Operando Insights into Nanoparticle Transformations during Catalysis

Nanostructured materials play an important role in today’s chemical industry, acting as catalysts in heterogeneous thermal and electrocatalytic processes for chemical energy conversion and the production of feedstock chemicals. Although catalysis research is a longstanding discipline, the fundamental properties of heterogeneous catalysts such as atomic structure, morphology and surface composition under realistic reaction conditions, together with insights into the nature of the catalytically active sites, have remained largely unknown. Having access to such information is however of outmost importance in order to understand the rate-determining processes and steps of many heterogeneous reactions and identify important structure−activity/selectivity relationships, enabling knowledge-driven improvement of catalysts. In the last decades, in situ and operando methods have become available to identify the structural and morphological properties of the catalysts under working conditions. Such investigations have led to important insights into the catalytically active state of the materials at different length scales, from the atomic level to the nano-/micrometer scale. The accessible operando methods utilizing photons range from vibrational spectroscopy in the infrared and optical regime to small-angle X-ray scattering (SAXS), diffraction (XRD), absorption spectroscopy (XAFS), and photoelectron spectroscopy (XPS), whereas electron-based techniques include scanning (SEM) and transmission microscopy (TEM) methods. In this work, we summarize recent findings of structural, morphological, and chemical nanoparticle transformations during selected heterogeneous and electrochemical reactions, integrate them into the current state of knowledge, and discuss important future developments.


INTRODUCTION TO IN SITU AND OPERANDO INVESTIGATION OF NANOCATALYSTS
One of the greatest impacts of catalysis research on humankind was the development of the Haber−Bosch process more than 100 years ago to produce ammonia for fertilizers on an industrial scale, as it contributed to increase the world food production. 1,2 Today, catalysis research involves various disciplines ranging from surface science investigations of atomically flat or tailored single crystal surfaces 3 and nanoparticles, 4 to engineering disciplines optimizing reactor design and operation for large-scale facilities in the chemical industry. Nevertheless, independently of the length scale in which the catalyst has to be operated or studied, the design of an optimized catalyst for improved efficiency, stability, sustainability and low cost demands extensive knowledge of the catalyst system under operando reaction conditions. Here, the structural/chemical evolution and deactivation under stationary as well as varying operating conditions must be known to estimate or predict the catalyst lifetime. Varying the operating conditions is especially relevant for (electro)chemical energy conversion processes to store intermittently provided electrical energy from renewable power sources. In many cases, the key to operate industrial scale facilities efficiently and cost effectively is not to achieve peak activity marks but to increase the catalyst lifetime. 5 At this stage, in situ and operando insights into nanocatalysts is of utmost importance, as here (varying) reaction conditions can be simulated while providing information on the nanocatalyst structure, morphology, composition, and chemical state under reaction conditions. The overall goal of identifying decisive properties for catalysts including the active sites under operating conditions, remains. It should be kept in mind that the as-prepared materials are only precatalysts that become activated under reaction conditions, often through drastic structural/chemical modifications. Therefore, efforts that have been undertaken in the past to correlate physicochemical properties of the as-prepared catalysts to a particular activity/selectivity trend must be redirected toward those that take into consideration the transformations that the active catalyst has undergone to adapt to the specific reaction conditions.
In this review, we discuss recent findings on in situ transformations of nanoparticles during heterogeneous gasphase catalytic and electrocatalytic processes. Studies based on various experimental techniques are presented, such as in situ scanning and transmission electron microscopy (SEM and TEM) applied under a gas-phase (E-) or liquid (L-) environment or scanning probe microscopy (e.g. atomic force microscopy (AFM)) to identify morphological and structural transformations. In situ X-ray-based techniques are presented, showing changes in the chemical state as well as in the local atomic structure via X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectroscopy. Insights into morphological and structural transformations were also provided by using small angle X-ray scattering (SAXS) and X-ray diffraction (XRD) obtained for either powder samples or (epitaxial) NPs and thin films in a grazing incidence (GI) configuration. The experimental findings were supported by theoretical studies based on not only density functional theory (DFT) and thermodynamics-driven NP reconstructions, especially with respect to adsorbate-induced changes of the equilibrium NP morphology, but also ab initio molecular dynamics (AIMD) calculations. We also exemplify the importance and strength of combining structural and morphological investigations with vibrational spectroscopy such as diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy 6 and Raman spectroscopy, allowing deep insight into adsorbate-induced NP modifications. Additionally, simultaneous product analysis using gas chromatography (GC) or mass spectrometry (MS) converts in situ into operando investigations. 7 In this manner, correlations between reaction mechanism/pathways and NP transformations can be established. Compositional changes, especially of the catalytically most relevant near-surface, are discussed on the basis of near-ambient-pressure (NAP) X-ray photoelectron spectroscopy (XPS).
In this work, relevant catalytic processes from an academic and industrial point of view, involving the activation of small molecules, are discussed, such as the oxidation of CO, H 2 O, and NO and the reduction of CO 2 and NO. Our review addresses mainly transformations of nanoparticle catalysts taking place under (near-)ambient-pressure conditions, although in sections where a limited number of such studies are available, some selected examples are also given involving lower pressures. In section 2 we focus on morphological transformation with respect to the NP size and shape under oxidizing and reducing reaction conditions. In section 3, we discuss the (bulk) chemical state and near-surface structure transformations followed by a summary of most recent studies on compositional changes and surface segregation 8 in NPs under reaction conditions. Finally, in section 4 we discuss challenges that catalysis research faces today and how these may be tackled in the years to come.

MORPHOLOGICAL CHANGES IN NANOPARTICLES
UNDER REACTION CONDITIONS: SIZE, SHAPE, SINTERING, AND REDISPERSION 2.1. Size Modifications of Nanoparticles under Adsorbate Exposure and Catalytic Reactions. Designing heterogeneous catalysts with improved catalytic performance involves being able to achieve a high density of catalytically active sites, which in a first approximation correlates with its physical surface area. Thus, nanostructuring is a typical pathway employed for catalyst optimization to increase the surface to volume ratio and mass-based surface area. In contrast, in a number of cases, catalyst degradation correlates with the growth of the nanoparticles under catalytic conditions, which is accompanied by the loss of surface area and active site density. 9 In situ techniques such as TEM, 10−20 XAFS, 21−25 XRD, 26−29 and SAXS 30−36 have been used to monitor the growth of metal NPs under gas-phase thermal catalysis and electrochemical reaction conditions, as seen in the in situ TEM images of Figure 1.
A versatile approach to study morphological transformations in nanocatalysts is to utilize size-controlled NPs homogeneously distributed on a support, since this allows the researcher to more easily investigate sintering and redispersion processes. 37 For small NPs and clusters, changes in their size and shape directly influence the mean coordination number (CN) of the nearest metal−metal distances in the NP. 38 Such modifications can be tracked via in situ/operando EXAFS, a technique that has been applied to various noble-metal catalysts such as Al 2 O 3 -supported Pt NPs during thermal NO oxidation 21 and Pd NPs during NO/CO cycling 22 as well as during NO reduction conditions. 23 Similarly, the NP growth under electrocatalytic conditions has been reported during electro-oxidation of isopropyl alcohol for Au NPs. 24 In some cases, sintering was studied using in situ TEM. 10−18 For example, the temporal evolution and dynamics of particle growth was shown for Pt NPs supported on zeolites under a variety of reaction conditions (CO oxidation, NO reduction via CO or H 2 , water-gas shift reaction). 18 The mobility and morphological stability of size-selected Pt NPs synthesized by inverse micelle encapsulation on SiO 2 thin films was also investigated (Figure 1a−f). 12 The ∼2.5 nm large preoxidized Pt NPs were exposed in situ to 1.33 mbar of either pure O 2 or pure H 2 from 25 to 800°C. Under oxidative conditions the NPs do not show any sintering up to 450°C, while above 650°C the formation of small PtO x clusters coexisting with larger NPs is assigned to a redispersion phenomenon favored by the formation of volatile PtO x species (Figure 1b). An additional decrease in the NP size is observed upon further annealing in O 2 at 800°C, suggesting the loss of material (via pumping in the E-TEM, Figure 1c).
In clear contrast, increasing the temperature under H 2containing conditions up to 650°C leads to PtO x reduction ( Figure 1e) and drastic sintering at 800°C via NP diffusioncoalescence (Figure 1f), also leading to 4−5 nm Pt 3 Si NPs through NP/support (Pt NP/SiO 2 ) interactions, as revealed by electron diffraction. Interestingly, subsequent exposure of the redispersed PtO x to H 2 at 450°C resulted in reagglomeration and larger metallic Pt NPs. The formation of Pt 3 Si can be explained by the decomposition of SiO 2 regions in the proximity of Pt NPs, leading to the direct interaction of the Pt NPs with Si from the support, which enables silicide formation. A similar process was identified for Au NPs which dig vertical channels into a SiO 2 support at temperatures up to ∼1000°C under UHV conditions, as revealed by AFM. 41 We note that, in the Pt/SiO 2 study, special care was taken to minimize the influence of the electron beam. This work especially illustrates the power of HRTEM studies in combination with electron diffraction to reveal not only morphological but also structural adsorbate-driven NP transformations and sintering phenomena.
Among the in situ XAFS studies, 21−24 a combinatory study applying DRIFTS and mass spectrometry gave especially comprehensive insights into the dynamic changes of the size and dispersion of Pd NPs supported on CeO 2 /ZrO 2 /Al 2 O 3 during alternating reaction conditions. 22 On switching from a CO-to a NO-containing atmosphere at 400°C and 2 bar, the initially oxidized Pd NPs reduce in the presence of CO under CO 2 evolution, forming a CO-covered Pd NP surface. Introducing NO initiates the Pd NP redispersion on the support, while CO adsorbates get catalytically converted to CO 2 and N 2 via NCO adsorbates as well as N and O species. After full conversion of the C-containing adsorbates, the Pd NPs oxidize in the presence of NO. Switching to a COcontaining atmosphere initiates the Pd NP sintering and the described cycle restarts. Oscillations of the Pd NP size within the cycle were determined from the Pd−Pd coordination numbers extracted from the in situ EXAFS data. This study nicely shows how dynamic processes and the mechanisms of chemical transformations can be revealed by concurrently tracking adsorbates via vibrational spectroscopy, reaction products via mass spectrometry, and structural/morphological transformations via XAFS.
We have to note that in situ/operando studies based purely on either local (SEM, TEM) or ensemble-averaging methods (XAFS, XRD) do not necessarily yield the correct global picture of the NP catalyst. In the case of microscopy methods, deep atomistic insight may be obtained from a spatially limited and not necessarily representative fraction of the sample. On the other hand, averaging spectroscopy and diffraction methods may result in wrong structural/chemical information if the NP properties are heterogeneous (lack of narrow size, shape, and composition distribution). In the case of XRD, approaches in data analysis can be applied that can account for and quantify asymmetric size distributions, as shown for Cu NPs on ZnO. 42 An elegant way to overcome these difficulties for NP catalysis research is to combine both local and statistical in situ/operando methods, yielding a more complete picture of the complex NP catalysts. This has been shown for instance for supported Pt NPs during ethylene hydrogenation 43 as well as CO oxidation in the presence of propene simulating the conditions of a diesel exhaust atmosphere. 44 In the former work, a universal microreactor that can be applied for STEM and XAFS experiments at a gas pressure of 1 atm

