YSZ (Yttria stabilized zirconia) composite electrode and ScSZ (Scandia-stabilized zirconia) based electrolyte supported cells [141]. These two technologies are distinguished by the thick layer that provides mechanical support to the cell and this leads to operating temperature and performance discrepancies, and thus, it is essential to be able to implement both approaches in SOECs with exsolved electrode materials [141]. In Figure 13 are summarized the current challenges in the field and the future research directions in conjunction with exsolution concept.

Advances in science and technology to meet challenges
Recent studies in the SOEC field suggests that there has been significant progress towards limiting the aforementioned issues. Some challenges were addressed by modifying the operating conditions For instance, reactant purification and operation at lower temperature has been shown to hinder poisoning, while the reversible operation could decrease delamination of the oxygen electrode [135][136][137]. Herein, however, we will focus on solutions that could emanate by exploring alternative electrode components.
Regarding the development of SOECs, in addition to fundamental understanding, two main directions for enhancing materials performance include modification or decoration of SoA electrodes, and development of new advanced electrode materials. In both directions, it appears that exsolution could have a pivotal role [3,[135][136][137]. Exsolved nanoparticles display remarkable stability, limited agglomeration while they have low tendency for coking, due to their firm anchoring to the ceramic surface. Furthermore, the method is faster, cheaper and gives better access to hardly reachable locations than conventional impregnation methods, and concurrently maintain exceptional control over the amount and distribution of deposited particles [141].

Fuel electrode
Compared to traditional Ni-YSZ electrodes, the perovskite oxides may display mixed ionic -electronic conductivity, superior redox activity and long-term operation stability. Moreover, the socketed nature of the particles offer superior resistance to carbon deposition and sulfur poisoning than their counterparts prepared by wet deposition methods [2]. Perovskites decorated with exsolved nanoparticle of Ni [142], Fe [143], Co [144], NiFe [145] and CoFe [146] have been successfully implemented on increasing electrochemical activity (catalytic activity and conductivity) and redox stability of fuel electrodes for co-electrolysis. It worth to note that perovskites (in contract with Ni-YSZ) can operate in the absence any gaseous reducing agent (e.g. H 2 , CO) hence decreasing the total cost and complexity of the process. Table 1 summarizes recent representative co-electrolysis studies with exsolved fuel electrodes and provides a comparison with Ni-YSZ as fuel electrode on both electrolyte and Ni-YSZ electrode supported cells [142,147]. The comparison is challenging due to difference in reaction conditions, as well as cell architectures which implies the need for the creation of a protocol for benchmarking the performance of different electrodes. Nevertheless, the table provides evidence that exsolved electrodes are competitive to SoA materials in terms of electrocatalytic activity even when no protecting gas atmosphere is used.

Oxygen electrode
Apart from fuel electrodes either in electrolysis or fuel cell operation, the concept has shown potential for improving the activity of oxygen electrodes [148][149][150] For instance, Ag exsolved nanocatalysts on perovskites have been reported as highly active oxygen electrodes for SOFCs [149,150]. The excellent performance in terms of electrocatalytic activity and stability is attributed to the synergistic effect of socketed Ag, which acts as a catalytic core to boost surface exchange rates, combined with the promotion of the charge transfer reaction at the Ag-oxide interface. Electrodes with such properties worth to be investigated as a possible solution to delamination, that it is observed in La 0.8 Sr 0.2 MnO 3−δ (LSM) based electrodes at long term operation.    [126]. This approach could be adapted to expand the concept of exsolution in both fuel and oxygen electrodes for boosting the overall performance and/or decrease the operating temperature.

Concluding remarks
As a powerful toolbox that could combine numerous elements and preparation methods, exsolution is already a proven option to solve a series of challenges in solid oxide co-electrolysis of CO 2 and H 2 O towards upscaling. The solutions identified primarily concern the cell assembly development, such as improved fuel or oxygen electrodes to enhance redox stability and decrease operating temperature towards developing cheaper cell stacks with long-term stability. The reported advances in materials fabrication and exsolution triggering methods (see chapter 1) may be combined with modern machine learning in order to develop high performance electrodes with tailored properties. It is critical that the future works in the field should focus not only on new materials development but also on evaluating exsolved nanoparticles in long-term durability studies in realistic conditions or even in the presence of poisonous compounds. Also, the area can be critically benefited by advances related to perovskite electrode-supported cells to enable high performance cells (where both electrodes will be nanoengineered with exsolution) with thin electrolytes similar to the Ni-YSZ supported cells which have been practically commercialized.

Status
Oxide-based solid-state proton conducting systems [151] show significant potential for energy conversion applications including fuel cells/electrolysis cells, membrane reactors, and sensors. Their ability to operate at intermediate temperatures (350 • C-650 • C) enables direct integration with important thermochemical processes, relaxes balance-of-plant requirements, and reduces degradation [152]. However, low operating temperatures lead to significant activity challenges, both for the oxygen reduction/evolution reactions facilitated by the air/steam electrode and for the fuel reduction/oxidation reactions facilitated by the fuel electrode. These challenges have led researchers to consider exsolution, first applied to oxygen-ion conductors, to increase electrochemical performance in protonic oxides as well. Proton-conducting air/steam electrodes are typically based on mixed ionic and electronic-conducting perovskites that incorporate catalytically active transition metal(s) (such as Fe, Co, Ni, and/or Mn) on the B-site. Proton-conducting fuel electrodes generally involve ceramic-metallic (cermet) composites of Ni metal and a ceramic proton conductor. Electrocatalytic activity is generally sufficient at higher operating temperatures but diminishes greatly at lower temperatures (especially <500 • C) [153]. Conventional electrode formation leads to relatively coarse microstructures and can trigger surface segregation processes that compromise activity, especially at lower temperatures [154][155][156].
Several strategies can enhance electrode performance, including infiltration [140], one-pot composite synthesis, patterned electrodes, and freeze casting. However, these approaches require additional processing steps and introduce constraints that add complexity and expense to cell fabrication. As a promising alternative, in-situ exsolution processes can greatly improve catalytic activity without requiring additional fabrication steps [13]. The approach involves the deliberate substitution of appropriate (typically transition-metal or noble-metal) dopants into a suitable oxide-based parent electrode which, when subjected to appropriate treatment conditions, triggers the exsolution of second phase nanoparticles derived from these dopants that decorate the electrode surface. These exsolved nanoparticles provide additional catalytic surface area and catalytic functionality and exhibit improved stability and cohesion with the parent electrode support relative to infiltrated, composited, or decomposition-derived nanoparticle catalysts. Thus, exsolution promises a new frontier to enhance the stability, functionality, and overall catalytic activity of protonic oxide devices.
Exsolution techniques have been applied to several protonic oxide systems. Ba 0.95 Ag 0.05 Co 0.4 Fe 0.4 Zr 0.1 Y 0.1 O 3−δ air/steam electrode can undergo an Ag exsolution process under oxidizing conditions through a high-temperature steam-mediated air annealing process [157] (figures 14(a) and (b)). Exsolution of Ni and/or Ni-alloy nanoparticles from various proton-conducting fuel electrodes under reducing conditions has also been demonstrated. As in other systems, these nanoparticles are observed to exsolve in a socket-like fashion, leading to excellent catalytic performance and enhanced durability. For example, a Ni/BaZr 0.8 Y 0.2 O 3−δ electrode with exsolved Ni nanoparticles achieved by in-situ conditioning of a solid-state reactive-sintered support demonstrated remarkable activity and stability under fuel cell operation with a variety of fuels over thousands of hours of testing [158] (figures 14(c) and (d)).

