Progress and challenges in structural, in situ and operando characterization of single-atom catalysts by X-ray based synchrotron radiation techniques

Abstract

Single-atom catalysts (SACs) represent the ultimate size limit of nanoscale catalysts, combining the advantages of homogeneous and heterogeneous catalysts. SACs have isolated single-atom active sites that exhibit high atomic utilization efficiency, unique catalytic activity, and selectivity. Over the past few decades, synchrotron radiation techniques have played a crucial role in studying single-atom catalysis by identifying catalyst structures and enabling the understanding of reaction mechanisms. The profound comprehension of spectroscopic techniques and characteristics pertaining to SACs is important for exploring their catalytic activity origins and devising high-performance and stable SACs for industrial applications. In this review, we provide a comprehensive overview of the recent advances in X-ray based synchrotron radiation techniques for structural characterization and in situ/operando observation of SACs under reaction conditions. We emphasize the correlation between spectral fine features and structural characteristics of SACs, along with their analytical limitations. The development of IMST with spatial and temporal resolution is also discussed along with their significance in revealing the structural characteristics and reaction mechanisms of SACs. Additionally, this review explores the study of active center states using spectral fine characteristics combined with theoretical simulations, as well as spectroscopic analysis strategies utilizing machine learning methods to address challenges posed by atomic distribution inhomogeneity in SACs while envisaging potential applications integrating artificial intelligence seamlessly with experiments for real-time monitoring of single-atom catalytic processes.

---

Yuhang Liu

Yuhang Liu is a Master jointly trained at the Shanghai Synchrotron Radiation Facility and Suzhou University of Science and Technology, under the supervision of Prof. Hong Bin Yang and Prof. Xiangjun Wei. His current research includes electrochemical energy conversion and in situ/operando spectroscopy studies.

Xiaozhi Su

Xiaozhi Su is a staff scientist at the Shanghai Synchrotron Radiation Facility (SSRF) of Shanghai Advanced Research Institute, Chinese Academy of Science. He is responsible for the daily operation of the XAFS beamline (BL14W1) at SSRF and also plays a key role in designing and constructing the XAS endstation for the China Petrochemical Corporation (Sinopec) at BL04W, as well as building the XES spectrometer at BL20U. His research mainly focuses on the computational modelling of core-level spectroscopy of solids and the design of in situ/operando electrochemical setups.

Xiangjun Wei

Xiangjun Wei is a professor at the Shanghai Synchrotron Radiation Facility (SSRF) of Shanghai Advanced Research Institute, Chinese Academy of Science, graduated with a Bachelor's Degree in materials science from Lanzhou University in 1990 and a PhD degree from Lanzhou University in 2006. He participated in the construction and operation of the XAFS beamline at SSRF and was responsible for the design and construction of the Dynamics Beamline (D-Line) at SSRF, which combined synchrotron radiation infrared with ED-XAS and realized the concurrent measurement of molecules, atoms and electronic structures at microsecond/millisecond time resolution. Now, he is the team leader of D-Line.

Hong Bin Yang

Professor Hong Bin Yang received his received BS (1998) from Lanzhou University and PhD (2008) from Fudan University. Thereafter, he moved to Nanyang Technological University and worked as a research fellow in the School of Chemical and Biomedical Engineering (2008–2016). In 2016, he joined the Suzhou University of Science and Technology. His active research interests focus on electrocatalysis, photoelectron chemistry and carbon-based catalysts. He has published over 170 original research papers in chemistry, materials and energy related journals. His research has been cited over 23 000 times with an h-index of 72 (Google Scholar).

Bin Liu

Professor Bin Liu received his bachelor/master of engineering degree from the National University of Singapore in 2002/2004 and completed his doctoral degree at the University of Minnesota in 2011. After spending a year as a postdoctoral fellow in the University of California Berkeley, he joined the School of Chemical and Biomedical Engineering at Nanyang Technological University as an Assistant Professor in June 2012 and was promoted to Associate Professor in March 2017. In February 2023, he joined the Department of Materials Science and Engineering at the City University of Hong Kong as a STEM Professor. His research focuses on photo(electro)catalysis and in situ/operando characterization.

---

1. Introduction

The development of green energy and low-carbon emission technologies is the key to a sustainable human society. In this regard, the application of renewable and sustainable energies such as solar, wind and tidal energies has been greatly expanded. It is worth mentioning that the transformation between neo-energy technologies and traditional industrialization largely depends on the development of innovative catalytic reactions. Therefore, an in-depth study of the catalytic reaction mechanism is extremely important for the innovation of catalysts and catalytic processes, which provides an effective means to promote the innovation of new energy technologies.^1,2 Over the past few decades, various strategies such as structural engineering, alloying, doping, etc. have been developed to improve the intrinsic activity of catalysts. It is worth noting that the essence of improving the intrinsic catalytic activity, regardless of the structure or morphology of the catalyst, mainly stems from the synergistic effect between the active site and its surrounding chemical environment. Therefore, we can consider the active atoms and their coordination environments as the active centers.^3–5 Meanwhile, the earth's resources of several highly active metallic species, especially noble metals such as Pt, Ir, Rh, etc., are limited. This leads to high prices for these noble metal substances and consequently researchers have to consider the financial profitability when developing highly active catalysts.⁶ SACs are some of the ideal solutions for this challenge, thanks to the apparent advantages of maximizing atomic utilization efficiency. We can regard the central atom with its coordination environment in a SAC as the smallest unit of the active site.^7–9

From an electronic structure perspective, as the size of a material decreases from bulk to nanometers/sub-nanometers or even single atom, its electronic structure changes from continuous energy bands to discrete energy levels, resulting in electron orbital quantization (Fig. 1a).^10–12 The quantized electron orbital of a single metal atom will be readily influenced by variations in its coordination environment when it directly interacts with heteroatoms on a substrate to form an active center.^13–17 Thus, the catalytic activity and selectivity of SACs could be optimized through adjustment of the coordination environments.

Fig. 1 (a) Schematic diagram showing the change in d-bands with the metal size. (b) Schematic diagram showing the coordination environment characteristics of SACs.

As early as 1960s, scientists observed that certain catalysts with low metal loadings exhibited complete dispersion based on chemical adsorption, indicating possibly the presence of atomically dispersed active sites. However, due to limitations in characterization techniques at that time, it was not possible to confirm whether these active sites were composed of single atoms or small clusters. With advancements in characterization techniques in the 1990s, it was proven that loading single metal atoms onto a support effectively enabled them to function as catalytically active sites. The introduction of XAS into research on SACs in the early 1980s led scientists to recognize its immense potential. In 1999, Iwasawa and co-workers reported atomically dispersed Pt/MgO catalysts and discovered their superior activity for propane combustion as compared to Pt (nanoparticles) supported on MgO catalysts.¹⁸ By utilizing EXAFS spectroscopy, the researchers determined the atomic-level dispersion of Pt and found reversible reconstruction between Pt₆ and Pt₁ under a reaction atmosphere, which likely accounted for the high activity in propane combustion. Subsequent studies employed acid washing to selectively retain only single-atom sites in the catalyst without any change in catalytic activity,^19,20 thus proving that these atomically dispersed metal sites were indeed the true catalytically active sites. This milestone had a significant impact on SAC research. In 2007,²¹ Adam et al. successfully synthesized a catalyst with atomically dispersed Pd (0.03 wt%) on Al₂O₃. Analysis of the absorption spectrum confirmed the presence of atomically dispersed Pd atoms that were bonded with lattice oxygen of Al₂O₃. Sub-angström-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) is used to directly observe the individual Pd atoms on Al₂O₃. In turn, the term “single-atom catalyst” was proposed by Zhang et al. in 2011,²² gaining significant attention in the fields of both heterogeneous catalysis and nanomaterials.^23,24 In general, SACs are characterized by isolated metal atoms with well-defined local coordination structures on a substrate, including a metal center, the first coordination shell and the second coordination shell, where the first coordination sphere resembles the ligand coordination environment in homogeneous catalysts (e.g., metalloenzymes) (Fig. 1b). Therefore, the coordination environment of the metal atomic centers is of great significance for understanding the origin of the activity of SACs.²⁵ For example, variation of ligands (such as pyrrolic N/pyridinic N or N/O/C) has an impact on the selectivity of the oxygen reduction reaction (ORR) on cobalt-based SACs with Co metal centers.²⁶