ACS Catalysis
Review was developed. The separate experiments can be linked by the product analysis that was conducted simultaneously. With this approach, it was revealed that the catalyst consists of atomically dispersed Pt species as well as NPs above 1 nm size in an ethylene-rich gas feed (H 2 :C 2 H 4 ratio of 1:3). The atomically dispersed Pt species were not detectable by STEM, which solely revealed the growth of the NPs, but Pt redispersion on the support was identified using XAFS. 43 In the latter work, the dynamics of the Pt NP redispersion on a CeO 2 support during an oxidative pretreatment, reductive activation, and CO oxidation catalysis was investigated using operando XAFS combined with E-TEM. 44 Similarly, the Pt NPs redisperse under an oxidative treatment at 500°C and Pt NPs form under reductive atmospheres at 250°C. Most importantly, the combined investigation leads to the identification of an optimized pretreatment procedure of the Pt NPs during which the propene concentration was cycled in a reductive activation step.
With respect to electrocatalytic conditions, significant morphological changes of NPs have been recently reported during electrochemical reactions via for example electrochemical AFM, STM, and scanning electrochemical microscopy (SECM). 39,40,45 In situ electrochemical L-TEM is technically more challenging and is not yet fully established, as there have been only a few studies of NPs under electrochemical conditions. 16,17,46 Recently, the degradation pathways of octahedrally shaped PtNi NPs during electrochemical potential cycling were investigated using in situ electrochemical L-STEM. 18 Therein, NP dealloying via Ni dissolution could be followed and quantified in the initial potential cycles. After Ni dealloying, the mobility and coalescence of the NPs during carbon corrosion conditions was the main structural modification observed, which was less severe under potential cycling in comparison to constant potential conditions. At constant high electrode potential, NP sintering via a facet-to-facet attachment was found to occur within 10 s. Additionally, the degradation of FePt NPs during the electrochemical oxygen reduction reaction was studied via electrochemical L-TEM under conditions that mimic fuel cell operation. 16 During electrochemical potential cycling, an increasing density of NPs on the carbon support was found as well as the formation of chain-or dendrite-like structures. Additionally, carbon corrosion as well as a potential-dependent variation of the particle size was revealed. This work shows the possibilities but also indicates the experimental difficulties, as reflected in the inhomogeneity of the processes due to local changes in the conductivity, as well as the possible effects of the electron beam itself, which are still largely unknown. In a different study, a surprisingly strong NP growth was observed under supposedly electrochemical conditions in a case when the NPs lost electric contact to the substrate (Figure 1g,h). 17 Notably, areas which remained electrically connected did not exhibit this large particle growth, indicating a very significant beam-induced particle growth. Nonetheless, we expect that the field of electrochemical L-TEM will continue to develop rapidly in the near future, allowing researchers to gain more indepth insight into NP growth, dissolution, and shape modifications under electrochemical conditions provided that the many technical and experimental issues currently faced by researchers in this field can be satisfactorily resolved. As a compromise for a real electrochemical flow cell configuration, insights into the morphological transformation of Pt NP fuel cell catalysts supported on carbon were gained using E-TEM under an O 2 and H 2 O atmosphere. 15 Therein, it was shown that the mobility of NPs leads to a particle growth via coalescence, which is preceded by a NP rotation to align the lattice planes of the individual NPs. It should be considered that any in situ TEM investigation should be benchmarked against the already successful identical location-TEM approach, 47−50 as well as alternative methods such as in situ (GI)SAXS studies. [32][33][34][35]51 For example, an in situ SAXS study nicely revealed the growth of Pt NPs under electrochemical potential cycles simulating fuel cell operating conditions. 51 Almost simultaneous in situ XAFS experiments (the spectra were recorded in the same setup within a couple of minutes) showed that the Pt−Pt distances within the Pt NPs increased as the particle size grew.
Utilizing in situ electrochemical scanning probe techniques, the reversible electrochemical formation and dissolution of Pt NPs on electropolished Pt could also be demonstrated with EC-AFM (Figure 1i−k). 40 The Pt surface was found to resist morphological transformations under mild electrochemical potential cycling. Increasing the upper potential limit up to 1.8 and 2.5 V leads to the formation of Pt NPs with mean heights of ∼3 and ∼10 nm, respectively. Similarly, the electrochemical roughening and nanoisland formation during potential cycling was studied on Pt(111) using EC-STM. 52 As CO 2 electroreduction (CO 2 RR) has received increasing attention in catalysis research in recent years and the majority of works utilize Cu-based electrocatalysts, 53−56 the morphological transformation of Cu-based nanocubes (NCs) during CO 2 RR was tracked using in situ EC-AFM and in situ XAFS. 39 The electrochemically prepared NCs were synthesized between ∼200 nm and ∼1 μm. Exposing the NCs to the electrolyte induced the formation of cracks on their surface. Already after 1 min under CO 2 RR conditions the NC size and morphology strongly changed (see Figure 1l−o). The cube edge length decreased by ∼10%, and the NC became smeared out. After three additional hours of electrocatalytic operation, the NCs appeared rather spherical and their size decreased again by about 10%. In the case of larger NCs, a pore structure became visible. In situ XAFS measurements showed that the size changes are linked to the reduction of the NCs to metallic Cu under the reaction conditions as well as the loss of Cl when the NCs are first introduced into the electrolyte, even in the absence of an applied potential.
The former in situ electrochemical AFM studies nicely illustrate the capability of the technique to track the morphological changes of well-defined surfaces and individual NPs under electrochemical and electrocatalytic conditions. From here on, the in situ combination of scanning probe techniques with optical, vibrational, or X-ray spectroscopy methods under electrochemical conditions will give a more comprehensive picture.
2.2. Shape Transformations of Nanoparticles under Reaction Conditions. In addition to the changes in the particle size, shape modifications of NP catalysts can tremendously influence the catalytic activity, as the active site might be linked to a certain defect motif or preferred surface termination. 57 It is well-known that the growth direction of NPs from clusters can be controlled by the adsorption of molecules such as CO, leading to sophisticated particle morphologies. 58 Heterogeneous catalysis, however, is based on the adsorption of molecules as either reactants or intermediates, which can affect the morphology (as well as surface composition) of nanocatalysts under reaction con-