Current and future challenges
Recent exsolution studies in oxygen-ion conductors have established general mechanistic insight [159], however little is known about potential mechanistic differences that might be associated with exsolution in protonic oxide systems. Exsolution represents a chemically driven heterogeneous phase transition, and involves four distinct physical processes of diffusion, reduction, nucleation, and growth. Typically, the chemical potential of oxygen in the gas phase surrounding the electrode (via the imposition of highly reducing conditions) provides the driving force for exsolution, although as previously discussed, the use of steam in mixed proton/electron-conducting oxides can induce exsolution under oxidizing conditions as well. Voltage bias, strain, topotactic conversion, and deliberate non-equilibrium cation deficiency can also serve as potential driving forces for exsolution [159]. For the case of oxygen reduction-driven exsolution, Zhu et al [27] has proposed a thermodynamic model accommodating both single and multi-cation exsolution processes. Generalized here for a simple perovskite undergoing single-cation exsolution of a reducible B-site dopant B ′ with oxidation state m, and assuming complete exsolution of the B ′ dopant: the critical oxygen partial pressure for exsolution (P Red O2 ) is: where a oxide is the activity of the oxide before exsolution, and a oxide-ex is the activity of the oxide after exsolution. Few oxide systems have sufficient thermodynamic information to establish a oxide and a oxide-ex , although TG experiments [27] have enabled estimation of these quantities in limited cases. Because thermodynamic information is limited and exsolution models remain underdeveloped, predicting which parent oxide and dopant combinations lead to exsolution under accessible conditions remains a challenge. This challenge is further complicated for multi-cation exsolution processes and for protonic oxides, where competition between more/less reducible cations and/or multiple charge-compensating defects can further complicate analysis. A germane example of this complexity is given by the aforementioned case of steam-driven exsolution in the mixed proton/electron-conducting Ba 0.95 Ag 0.05 Co 0.4 Fe 0.4 Zr 0.1 Y 0.1 O 3−δ system.
It is hypothesized that proton uptake in steam-rich environments triggers the exsolution of the highly reducible noble metal dopant (M) with metal cation vacancies created for charge compensation: Importantly, steam is required to trigger exsolution-changes in oxygen partial pressure alone are insufficient. Also, the process only occurs in 'triple-conducting oxides' , special materials which accommodate the transport of three mobile charged defects-protons, oxygen vacancies, and electron holes. The steam-mediated exsolution process is not observed to occur in more conventional oxygen vacancy/electron-hole mixed conducting oxides. This example demonstrates how the unique steam-modulated defect chemistry in protonic oxides represents not only a challenge for implementing exsolution processes-but also an opportunity-as it provides new pathways and mechanisms for exsolution not available in conventional oxygen-ion conducting systems.
Beyond improved mechanistic understanding and the consequent ability to predict suitable parent oxide and dopant systems amenable to exsolution, control over exsolved nanoparticle size and distribution represents another significant challenge. The relative rates of nucleation and growth for exsolved particles depend on many intrinsic and extrinsic properties including surface energies/ wetting angles, cation diffusion rates, surface curvature/orientation, grain boundary densities, and mechanical stress/strain. Exsolved particle size and density depend on the strength of the exsolution driving force and exposure time, with larger driving forces and higher exposure times generally leading to larger particle sizes. Furthermore, the shape of exsolved nanoparticles can also be unpredictable. For example, mixtures of exsolved particles composed of Ni and Co can change shape from spherical to square or faceted based on the Kirkendall effect [160]. Recent work has applied classical nucleation and growth theories to describe exsolution, with strain, reactant-concentration, and diffusion-limited growth models proposed [64]. More work in this area is clearly warranted.
Exsolution in protonic systems brings additional challenges not typically present in conventional oxygen-ion conducting systems. Protonic-ceramic devices generally operate at lower temperatures than conventional oxygen-ion conducting devices. Thus, to trigger and control exsolution, initial conditioning may be required at temperatures far higher than the designed operating temperature of the device, with potentially unknown or unexpected consequences. Furthermore, most protonic oxides involve high concentrations of Barium (often the dominant A-site constituent), which can lead to chemical stability issues during exsolution, as Ba surface segregation or Ba-based second phase decomposition products may compete against or complicate B-site cation exsolution under certain conditions (especially under high steam or CO 2 -rich conditions) [161].

Advances in science and technology to meet challenges
To further advance scientific understanding, detailed mechanistic studies leveraging advanced (especially in-situ) characterization techniques are needed. Physiochemical analysis and computational calculations are also desired to provide fundamental thermodynamic, defect-chemistry, and microstructural information to quantitatively model and predict exsolution behavior. As identified earlier, special attention must be given to the mixed-ionic conduction character of most protonic oxides, as multiple species can play a role in mediating the exsolution process.
Powerful characterization techniques are needed to probe the fundamental mechanisms of exsolution, including synchrotron-based techniques (structure, bulk chemistry), in-operando XPS (surface chemistry), and in-operando TEM (surface structure, nanoparticle morphology, nucleation and growth kinetics). Synchrotron-based in-situ x-ray absorption near edge structure and XRD can reveal the structural evolution and coordination of exsolved metal nanoparticles [162]. In-operando electron microscopy can also track exsolved particle evolution in real-time (and with temperature) providing unprecedented insights into the kinetics of exsolution ( figure 15(a)) [163]. The reader is directed to section 2 of this roadmap for further information. These intensive analytical techniques have not yet been not adapted to protonic oxides due to the complex multi-species conduction behavior and challenges working with water vapor in most analytical systems.
Computational modeling is also essential to inform the rational design of exsolution processes, and is discussed in more detail in section 3 of this roadmap. Basic trends in exsolution propensity from a perovskite lattice have been successfully predicted through DFT, see figures 15(b) and (c), with co-segregation energy between the host lattice and the exsolved metal as an important predictive factor [15]. Hofu similarly predicted favorable exsolution of bimetallic nanoparticles from a host lattice in an oxide-based CO 2 electrolysis cell by calculating segregation energies [164]. Unfortunately, DFT calculations are confined to relatively small systems (at most a few hundred atoms) and cannot easily incorporate important larger-scale features such as surface curvature or extended defects (grain boundaries, dislocations). To address this issue, the PFC model can potentially link the atomic and mesoscopic scales [165]. Combining PFC models with in-situ characterization may by particularly well suited for understanding and predicting exsolution dynamics.