Although there are many characterization techniques available that can be used to obtain structural information at the atomic level, conventional characterization techniques cannot precisely determine the coordination structures of active centers in SACs, which greatly limits the in-depth study of the structure–performance relationship in single atom catalysis.^27–29Fig. 2 summarizes the common characterization techniques and their advantages/limitations for probing structural information as compared to synchrotron radiation techniques. X-ray diffraction (XRD)/small-angle X-ray scattering (SAXS) is unsuitable for detecting the structure of materials with sizes below sub-nanometers. Infrared (IR) and Raman spectroscopies can be used to probe surface functional groups of SACs as well as the evolution of reaction intermediates during catalytic processes. Additionally, by coupling with probe molecules, IR spectroscopy can provide a complementary sample-averaged assessment of a metal structure's uniformity and distribution. However, probe molecule-based characterization is an indirect method and may not be applicable to all metal species, while the controversial peak assignment and potential probe molecule-induced reconstruction pose challenges for in situ structural analysis. Meanwhile, IR and Raman spectroscopies cannot provide structural information of SACs, as substrates of SACs produce too strong signals in the infrared and Raman spectra, which will mask the local single-atom coordination signals. Electron paramagnetic resonance (EPR) is highly suitable for studying paramagnetic SACs, and nuclear magnetic resonance (NMR) can provide atomic level information about local chemical environments in SACs. However, SACs with complex coordination environments often complicate the EPR and NMR spectral interpretation due to signal overlap and broadening. Electron microscopy can be used to probe the local topography and geometry of materials, but for SACs, HAADF-STEM with ultra-high resolution can show only a few “bright spots” to prove the presence of single atoms.³⁰ X-ray photoelectron spectroscopy (XPS) can provide valence information of SACs, but the poor signal-to-noise ratio of the spectra recorded for trace elements does not provide more accurate information.

Fig. 2 Overview of characterization techniques, including the advantages/limitations of the common characterization techniques and the structural information that can be obtained through synchrotron radiation characterization techniques.

Compared with the above-mentioned common characterization techniques, X-ray based synchrotron radiation techniques can provide clearer structural information of SACs.^31,32 Since the introduction of XAS into SAC research in the early 1980s,³³ it has become an increasingly important technique for characterizing the coordination structure of SACs and tracking evolution of single-atom active sites in catalytic reactions. By using high-energy X-rays to interact with the sample to produce a specific energy level transition and measuring the absorbed or scattered radiation intensity changes, information about the atomic structure, charge state and interaction between elements can be obtained. The widespread use of XAS has provided researchers with a more comprehensive means to obtain structural information such as oxidation states, coordination numbers, metal–ligand bond lengths, symmetries, etc. However, geometric and electronic structure determination by XAS is an inverse problem, which is open to subjective interpretations on whether a proposed structural model is statistically valid and differentiable from other structures. Consequentially, there is always the potential for ambiguity in conclusions regarding structures of metal species in a sample, especially that the oxidation state and symmetry information obtained from XANES is only an average signal, and elements with similar atomic numbers are indistinguishable by analyzing the backscattering distances between atoms. This is because these elements are located close to each other in the periodic table, and they have similar charge densities and scattering cross-sections. It is impossible to precisely distinguish different ligand atoms in the coordination environment of SACs, especially some elements with similar atomic numbers such as C, O, and N, due to the fact that the obtained information from EXAFS only represents the average backscattering distances between atoms. Therefore, in the process of analyzing the structures and properties of SACs, although higher spectral resolution can provide more detailed information, to a large extent, XAS data and analysis methods such as fitting and computational simulation can only provide possible structural information, for which more complementary characterization techniques are needed. For example, X-ray emission spectroscopy (XES) is highly sensitive in determining elements, valence states, orbitals, spins, etc., allowing precise resolution of occupied orbitals and nearby ligands, especially for ligands with similar scattering distances, such as C, N and O, providing more credible structural information of supported SACs.³⁴ However, due to the low concentration characteristics of SACs, the electronic signals collected are weak, thereby necessitating a time-consuming process to obtain accurate and reliable results. Furthermore, resonant inelastic X-ray scattering (RIXS) (including resonant X-ray emission spectroscopy (RXES)), which is a combination of XAS and XES, can provide more hierarchical and higher-dimensional data than traditional XAS and XES.^35–37 This technique not only offers insights into charge transfer processes within atoms or molecules, but also reveals important features such as excited states, spin–orbit coupling, and lattice dynamics in materials.³⁸ However, RIXS is limited to samples with high atomic numbers or heavy elements, and it also requires a deep understanding of theoretical models and algorithms for accurate interpretation. It is important to keep in mind the limitations of these techniques when conducting structural studies and actively seek complementary methods to obtain more comprehensive and accurate results.

By perturbing a chemical system followed by observing the system's response, it is possible to provide an understanding of the reaction kinetics and mechanism beyond that can be achieved from steady-state measurements. This gives rise to the field of in situ/operando methods for studying kinetic events.^39–42 To further explore and understand the formation of catalytically active sites in SACs and also the changes in their geometry and electronic structure during reaction, in situ/operando experiments offer undoubtedly the most effective method.^43,44 Techniques such as XAS, XES and RIXS/RXES enable precise measurements of valence electrons, which are important for understanding the origin of reaction activity at the electronic level under in situ/operando conditions. Moreover, we like to highlight an energy-dispersive X-ray absorption spectroscopy (ED-XAS) synchrotron radiation technique based on an oscillator light source that can achieve microsecond time resolution. Thanks to the unique advantages of this technique of spatial and temporal scale synchronization, in situ/operando reaction kinetics studies can be conducted by coupled multi-spectroscopy techniques (such as infrared spectroscopy, mass spectrometry, etc.). In this review, we will summarize the recent advances in synchrotron radiation-based methods for characterizing SACs, focusing on the spectral analysis strategies for SACs using synchrotron radiation-based XAS, XES, and RIXS/RXES. Next, we will highlight the limitations of synchrotron radiation-based techniques in determining the coordination environments of SACs. Finally, we will discuss the future challenges and opportunities of ED-XAS techniques for in situ/operando characterization of SACs under reaction conditions.

2. Hard X-ray absorption spectroscopy (hard-XAS)

X-ray based synchrotron radiation spectroscopy is considered as one of the most powerful and advanced characterization techniques that can provide information on the electronic/geometric structure and elemental composition of materials,⁴⁵ among which, hard-XAS has become an indispensable tool for studying SACs, playing a crucial role in identifying the structural characteristics of single-atom centers. Hard-XAS is also regarded as the most ‘traditional’ X-ray based synchrotron radiation technique for characterizing SACs. Consequently, hard-XAS has been extensively reviewed in the current literature. Building upon this foundation, we aim to elucidate the correlation between structural changes of SACs and their intricate spectral features observed by hard-XAS to systematically summarize and supplement the existing knowledge and provide a comprehensive understanding of how structural modifications influence the electronic states and chemical reactivities of SACs.

During XAS measurement, the energy of the incident X-rays resonates with the electrons in the core energy levels of an element, and this results in excitation of inner shell electrons to unoccupied continuum energy levels. With increasing energy of the incident X-rays, there is a sharp increase in the number of electrons that contribute to absorption excitation, and this specific energy is referred to as the absorption edge (Fig. 3a).^46,47 The edges are designated, in part, based on the excitation of specific core electrons: the K-, L-, and M-edges correspond to principal quantum numbers n = 1, 2, and 3, respectively. For instance, excitation of a 1s electron occurs at the K-edge, while excitation of a 2s or 2p electron takes place at the L-edge. The energy of the absorption edge is directly correlated with the core energy level of the element, making XAS an element-specific technique. The XAS spectrum can be broadly categorized into three distinct regions. Firstly, the pre-edge region refers to the low-energy range before the absorption edge. This region provides valuable information about the electronic structure of atoms or ions (Fig. 3b). Moving on to X-ray absorption near edge structure (XANES), this region encompasses a higher energy range just after the absorption edge, which is characterized by fine spectral features that arise from transitions between different electronic states of atoms or ions. The XANES region offers insights into local atomic arrangements, oxidation states, and coordination environments within materials. Lastly, the extended X-ray absorption fine structure (EXAFS) region surpasses the XANES range, exhibiting intensity oscillations resulting from interference effects caused by multiple scattering events of photoelectrons with neighboring atoms or ions. Through analysis of these oscillations, researchers can acquire intricate structural information such as bond lengths and coordination numbers.