ACS Catalysis
Review ditions. Not only the most stable surface termination but also the number of edges and corners can be decisive for the catalytic function. 59,60 In this respect, the nearest-neighbor approach takes into account the catalytic activity of distinct metal adsorption sites by their coordination environment. A generalized coordination number can be formulated for each site by counting the number of atoms in the first coordination shell weighted by their deviation from the expected ideal coordination environment, such as a coordination number of 12 for a metal atom in the face-centered-cubic structure. This generalized coordination number was found to be correlated to the catalytic activity of distinct terminations and defect motifs of metal surfaces and NPs. 61,62 On the basis of the nearest-neighbor approach, the influence of the ensemble of sites on Pt NPs of various shapes on the overall CO oxidation kinetics has been investigated by kinetic Monte Carlo simulations. 67 The authors showed that the kinetic coupling between surface sites, e.g. on the NP edges between (100) and (111) facets, has to be taken into account and that isolated active sites are not sufficient to describe and explain the catalytic activity of a NP. Generally, a wide distribution of surface sites seems beneficial for the CO oxidation catalysis according to these theoretical calculations.
Shape transformations of NPs and, especially, adsorbateinduced phenomena have been investigated using not only experimental techniques such as in situ E-TEM but also theoretical calculations or combined studies. 20,59,60,63−66,68−81 For example, the shape change of a Pt NP (∼20 nm) on changing from reducing (H 2 ) to oxidizing (O 2 ) conditions in the millibar regime was experimentally studied using E-TEM. 69 In vacuum and under an H 2 atmosphere, the NP morphology is close to the equilibrium (Wulff) shape of a truncated octahedron, but changing to an O 2 atmosphere leads to a roughening of the (111) facets between the (100) facets. These changes decrease the (111)/(100) ratio of facets and leads to a rounded morphology. In contrast, an in situ STEM study in a gas cell at 1 bar of O 2 pressure at 350°C revealed a facet-dependent oxidation of Pt 3 Co NPs. 65 Figure 2a,b, the Ptterminated (100) facet restructures, forming low-coordination sites within less than 1 min, whereas the CoO-terminated (111) facets retain their atomically flat termination. The Co− O termination of the (111) facets results from a lower surface free energy, as determined by DFT calculations.
The morphology of pure Pt 63,70 and pure Au NPs 20,71,72 was investigated under CO oxidation conditions. In the case of the pure Pt NPs, the truncation of the as-prepared octahedron decreased in the presence of a CO-covered surface above 127°C, as revealed by a combined in situ STEM and DFT study. 63 In contrast to the O 2 -containing atmosphere described above (CO + O 2 ), the Pt(111) facets remain intact when CO

ACS Catalysis
Review is adsorbed, but the (100) facets were roughened into stepped (210) and (310) facets, Figure 2c. The higher index facets exhibit a lower surface free energy at high CO coverages in comparison to the (100) and (111) facets, and the shape determined was confirmed by theoretical calculations (Figure  2d,e).
In a different in situ TEM study, an oscillatory morphology change of Pt NPs involving a dynamic refaceting at ∼450°C was revealed using a micro flow reactor which allows studies at 1 bar and elevated temperatures with simultaneous detection of the reaction products. 70 The fraction of the (111) facets on the Pt NP constantly alternates, leading to an oscillation between a rather faceted and a rather spherical NP morphology. In parallel, the reaction rate of CO oxidation oscillates, as revealed by calorimetry and reaction product analysis. This work furthermore shows that combined operando studies also tracking reaction products are especially important for an in-depth understanding of the reaction mechanisms in nanocatalysis.
In the case of CeO 2 -supported Au NPs under stationary CO/air conditions, a reconstruction of (100) facets was determined at 0.45 mbar and room temperature because the outermost atomic layers form an undulating hexagonal lattice with shorter lateral Au−Au distances. 20,71 Additionally, a decreasing CO partial pressure in air (3 mbar) led to a gradual loss of faceting of the Au NP from a polyhedral shape in a CO/ air mixture to a spherical shape in air. 71 A study of the dynamics of Au NPs/CeO 2 combining E-TEM and theoretical calculations revealed that under CO/O 2 gas mixtures the 4 nm NPs reconstruct, whereas smaller Au NPs (<2 nm) decompose, leading to single-site AuCO species on the CeO 2 (111) surface. 72 The fast dynamics of the AuCO extraction process was simulated using DFT and ab initio molecular dynamics (AIMD) calculations and showed that 2D clusters transform within 25 ps to 3D clusters. In this study, the influence of the beam dose rate effect on the dynamics of the restructuring was considered. Generally, the presence of the electron beam during the experiment is expected to be detrimental to the adsorption of CO on the NP, since it gives rise to an electron energy transfer leading to CO desorption and/or a knock-on mechanism.
The mobility of surface atoms under reaction conditions was additionally shown for nanoporous gold. 64 Nanoporous Au is an important model system in catalysis research simulating Au NPs and enables the study of selected structural and morphological transformations on sub-nanometer and atomic length scales during catalysis. Due to the absence of NP mobility in nanoporous Au, these in situ experiments can be performed in greater depth in comparison to those on supported Au NPs. An atomic motion of Au along surface steps only occurred under CO oxidation conditions, but not in the pure gases or under vacuum. Thus, it was concluded that such motion was caused by the generated heat of reaction. Interestingly, planar defects in the Au lattice such as twin boundaries that reached the surface of a nanopore were found to limit the atomic motion, Figure 2f,h, leading to a preferred growth of the nanopores. This study nicely demonstrates how atomic insights extracted from in situ electron microscopy can contribute to the improvement of the catalyst design by tailoring structural and morphological properties.
Shape changes of NPs have been predominantly investigated using advanced microscopy methods, but morphological modifications can also be assessed by X-ray scattering and diffraction techniques. Although these X-ray-based techniques rely on structural models and a fitting procedure in contrast to the direct assessment of the NP morphology in microscopy, they are highly important due to their more straightforward utilization as in situ or operando methods under more realistic conditions at higher temperatures and pressures. For example, Rh NPs have been investigated using in situ surface X-ray diffraction under O 2 dissociation and CO oxidation conditions. 82 Therein, it was shown that the epitaxial Rh NPs supported on Mg(001) reversibly change their (001)/(111) faceting ratio. O 2 dissociation at 327°C led to the growth of the (001) facets, whereas CO oxidation was accompanied by a growth of the (111) facets. The NP size was found to remain stable, but the formation of an O−Rh−O surface termination was confirmed by ex situ HRTEM.
In addition to the experimental studies, theoretical calculations have been widely applied to predict the shape changes of NPs induced by specific adsorbates and/or under reaction conditions. These theoretical studies are mostly based on first-principles calculations, e.g. of the surface free energies using DFT, and Wulff−Kaichew constructions. These calculations yield the equilibrium shapes of NPs as well as smaller clusters and have been applied to oxides and metals not only under noncatalytic 72−76 but also under reaction conditions. 66,76−81 These theoretical studies allow a straightforward variation of the reaction conditions in comparison to the experimental studies and can lead to predictive statements: e.g., on the catalyst stability. 83 The equilibrium shapes of metal NPs such as Cu, Pd, Pt, and Au have been investigated in various (reactive) gas atmospheres such as NO, CO, and H 2 O by applying a multiscale structure reconstruction model which combines Wulff−Kaichew constructions, Langmuir adsorption isotherms, and DFT. 66,80,81 For example, Meng et al. studied the equilibrium structures of 10 nm Pd, Pt, and Rh NPs in COand NO-containing gas mixtures at various pressures and temperatures (Figure 2i). 66 The calculations showed not only that the temperature and pressure determine the NP shape but also that an increasing NO content leads to a growing fraction of (110) facets, while the fraction of (100) and (111) decreases. In a more recent study, the interplay between Cu NP and the ZnO support as well as between Pt NP and SrTiO 3 support was taken into account 78 and contrasted with experimental findings based on in situ E-TEM.
These studies exemplify the morphological evolution of NPs under reaction conditions and how theoretical calculations can contribute to the understanding of such transformations. For example, they show how NPs with certain surface terminations can be rationally designed and tailored by exposing them to specific adsorbates.
Unfortunately, the high number of atoms involved in NP− support ensembles limits the extent of calculations that can be performed because of the high costs of the theoretical calculation. Therefore, many theoretical studies of NP− support ensembles consider metal (oxide) clusters on a few monolayers of support material. 62,72,79,83,87−92 In this respect, it was shown that the surface terminations of the (100) and (110) facets of the γ-Al 2 O 3 support under reaction conditions determine the most stable morphology of Pd 13 and Pt 13 clusters. 79 On the bare γ-Al 2 O 3 (100) surface, a biplanar cluster is favored, whereas surface 3D-like clusters are more stable on the partially hydroxylated γ-Al 2 O 3 (110). Experimentally, similar trends have been found for Al 2 O 3 -supported Pt 22 clusters studied at various H 2 pressures by using in situ