Concluding remarks
Significant enhancements in electrochemical performance and durability are achieved by employing in-situ exsolution-based strategies in proton-conducting electrode materials. The presence of multiple charge-compensating defects (protons, oxygen vacancies, electron holes) complicates exsolution behavior. In particular, the steam-modulated defect chemistry in protonic oxides challenges mechanistic understanding but also provides new exsolution pathways not available in oxygen-ion conductors. While analogues from oxygen-ion conductors may provide a baseline for modeling exsolution in protonic oxides, additional insight from in-situ/in-operando characterization, and new mechanisms that capture the role of protonic defects are needed. Here, atomic level (e.g. DFT) and the mesoscale (e.g. PCF) modeling can contribute by informing basic defect-chemistry thermodynamics as well as the longer-range kinetic phenomena that determine exsolution dynamics. There are plenty of opportunities to contribute to this exciting and expanding field of research.

Status
Methane reforming encompasses a series of industrially relevant reactions that transform this hydrocarbon into CO and H 2 , a mixture known as syngas, which could be further transformed into liquid fuels in Fischer-Tropsch units. These reactions are steam reforming (SMR), dry reforming (DRM), bi-reforming (BRM) and tri-reforming of methane (TRM). In SMR, methane reacts with steam according to equation (1). The CO produced in SMR is further transformed via the water gas shift (WGS) reaction (equation (2)) to produce more H 2 , and importantly CO 2 , which is removed through pressure swing absorption to produce pure H 2 . SMR is still nowadays the major technology for hydrogen production worldwide. In DRM, CH 4 reacts with CO 2 to produce syngas in a H 2 :CO molar ratio equal to 1 (equation (3)). Deviations from that ratio are normally caused by the Reverse WGS (RWGS) reaction (equation (4)). In BRM, methane reacts with both H 2 O and CO 2 , resulting in syngas with a H 2 :CO ratio of 2 (equation (5)), which is ideal for methanol synthesis [166]. Finally, oxygen can be also incorporated as reactant to adjust the final H 2 :CO in the so-called TRM, which is a combination of steam (equation (1)) and DRM (equation (3)), with methane partial oxidation (equation (6)), Among these technologies, DRM has the advantage of using CO 2 as reagent. In comparison with SMR, using DRM will suppose a decrease of 0.5 Gt year −1 of CO 2 [167]. In addition, dry, bi and tri-reforming show great promise in the transformation of biogas (a mixture of CH 4 , 50%-75%, and CO 2 , 25%-50%) into syngas [168]. All these processes are endothermic and, therefore, require operation at high temperatures, normally above 700 • C. These harsh conditions generally cause sintering of the catalyst particles, conventionally Ni supported in oxides such as Al 2 O 3 , CeO 2 or MgO. In addition, Ni catalysts are prompt to form carbon depositions, which can cause catalyst deactivation and reactor clogging. Nobel metal catalyst are more resistant to carbon depositions; however, their high cost has prevented its industrial commercialization.

Current and future challenges
Based on the above, it is paramount to develop more stable metal-supported catalyst with higher resistance against particle sintering and coking, and high selectivity to syngas, avoiding side reactions. For that reason, nanocatalysts prepared via the exsolution method have gained considerable attention for methane reforming reactions, especially for DRM of methane. This technology has not reached the industrial level yet and, therefore, the design and development of more efficient and robust materials remains a fundamental aspect to achieve commercialization. Exsolution, thus, emerges as a promising materials design route, since exsolved nanoparticles show robustness and low tendency to carbon formation [2]. Motivated by the remarkable performance of exsolution-functionalized fuel electrodes in SOFCs and electrolyzers, perovskites decorated with exsolved nanoparticles were successfully tested for DRM. Probably, the first work fully devoted to DRM was by Zubenko et al, which exsolved Re-Ni-Fe alloys from LaFeO 3 perovskites, with CH 4 and CO 2 conversions close to 100% at 900 • C [169]. However, researchers have applied the exsolution concept to other oxides that have been previously used as supports in DRM. Thus, in the last years the study of exsolution has been extended to other oxides rather than perovskites, which has been the most widely oxide host, especially in the solid oxide cell community. For methane reforming, exsolution has been tested in a variety of oxide supports, such as, pyrochlores [162], yttria-stabilized zirconia [170], MgO [171], spinels [172], ceria [97], double perovskite [173] or RP phases [174]. All these studies targeted at long-term catalytic stability while assuring high resistance to sintering and low coking, paying special attention to the fundamental aspects of nanoparticle exsolution and structure-property relationships.