Fig. 3 (a) Nomenclature of X-ray absorption edges and their relationship with the corresponding excitation of core electrons. (b) The XAS spectrum consisting of the absorption edge and a series of oscillatory structures. The conventional division between XANES and EXAFS regions is indicated by different colors. The inset shows the schematic diagram of the X-ray absorption process. (c) Orbitals showing electronic transitions and life-time broadening. Copyright 2010, Royal Society of Chemistry.³⁶ (d) Summary diagram of SAC structure information obtained by hard/soft-XAS.

The extremely low atomic concentrations of SACs pose great challenges in the accurate characterization of their structures using traditional techniques. The extended-edge region in hard-XAS often exhibits insufficient spectral resolution for SACs due to the limited number of atoms present. Additionally, the K-edge XAS of SACs has very a short core–hole lifetime, which can further obscure many spectral details (Fig. 3c).^36,48 To address this issue, researchers are constantly exploring new methods to further enhance the accuracy of XAS for determining specific atoms in SACs. For example, researchers have turned to high-energy resolution fluorescence detection X-ray absorption spectroscopy (HERFD-XAS). One key advantage of HERFD-XAS is its ability to provide a much higher resolution than the natural line width when measuring X-ray fluorescence. By achieving better energy discrimination, HERFD-XAS allows researchers to distinguish subtle differences in the electronic structure and chemical environment within SACs with greater accuracy.^49–52 As depicted in Fig. 3c, the fluorescence lines at identical absorptions result in longer-lived core–hole states, which can effectively mitigate the broadening effect and improve spectral resolution. With greatly improved energy resolution, HERFD-XAS enables researchers to resolve the fine features of SACs. This not only enhances our understanding of the electronic properties and catalytic mechanisms of SACs, but also provides valuable insights into their performances for potential applications. Additionally, other advanced technologies such as high-energy radiation beam lines can also amplify XAS signals for more detailed and accurate descriptions of elemental environments and their influence on SACs.⁵³

2.1 Probing isolated single atoms

Most of the research works reported that in the Fourier transform of the EXAFS (FT-EXAFS) spectra of SACs, only contributions from short-range scattering paths resulting from metal–heteroatom interactions can be observed,⁵⁴ while peaks associated with metal–metal or metal–heteroatom–metal scattering paths were not detected,^55,56 which is considered to be a purely qualitative feature.^57–60 However, most relevant literature lacked rigorous quantitative analyses to evaluate the validity of the conclusion that EXAFS could prove the absence of aggregated metal (or metal oxide) structures. This claim may have some omissions for detecting disordered or mixed materials with different sizes (such as subgroups of metals and clusters of metal oxides) in some samples. For example, due to the inherent characteristics of low-polymeric ReO_x and Re₂O₇ solid structures, even when the loading amount of Re was as high as 20 wt% (Fig. 4a), XAFS could only display a distinct first shell peak and was not able to detect scattering features related to the second shell peak associated with Re–O–Re (Fig. 4b and c).⁶¹ Therefore, relying solely on EXAFS data to determine isolated single atoms is unreliable and other complementary characterization techniques are needed from different perspectives to confirm the exact structure of SACs.⁵⁴ For example, for a series of SACs (including Mn, Fe, Co, Ni and Cu),^62–64 the corresponding FT-EXAFS spectrum showed only one apparent characteristic peak, and the XRD, XPS and HAADF-STEM measurements confirmed the absence of metal/metal oxide clusters/nanoparticles. Based on the above results, it can be inferred that the peaks in the EXAFS spectra can be attributed to the M–N coordination in the first shell.^65,66 Meanwhile, in another example containing 60% Pt single atoms and 40% Pt oxide clusters, the EXAFS fitting result according to atomically dispersed Pt still gave a good R-factor (<1%) (Fig. 4d).⁶⁷ Here, we list a few limitations in applying EXAFS to the SAC study. (1) SACs typically have a very low concentration of the active metal sites, leading to the weak EXAFS signal. This low signal intensity can result in poor data quality, making it difficult to extract reliable structural information. (2) The weak signal from SACs can be easily overwhelmed by background noise, which will complicate the data analysis and reduce the accuracy of the structural parameters obtained. (3) The atomically dispersed metal centers in SACs cannot exhibit a sole coordination environment, and in most cases, the metal atomic center possesses a range of different coordination environments. This complexity can lead to overlapping features in the EXAFS spectra, making it challenging to deconvolute and accurately interpret the data. (4) EXAFS is sensitive to the local atomic environment around the absorbing atom and is only able to separate backscattered atoms by path length criteria, giving information on the average interatomic distance and coordination number between the absorbing and nearby backscattered atoms (the second row to the right of Fig. 3d),^68–71 typically within a few angstroms. While this provides detailed information about the immediate coordination sphere, it offers limited insight into the long-range order and overall structure of the catalyst. (5) The support material in SACs can contribute significantly to the EXAFS signal, complicating the analysis. Distinguishing between the contributions from the single metal atoms and the support requires careful experimental design and sophisticated data processing techniques. (6) Accurate interpretation of EXAFS data often relies on comparison with well-characterized reference standards. For SACs, suitable reference compounds may not always be available, making it difficult to draw definitive conclusions about the local structure. (7) In some cases, multiple scattering effects can complicate the EXAFS spectra, making it more difficult to extract accurate structural information. This is particularly relevant for systems with complex geometries or multiple scattering paths. (8) Quantitative analysis of EXAFS data to determine precise bond lengths, coordination numbers, and disorder parameters can be challenging, especially for systems with low signal-to-noise ratios and complex coordination environments. Nevertheless, EXAFS analysis and EXAFS data fitting have now become a routine method to be used to determine the metal coordination environment in SACs. Modeling the Fourier-transformed χ(k) data allows for obtaining information such as the coordination number, bond length, and disorder term (σ²) associated with a given photoelectron scattering path. The data fitting quality is assessed using metrics such as the R-factor (the sum of squared differences between each data point and the corresponding fitted result divided by the sum of squares of each corresponding data point) or reduced χ² (more commonly used).^72–74 However, due to the close atomic distances between certain elements (such as N, O, and C), it can be challenging to determine the exact coordination numbers of these elements in the metal center coordination environment. Some literature reports reported relevant fittings about ligand atoms (N/O/C). For instance, in the Co–N–C SAC study, 4 and/or 5-fold coordination structure information was obtained by best-fit analysis.⁷⁵ An excellent fit could be obtained without considering the Co–Co backscattering signal and the main contribution to the EXAFS signal is given by the Co–N and Co–O bonds of the first coordination shell (Fig. 4e). The corresponding fit in Fourier transform space shows that the peak at 1.45 Å is associated with Co–N and/or Co–O first shell layer contributions, while the peak at 2.44 Å is ascribed to Co–C (Fig. 4f).⁷⁶ Zhu et al. used Cu–N, Cu–O, and Cu–C reflection paths for optimal fitting analysis and proposed a coordination structure model where the Cu center is coordinated with two N atoms and two O atoms (Fig. 4g).⁷⁷ Despite these fitting results being very refined, we must still recognize that the structural information obtained solely from analyzing the average signal of EXAFS can only serve as an ideal fitting model with certain rationality, and there is still a lack of exact evidence to determine the results, so it is difficult to fully determine the type of backscattering atoms solely through FT-EXAFS fitting. By contrast, distinguishing between atoms with significantly different atomic numbers is easier and much more reliable. Using the least-squares EXAFS fitting method to extract the structural parameters of Cu-SAC demonstrated that the central Cu atom had a first shell coordination number of 4, which was directly connected by one S and 3 N atoms with average bond lengths of 2.32 Å and 1.98 Å, respectively (Fig. 4h).^78,79 Although the high-quality EXAFS data mentioned above can provide more detailed spectral features, the uncertainty of active sites set an upper limit on the accuracy of SACs’ structural interpretation, largely representing only the average coordination environment of the atomic center. Therefore, other complementary characterization techniques as well as theoretical simulations are needed to verify the structural information obtained from EXAFS. For example, XRD, XPS, XES (will be discussed below), reaction mechanism studies based on density functional theory (DFT) and machine learning are now commonly used to explain the catalytic performance of active sites determined by XAS, while excluding contributions from observed but inactive bystander species in the spectra.