ACS Catalysis
Review XAFS, Figure 3a,b. 84 Increasing the H 2 pressure up to 21 bar led to an increased H surface coverage on the Pt NPs that resulted in a 2D to 3D shape transformation, as concluded from the increased Pt−Pt CNs.
In all in situ and operando investigations, the influence of the probing method should be considered and minimized as much as possible. This is especially valid for in situ TEM investigations, as already pointed out for the electrochemical L-TEM studies. For example, in the case of manganites under E-TEM conditions, cationic movement was revealed during electron bombardment. 93 Drastic morphological and structural changes can be induced by the electron beam. For instance, Co NPs were found to oxidize under 0.2 μbar of oxygen at 300°C only under electron beam irradiation. 94 Irradiating the metallic Co NPs (∼20 nm), which were covered by graphite layers, with an electron beam at various intensities led to the formation of hollow Co 3 O 4 NPs. The initial stages of the oxidation showed cracking of the graphite followed by the diffusion and oxidation of the Co atoms to CoO domains. This oxidation initiates the migration of Co atoms and the formation of cavities within the particle. Prolonged illumination was found to increase the size of the cavity, leading to a shell thickness of ∼4 nm after 20 min. The Co NP was also found to oxidize at room temperature under electron irradiation, but with a significantly lower rate. These studies exemplify the effect that the electron beam can have on NPs and highlights the importance of differentiating between the electron beam induced effect and changes induced by the nonambient conditions during in situ E-TEM experiments. However, also in the case of Cu NPs, the formation of hollow Cu 2 O NPs under ambient ex situ conditions was found. 95,96 In addition to shape changes of the NP catalyst itself, the interaction between two (NP) phases can lead to significant morphological changes under reaction conditions. A prominent process is the strong metal−support interaction (SMSI), during which a typically reducible material such as ZnO, CeO 2 , TiO 2 , or Fe 3 O 4 , usually present as a support, grows above the primary NPs, forming an overlayer and changing the NP shape significantly. 85,86,97−99 This shape transformation and modified surface composition can have a favorable or a detrimental effect on the catalytic function of the primary NP. 100 The SMSI has been investigated in the case of high-pressure CO 2 hydrogenation catalysts such as Cu/ZnO/Al 2 O 3 and Rh NPs deposited on TiO 2 . High-pressure CO 2 hydrogenation is the industrial process to produce methanol. One of the pioneering works on E-TEM for the investigation of NP catalysts was performed on Cu NP supported on ZnO/Al 2 O 3 . 10 In the asprepared state, the catalyst NPs consist of CuO, which have to be activated by a reductive treatment prior to catalytic operation. During this activation treatment the CuO x domains were found to interact with the ZnO, leading to a graphite-like ZnO overlayer on the Cu NPs. 86 These overlayers are in

ACS Catalysis
Review contrast to the typical wurtzite-like ZnO domains, as they consist of stacked graphene-like layers in which the Zn and O planes of wurtzite-like ZnO are merged. 101 Importantly, Zn ion migration onto the Cu NPs has been found to play an essential role in the catalytic function. 102,103 Zn δ+ ions act as adsorption sites for the intermediate oxygen species during CO 2 reduction, and their proximity to Cu steps formed due to stacking faults in the Cu lattice leads to high methanol yields. 102 Thus, here the SMSI effect plays a key role in the active site formation of these industrial catalysts.
However, SMSI leading to overlayer formation and NP encapsulation can also limit the accessibility of adsorption sites, as shown for Rh NPs/TiO 2 via in situ TEM using a closed gas reaction cell operating at atmospheric pressure. 85 A thermal treatment at 550°C in H 2 leads to a compact impermeable Ti 3+ oxide overlayer, whereas a reactant-permeable Ti 3+/4+

ACS Catalysis
Review oxide overlayer was formed at 250°C in a CO 2 -rich gas stream (Figure 3c−e). Adsorbed HCO x on the support was identified by using in situ DRIFT spectroscopy, and this adsorbate was attributed to initiate oxygen vacancy formation and Ti atom migration onto the NP surface. Interestingly, the presence of the adsorbate-mediated SMSI was found to lead to CO as the main reaction product in comparison to CH 4 on the uncovered NPs.
2.3. Sintering of Nanoparticles: Temperature and Adsorbate Effects. Sintering effects of CO 2 hydrogenation catalysts were also investigated using in situ and operando methods. The morphological and structural evolution of Cu and ZnO NPs of Cu/ZnO/Al 2 O 3 catalysts during methanol formation has been investigated by using operando neutron diffraction. 104 The catalyst morphology and the structural defects of the Cu NPs remain stable under reaction conditions, and the incorporation of reactive species in the bulk of the NP could be excluded. In a more recent study, Cu and ZnO NPs supported on Al 2 O 3 were found to sinter under nonstoichiometric reaction conditions. 42 In contrast to the optimal CO/CO 2 /H 2 conditions leading to the highest methanol yield, the Cu NPs grew slightly and isotropically in a CO/H 2 and CO 2 /H 2 gas feed (Figure 4a−c). In the CO 2 /H 2 feed the Cu NP growth was more prominent, and the ZnO NP also grew significantly along the hexagonal base, leading to a platelet-like morphology. The growth of the NPs was attributed to the presence of H 2 O, which was detected as a reaction byproduct. H 2 O was assigned to cause Cu segregation from the ZnO and thus contribute to the sintering of the NPs. Under optimal CO/CO 2 /H 2 conditions, the water-gas shift reaction minimizes the H 2 O concentration, as revealed by in situ DRIFT spectroscopy, and thus leads to the structural and morphological integrity.
In the case of the CO oxidation reaction, the sintering of Rhbased NPs of different shapes was tracked under near-ambientpressure (NAP) conditions using operando grazing-incidence (GI)XRD with simultaneous reaction product tracking (Figure 4d−f). 26 Bragg peak maps and line scans of epitaxially grown Pt 1−x Rh x alloy NPs on Al 2 O 3 (0001) were recorded under various catalytic conditions. The particle height at 277°C increased under CO oxidation conditions in comparison to noncatalytic conditions, as extracted from the finite height (Laue) oscillations showing a composition-dependent contraction. The main findings revealed that particle sintering and the decrease in particle coverage are significantly reduced when Rh is incorporated into the Pt NPs, Figure 4d−f. Due to the lattice mismatch between the Rh-rich NP and the Al 2 O 3 (001) support, the NP−support interface is minimized, leading to an initial rather 3D-like particle shape. This shape is closer to the equilibrium, and thus, the metal−support interaction resulted in a reduced degree of sintering. This has been previously shown in the case of CeO 2 -supported NPs. 105 We note that applying GIXRD to epitaxially grown NPs can provide insights into the NP morphology under catalytic conditions and is thus a very powerful tool to track structural transformations of model NP catalysts.
Combined insights into the chemical state and local atomic structure as well as NP and domain size can be obtained from EXAFS analysis of NPs in which the surface to bulk atom ratio is sufficiently large. At a size below 3 nm of metallic NPs the metal−metal CN decreases from the bulk-like value. 38 This approach has been applied to show the morphological transition of small Pt NPs on an Al 2 O 3 support under high temperatures and various gas atmospheres. Coarsening of the Pt NPs was identified in H 2 , in analogy to morphological changes observed for similar Pt NPs in Figure 1d−f with in situ TEM. The coarsening of Pt NPs grown on TiO 2 (110) by physical vapor deposition and by an inverse micelle encapsulation route was studied using STM after various thermal treatments and pretreatments of the support. 106 It could be shown that the smaller PVD-grown Pt NPs are more strongly bound to a prereduced support and that in general the larger micellar Pt NPs display stronger resistance against sintering.
During catalysis, in situ XAFS studies have been performed on Pt NPs during the thermal oxidation of nitric oxide 21 as well as on Au NPs during electrochemical 2-propanol oxidation. 24 Ultrasmall Pt NPs with size below 1 nm were prepared on Al 2 O 3 supports using the inverse micelle encapsulation route. 21 The particle size (diameter) was stable, as revealed by ex situ TEM (Figure 4g), but differences in the Pt−Pt CNs as determined from fitting EXAFS spectra were found. The discrepancy between TEM and EXAFS analysis was attributed to a different NP shape after catalytic operation. Interestingly, a change in NP morphology in an NO/O 2 stream below the catalytic onset was observed. The Pt−Pt CN was found to decrease significantly, indicating that the Pt NPs redisperse and/or flatten on the support and no significant sintering of the NPs appeared during catalytic testing. Above the onset temperature for NO oxidation (>150°C), the Pt−Pt CN increased, indicating the formation of more 3D Pt NPs in the catalytically active state. Similarly, Au NP electrocatalysts for 2propanol oxidation were found to sinter with increasing reaction time, which resulted in deactivation as revealed by the analysis of CNs extracted from operando EXAFS data. 24 Furthermore, sintering of catalysts plays a crucial role in molecular (electro)catalysis research, as the formation of NPs or nanoclusters increases the heterogeneity of the catalyst and thus complicates the extraction of reliable conclusions with respect to the catalytic mechanism and the formation of the catalytically active state. There are numerous examples in the literature reflecting the longstanding discussion on the nature of the catalysis on initially molecular catalysts, 107−109 and an increasing number of methodological works on molecular (electro)catalysts has been published. 109,110 Some examples include Co-based polyoxometalates (POM) such as Co 4 -and Co 9 -POMs, which have been widely studied in molecular photo-and electrocatalysis. 108,111−116 In particular, the molecular integrity of Co 9 -POM was confirmed under photocatalytic phosphate-free water oxidation conditions by using in situ SAXS and pair distribution function (PDF) analysis, although with increasing phosphate concentration nanometer-sized Co phosphate domains were obtained. 114,115 Similarly, NP formation was unraveled in the case of Co-based complexes with a metal-free ligand environment during electrocatalytic HER. 117 On the basis of this study, the formation of Co NPs from a [Co(dpg) 3 (BF) 2 ] + precursor complex was studied via in situ XAFS, showing that a typical Co NP consisted of clusters as small as 1 nm in size. 118 This study shows the relevance of organometallic complexes as precursors to form single-site, NP, and nanocluster catalysts. Recently, it has been shown that NP catalysts supported on SiO 2 decorated with single-site metal ions are highly active thermal catalysts. 119−122 For instance, increased activity and selectivity toward methanol during CO 2 hydrogenation was achieved for Cu NPs on SiO 2 when Zr(OSi(OtBu) 3 ) 4