Advances in science and technology to meet challenges
The benefits of nanoparticle exsolution against more conventional methods, such as impregnation, were soon reported. Naeem et al prepared catalyst based on Ru exsolution from Sm 2 Ru 0.2 Ce 1.8 O 7 pyrochlores, and compared its activity for DRM versus Ru/Sm 2 Ce 2 O 7 catalysts prepared by impregnation and borohydride reduction (figures 16(a) and (b)) [162]. In figure 16(a), it can be seen that over the course of a 50 h catalytic test, the only material with stable performance was the catalyst prepared via exsolution, exhibiting a 70% of CH 4 conversion. The high temperatures, at which some oxide hosts, e.g. perovskites, are synthesized to obtain pure phases, detrimentally affect to the material's textural properties. For instance, double perovskites used for DRM with exsolved ternary alloyed nanoparticles have specific surface areas (SSAs) that lie between 1.2 and 2.3 m 2 g −1 [175], whereas for high-performance MgO based catalysts exhibit SSA values around 70 m 2 g −1 [167]. Thus, increasing the SSA of the host oxides, while guaranteeing a pure phase and the ability to exsolve nanoparticles remains then a major challenge in the field. A solution to overcome this problem was recently reported by Xiao and Xie, which fabricated porous single-crystalline CeO 2 monoliths from where Ni nanoparticles were exsolved (figures 16(c)-(e)) [97]. This strategy led to SSA values between 25 and 30 m 2 g −1 an order of magnitude higher compared to the aforementioned perovskites. The exsolved Ni nanoparticles presented high dispersion and resistance to sintering after a catalytic test of 240 h ( figure 16(c)), in which the CH 4 conversion values were very stable ( figure 16(e)). Xiao and Xie, studied several Ni contents, which is limited by the low solubility of Ni in the CeO 2 lattice, finding that CeO 2 with 2.31%wt of Ni reached almost 100% CO 2 conversion at 650 • C [97]. These results illustrated the prospect of nanoparticle exsolution in porous oxides.
An additional attractive feature of exsolution is the possibility of exsolving metallic alloys in a facile manner [176]. This is especially linked to the ability of perovskite oxides to host several transition metal cations in the B-site of the crystal lattice while maintaining a pure phase. These cations can recombine under certain exsolution conditions under metallic alloys, which can unlock unprecedented catalytic properties compared to single-element metallic nanoparticles. Several works have reported exsolution of metallic alloys for DRM, e.g. Re-Fe-Ni [169], FeNi 3 [173,177] or Fe-Ni-Co [175]. In this last case, Joo et al fabricated these ternary metallic alloy nanoparticles via topotactic exsolution in PrBaMn 1.7 Co 0.1 Ni 0.2 O 5+δ, infiltrated with a solution of Fe precursors to obtain the multi-elemental alloy ( figure 16(f)). Importantly, the authors observed that the ternary exsolved alloy outperformed the bimetallic CoNi and metallic Ni and Fe exsolved nanoparticles as DRM catalysts ( figure 16(g)) [175]. Based on DFT calculations, Joo et al ascribed this improvement to an increase charge donation ability for the ternary alloy, which facilitates the dissociation of CO 2 [175].
Finally, a third approach (together with oxide porosity and exsolved metallic alloys) that exhibits great potential is shape-controlled nanoparticle exsolution [178]. Very recently, Kim et al reported on the possibility of changing the morphology of the exsolved nanoparticles by adjusting the exsolution conditions [178]. Namely, by increased time (24 h) and temperature (900 • C-1000 • C) it was possible to obtained pyramidal exsolved Ni nanoparticles [178]. These shape-controlled exsolved nanoparticles exhibited higher resistance to coking than the spherical shape ones, and higher CH 4 and CO 2 conversion over the course of a 390 h DRM catalytic test [178].

Concluding remarks
Exsolution enables the formation of highly stable nanoparticles with remarkable resistance to sintering and coking, the two main issues that catalysts face in methane reforming reactions. This fact has prompt the development of advanced catalysts with improved performance over prolonged operation. Recent advances in which oxide porosity is increased and nanoparticle shape and composition are easily controlled by exsolution can contribute to expand the application of exsolved nanocatalysts in DRM reactions, which still seeks for more efficient and robust materials to reach industrial development.

Status
To achieve a future where both economy and social prosperity are sustainable, independence from depletable resources is indispensable. Thus, it is necessary to establish a closed carbon cycle for industrial and energy processes. One way is to capture CO 2 from flue gas and transform it back into value added chemicals (CCU-carbon capture and utilization) [179]. In addition to general technologies for CO 2 utilization [180], exsolution materials represent a novel promising alternative [44]. The compositional flexibility of e.g. perovskites opens the possibility for a materials design approach [181].
One of the core strengths of exsolution catalysts for CO 2 utilization is their bifunctional nature: the exsolution process creates a surface decorated with metal nanoparticles embedded in a highly reactive oxide surface [44]. This makes them suitable for bifunctional reaction mechanisms, which are often required for CO 2 valorization; with one type of active centers for CO 2 adsorption and activation, e.g. in the form of oxygen vacancies, located at the oxide support, and a second type for hydrogen or methane dissociation on exsolved nanoparticles, e.g. metallic Ni, Fe, or Co ( figure 17). Furthermore, the close vicinity of both types at the interface between host lattice and exsolved nanoparticles is crucial for the enhanced performance. Prominent examples for CO 2 utilization processes which benefit from this bifunctionality are RWGS, (equation (7)), DRM of methane (DRM, equation (8)), or methanol synthesis (equation (9)): For thermodynamic and selectivity reasons, the first two processes are operated at high temperatures. For RWGS, higher operation temperatures (i.e. 700 • C-800 • C) shift the equilibrium to the product side. Even higher reaction temperatures (800 • C-1000 • C) are required to shift selectivity towards DRM [182]. Methanol synthesis is operated at much lower temperatures (200 • C-300 • C), but requires high pressures (>30 bar) to shift the equilibrium to the products [183].
Perovskite oxide catalysts are prime candidates for the high temperature reactions (RWGS + DRM), due to their enhanced thermal stability and tunable exsolution properties. Their rich surface oxygen chemistry (i.e. easy formation and refilling of oxygen vacancies) furthermore prevents catalyst deactivation by e.g. coking or carbon nanotube formation (via oxidation of deposited carbon). From an industrial point of view, RWGS is the preferred CO 2 utilization method, as it can be integrated in already established large-scale processes [184]. Moreover, exsolution catalysts display high intrinsic selectivity towards RWGS.