Fig. 4 (a) K²-Weighted χ(k) EXAFS and (b) the corresponding magnitude of the Fourier transformed EXAFS taken over Δk = 3.9–14 Å⁻¹ for various weight loading Re/SBA-15 catalysts after in situ oxidation. The inset of (b) shows the corresponding proposed model. (c) k²-Weighted magnitude of the FT EXAFS (Δk = 3.9–14 Å⁻¹) of Re₂O₇ with the crystalline structure shown in the inset. (d) Modeled FT EXAFS data for mixtures of the Pt/CeO₂ SAC site and oxidized Pt₁₄O₃₇ cluster. Copyright 2023, American Chemical Society.⁵⁵ (e) K²-Weighted k-space spectra of Co_0.5 with a CoN₄ moiety having one oxygen atom in the axial direction. Curves from top to bottom: Co–N, Co–O and Co–C γ⁽²⁾ two-body signals and the N–Co–N γ⁽³⁾ three-body signal included in the fit, the total signal (red line) superimposed on the experimental one (black dots). (f) The fit in the Fourier-transform of the EXAFS spectra of Co_0.5 with a CoN₄ moiety having one oxygen atom in the axial direction. The purple, blue and red balls represent cobalt, nitrogen and oxygen atoms. Copyright 2017, Springer Nature.⁷⁵ (g) FT k³-weighted Cu K-edge EXAFS spectra of S-Cu-ISA/SNC and the references. Copyright 2020, Springer Nature.⁷⁸ (h) EXAFS fitting curves of Cu-CDs in R space using backscattering paths of Cu–N, Cu–O, and Cu–C. Copyright 2021, Springer Nature.⁷⁷ (i) and (j) Schematic view of two different coordination configurations for one type of backscattering atom (denoted as S1) located at two distances (so-called two shells) and two types of backscattering atoms (denoted as S1 and S2) overlapping at one distance (so-called one shell) around the central absorbing atom (denoted as O). The simulated EXAFS-FT moduli and EXAFS-WT distribution corresponding to the two coordination configurations, respectively. Copyright 2019, Royal Society of Chemistry.⁸⁰

Additionally, the wavelet transform (WT) of EXAFS can identify the contribution of each backscattered atomic path in both R-space and K-space, providing elemental properties and radial distance information.^81–83 The distinction between heavier (metal atoms) and lighter (C, N and O, etc.) backscattering atoms can be greatly simplified by the information obtained from the WT of EXAFS in R-space and k-space. The horizontal axis in the WT plot of EXAFS shows the wave vector number k, which is the key to distinguishing different kinds of coordination atoms, where atoms with smaller atomic numbers have weaker abilities to scatter photoelectrons and their strongest oscillations appear in the lower k-space part, while the strongest oscillations for atoms with larger atomic numbers appear in the higher k-space part (Fig. 4i and j).^39,80 Therefore, performing wavelet transformation on high-quality EXAFS data (with high resolution and signal-to-noise ratio) may even provide theoretically possibility for distinguishing coordinating atoms of C, N, and O in the coordination environment of SACs.

2.2 Exploring electronic states and geometric configurations of single atoms

XAS signals at the energy of interest are very sensitive to materials’ atomic, electronic and magnetic structures.⁸⁴ As a result, the XANES spectral shapes and edge energy positions are sensitive to the ligand type and the three-dimensional coordination environment around the central absorbing atom, whose edge shoulder features sometimes introduce a square planar configuration of the ligand atom that can be used to better identify the atomic site structure (the right of Fig. 3d).^85–87 As shown in Fig. 5a, the change in the oxidation state of a single Fe atom in the Fe-SAC could be clearly inferred by comparing the Fe k-edge energy positions.⁸⁸ For 3d transition metal single atoms, two important pre-edge features can usually be observed. The lower energy peak is due to the quadrupole-allowed 1s → 3d transition, whose intensity is proportional to the ligand deviation of the photo-absorbers, while the peak with higher energy comes from the 1s → 4p_z smooth transition and is a fingerprint of the square-planar configuration with high D_4h symmetry. For instance, the XAS spectra of A-Ni-NG and A-Ni-NSG with a similar D_4h symmetry to nickel phthalocyanine (NiPc)^89,90 show features on the edge of the XANES spectra, indicating distorted D_4h symmetry of Ni atoms in A-Ni-NG and A-Ni-NSG (Fig. 5b).^91,92 In addition, the relative intensity ratios (I_C/I_D) of the C (1s → 4p_x,y transition) to D (multiple scattering process) peaks in A-Ni-NG and A-Ni-NSG are significantly larger than that in Ni(II)Pc, and the increase in I_C/I_D further reflects the intensity of the oxygen reduction reaction.^93,94 More accurate structural information of SACs can be obtained by combining spectroscopic analyses and theoretical calculations.⁹⁵

Fig. 5 (a) XANES spectra at the Fe K-edge of various samples. Copyright 2019, Springer Nature.⁸⁸ (b) Ni k-edge XANES spectra of A-Ni-NG, A-Ni-NSG and NiPc, where peak A represents 1s → 3d transition, peak B represents 1s → 4p_z transition, and peaks C and D represent 1s → 4p_x,y transitions and multiple scattering processes, respectively (the inset shows the expanded pre-edge region; the grey shaded areas show the intensity of the 1s → 3d transition). Copyright 2018, Springer Nature.⁸⁹ (c) Comparison between the experimental Ru K-edge XANES spectrum of Ru-SAS/SNC and the theoretically simulated XANES spectrum. Copyright 2022, American Chemical Society.⁹⁶ (d)–(f) Comparison between the experimental K-edge XANES spectrum of Co_0.5 (black hollow circles) and the theoretically simulated XANES spectrum (solid red lines). Insets show the models used for theoretical simulation: cobalt (purple), nitrogen (blue), oxygen (red) and carbon (gray). (g) A table showing the XANES structural parameters. Best-fitting structural parameters based on the structures proposed in Fig. 5d–f. Bending is the angle between the Co–O vector and the O–O bond and R_sq is the residual function. Copyright 2017, Springer Nature.⁷⁵

To find whether sulfur is attached to the nuclear Ru atoms or the peripheral N atoms, Xin et al. developed an optimized DFT calculation model for XANES simulation.⁹⁶ In the Ru K-edge XANES spectral simulations based on the Ru–N₄, Ru–N₃S and Ru–N₄–S fraction models embedded in graphene sheets, the simulated spectrum based on the Ru–N₄–S model matched well with the experimental spectrum (Fig. 5c). Thus, the bond between the S and N atoms initiated a slight asymmetric expansion of the Ru–N bond, achieving an indirect electronic modulation of Ru from the second shell layer. The changes in coordination structures (such as the coordination number and intermediate adsorption) caused by external factors are important for the metal centers of SACs. Jaouen et al. determined the detailed structure of CoN_xC_y in the Co–N–C SAC by XANES measurement and DFT calculations based on the previously discovered active site geometry in Fe SACs.^75,97 The results showed that the Co–N bond lengths of square planar CoN₄C₁₂, porphyrins CoN₃C_10,porp and CoN₂C₅ were all 1.96 Å when O₂ was absent (Fig. 5g). When O₂ was end-on adsorbed and there was no out-of-plane cobalt displacement, the Co–N bond lengths of CoN₄C₁₂ and CoN₃C_10,porp were 1.97 and 1.99 Å, respectively (Fig. 5d and e). On the other hand, CoN₂C₅ representing half of the porphyrin fraction was bound to the end of O₂ with a Co–O bond length of 1.90 Å. O₂ and CoN₂C₅ motifs were in the same plane, and the Co–N bond length was 2.01 Å (Fig. 5f). In summary, XANES contains more structural information, which can provide a more accurate qualitative description in terms of the coordination structure, electronic orbitals, energy band structure and multiple scattering. The structural information, such as the coordination atom, coordination number, and oxidation states, can be further obtained by a least-squares EXAFS fitting method and DFT modeling.