ACS Catalysis
Review to prepare isolated Zr 4+ sites on SiO 2 . 120 Beneficial effects have also been found during methanol synthesis for Cu NPs on Ti 4+ -modified SiO 2 . 121 In the case of the Zr 4+ -modified SiO 2 , in situ XAFS revealed that the Zr sites are structurally stable under reaction conditions and do not form Zr-containing NPs or lead to Cu−Zr alloying. 120 In addition, bimetallic Ga−Pt NP catalysts prepared from single sites on SiO 2 have been shown to be highly active catalysts for propane dehydrogenation. 122 In situ XAFS showed that under reducing conditions a GaPt alloy is formed from the single sites, but a significant fraction of the Ga remains isolated on the SiO 2 surface.
These studies highlight the importance of in situ/operando investigations in the field of molecular catalysis research and the conceptual similarities to well-known electrocatalysts weakening the border to heterogeneous catalysis. We note that the same methodological pitfalls have to be considered when single-site catalysts are being prepared and investigated.

CHEMICAL CHANGES IN NANOPARTICLES UNDER REACTION CONDITIONS
In contrast to the morphological investigations that are most directly studied using electron or scanning probe microscopy methods, the catalyst chemical state is mostly studied using spectroscopic methods. In catalysis research, environmental, in situ/operando X-ray-based techniques have been established for many years 27,28,32,36,123−131 and thus are technologically more established than in situ/operando electron microscopy studies. Herein, XAFS in the hard X-ray regime covering 3d transition metal K-edges as well as 5d L-edges plays an important role in NP catalysis, as it can yield information on the metal oxidation state as well as metal−ligand and metal− metal distances in the coordination shells, together with providing further insight into NP−adsorbate and NP−support interactions. Furthermore, surface-sensitive diffraction as well as in situ XPS on (epitaxial) NPs can yield valuable insights into the chemical state of surface and near-surface regions, although the latter has been further developed than in situ electron microscopy. The mechanism of NP oxidation of catalytically relevant metals for oxidation reactions such as Pd and Pt has been widely studied using in situ methods. 35,126,132−139 In the case of Pd NPs the surface oxidation of large truncated Pd octahedra (edge length of 20 nm) and spherical Pd NPs (diameter of 7 nm) was investigated in an E-TEM study, Figure 5a−h. 132 The oxidation of the Pd octahedra was found to start on the step edges of the (111) surface, leading to up to two monolayers of tetragonal PdO which then grew laterally (Figure 5a−d). The vertical growth showed a significantly lower rate, and the adjacent (100) and (1-1-1) surfaces remained unchanged. In the case of a rather spherical Pd NP, the oxidation started at the vertex sites between two (111) facets and the PdO spread on one of the (111) facets (Figure 5e−h). The oxidation might preferably start at low-coordination sites such as step edges or vertex sites and then lead to facet oxidation. Here, we have to note the differences between the oxidation of PtRh 2 during NAP CO oxidation conditions and the oxidation of Pd nanocrystals in an E-TEM. In the first case, the change of the surface chemical state starts from molecular oxygen, whereas in the second case it starts from O − radicals generated by the electron beam and thus it significantly differs from the catalytic conditions. The latter highlights the importance of understanding electron-beam/environment/sample interactions and the control of the beam flux during catalysis experiments.
Within the framework of hard X-ray-based operando techniques, special configurations including grazing incidence and/or nanostructured material systems (high surface to volume ratio) can be used to overcome its large penetration depth (bulk sensitivity) and thus to gain surface insight. 21,140−143 As already pointed out above, growing NPs epitaxially on single-crystal surfaces 26,82,143 or exposing ligandfree colloidal NPs to high-temperature thermal treatments to achieve epitaxial NP/support interfaces 144 can result in atomic layer sensitivity in particle height determination under catalytic conditions with these methods. This approach has also been used for investigating the catalytically active state of oxide-supported PtRh 2 NPs during near-ambient-pressure (NAP) CO oxidation. 143 Additional Bragg peaks in the reciprocal maps were identified that were attributed to the surface oxidation of the (001) facets of the NP. The formation of the O−Rh−O surface termination is accompanied by an increased CO 2 evolution rate, as determined by online mass spectrometry shown in Figure  5i,j. Here, dynamic gas-switching experiments changing the stoichiometry of CO and O 2 showed that the presence of the oxide surface termination on the (001) facets accompanies the CO 2 evolution rate, whereas the chemical state of (111) facets showed slower dynamics and was not correlated to the catalytic turnover. Interestingly, increasing the O 2 content in the gas feed from understoichiometric to stoichiometric conditions increased the CO 2 evolution rate more strongly than expected, which suggests that the reaction does not follow a simple Langmuir−Hinshelwood mechanism. Furthermore, driving the reaction under overstoichiometric O 2 conditions, which induces the oxidation of the (111) facets, did not increase the catalytic turnover. Thus, it was discussed that the dominating catalytically active sites are not located on the NP facets but at metallic sites at edges and corners presumably adjacent to the oxidized PtRh(001) facets that could lower the CO oxidation barrier. This work furthermore elucidated the need of combined structural and reaction product characterization to gain insights into catalytically relevant structural motifs.
In the case of NO oxidation over small but differently shaped Pt NPs, 21 the degree of Pt oxidation as determined by in situ XAFS analysis was found to be correlated with the catalytic activity. The Pt NPs were found to be oxidized above the onset temperature for NO oxidation (>150°C), and a fraction of PtO x remains present in the catalytically active state also at higher temperatures. Comparing their degree of oxidation with their catalytic properties revealed that the lower fraction of PtO x present for the NPs with 3D-like rather than 2D-like morphology yields a higher conversion and rate constant. Thus, Pt oxide formation deteriorates the catalytic conversion of NO.
In electrocatalysis, Pt-based NPs play an important role as fuel cell catalysts due to their high activity for the hydrogen evolution reaction and oxygen reduction reaction (ORR). In the case of the ORR, the formation of Pt−O bonds of optimum strength is important to yield high catalytic activity. 62,145 Additionally, Pt oxide formation plays a crucial role in electrocatalyst degradation due to the reductive dissolution under the ORR conditions of initially formed Pt 2+ species. 146,147 The electrochemical oxidation of Pt NPs has been widely studied for single-crystal surfaces and NPs using in situ X-ray absorption and diffraction. 35,126,133,134 For example, the electrochemical oxidation of size-selected Pt NPs (∼1.2 nm) was investigated using in situ high-energy resolution fluorescence detection (HERFD) XANES recorded at the Pt L 3 -edge (Figure 5k). 126 The oxidation was tracked by recording Pt L 3 XANES spectra during potential step experiments between ∼0 V and ∼1.2 V, and two distinct features attributed to metallic Pt and to Pt n+ species were considered. Peak fitting suggests that the Pt NPs are metallic up to ∼0.9 V, while OH x coverage increases with the potential. Above 0.9 V, the Pt NPs oxidize to form PtO-like species concomitant with metallic Pt. Above 1.1 V the contribution of the metallic Pt feature decreases significantly, suggesting its transformation to PtO x species. Comparing the experimental data to calculated XANES spectra led to the conclusion that at 1.2 V half of the Pt atoms are oxidized and can be best described as Pt 2+δ species. In a different study, the oxidation behavior of Pt NPs with a mean size of ∼2.6 nm was investigated using a combination of in situ XAFS at the Pt L 3 and XRD, and a special focus was placed on the structural response at potentials above 1.5 V. 133 Therein, only the upper two Pt layers of the NPs oxidize up to 1.5 V, leading to a lower structural coherence length of the Pt metal domains, as revealed from the broadening of the diffraction peaks. Above 1.5 V the anodic dissolution of Pt ions into the electrolyte was concluded from the decreasing diffraction peak intensity as well as edge height of the XANES spectra. These studies show how X-ray-based techniques can give insights into changes in the chemical state of NP electrocatalysts.
The evolution of the structure and chemical composition of O 2 -plasma-pretreated micellar Au NPs was investigated via operando XAFS during 2-propanol electro-oxidation. 24 During electrochemical conditioning and 2-propanol oxidation, the XANES features typical of oxides present in the as-prepared samples as well as the Au−O typical peaks in the Fourier transform of the (FT-)EXAFS spectra disappeared, while the metallic Au−Au CNs increased. These findings also showed