Current and future challenges
Due to high necessary reaction temperatures for efficient RWGS and DRM, requirements regarding properties and stability of the exsolution materials are demanding. The composition of e.g. doped perovskites has a huge influence on their lattice stability, surface chemistry, and exsolution properties. Different material classes (e.g. titanates, ferrites, cobaltites etc) feature different base catalytic properties and stabilities. Commercial WGS catalysts are for instance Fe 2 O 3 /Cr 2 O 3 mixtures, making particularly ferrites interesting for RWGS. The proper choice of the A-site cation (e.g. redox-capability) with additional A-site doping can significantly improve the oxygen surface chemistry (i.e. vacancy dynamics) for CO 2 activation, figure 18(A). B-site doping for preferential exsolution of e.g. Co (also highly active for RWGS) can further improve catalytic performance. In summary, a huge parameter space needs to be examined to optimize the catalytic properties via a materials design approach [4]. Predictions via theoretical modeling (including machine learning and DFT) to efficiently determine the optimum between material stability and catalytic performance are of increasing importance [185].
For industrial applicability, exsolution catalysts additionally must be stable in varying gas environments, as in large-scale reactors gas compositions and temperatures, and hence chemical potentials, change across the catalyst bed (from product-lean at the entrance to product-rich at the end of the reactor). For DRM reactions at high conversions (i.e. high H 2 content), this can lead to strongly reductive conditions that might result in decomposition of the host lattice. Even without complete decomposition, A-site cation segregation is often observed at high temperatures, thus causing deactivation (e.g. carbonate formation, figure 18(C)). How nanoparticle exsolution is achieved constitutes an important question: Size, shape and distribution of the formed nanoparticles strongly depend on process parameters, such as temperature, time, and gas phase (i.e. either reductive pre-treatment or in-situ exsolution in a reaction environment) [186]. This can have a huge effect on the achievable CO 2 utilization efficiency, as has been shown by recent studies that prove the beneficial effect of in-situ exsolution on the reactivity [44,187]. Also, additional growth of nanoparticles during operation is possible. For structure sensitive chemical reactions this needs to be considered.
Another particularly important challenge is to increase the active surface area of exsolution catalysts. Classical synthesis routes, e.g. Pechini synthesis, sol-gel methods, or solid-state reactions, lead to materials with relatively low surface areas (below 10 m 2 g −1 ). Usually, high temperature calcination steps are required to achieve phase purity; however, this decreases the surface area. Therefore, novel synthesis routes that enable the production of larger quantities of high surface area (50-100 m 2 ) exsolution catalysts need to be developed.

Advances in science and technology to meet challenges
In-situ or operando studies (e.g. spectroscopy, diffraction, microscopy) are required to optimize exsolution catalysts with respect to surface structure and composition and to correlate structure and surface chemistry, figure 18(D). However, the high operation temperatures of RWGS and DRM make these investigations particularly challenging. Hence, customized measurement cells have to be developed that allow precise control of the process parameters (even at high temperatures-e.g. above 800 • C-and in strongly reducing gas atmospheres; exact determination of the temperature at the sample surface is often particularly challenging) and still deliver excellent in-situ/operando data, which is needed to monitor possible intermediate species, reaction pathways, and possible deactivation phenomena [188].
Detailed studies regarding long-term performance and stability of exsolution catalysts under realistic operating conditions (high T and high p) are also missing; however, this information is crucial for further development of industrially relevant catalysts. Especially the behavior under realistic gas streams (often complex mixtures or containing contaminants) needs to be addressed for efficient application of industrial scale CO 2 utilization processes [189], e.g. issues connected to sulphur poisoning.
Transitioning from fundamental scientific investigations to first real applications necessitates additional prerequisites to be met. As exsolution properties can strongly vary with small compositional variations, e.g. Figure 18. Opportunities and challenges of perovskite-based exsolution catalysts used in CO2 utilization. Top: Changing the A-site base element and/or B-site doping allow tuning of catalytic properties (e.g. performance optimization for RWGS reaction; (A). The electronic structure of the catalysts can be influenced via doping as well: depending on the respective dopants, the Fermi level (E Fermi ) can be shifted up or down (see densities of states calculated with density functional theory shown in (B). Bottom: the same material (a Co-doped perovskite) shows differently sized exsolved nanoparticles (highlighted with red circles in SEM images; (C), depending on the conditions during exsolution (C1: lower temperature, C2: higher temperature). Additionally, unwanted triangular carbonate crystallites can be seen in C2 (blue marking), which formed during high temperature RWGS reaction. Operando XRD data (taken during RWGS at a synchrotron facility; (D) support those findings and confirm the formation of additional phases (metallic/oxidic nanoparticles, unwanted carbonaceous species) with increasing temperature. stoichiometry, A-site deficiency, or lattice strain [13], novel synthesis strategies are needed that enable highly reproducible production of doped perovskites; especially on an industrial scale. A promising approach is e.g. synthesis via flame spray pyrolysis [190], which would produce catalysts with high surface area as well. As a further step, the powder materials usually used in fundamental studies need to be transformed into structured catalysts that can be utilized in industrial processes (e.g. by mixing and supporting the perovskite on high surface area alumina) [191]. Furthermore, the issue of integrating exsolution catalysts for CO 2 utilization into existing industrial processes needs to be addressed.
For catalyst optimization, comparable data is urgently needed, especially regarding catalytic performance of different materials-for instance in the form of unified turn over frequencies or specific activities with detailed documentation of reaction and exsolution parameters.
In terms of industrial applicability, CL is a further interesting topic, as it can be paired with discontinuous CO 2 -intense processes [192]. Although the rich redox chemistry of exsolution catalysts makes them promising materials, detailed performance studies for extended looping is rare. Especially the continued phase changes (and the more extreme conditions, e.g. pure H 2 instead of a mixture) may lead to pronounced performance loss over time due to sintering and phase separation processes.
Concluding remarks CO 2 utilization via exsolution catalysts is a rapidly emerging field. The reasons are diverse and include: (i) High temperature stability of the host lattice (e.g. perovskites) which is necessary to withstand the high operation temperatures of CO 2 conversion processes, as required for RWGS and DRM. (ii) Intrinsic CO 2 activation capability of the reducible host lattice through surface oxygen vacancies. (iii) Catalyst surfaces with exsolved nanoparticles with highly improved H 2 or CH 4 dissociation on the metallic nanoparticles. (iv) A bifunctional reaction pathway as a result of the combination of (ii) and (iii).
(v) A materials design approach via selective doping of the A-and B-site of the host lattice enables tailoring catalysts materials to industrial process conditions (e.g. optimized operating temperatures and pressures).
However, several challenges (e.g. long-term stability of the catalysts under operating conditions) have yet to be met on the pathway to industrial applications. First, novel synthesis routes need to be developed that allow the synthesis of larger quantities of high surface area materials. Especially achieving and maintaining high surface areas is challenging. Finally, exsolution materials have to be transformed into a shape (structured catalysts) that is applicable in industrial processes.