2.3 Tracing dynamic structural evolution of SACs under reaction conditions

The application of in situ/operando hard X-ray absorption spectroscopy enables observation of the chemical state evolution of the metal center in SACs,^98–100 thereby facilitating a deeper understanding of reaction mechanisms and exploration of the structure–property relationship in SACs.^101–103 It is worth noting that, during in situ experiments, the complexity of interactions between sample surfaces and environments needs to be taken into account, especially in liquid-phase electrocatalytic systems where there are more interfering factors on the spectral signal intensity compared to thermal or photochemical catalytic reactions.^104,105 Moreover, compared to the transmission mode, the fluorescence mode is more suitable for these complex mixed in situ experimental systems, and its high sensitivity is also applicable for detecting low/trace concentrations of metals.^106–109 Hu et al. discovered a correlation between the chemical state changes of active centers in SACs and their reaction activities through electrochemical operando XAS.¹¹⁰ Comparing the dry and OCP data in Fig. 6a and b, it can be seen that the influence of the liquid phase environment on the spectral resolution is relatively obvious. Nevertheless, thanks to the advantages of beamline stability at SPring-8, this provides a certain guarantee for the quality of the in situ XAS data; it is still possible to clearly trace the chemical state and dynamic structural evolution of single-atom catalysts under different reaction conditions. The Fe–N–C site, maintaining a +3 oxidation state during electrocatalysis, exhibits faster CO₂ adsorption and weaker CO adsorption compared to traditional Fe²⁺ sites. When an applied cathodic potential reached −0.5 V (vs. RHE) or higher, the Fe K-edge of Fe³⁺–N–C exhibited a shift towards lower energy, indicating reduction of Fe³⁺ (Fig. 6a and b). Simultaneously, as the Fe³⁺ site was reduced, the coordination number of the first-shell ligands around the Fe single-atom center (Fe–N/Fe–C) decreased from 4 to 3 (Fig. 6c and d), accompanied by a significantly decreased CO₂RR activity. Liu et al. studied the chemical activation of CO₂ molecules on Ni(I) active sites and probed the structural evolution of the Ni active sites in electrochemical CO₂ reduction to CO using operando XAS, XPS and valence band spectroscopy (Fig. 6g).⁸⁹ The increase in the Ni oxidation state observed in a CO₂ atmosphere may result from formation of unpaired electrons in the 3d_x²−y² orbital and the spontaneous charge transfer from Ni(I) to the carbon 2p orbital in the presence of CO₂, forming CO₂^δ− species (Fig. 6e). While the Ni K-edge of an atomically dispersed Ni catalyst (A-Ni-NG) shifted to lower energies during the CO₂ reduction reaction (CO₂RR), indicating the recovery of the low oxidation state Ni after one cycle of CO₂RR. The intensity of the main peak at about 1.45 Å in the FT-EXAFS spectrum increased slightly, which could be attributed to the contribution of the Ni–C bond that overlapped with the Ni–N bond. The Ni–N scattering path increased by about 0.04 Å due to the redistribution of electrons in the Ni 3d orbital between the 4 N ligands and the Ni–C bond (from the adsorbed CO₂), which caused the expansion of the Ni–N bond (Fig. 6f). Additionally, operando XAS was also applied to trace the dynamic reconstruction of SACs to form nanoparticles during the oxygen reduction reaction (ORR) and nitrogen reduction reaction (NRR).^62,111–113

Fig. 6 (a) and (b) The first derivative of the Fe K-edge XANES spectra of Fe³⁺–N–C and Fe²⁺–N–C. (c) R-space Fe K-edge EXAFS spectra of Fe³⁺–N–C. (d) The fitted first-shell coordination number of Fe (Fe–N and Fe–C) for Fe³⁺–NC (black) and Fe²⁺–N–C (red) under different conditions. Copyright 2019, AAAS.¹¹⁰ (e) Normalized Ni K-edge XANES spectra of A-Ni-NG recorded at various biases in 0.5 M KHCO₃ aqueous solution at room temperature in Ar or CO₂. (f) Fourier transform magnitudes of EXAFS spectra of A–Ni–NG under open-circuit voltage bias in Ar and CO₂ and at −0.7 V (versus RHE). (g) Structural evolution of the Ni active site in electrochemical CO₂ reduction.

In summary, it seems that hard-XAS is an indispensable technique for characterizing the structural information of SACs. In situ/operando hard-XAS offers a powerful technique to observe the dynamic evolution process of SACs, thus exploring the reaction mechanism and the structure–performance relationship of single-atom catalysis in greater depth. However, due to the limitations of the hard-XAS technique, key information should be further verified by other complementary characterization techniques and theoretical simulations (e.g., precise identification of C, N, and O, confirmation of metal center valence states, etc.). Additionally, the poor signal-to-noise ratio of the experimental data measured under in situ/operando conditions poses a major obstacle to the comparison of the fingerprint information. Therefore, optimization of synchrotron beam stability, improvement of in situ/operando experimental conditions and development of a new generation of synchrotron radiation techniques are key aspects for future exploration of single-atom catalysis.

3. Soft X-ray absorption spectroscopy (sXAS)

In recent years, researchers have discovered that light element atoms/surface functional groups in the coordination environment of SACs can influence the metal centers in SACs and sometimes may even directly participate in the chemical reactions. sXAS (soft (0.1–2.0 keV) and tender (2.0–7.5 keV)), as a surface probing technique, provides a crucial experimental tool for comprehending the electronic states of metal centers in SACs (L-edge sXAS of transition metals), as well as elucidating their coordination environment and the role played by functional groups on catalysts’ surfaces (K-edge sXAS of lighter elements such as C, N, O, etc.) (the left of Fig. 3d)^49,114,115 Compared to hard-XAS, sXAS is also a bulk technique, but it can detect specific absorption behaviors with high sensitivity, reflecting them in the signal intensity and shape of the spectrum, which is beneficial for understanding the possible interactions between elements in the catalyst and the location and shape of the active sites. Moreover, in situ/operando sXAS remains highly challenging for studying SACs. Firstly, due to the highly dispersed and low single-atom concentration characteristics of SACs, more precise and sensitive detection methods are required to obtain accurate and reliable data. Secondly, multiple competing or synergistic interactions may exist in SAC systems, leading to the formation of complex reaction networks. Therefore, a comprehensive combination of various characterization techniques is needed to fully reveal the underlying reaction mechanisms. Additionally, since sXAS has to be operated under vacuum conditions, considerations must be given to the design of in situ reactors and external factors such as reaction environments that may affect sample's stability and signal intensity.¹¹⁶

3.1 Distinguishing light element structures

Notably, sXAS provides the possibility to detect the structure change of SACs near a ligand environment with adsorption/desorption of reaction intermediates during the reaction process, which is beneficial for better understanding the structure–performance relationship and reaction mechanism of the catalyst. K-edge sXAS is one of the direct methods to probe light elements in SACs. Chen et al. conducted a systematic study of the electronic structure and coordination environment of single-Fe-atom catalysts (Fe–N_xC_y).⁸⁸ The C K-edge XAS spectra of Fe–N_xC_y show C–C π* (∼285–286 eV) and C–C σ* (∼293–295 eV) peaks from sp²-hybridized carbon, as well as C–N–C (∼287 eV) peaks, indicating the presence of defect sites in the carbon lattice (Fig. 7a).^117,118 As the coordination environment of Fe–N_xC_y changes, both C–C π* and C–C σ* undergo a corresponding energy shift, indicating a change in the C–C lattice. As shown in Fig. 7b, the a (398.4 eV), b (399.6 eV), and c (400.9 eV) peaks in the N K-edge XAS spectra represent the π* transitions of pyridine N, pyrrolic N and graphite N, respectively, and the d peak represents the excitation of σ*, forming a C–N–C bond or C–N bond. The intensity of the a and b peaks decreases significantly with the changing coordination environment from the N-ligand to the C-ligand. Simultaneously, Fig. 7c shows the Fe L-edge XAS spectra, from which, it can be found that with increasing C-ligands in the coordination environment, the L₃ and L₂ peaks are gradually negatively shifted and the L₃/L₂ ratio is significantly reduced, which indicate a decrease in the number of unpaired electrons in Fe and an increase in electron density around Fe. Additionally, some articles compared the sXAS spectra before and after the reaction. It should be noted that some single-atomic species exhibit reversible behaviors, and their structures and chemical states under reaction conditions may differ from those before and after the reaction. Interestingly, information about functional groups on catalysts’ surfaces and coordination structures of metal centers can also be obtained through comparing K-edge sXAS of light atoms before and after the reaction. This information is important for deducing the change in the coordination environment and the structure–performance relationship of SACs as most of these light-element atoms on the substrate are irreversible after the reaction. For example, Liu and co-workers found that O* and OOH* intermediates adsorbed on the C atoms near graphitic-N after the ORR caused the distortion of the heterocycle,¹¹⁹ while the half-maximum full width of the pyridinic-N peak increased and formed a new peak on the high-energy side after the OER (Fig. 7d), which indicated that the n-type doping of graphitic-N and p-type doping of pyridinic-N was responsible for the ORR and OER, respectively. The investigation of the change in surface chemistry during ethylbenzene dehydrogenation showed a decrease in pyrrole-N and an increase in graphite-N after the reaction, which clearly reflected the change in the stability of pyrrolic-N and the graphitization of the N–C framework during ethylbenzene dehydrogenation (Fig. 7e).^120,121 In the O K-edge XAS spectra, the A (C=O π*), B (Co–O), C (C–OH, the fingerprint of OH groups), D (C–O σ*), and E (C=O σ*) peaks are located at ∼532 eV, ∼532.8 eV, ∼535.7 eV, ∼540 eV, and ∼544 eV, respectively (Fig. 7f). The C–O and C–OH bond lengths of the catalyst after the reaction increased significantly, and the reacted catalyst showed a distinct peak at ∼532.8 eV, which might be due to the transitions of O 1s to the hybridized states of Co 3d.¹²² The results provided direct evidence for the generation of Co–O ligands in the coordination environment of Co–N_x SAC during ethylbenzene dehydrogenation. To sum up, the reversibility of species other than the metal active centers in SACs can be largely disregarded, and thus their changes before and after reaction provide us with a valuable perspective to gain a deeper understanding of the chemical state changes, microstructural evolution and functionality of light elements in complex catalytic systems. This is of great significance for studying the active sites and surface states of catalysts.