ACS Catalysis
Review that, despite the initial atomic oxygen exposure leading to the formation of Au 3+ species, metallic Au is the active species for 2-propanol electro-oxidation to CO 2 .
For the oxygen evolution reaction (OER), Ir-based as well as Co-and Ni-based materials are state of the art catalysts in acidic as well as in neutral and alkaline electrolytes, respectively, and they have been widely studied using in situ techniques. 25,130,131,148−158 In a recent study including in situ XAFS, the local atomic Co structure in nanostructured and NP electrocatalysts was found to unify under OER conditions. 155 For example, pyramidally shaped wurtzite-like CoO NPs oxidize during the OER toward a CoO x (OH) y primarily containing octahedrally coordinated di-μ-oxo-bridged Co 3+ ions. This oxidation process is accompanied by a significant roughening of the NP facets after the OER. The oxygenevolving state is characterized by a contracted Co−O bond length. Furthermore, in situ XRD has been applied on CoOcovered Co 3 O 4 nanocubes under technologically relevant alkaline OER conditions and a reversible formation of a crystalline CoOOH adaption layer was unraveled, Figure 6a− c. 157 These results are in agreement with findings on nanoporous Co 3 O 4 films showing a reversible decrease of the Co 3 O 4 nanocrystallite coherence length during the OER. 154 These and other works applying in situ Co K-edge XANES spectroscopy suggest changes in the charge of the Co ions and the presence of Co 3+δ during the OER. 130,131,[154][155][156]158 Similarly, an in situ XAFS investigation on Ir-based NPs suggests that Ir ions oxidize during the OER to an oxidation state formally higher than 4+. 148,152 Nevertheless, the presence of reactive oxygen species (O − ) has also been reported for Irbased electrocatalysts under OER conditions by combining in situ O K-edge XANES spectra and DFT calculations. 150,151 Similar changes were revealed in Co-based catalysts by in situ XPS based on the dip-pull method. 160 Thus, the location of oxidation equivalents on metal oxide NPs which are necessary for water oxidation and the OER is still under debate. The latter highlights the necessity of further in situ studies on model NP catalysts under OER conditions.
All of these findings suggest that the catalytically relevant structure of the OER catalysts is not well described by idealized single-crystal surfaces but rather has a disordered and/or dynamic near surface. This is even further supported by findings on the highly OER-active Ba 0.5 Sr 0.5 Co 0.8 Fe 0.2 O 3-δ (BSCF) NPs with a diameter of ∼150 nm investigated using E-TEM under NAP conditions to mimic the OER (Figure 6d− f). 159 Although using water vapor as reactant and studying ebeam induced reactions do not represent the "real" catalytic conditions, the BSCF NPs are prone to structural oscillations on the nanoscale during e-beam-induced oxygen evolution, as shown in Figure 6d. Gas bubble formation could be linked to the dynamics of the lattice in the near surface of the NP via electron energy loss spectroscopy (EELS). Significantly

ACS Catalysis
Review suppressed or absent oscillations were found for other less OER-active Co-based perovskites such as SrCoO 3−δ , shown in Figure 6e, and LaCoO 3 , shown in Figure 6f.
All these in situ works on 3d transition-metal-based OER catalysts revealed that the (near) surface of these nanocatalysts is not identical with the as-prepared single crystal state but rather is disordered during water oxidation and oxygen evolution. Structural flexibility is either a necessity or the disorder a result of the reaction.
3.1.2. 3d Transition-Metal-Based Nanoparticle Catalysts under Reducing Reaction Conditions. In addition to noblemetal electrocatalysts, 3d transition-metal-based NPs and nanocatalysts have received strong attention due to their abundance and catalytic activity. Two material classes based on 3d TM NPs excel by their relevance for industrial processes based on thermal catalysis: Cu NPs supported on ZnO for methanol synthesis during CO 2 hydrogenation 86,102,161,162 and supported Co NPs for Fischer−Tropsch synthesis (FTS) of hydrocarbons. 163−176 Furthermore, increasing efforts have been invested in the investigation of Cu-based NPs for the CO 2 RR to efficiently produce multicarbon products such as ethanol and ethylene. 53−56 In situ studies have been applied to investigate the chemical state and redox chemistry of Cu NPs under various conditions. 162,177−182 For Cu-based NPs, particular attention has to be paid to oxide formation under ambient conditions. 95 Thus, the transition from ambient to reaction conditions involves a chemical or electrochemical reduction: e.g., a reductive pretreatment is used as part of the catalyst activation protocol for high-pressure CO 2 hydrogenation. 86 When this is taken into account, the redox chemistry of Cu-based NPs was investigated using E-STEM, showing that Cu NPs oxidize and the formed Cu 2 O NPs reduce in a very similar fashion, although with significantly different kinetics. 177 The Cu NP oxidation starts at a single point and proceeds with an oxidation front along the NPs, as shown in Figure 7a. During Cu 2 O reduction under H 2containing NAP conditions, oxygen vacancies migrate into the bulk of the NP and accumulate at the Cu 2 O/Cu interface. A correlation between the accumulation of O vacancies and a lattice contraction leading to a collapse of the crystal lattice above a vacancy concentration of 50% was identified. 178 The redox chemistry of chemically prepared Cu 2 O NCs and their light-induced structural transformation were investigated to illuminate the degradation pathway of this photocatalyst. 179 Therein, the photoreduction of Cu 2 O to metallic Cu under UV light irradiation and in the presence of 5 mbar of water vapor led to an ∼25% decrease in the NC size. Interestingly, a gradual transition was identified, as both phases coexist in the same NC within the reduction. In the case of industrial CO 2 hydrogenation catalysts, a similar gradual reduction of CuO NPs supported on ZnO/Al 2 O 3 was revealed using in situ XAFS. 180 The calcined NPs can be reduced to metallic Cu in 10% H 2 /He at 150°C within less than 30 min, Figure 7b,c. The reduction was reported to proceed via Cu(I) formation with an interstitial O in the Cu lattice forming a substoichiometric Cu oxide.
A similar reaction process was determined for Cu NPs supported on or embedded in ZnAl 2 O 4 , as revealed by a combined TPR and in situ XANES spectra recorded at the Cu L 3 -edge. 162 In a reactant gas mixture for methanol synthesis, the reduction process also proceeds via a lower fraction of intermediate Cu + species. 181 Furthermore, the reduction