Exsolution for CO oxidation
Evangelos I Papaioannou School of Engineering, Newcastle University, Newcastle, United Kingdom

Status
The reaction between CO and O 2 (CO oxidation reaction) is seemingly straightforward and is one of the most widely studied chemical reactions. Although seemingly uncomplicated, it still attracts wide scientific interest, not only because it serves as an excellent model reaction for surface science investigations but also because CO oxidation finds a wide range of applications in air purification, automotive exhausts, fuel cells, gas sensors and lasers among others [193]. Up to now noble metals like platinum, palladium and iridium are frequently used as catalytic materials for CO oxidation while transition metals and metal oxides were proposed as lower cost alternatives [194]. Redox exsolution has emerged over the past decade as an alternative and simple synthetic platform for supported nanoparticles synthesis. Exsolved nanoparticles demonstrate enhanced activity across a wide range of application areas combined with high stability over thermal agglomeration and poisoning as opposed to their deposited analogues, owing to their partially socketed nature which results in a strained particle-oxide interface [2]. The recent years a number of different types of design elements (such as crystal structures, host lattice and exsolvable elements) and synthesis routes have been reported aiming for high catalytic activity and stability at a variety of operating conditions. Particularly for the CO oxidation reaction Co, Fe and Ti have been incorporated as B-site host lattice elements, while Ni, Co, Rh, Ir and Pd have been reported as exsolvable elements [160,[195][196][197][198][199][200]. The solid state and the sol-gel methods were the two most prominent synthesis techniques used with the aim to achieve either strict control of stoichiometry (a critical parameter for preparation of materials with large amount of point defects) or phase formation at low temperatures (∼700 • C) when smaller grains are desired [196]. Recently the redox exsolution efforts in CO oxidation have focused on the design of compositionally diverse transition metal oxide nanoparticles capable of achieving site activities previously observed in their noble metal analogues. Simultaneous exsolution of two or more reducible metal species or post synthesis morphological and chemical transformations are two representative examples of novel methods to unlock superior functionality and stability of such catalyst systems [160,176]. This concept paved the way for a step change in the design of earth-abundant metal catalysts rivaling noble metals for the CO oxidation reaction.

Current and future challenges
There is no doubt that the redox exsolution concept has demonstrated that creating and manipulating confined particles enhances the intrinsic site activities of base metals by several orders of magnitude, leading to nanostructures capable of rivaling platinum-group catalysts on a per-site or per-weight basis for CO oxidation [160,199]. Understanding and controlling the physical and chemical properties of exsolved nanoparticles at the surface as well as the bulk of functional oxide supports is critical for further enhancing the activity and stability. Recent studies demonstrated that the shape and chemistry of exsolved nanoparticles can be tuned by controlling the gas environment the particles are exsolved from [11,160]. Even though controlling extrinsic parameter for particle growth is intriguing, more research is required to model the resulting structures and establish links between specific facets and catalytic activity. For example, even though the mechanism of CO oxidation is well studied in conventionally prepared supported nanoparticles, i.e. a balance between Mars Van Krevelen mechanism on the interface and Langmuir-Hinshelwood mechanism on the metal surface, there are currently no reported studies for the exsolved analogues to interpret the higher catalytic activity apart from limited reports on strong metal-support interactions which would facilitate ion and electron transport at strained interfaces [101].
Furthermore, it has been recently shown that exsolved nanoparticles can be dispersed throughout the bulk of oxide supports unlocking new dimensions in materials functionalization [201,202]. These developments are very important as strain engineering can be tuned via particle immersion in a controlled manner [28]. Nevertheless, there is still a great deal of work that needs to be done for understanding the links between lattice/particle strain and active sites for the CO oxidation reaction. Even though the developments on the fundamental studies on redox exsolution are profound, the scale-up of the exsolution method so that the exsolved catalyst structures can be used in large-scale catalytic applications is not trivial and present major challenges. Firstly, very few studies have focused on low loadings of the catalytic element present on the B-site and usually only a small fraction of these elements can emerge on the surface as exsolved nanoparticles [2]. Also, the internal surface area of the supporting oxide used for exsolution tend to be low, so once scaled up, the available SSA will be low and hence impact the resulting catalytic activity. Finally, in the large majority of studies, the exsolution takes place at high temperatures (>600 • C) and under highly reducing conditions (99.9%+ purity H 2 ), making the cost for upscale forbidden for further industrial implementation. , before (points 2-4) and after (points 4-5) the CO kinetic experiment: (a) CO2 production rate and nominal turnover frequencies (nTOF) from the CO oxidation reaction before and after the CO kinetic experiment, as a function of temperature (light-off experiment). (b) CO2 production rate as a function of the partial pressure of CO (yCO) at 520 • C. (c)-(e) microstructure evolution showing: c as-prepared metal particles; d particles after the light-off experiment; € morphological transformations during the CO kinetic experiment. Reproduced from [160]. CC BY 4.0.

Advances in science and technology to meet challenges
Further scientific advances are required in the areas of modeling, operando and in-situ techniques for mechanistic understanding of particle growth as well as morphological and chemical transformations, the use of model systems under realistic conditions and manufacturing advances for scale-up. A holistic approach to modeling of the resulting nanostructures is crucial if one was to link materials characteristics (such as grain properties, surface and bulk structure and defects, particle properties etc) with intrinsic and extrinsic control parameters. Recent work has investigated the exsolution of individual nanoparticles with high spatial and temporal resolution and the results revealed the mechanisms that underpin the socketing and straining of the exsolved nanoparticles [11]. High resolution scanning electron microscopy in conjunction with nanoparticle tracking and kinetic experiments (figure 19 [160]) has shown that exsolved particles do not re-dissolve back to the parent oxide and this can serve as a platform and alter their structure, composition and functionality while preserving their initial spatial arrangement for the CO oxidation reaction [160]. Such techniques, although complicated and expensive, could be used more routinely to provide insight into the in-situ formation and morphological and chemical transformations of exsolved nanoparticles.
New methods need to be developed to address the challenges for scaling up exsolution. It is likely that thin-film heterostructures on low cost and high surface area supports can present several advantages as opposed to traditional synthesis methodologies. Exsolution from thin films if successful can achieve more efficient use of precursor materials and lower synthesis temperatures. To date only a few preliminary studies on this area have been reported [66]. The use of non-thermal atmospheric plasma as the driving force for exsolution can also be explored to induce exsolution in milder and more economical conditions which are more favorable for industrial implementation of the exsolution method.

Concluding remarks
Undoubtedly the scientific advances in fundamental research and understanding of the exsolution mechanisms and structure-property relationships has demonstrated that exsolved systems with enhanced activity, stability as well as emergent functionality can satisfy the need for low cost and efficient catalysts for the CO oxidation reaction. In addition, with further simplification of the exsolution process requirements it is also possible to pave the way for industrial implementation of this revolutionary materials design concept.