Fig. 7 (a)–(c) XANES spectra of Fe–N_xC_y SAs/N-C at the (a) C K-edge, (b) N K-edge, and (c) Fe L-edge. Copyright 2019, Springer Nature.⁸⁸ (d) N K-edge XANES spectra of the N-GRW catalyst. Copyright 2016, AAAS.¹¹⁹ (e) N K-edge XANES spectra of the fresh and used CoN_xO_y-900 catalyst, (f) O K-edge XANES spectra of the fresh and used CoN_xO_y-900 catalyst. Copyright 2019, American Chemical Society.¹²¹

3.2 Probing oxidation, spin and orbital states

L-edge sXAS relates to a 2p core-level electrons being excited to unoccupied valence orbitals of transition metals (2s²2p⁶ → 3d), thus directly probing the covalence in transition metal coordination complexes with better resolution.¹²³ It is worth noting that the intensity of the transition oscillator in the L₁ quadrupole transition edge (2s²3dⁿ → 2s¹3dⁿ⁺¹) is hundred times weaker than the strong electric dipole transition in the L_3,2 edge (2p⁶3dⁿ → 2p⁵3dⁿ⁺¹), so we usually collect transition metal L_3,2 edge spectra. The transition (2p⁶3dⁿ → 2p⁵3dⁿ⁺¹) leads to the 2p⁵ core configuration, which splits into two states at J = (1 + s) = 3/2 and J = (1 − s) = 1/2 due to spin–orbit coupling.¹²⁴ The transitions at low energy involving 2p_J = 3/2 manifolds are called L₃ edges, and the jumps involving 2p_J = 1/2 manifolds are called L₂ edges.¹²⁵ An important feature of L-edge sXAS is that the strong interaction between the nuclear holes and the 3d orbitals produces a multiplet structure. The shape and peak centroid shift of L-edge sXAS can reflect the valence state distribution of active sites in the catalyst (the top to the left of Fig. 3d). For instance, as nickel is oxidized from Ni¹⁺ to Ni²⁺, Ni³⁺, and Ni⁴⁺, the average Ni L₃-edge absorption centroid shifts to higher energy positions.^126–129 Therefore, this feature has the potential to determine the correlation between the position of absorption centroid of L-edge sXAS and the valence state and the relationship between the shape of the multiplet structure and its spin state and symmetry. These fine features of sXAS undoubtedly provide a clear clue for inferring a series of characteristics in SACs, especially in evaluating the homogeneity of active centers in SACs by comparing them with standard samples, such as molecular catalysts with similar coordination environments to SACs. But the biggest challenge lies in selecting an appropriate standard sample, which is a crucial issue that requires us to have a preliminary understanding of the structure of catalyst's active centers. The high-spin (HS) Ni²⁺ spectrum and the low-spin (LS) Ni²⁺ spectrum have distinctly different spectral multiplets. The ionic Ni³⁺ multiplet feature appears on the low-energy side of the L₃ peak, while the covalent Ni³⁺ peak is broader and shows no apparent multiplet structure. The Ni¹⁺ spectrum does not exhibit a multiplet structure because the d-shell layer is full in the final state.^130,131 Besides, the electronic structure of central metal atoms in SACs can be significantly influenced by the surrounding ligands. For example, metal-to-ligand charge transfer (MLCT) can be induced by the CN⁻ ligand. Hocking et al. identified distinct spectral features of the Fe L-edge in K₄[Fe(CN)₆] and K₃[Fe(CN)₆] arising from contributions of the ligand π* orbitals due to MLCT.¹³² The similar phenomenon of delocalization of Fe d-electrons into the porphyrin ligand has also been found by Hocking and co-workers in low-spin Fe heme.¹³³ Zhou et al. confirmed the in situ changes of ligands around Co and Fe from N/C to O in MOF catalysts during the OER process through charge transfer multiplet model calculations.¹³⁴ Thus, L-edge sXAS can provide information to distinguish the changes in the ligand environments through precise analysis of the electronic structure of metal ions. The spin states of 3d transition metals significantly influence their L-edge sXAS spectra. For example, the L₃/L₂ ratio of the Fe L-edge is greater than 2/1 for the HS configuration, but smaller than 2/1 for the medium-spin (MS) configuration.¹³⁵ Miedema et al. calculated the L₃/L₂ ratio of iron phthalocyanine (FePc) in O₂ to be 3.76, corresponding to 30% MS Fe²⁺ and 70% HS Fe³⁺.¹³⁶ The electronic structure of Fe was analyzed by Fe L-edge XANES, where the L₃/L₂ ratio of Fe in FePc was 3 and that in Fe–N–C (950 °C) was 3.61, revealing predominant HS Fe³⁺ in Fe–N–C (950 °C).^137–139 Additionally, the combination of sXAS and theoretical calculation offers the possibility to identify the adsorption structures of adsorbates on transition-metal SACs, which helps to gain insights into the adsorption forms of reaction intermediates and thus the reaction mechanism. Ding et al. immobilized cobalt phthalocyanine (M-CoPc) and binuclear CoPc (B-CoPc) on nitrogen-doped carbon (NC) via π–π interaction at room-temperature (M-CoPc-RT and B-CoPc-RT), followed by heat treatment at 400 °C in an Ar atmosphere to prepare Co-SAC and Co-DAC (M-CoPc-400 and B-CoPc-400) (Fig. 8a).¹⁴⁰ Both KIE and DFT results indicated energetically favorable hydrogenation of CO reduction intermediates over B-CoPc-400. In order to uncover the root cause of the rapid hydrogenation over B-CoPc-400 in the CORR to methanol, Co L-edge XAS measurements were performed to distinguish the electronic structures of M-CoPc-400 and B-CoPc-400. The Co L-edge XAS spectra of M-CoPc-RT and M-CoPc-400 obtained experimentally and via conformational interaction cluster model simulation¹⁴¹ showed the same characteristics as molecular CoPc (Fig. 8b). Three features appeared in the energy region of the Co L₃-edge: the A₁ peak could be attributed to the transition from 2p_3/2 to 3d_z² orbitals, and A₂ and A₃ peaks originated from the transition to 3d_x²−y² orbitals.¹⁴² The ground state of Co²⁺ in both M-CoPc-RT and M-CoPc-400 is ²A_1g (d_x²−y²⁰d_z²¹d_xy²d_xz,yz⁴) symmetric with a total spin of S = ¹/₂ (Fig. 8c). The Co L-edge XAS spectrum of B-CoPc-RT is similar to that of M-CoPc-RT and M-CoPc-400, indicating that the Co site in M-CoPc-RT has the same ²A_1g (d_x²−y²⁰d_z²¹d_xy²d_xz,yz⁴) symmetric ground state, which is the low-spin (LS) state (Fig. 8d). However, the L-edge XAS line shape of Co²⁺ in B-CoPc-400 changed considerably; the L_2,3-edge shifted to a lower binding energy and the contribution of the transition from the 2p to the 3d_z² orbital was significantly reduced, which induced a severe distortion of the coordination environment around Co and the ground state of the electronic structure changed from LS (b_1g (d_x²−y²⁰), a_1g (d_z²¹), e_g (d_xz², d_yz²), b_2g (d_xy²)) to HS (b_1g (d_x²−y²¹), e_g (d_xz¹, d_yz¹), a_1g (d_z²¹), b_2g (d_xy²)), corresponding to a symmetry change from a planar D_4h to a distorted D_4h symmetry of Co²⁺ (Fig. 8d). Based on the charge density difference calculation and molecular frontier orbital interaction analysis of *CO-LS-Co²⁺ and *CO-HS-Co²⁺, the weaker 3d_z²-5σ and more electron transfer from Co to *CO via π back-donation (via d_xz/d_yz-2π* bond) over HS-Co²⁺ led to smaller n_CT (net charge transfer) from CO to the Co site, enabling more electron accumulation on the 2π* orbital of *CO, due to the antibonding feature of the d_xz/d_yz-2π* bond, which would effectively promote the hydrogenation of CO reduction intermediates (Fig. 8f and g).