ACS Catalysis
Review process was found to be size-dependent for Cu clusters supported on hydroxylated Al 2 O 3 , as studied with in situ grazing-incidence XANES, and the reduction temperature was found to decrease in the order Cu 3 > Cu 4 > Cu 20 . 182 Notably, the onset for methanol formation follows the reduction temperature.
Under the actual CO 2 hydrogenation conditions, the Cu and Zn oxidation states and local atomic structures did not change significantly. 180 These findings are in agreement with the structural integrity of the Cu NPs as revealed in an in situ diffraction study. 104 The second highly relevant heterogeneous reaction where chemical state changes in the NPs have been reported is the Fischer−Tropsch synthesis. Typical FTS catalysts consist of Co NPs supported on Al 2 O 3 , SiO 2 , carbon, and TiO 2 . 183 In situ methods, mostly combining XRD and XAFS, have been used to study aspects related to the catalyst activation, the active catalyst state, and deactivation pathways for the Co NPs. 163 . 166−168 The activated Co-based Fischer−Tropsch catalyst consisted of both hexagonal-close-packed (hcp) metallic Co NPs with a size of ∼2.5 nm and face-centered-cubic (fcc) Co NP swith a size of 20 nm present as primary and secondary phases, respectively. 168 The coexistence of Co in fcc and hcp phases has also been reported in other in situ XRD and E-TEM studies. 166,169 Under the reaction conditions, the Co NPs were found to remain in the metallic state and resist sintering within the first 10 days on stream investigated. 164,165,168,172 Co 2 C formation was found during high-pressure FTS and under a pure CO atmosphere accompanied by a decreasing fraction of fcc Co, while the fraction of hcp Co remained constant. 168,171,173 As Co 2 C is inactive for FTS, its formation is considered to be part of the degradation pathway of Co NP FTS catalysts. This is in contrast to the case of Fe-based FTS catalysts, in which Fe 2 C is found to exhibit high catalytic activity. 184 Thus, in situ investigations on Co-based NP catalysts for FTS provided valuable insights on the reduction mechanism of Co oxides and on the structural integrity of the NPs under reaction conditions.
In the CO 2 RR, the catalyst structure and morphology are decisive for the catalytic activity and selectivity, as small and/or defective NPs have been found to favor the HER. 56 The presence of subsurface oxygen or substoichiometric Cu oxides under the reaction conditions is under debate, but has been suggested to lead to high ethylene/ethanol yields. 185 Therefore, it is highly important to identify the chemical state, the structure, and the catalyst morphology under CO 2 RR conditions. For this purpose, CO 2 RR electrocatalysts have been investigated using a variety of in situ and operando techniques, such as Raman spectroscopy 7,125,[186][187][188][189][190] as well as XAFS at the Cu L 3 -and Kedges. 39,125,192−195 Fewer in situ/operando studies on NP catalysts are available for the electrochemical CO 2 RR in comparison to long-studied reactions such as methanol and Fischer−Tropsch synthesis. Therefore, in the following, we cover the chemical state changes of NP and nanostructured Cu-based electrocatalysts under CO 2 RR conditions. In a recent study, the evolution of the chemical state and local atomic structure was investigated for electrochemically prepared Cubased nanocubes (NCs) supported on C during CO 2 RR. 39 In situ XAFS and quasi in situ XPS showed that their primary constituents in the as-prepared state are Cu 2 O and Cu−Cl species in the bulk and at the surface. As described previously, immersing the NCs in the electrolyte not only significantly changed the morphology but also removed the Cu−Cl species. During the CO 2 RR, the NCs supported on C are completely reduced to metallic Cu (Figure 8a), which is in agreement with other Cu-based NCs studied. 194 In contrast, those NCs grown on a Cu foil still show Cu 2 O surface species by quasi in situ XPS after 1 h of the CO 2 RR. 39,196 The latter finding is in agreement with results for plasma-treated Cu surfaces during the CO 2 RR, for which the Cu oxide domains partially remain,

ACS Catalysis
Review as probed by operando EXAFS and corroborated by ex situ STEM-EELS. 195 For longer reaction times, no conclusive statement could be made on the basis of the operando XAFS data, since the bulk metallic signal dominated and this method is not sensitive to very thin surface oxide layers, even in the grazing geometry used in this study. Interestingly, online product analysis shows that the Faradaic efficiency and thus the selectivity toward ethylene benefits from the oxidative plasma treatment, while the yield of methane almost vanishes. This work shows that the catalytic function can be tuned by applying a plasma activation pretreatment leading to a modified catalyst morphology and chemical state. The presence of subsurface or adventitious oxygen was identified in oxide-derived Cu electrocatalysts by using in situ NAP-XPS based on the dip-pull method. 191 Even under reductive CO 2 RR conditions, the adventitious oxygen signal in the O 1s XP spectra remained, whereas no spectroscopic fingerprint of Cu 2 O and CuO species was found. Thus, this indicated the absence of distinct CuO x phases in the reduced state and the absence of any Cu 2+ species was confirmed by Cu 2p XP spectra. Generally, we note that the approach of utilizing an electrochemical flow cell with a graphene electron-transparent window for in situ electrochemical X-ray photoelectron spectroscopy should be favored, because it allows stationary, realistic catalytic conditions by minimizing mass transport limitations and radical or product accumulation via constant electrolyte exchange. The presence of Cu + species under CO 2 RR conditions was furthermore shown using in situ Cu L 3edge XANES spectra (Figure 7b). 193 Therein, a gradual transition from Cu 2+ to Cu 0 with decreasing potential under quasi-stationary conditions was revealed, but it was shown that complete electrochemical reduction during the CO 2 RR takes 1 h.
Another study highlighted possible deactivation processes hampering the electrochemical reduction of CuO in the presence of carbonate ions. 192 The formation of Cu carbonates can passivate CuO electrocatalysts, as shown by XANES at the Cu L 3 -edge, leading to negligible catalytic activity. In the case

ACS Catalysis
Review of Cu 2 O and metallic Cu electrocatalysts, the deactivation process was not seen and in both cases metallic Cu was found to be the dominant species under CO 2 RR conditions. Combining the structural and morphological changes with the formed reaction products shows the detrimental effect on C 2 −C 3 selectivity of having small Cu nanocatalysts or those completely free of oxygen. These studies likewise show the importance of having combined morphological, structural, and chemical state insights into the catalyst properties under reaction conditions by using various in situ techniques to understand the catalyst function and its relation to the structure and morphology. However, the variety of partially contrary experimental findings on the presence of Cu + or substoichiometric Cu oxides under CO 2 RR reaction conditions on differently synthesized samples, as well as insufficient theoretical descriptions, has prevented to date the formulation of unambiguous statements on their mechanistic role. The latter discrepancies also highlight the need of using a variety of complementary but synergetic techniques operating under reaction conditions for an understanding of electrocatalytic processes, since the probe depth and sensitivity to the chemical state or the presence of amorphous versus crystalline species that might be extracted from the distinct methods will be different.