Status
Chemical looping (CL) is a process in which individual reactions are divided into sub-reactions, either in different reactors, or in the same reactor but separated by time, with the reaction being mediated by a carrier material. Here we focus on the most common type, an oxygen carrier material (OCM). CL employing OCMs has received increased attention for hydrogen production as an alternative to the conventional process of steam methane reforming, followed by shift and separation, because of the ability of CL processes to produce separated product streams ( figure 20).
A suitable OCM needs to possess a number of characteristics including but not limited to, oxygen exchange ability, oxygen storage capacity, redox stability, catalytic activity and poison, sintering and coking resistance. The OCM is often a metal/metal oxide (M/MO) system that undergoes reversible redox reactions and displays suitable thermodynamics for the application. More complex materials such as non-stoichiometric oxides have been the subject of interest permitting conversions in excess of conventional 'mixed' reaction equilibria [203]. Mixtures of the two above categories can also be employed. However, OCM deactivation due to agglomeration, sintering, carbon deposition and matrix mismatch under redox conditions must be considered.
Recently the concept of exsolution has been used to address some of the above requirements. Exsolved systems exhibit redox stability, are catalytically active and possess resistance to poisons, sintering and coking [2]. Reversible exsolution and reincorporation of metal nanoparticles in the perovskite structure is thought to be the key to increase redox stability [204][205][206], avoid particle sintering and deactivation [207,208] and provide improved activity due to possible synergetic effects between the particles and the oxide support material [44]. In contrast, in irreversible exsolution the stable host matrix together with the exsolved particles form a metal-ceramic composite that due to the crystallographic alignment and socketing of the metallic particles overcomes the traditional limitations posed by deactivation due to matrix mismatch [209].
While exsolution has been used to decorate the surface of the host matrix, recently titanates were used to exsolve epitaxially aligned particles not only on the surface but also throughout the bulk [201,202]. For these structures, while the perovskite remains responsible for mediating oxygen exchange between the surface and the embedded nanoparticles, its oxygen transport capabilities are modified, in this case favorably, by the introduction of strain via the bulk particles (figure 21).

Current and future challenges
The challenges in achieving a leap forward in the area of exsolution for CL belong in three main areas: design and synthesis of new materials, characterization techniques and long-term stability. All should ideally be supported by computational modeling.
Many materials that have been used for CL processes take advantage of a M/MO phase change for the oxygen capacity needed, which is usually not easily tunable to the process itself. Recently, non-stoichiometric oxides have been employed in order to provide more favorable thermodynamic strategies. However, such materials usually lack the capacity of a M/MO phase transition. Additionally, activation of hydrocarbon fuels, i.e. CH 4, is often very slow and becomes the rate limiting step. Hence, combining a non-stoichiometric oxide with a catalyst could allow for systems that demonstrate these desired properties. This approach comes with challenges in terms of stability, due to the presence of separate phases, different deactivation mechanisms and interphases. Exsolution can be used as a technique that can introduce the catalytic activity needed within a non-stoichiometric perovskite, in what could be considered to be a single material. However, preparation of exsolved systems has so far been dominated by solid-state methods which leads to low surface areas. Hence, new synthesis techniques are needed that would allow for tunable surface and microstructural characteristics. This could in turn allow for increased surface exchange kinetics leading to potentially lower operation temperatures.
The lack of thermodynamic and kinetic data for exsolved systems in CL both when it comes to controlling key parameters i.e. particle size and population of said materials but also their thermodynamic behavior when used as OCMs is inhibiting wider application. In order to acquire this missing data, we need to invest in developing correlated in-situ characterization techniques. This will allow us to monitor the way the materials function and evolve during operation and allow us to set the basis of designing materials tailored to specific applications.  However, for all the above cases it will still be difficult to establish a well-defined set of rational design principles unless we aim to further develop computational modeling.
Lastly, the use of exsolved materials at scale is pretty much still at its infancy. By default, exsolution would allow for protecting against deactivation mechanisms such as agglomeration, carbon deposition, sulphur poisoning. However, long term studies under realistic conditions to evaluate the structural and chemical evolution of the materials are yet to be explored.

Advances in science and technology to meet challenges
Advances needed to address the challenges are divided into three main categories; understand, modify, and demonstrate.
To understand the origin of materials' performance, advances are required. In-situ techniques such as XRD (bulk composition, dynamic changes during operation), XPS (surface species-catalytic activators), extended X-ray absorption fine structure (EXAFS) (coordination environment) especially at synchrotron sources coupled with detailed thermodynamic and kinetic data as well as the development of computational modeling at all length-scales, from materials design to process, would allow us to establish a detailed set of rational design principles [210]. The use of correlated in-situ techniques would provide an insight into the possibility of exsolution under reaction conditions and appropriate oxygen potentials. Additionally, different surface techniques that measure oxygen exchange as well as surface exchange and diffusion, would be of high interest because they would provide an insight into improving the design of these systems. Such techniques could include but are not limited to the use of secondary ion mass spectrometry or nuclear reaction analysis to determine isotopic exchange [211,212], conductivity relaxation experiments to determine ionic conductivity or in situ optical transmission relaxation coupled with impedance spectroscopy for oxygen surface exchange kinetics determination [213]. New methods might also need to be developed for measuring kinetic rates since in such materials these could potentially be dominated by interfaces [214].
Additionally, employing extrinsic techniques to control the crystal faces on the surface i.e. particle shape would make possible the design of highly active, selective and tailored materials [11,39,186], enabling a functional diversity that will allow for more efficient CL processes. As mentioned above, preparation method plays a key role in the nano and microstructure of these materials and modification of the current synthesis methods would impact on the microstructure of the backbone matrixes. This in turn would lead to the development of appropriate structures, i.e. smaller grain size, hierarchically ordered perovskites, in order to facilitate not only extent of exsolution, which is important for any application, but also oxygen exchange essential for CL processes [214]. Additionally, developing synthetic methods to prepare large batches of exsolved materials controllably, would facilitate their transition to pilot scale and ultimately commercial application.
Lastly, the integration of machine learning models with the above techniques may help uncover important structure-property relationships and guide further exploration and upscaling of these materials. The use of high-throughput testing would also accelerate the use of those materials at lab and pilot scale and eventually the transition to commercial processes [215].

Concluding remarks
There is no doubt that the use of exsolved materials in CL processes will lead to smarter more efficient chemical conversions for a sustainable future. However, challenges still remain. The key to exploitation is a combination of materials engineering and long-term real-life testing. The development of in-situ characterization techniques as well as computational modeling will aid the design of improved materials with low cost, high activity and cyclability for clean hydrogen production via CL.