Fig. 8 (a) Schematic of the synthesis process of M-CoPc-RT/400 and B-CoPc-RT/400. (b) and (d) The experimental and simulated cobalt L_2,3-edge XAS spectra of M-CoPc-RT/400 and B-CoPc-RT/400, respectively. (c) and (e) 3d orbital diagrams of M-CoPc-RT/400 and B-CoPc-RT/400. (f) and (g) Interactions between CO molecular frontier orbitals (5σ and 2π*) and the 3d orbital of (f) LS-Co²⁺ and (g) HS-Co²⁺ sites. Black dots in (f) and (g) indicate the electrons from the electrode under cathodic bias. Copyright 2023, Springer Nature.¹⁴⁰

3.3 Tracing structural evolution of the catalytic surface

In situ/operando sXAS experiments are attractive to researchers in the field of SACs, while due to the high absorption cross-sections of C, N and O atoms in air, the decay length of soft X-rays is much smaller than that of hard X-rays.^143–145 Therefore, the development of in situ/operando reactors and the use of sXAS under near-ambient pressure conditions have a significant impact in the field of catalysis.^146,147 Notably, in situ/operando sXAS also has the potential to reveal the hydrogen-bonded structure of liquid water and the local structure of aqueous solution, where solute organic molecules are measured at the C and N K-edges and solvent water is observed at the O K-edge.^148–152 Different functional groups composed of the same elements can also be distinguished via the energy differences in sXAS peaks.^153,154 Besides, due to the excessively harsh conditions of in situ sXAS experiments, researchers simulate the properties exhibited by catalysts in normal working environments through near-ambient pressure experiments. The atomic dispersion of single P atoms on graphene (P-SAC-NG) was confirmed by HAADF-STEM,¹⁵⁵ in which most of the bright spots corresponding to single P atoms were distributed at the edges of graphene (Fig. 9a). The size of the bright spots is significantly smaller than that of a single metal atom on a carbon substrate.¹⁰⁶ The bonding pattern of CO₂ over the P-SAC-NG catalyst could be obtained from the changes in the P K-edge XANES spectrum before and after the CO₂RR. Differential spectral analysis showed that due to the 1s transition to the P–C (σ*) and P–O (π*) antibonding orbitals, the intensity of which increased at the low energy side and decreased at the high energy side, suggesting that CO₂ adsorbed at the P site with a P–CO₂^δ− configuration (Fig. 9b). Together with in situ spectroscopy measurements and DFT calculations, it was found that the electrochemical reduction of CO₂ at the single P atomic site was initiated by a structural transition from 2C–P=O(OH) to 2C–P=O via proton-coupled electron transfer, which could effectively reduce the activation energy of CO₂ at the P atomic site to form a 2C–P(CO₂)^δ−=O intermediate (Fig. 9c).¹⁵⁶ The biggest challenge in achieving in situ sXAS measurements is the design of in situ reactors, which needs to consider multiple factors: (1) selecting materials with good high temperature resistance, corrosion resistance, and sealing performance to construct the reactor, to withstand intense X-ray radiation and the impact of chemical reagents while effectively reducing background noise and (2) setting up a reasonable and easily controllable temperature and pressure (including other experimental parameters) adjustment device inside the reactor, equipped with corresponding monitoring instruments and an automated control system. In 2013, Gericke et al. designed a thermal catalytic Berkeley-type cell to study specific atomic oxygen species over thin film (coated on a window) and powdered silver epoxidation catalysts coupled with total electron yield (TEY) and fluorescence yield (Fig. 9d).¹⁴⁶ A gap of only a few tens of millimeters between the powder and the window was used to minimize the gas signal. Besides, Simonov et al. reported a custom-made flow cell that enabled high-quality in situ sXAS/RIXS electrochemical experiments (Fig. 9e).¹⁵⁷ The incident X-rays pass through a 100 nm Si₃N₄ membrane and a 20 nm Au layer, interact with the catalyst, and generate emitted/scattered photons that are measured with energy resolution. The catalyst comes into contact with an electrolyte. The potential applied to the catalyst/Au layer is controlled by a traditional three-electrode electrochemical setup. Chen et al. monitored the oxidation state and charge transfer between catalysts and reactants using a similar in situ electrochemical reactor as mentioned above.¹⁵⁸ By applying different potentials to Co L_3,2-edge sXAS during the ORR, the shift of the Co peak position could be observed within a specific bias voltage range (Fig. 9f), which indicated that Co was partially oxidized, and the shift could be attributed to the summation of the different Co²⁺-oxo intermediates, from which, the charge transfer mechanism between active Co sites and adsorbed intermediates was inferred. Further FEFF calculations confirmed that the bonding between O–O and O with Co caused changes in the oxidation state of Co, resulting in an upward shift of the Co L-edge (Fig. 9g).¹⁵⁹ The evolution of the Co²⁺-oxo intermediate state led to charge transfer from Co to O atoms, causing the maximum shift. This was evident in the increase of peak intensity at the Co L-edge under bias voltages ranging from 0.8 to 0.4 V vs. RHE. Subsequent mass transfer would restore the displacement of the Co peak, revealing that O₂ was adsorbed on the catalyst's surface as a negatively charged intermediate species, leading to a 4e⁻ ORR process.¹⁶⁰ Through analyzing the voltage- and time-dependent changes in the in situ O K-edge XAS spectra, Zhou et al. observed the dynamic valence state transition of cobalt from +3 to +4 under OER conditions with undisturbed crystal structure and a low spin state of cobalt.¹⁶¹ This work highlights the importance of high valent metal sites in the OER through strong M–O hybridization and reveals in situ sXAS as a powerful tool for identifying valence state change of the catalyst under reaction conditions.¹⁵⁹ Chou et al. conducted sXAS to explore the evolution of the electronic structure of a defect-rich catalyst in the OER and found that defects played an important role in improving the catalytic activity of the OER.¹⁶²

Fig. 9 (a) HAADF-STEM image of P-SAC-NG. Scale bar: 2 nm. Inset shows the structure of P-SAC-NG. Black, gray, red and yellow spheres represent C, H, O, and P, respectively. (b) P K-edge XANES spectra of P-SAC-NG before and after the CO₂RR (P 1s → σ* of P–C and P 1s → π* of P–O). (c) Schematic of the CO₂RR process on P-SAC-NG. Copyright 2023, Royal Society of Chemistry.¹⁵⁵ (d) Schematic diagram showing an in situ thermal catalytic reactor. Copyright 2013, American Institute of Physics.¹⁴⁶ (e) Schematic drawing of a flow electrochemical cell used for in situ sXAS experiments.¹⁶³ Copyright 2019, Wiley.¹⁵⁷ (f) Co L_3,2-edge XAS spectra of the Co–N_x/C catalyst at various operando biases. (g) FEFF calculated Co L-edge with different oxygenate species. Copyright 2020, Springer Nature.¹⁵⁸