(Near-)Surface Compositional Changes and Segregation under Reaction
Conditions. The activity of NP catalysts can strongly depend on their composition not only in the near surface but also in the bulk. For example, it has been shown that core−shell NPs exhibit a significantly higher catalytic activity as fuel cell catalysts. 29,197−202 Therein, the lattice strain in the NP shell induced by compositional differences to the bulk varies the electronic structure and dband center. The metal d-band center directly influences the bond strength of adsorbates and reaction intermediates, and thus tuning the electronic structure of the surface can lead to improved catalytic activity. 203−205 Electrochemical potential cycling is often applied as an accelerated stress test (AST) for fuel cell catalysts, but it can induce substantial changes in the near-surface composition of Pt-based alloy NPs and the formation of core−shell NPs. 204,206−208 For example, in situ anomalous SAXS and pair distribution function analysis was applied to study the segregation and atomic ordering in PtNi NPs under electrochemical potential cycling (Figure 9a,b). 207 The electrochemical potential cycling was applied as a dealloying procedure for activation. In this study, the formation of a Ni-free shell was tracked and the transformation of the bulk PtNi solid solution into an ordered PtNi alloy revealed. In the case of octahedral Au NPs with a Pd shell, electrochemical potential cycling diminishes the core−shell structure and leads to a more homogeneous distribution of elements within the entire NP. 204 In the case of spherical PtCo NPs under gas-phase oxidizing and reducing conditions in an E-TEM, Co surface segregation and the formation of CoO islands was revealed. In the intermediate step, strained CoO films formed at the NP surface. Under reductive conditions, the Co atoms migrate into the bulk of the NP, leading to a pure Pt termination. Adsorbate-induced segregation was theoretically studied using DFT for an Au/Pd bimetallic surface in the case of CO. 211 In contrast to the bare Au surface on Pd, Pd migration to the surface and clustering were stronger in the presence of adsorbed CO.
The presence of Au vacancies on the surface significantly facilitate the Au−Pd swapping process, and the larger Pd NPs bind O 2 rather than CO. In the case of Pd−Ag alloys, the segregation process has been investigated by means of DFT during O 2 , CO, and C 2 H 2 adsorption. 209,212 Therein, it was revealed that acetylene adsorption reorders the Pd−Ag surface under simulated reaction conditions relevant for catalytic acetylene selective hydrogenation. In a combined experimental surface science and theoretical work, it was shown that on Ag(111) nanoislands of an AgPd alloy are formed and that they are Ag-terminated under UHV conditions (Figure 9c). 209 At elevated temperatures under CO conditions, Pd irreversibly migrates to the surface accompanied by an energetic stabilization of the Pd−CO bond. A similar process was identified under O 2 -containing conditions. Thus, this atom migration process is assigned to be the catalytic activation process. This study nicely shows how, by a combination of experimental (surface science) techniques and theoretical calculations, atomic segregation processes can be revealed and explained.
Segregation is also attributed to cause the deactivation of nanoporous Au due to residual Ag atoms. During nanocoarsening under CO oxidation conditions, the residual Ag atoms in these structures were found to rearrange and segregate to the surface, forming AgO x islands. 64 Ag surface segregation is enhanced by an ozone-based treatment of the nanoporous Au, and the chemical state of the oxide nature of the Ag-containing islands was identified using EELS. 68 These domains were discussed to provide O 2 dissociation sites for CO oxidation, and distinct O species were described after the activation procedure on the basis of NAP-XPS data.
The segregation of bimetallic NPs under gas-phase reaction conditions was investigated for CuNi NPs using NAP-XPS under simulated CO 2 hydrogenation conditions ( Figure  9d). 210,213 Therein, Ni surface segregation was identified in the case of an O 2 -containing atmosphere and is even stronger in a CO 2 + H 2 reaction mixture after reductive pretreatment. The presence of H-containing adsorbates as well as CO was attributed to cause the Ni segregation under simulated reaction conditions. In contrast, Cu surface segregation was revealed in a reaction mixture containing CO 2 + CO + H 2 , and DFT calculations suggested that CH 3 O, a stable reaction intermediate, could induce a reverse segregation energy for Cu and Ni under reaction conditions. 213 Here, we note that in situ investigations on the compositional changes and metal segregation under reaction conditions are especially important because the composition is prone to change even under ambient conditions due to the differences in the oxophilicity of the elements.

OUTLOOK
The presented in situ and operando studies focusing on NP transformations under gas-phase thermal and electrochemical catalytic conditions exemplify the insights that can be achieved with respect to the NP morphology, structure, chemical state, and composition using today's state of the art experimental tools. However, a more comprehensive picture of a working catalyst is still necessary for the quest of the active site as well as to go beyond the active site towards the entire catalytic system, extracting design principles for more active and/or stable catalysts. Therefore, improved experimental studies with respect to their methodology and the performance of the applied methods are still needed.

Review
First, we showed that the combination of operando techniques probing different properties of the catalysts such as the structure/morphology and the adsorbates as well as the reaction products in combined experiments yields a comprehensive picture of the active catalyst. However, these insights have been mainly limited to the investigation of NPs under gas-phase thermal reaction conditions. In the years to come, additional effort should be made in order to have such combined in situ/operando methods more extensively applied to the study of electrochemical catalytic processes. Combining local microscopy methods revealing the NP morphology with statistical methods such as XAFS or XRD showing the catalyst structure would significantly contribute to gaining more indepth insight into the heterogeneity of NP properties in the working catalysts. 43 Furthermore, we note that the simultaneous detection of reaction products and the determination of rates during in situ experiments, transforming them into operando investigations, is essential to identify function−property relationships. However, in the field of electrocatalysis this approach has not been established to the same extent as it has in thermal catalysis, and often the catalyst selectivity is analyzed in separate laboratory-based experiments. Here, the online analysis of gaseous products using gas chromatography and/ or parallel liquid product analysis using high-pressure liquid chromatography or mass spectrometry should be combined with operando structural and vibrational spectroscopic methods such as XRD, XAFS, IR, or Raman spectroscopy for the concomitant analysis of adsorbates, reaction intermediates, and structural/chemical modifications of the working catalyst. 7 In reactions such as the electrochemical water splitting or oxygen evolution, the number of possible reaction products is strongly limited. Nevertheless, the simultaneous detection of dissolved metal ions using inductively coupled plasma mass spectrometry (ICP-MS) would be ideal to gain further insight into the NP catalyst response to varying redox reaction conditions and its deactivation pathways via metal dissolution or NP detachment. 214 The presence of a liquid medium in electrochemistry significantly enhances the rate of removal of unstable intermediates from the NP surface, which emphasizes the need for additional operando stability studies to monitor the evolution of the NP loading on the support.
Additionally, operando experiments on NP catalysts based on X-rays are mainly conducted at synchrotron facilities, which limits the available experimental time and hampers the researchers' ability to conduct long-term stability studies on technologically relevant time scales. As noted above, these studies are essential to study the evolution of NP catalyst properties over the (whole) catalyst lifetime. This is especially important for highly oxygen sensitive catalysts such as metal NPs, which cannot be taken out of the reaction conditions prior to the end of the catalyst lifetime. Therefore, the development of additional suitable laboratory-based X-ray setups for operando catalyst characterization is primordial in order to open up the possibility of performing long-term durability experiments in house, extending the characterization time scales from a few days to weeks or months. This has already been implemented in operando XRD studies of NP catalysts under FTS conditions. 168 It is expected that laboratory XAFS setups and the implementation of suitable operando cells will speed up this needed development. 215,216 Generally, it is also clear that conventional in situ/operando investigations are limited with respect to the identification of the catalytically active site on NPs. Usually, the long-lived state in the catalytic cycle is probed as a majority species on the active sites, which is by definition the state preceding the ratedetermining step. The actual active states at which the crucial reaction step(s) proceed are most likely minority or trace species with a significantly shorter lifetime. Furthermore, the heterogeneity of most catalysts impedes the clarity of the obtained experimental observations. To fundamentally understand chemical reactions, more studies focusing on size-and shape-controlled NPs should be conducted so that activity-/ selectivity-determining properties can be more easily isolated. In this case distinct trends in NP properties such as the chemical state of surface and bulk atoms could be extracted from a series of measurements on differently sized NPs. Timeresolved in situ methods such as pump−probe experiments will allow the tracking of adsorbates, (electronic) structure, and composition changes during (electro)chemical reactions on millisecond to femtosecond time scales, leading to an improved understanding of active and spectator sites on a NP surface or its support. Some works have already given important insights in this field, 217−219 but many of these methods need to be adapted to investigate a wider variety of NP catalysts under different reaction conditions. In this respect, ongoing advances in the development of experimental setups at X-ray free electron lasers and synchrotron facilities are expected to accelerate the fundamental understanding of catalytic processes at the atomic scale.

CONCLUSION
In this work, we have presented a variety of in situ and operando studies revealing morphological, structural, and chemical state transformations of nanosized heterogeneous catalysts. The presented works mainly cover redox transitions of the nanoparticles under H 2 -and O 2 -containing atmospheres as well as the active state under gas-phase and electrochemical reaction conditions that are important within the framework of chemical energy conversion such as CO 2 hydrogenation and electroreduction as well as the oxygen evolution reaction. We presented various X-ray-based in situ/operando studies which revealed the transformation of NP catalysts under thermal and electrochemical reaction conditions with respect to crystallinity, chemical state, and local atomic structure. These studies provide a representative picture of the catalytically active state by probing a statistically relevant number of catalyst nanoparticles. Nevertheless, these techniques are not suitable to detect minor or trace species that can be of greater importance for the catalysis than the majority of spectator species, e.g. as potentially the Cu + species during CO 2 RR. Deep insight into the atomic processes on the surface of selected NPs could be obtained from environmental (S)TEM studies. Nevertheless, in these studies, the low reactant concentration hampered a comparison to realistic reaction conditions and it seems as if obtainable lateral resolution and reactant concentration, in the gas phase as well as the liquid phase, has to be compromised. In this respect, verifying the findings from environmental electron microscopy with either parallel statistical in situ/ operando insights or ex situ property−function correlations leads to a more comprehensive picture of the catalyst under reaction conditions. We think that in situ/operando studies have already given important insights in catalysis research and their further methodical progress, especially with respect to investigations of the catalyst dynamics, will allow researchers to

ACS Catalysis
Review increase their fundamental knowledge of the catalytically active state of nanoparticles.

Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Notes
The authors declare no competing financial interest.