Status
The photocatalytic reactions generally involve, sequentially, the generation, separation, and transferring of photocarriers (e − and h + ). The large time-scale differences among these processes have raised severe challenges to modulate the photocarrier flow smoothly. For instance, the photocarrier separation normally occurs at a time scale of picoseconds to microseconds (10 −12 s-10 −6 s) while photocarrier transfer at the surface takes milliseconds to seconds (10 −3 s-10 • s) [216]. This renders strong implications for the photocatalysts to encompass different functional units which work synergistically. For instance, a photocatalyst normally comprises a semiconductor substrate for photocarrier generation and supported nanoparticles as cocatalysts to facilitate interfacial charge transfer [217]. Presumably, a strong interconnection between different functional units is indispensable to keep charge flows but usually cannot be guaranteed through conventional methods [218]. Directly growing nanoparticles from their supported substrates, known as exsolution processes [13], offers an ideal means to fabricate nanoparticles with strong interactions with the semiconductor matrix. This has been ascribed in part to the crystallographic alignment between the exsolved nanoparticles and semiconductor matrix and in part to the peculiar microstructures where exsolved nanoparticles are partially socketed into the supporting substrate [3]. In 2015, Xu et al introduced this exsolution tactic to fabricate AgTaO 3 anchored with Ag-nanoparticles (i.e. Ag@AgTaO 3 ) [219]. The exsolved Ag nanoparticles have a size of around 40 nm whose plasmonic frequency falls into the visible light region and are capable of delivering hot electrons to the AgTaO 3 substrate through localized surface plasmon resonance (LSPR) ( figure 22(A)). The Ag@AgTaO 3 hybrid system not only owns visible light absorption but also shows improved photocatalytic activity for H 2 -evolution under both UV and visible light illumination. Following these findings, many Ag-exsolved photocatalytic systems have been fabricated and have demonstrated intriguing performance for different applications, e.g. photodegradation, water splitting, CO 2 reduction, etc [220][221][222]. Apart from noble-metal Ag, non-noble metals have also been exsolved from semiconductor substrates as hybrid photocatalytic systems. For instance, Cu nanoparticles have been exsolved from Cu 2 O which contribute to a much-improved CO 2 photoreduction activity even under near-infrared irradiation ( figure 22(B)) [223]. Likewise, Bi nanoparticles can be exsolved from BiOCl and deliver enhanced photocatalytic activity and stability for NO removal (figure 22(C)) [224]. Currently, the research on nanoparticles exsolved from semiconductor matrix is at its infancy but has demonstrated exciting results in different photocatalytic applications. With ongoing exploration and understanding of the mechanism underpinning semiconductor exsolution, one can foresee flourishing outcomes from photocatalysis involving exsolved nanoparticles.

Current and future challenges
Although exsolved nanoparticles have shown promising applications in the field of photocatalysis, there are many questions and challenges to solve. Firstly, the role of exsolved nanoparticles is yet to be clarified during photocatalytic processes. Previous investigations have suggested that the exsolved metal nanoparticles participate in photocatalytic reactions through their LSPR. The LSPR can either inject hot electrons into the semiconductor substrate or can enhance photocarrier generation by an amplified near-field electromagnetic field and resonant photon scattering. As metal nanoparticles are also known as active catalysts for chemical reactions, it is unknown whether these metal nanoparticles also collect photocarriers from the semiconductor substrate and promote interfacial charge transfer (i.e. as a cocatalyst). It should be emphasized here that the charge flow direction is distinct for the two mechanisms: electrons migrate from metal nanoparticles to the semiconductor in LSPR and conversely in cocatalyst. Secondly, the type of nanoparticles that can be exsolved and semiconductor substrates suitable for exsolution awaits further exploration. Currently, nanoparticles exsolved from semiconductor substrates are mostly metals that can be easily reduced, e.g. Ag, Cu, Bi, etc. Only a few non-metal nanoparticles have been exsolved from semiconductors including WS 2 and Mn 3 O 4 [225,226]. It would be of great interest to extend the categories of nanoparticles exsolvable to meet the diverse applications in the field of photocatalysis. On the other hand, the semiconductor substrates used to exsolve nanoparticles need more investigation. In particular, the tolerance and capability of different crystal structures for exsolution have not been systematically studied. This is critical for semiconductor photocatalysis as any structural changes such as non-stoichiometry, defects, strains, etc would dramatically modify the fundamental properties of the semiconductor, i.e. optical absorption, charge recombination, charge mobility, etc. Thirdly, the strategies to precisely control nanoparticle exsolution from semiconductors should be extended. From the viewpoint of photocatalysis, the activity would substantially hinge on the exsolved nanoparticles in terms of their location, density, particle size, crystal facets exposed, etc.

Advances in science and technology to meet challenges
The photocatalysts containing nanoparticles exsolved from semiconductor substrates are essentially heterogeneous multiphase systems. Knowledge about the location of photocarriers within the systems would be very helpful to understand the roles of different functional units and provides some clues to disclose the underlying mechanisms. There are some chemical and physical techniques that are spatially resolved for the detection of photocarriers. For instance, photo-deposition techniques can be used to identify the position of different photocarriers. MnO x and CoO x can be photo-deposited to the place where holes are accumulated whilst Pt and Ag are frequently used to probe the location of electrons [227]. In addition, recent advances in spatially resolved surface photovoltage (SPSPV) method have shown powerful capability in mapping the charge distribution at the nanoscale over heterogeneous photocatalysts and in revealing their charge separation mechanism [228] ( figure 23(A)). It would be very informative to apply the SPSPV method to study photocatalytic systems with nanoparticles exsolved from semiconductor substrates. With regard to the control of exsolution, a number of new synthetic methods or strategies have shown great potential to exquisitely tailor the composition and microstructures of nanoparticles exsolved. For instance, modifying the driving force needed for exsolution serves as a useful tool to modify the microstructures. Alternatively, some post-treatments can be used to prepare nanoparticles that cannot be directly exsolved from semiconductors [225] ( figure 23(B)). In the meantime, altering the properties of the semiconductor substrates can add another dimension to the design target of nanoparticles exsolved. In particular, perovskite-based semiconductors are well-known for their flexibility in structure and composition which provides copious possibilities in shaping the exsolved nanoparticles [159,222].

Concluding remarks
Exsolution of nanoparticles from semiconductors is an appealing phenomenon and is very useful in fabricating photocatalytic systems which involve different functional units with strong interconnection. Compared with conventional nanoparticles, the exsolved ones clearly have many advantages such as good dispersion against agglomeration, robustness against poisoning, exceptional durability and stability under various reaction conditions, etc. These desirable properties have endowed exsolved nanoparticles with broad applications in the field of photocatalysis. Although the properties of exsolved nanoparticles need further investigation, previous encouraging results have proven their usefulness in achieving superior photocatalytic activities. With the ongoing investigation of their catalytic mechanisms with the state-of-art analytic techniques as well as the development of various new preparation protocols, we ought to witness more exciting outcomes in the photocatalytic systems involving nanoparticles exsolved from semiconductors.