Compared with thermal and electrocatalytic reaction systems, in situ photocatalytic experiments are often more readily achievable, and the external light source can greatly reduce the difficulty in designing in situ photocatalytic reaction cells (Fig. 10a). Zheng et al. performed in situ sXAS to explore the photogenerated current in CeO₂/Ni-G throughout the photocatalytic N₂ fixation process.¹⁶⁴Fig. 10b shows the Ce M-edge XAS spectra, in which the peaks at 882.6 and 900.5 eV correspond to 3d_5/2 to 4f (M₅) and 3d_3/2 to 4f (M₄) transition, respectively, with Y and Y′ being satellite peaks for Ce⁴⁺ species.¹⁶⁵ The M₅/M₄ peak intensity ratio is often used to determine the electron density of the unoccupied state of Ce.^166,167 The M₅/M₄ peak intensity ratio of CeO₂/Ni-G increased from 0.835 to 0.849 under light irradiation, indicating a slight increase in electron density of the unoccupied state of Ce. In the Ni L-edge XAS spectra (Fig. 10c), the two main peaks of 2p_3/2 → 3d (L₃) and 2p_1/2 → 3d (L₂) excitations are located at 852.5 and 870.3 eV, respectively. The electron density of the unoccupied state of Ni is negatively correlated with the sum of the L₂ and L₃ peak intensities.¹⁶⁸ The sum of L₂ and L₃ peak intensities for CeO₂/Ni-G significantly decreased from 5.61 to 3.13 under light irradiation, suggesting that photoexcited electrons preferred to accumulate at the Ni rather than Ce sites. The electron density of the unoccupied state of Ce in CeO₂/Ni-G under a N₂ atmosphere remained almost unchanged under light irradiation, indicating a weak interaction of Ce with the N₂ molecule. On the other hand, the sum of Ni L₂ and L₃ peak intensities increased to 4.01 in a N₂ atmosphere under light irradiation, indicating transfer of electrons from the unoccupied state of Ni to the N₂ molecule. In addition, the C K-edge XAS spectra displayed three characteristic peaks (Fig. 10d), π* (C=C) at 285.0 eV, π* (C=O) at 287.5 eV and δ* (C–C) at 291.2 eV, which showed no noticeable changes during photocatalytic N₂ fixation, suggesting that rGO acted as an electron transfer channel rather than directly participating in the catalytic reaction.¹⁶⁹ In the N K-edge spectra, the peak at ∼401 eV is assigned to the 1s→ π* transition and the broad peak centered at 407 eV is attributed to the σ* resonance (Fig. 10e).¹⁷⁰ It is noteworthy that the peak intensity is sensitive to the density of the vacant part of N and increase in electron density of the unoccupied state leads to a lower intensity.¹⁷¹ The N peak intensity is significantly lower compared to that in the N K-edge XAS spectrum of the initial N₂, suggesting that the filling of the unoccupied state for the p-orbital of the N atom is caused by the hybridization of the 1s orbital (H) and 2p orbital (N) during the N₂ → NH_x transition. Using this theoretical approach for spectral interpretation can better provide information about the ligand environment and crystal field effects in catalysts, which helps to provide insights into the structure–effect relationships and rationalize the distinct catalytic behaviors.^172,173

Fig. 10 (a) Schematic illustration showing an in situ photocatalytic sXAS setup. In situ XAS spectra of the (b) Ce M-edge, (c) Ni L-edge, (d) C K-edge and (e) N K-edge for CeO₂/Ni-G. Copyright 2022, American Chemical Society.¹⁶⁴

In summary, the high surface sensitivity of sXAS provides valuable insights into the part of the catalyst where chemical reaction takes place, which is crucial for heterogeneous catalysts such as SACs. Although in situ/operando sXAS measurements have gradually become implementable at large synchrotron radiation facilities around the world, some key issues, especially the design of in situ reactors and signal interference phenomena, still need to be further optimized. In addition, both sXAS and hard-XAS techniques provide valuable electronic and structural information of SACs, which are considered highly advanced by many researchers. However, the accuracy of the obtained information is still insufficient to support studies on the intrinsic properties of SACs. Therefore, continuous optimization and development of XAS as well as complementary characterization methodologies and theoretical calculations are particularly important.

4. Nonresonant X-ray emission spectroscopy (XES)

XES is used to investigate the occupied electron energy levels and provides valuable insights into the electronic structure, charge/spin density, and ligand properties.¹⁷⁴ In recent years, XES has gained increasing popularity in studying metalloproteins, ligand complexes, inorganic catalytic centers, etc.³⁵ Although SACs are non-homogeneous, their active centers share similarities with homogeneous molecular catalysts in terms of ligand environment structures, and many SACs have been constructed by fixing metal complex molecules on supports.^175–177 Therefore, the methods and techniques used to study homogeneous molecular catalysts can also be applied to SACs. However, it should be noted that there is a photon shortage issue in XES practical testing, which results in a poor data signal-to-noise ratio.³⁴ This becomes particularly challenging for accurately detecting weak signals, especially for SACs with low active site concentration. In order to collect sufficient photon signals, it often takes at least several hours to acquire one spectrum. Exploring the ligand environment structure of SACs directly contributes to understanding the origin of their catalytic activity. For this purpose, XES technology is undoubtedly one of the most suitable characterization tools. This review primarily focuses on discussing two aspects: electron orbital properties and coordination environment structure regarding the fine spectral characteristics observed through XES. By combining these findings with recent advancements related to SACs, we aim to provide a comprehensive summary of XES and highlight the future expectations for applying XES under in situ conditions.

In nonresonant-XES, the system under study is excited by X-rays with specific energy to remove the inner shell layer (core) electrons of the element of interest, which leaves the absorbing atom in an excited state with a core–hole, and then immediately filled by valence electrons from the outer layers.^178–180 In de-excitation, photons of electrons or fluorescent X-rays are emitted, and the fluorescence energy of the photons is measured in the XES. If the electron is removed from the K shell layer (1s orbital), the resulting fluorescence is called K fluorescence; if the electron is removed from the L shell layer (2s or 2p orbital), the resulting fluorescence is called L fluorescence, and so on. Depending on the intensities, the emission lines are called α, β, γ, etc., with α being the most intense (Fig. 11a). The most intense K fluorescence, Kα, is produced by a hole in the 1s orbital being filled by an electron in the 2p orbital.¹⁸¹ Each emission line shows a fine structure with the following main effects: (1) the interaction of the electron spin with own orbital momentum:¹⁸² the fluorescence yield of spin–orbit interaction is strong for the Kα line (∼10 eV) and quite weak for the Kβ line (<1 eV), but the line-shape variations due to changes in spin in the 3d shell are stronger in Kβ (the interaction between valence electrons with the 2p orbital is comparatively weaker than that with the 3p orbital).¹⁸³ Thus, traditionally, Kβ has been preferred for spin-state studies. (2) The electron–electron interactions: the electron–electron interactions can occur within the valence shell layer or between the core electrons (i.e., 2p, 3p) and the valence electrons, making the Kα, β lines sensitive to the electronic configuration of the valence shell layer.^184,185 Taking the XES spectrum of Cr presented in Fig. 11b as an example, the Kα line is dominated by the 2p spin–orbit splitting (Kα₁, Kα₂), while the spin–orbit splitting in the Kβ line is not sufficient to separate the 3p energy level, so the Kβ₁ and Kβ₃ lines merge and are visualized as a single peak.^186,187 Strong interactions between the 3p and 3d orbitals separate the Kβ_1,3 and Kβ′ spectral features.¹⁸⁸ Additionally, the Kβ lines show weak satellites on the high-energy side, which come from transition between the valence orbitals and 1s orbital of the metal nucleus, denoted as Kβ_2,5 and Kβ′′.^189,190

Fig. 11 (a) Energy levels and different fluorescence emission lines. (b) XES spectrum of Cr₂O₃ showing the Cr K fluorescence lines (blue): CTC-Kα, CTC-Kβ and VtC-Kβ. In the plot, the intensity is normalized to the Kα₁ maximum. The insets show an expanded view for CTC-Kβ and VtC-Kβ. The top schemes show the origin of transitions after photoionization in a simplified one-electron picture. Black lines indicate core states (1s, 2p, 3p), while the rectangles symbolize the valence and conduction bands (VB and CB). Copyright 2014, IOP Publishing Ltd.¹⁹⁰ (c) Summary diagram of SAC structure information obtained by XES.

4.1 Exploring spin states and electron orbitals

The Kβ_1,3 and Kβ′ lines in XES can determine the electronic structure of the 3d energy level, thus providing information about the oxidation state, the spin state and the covalency of the metal–ligand resulting from 3p–3d exchange coupling (Fig. 11c).^{186,191–193} Bergmann et al. observed differences in energies of the Kβ_1,3 peak positions of two inorganic Mn complexes Mn^IICl₂ and Mn₂^III,IVTerpy, which reflected different numbers of unpaired 3d electrons in the two high-spin compounds.¹⁹⁴ Tirao and co-workers determined the nominal oxidation state of Mn by fitting and quantifying the XES spectrum, demonstrating the capability of XES to characterize chemical states.¹⁹⁵ Compared to XANES, XES exhibits a stronger linear relationship between energy shift and the oxidation state with less dependence on structure.¹⁹⁶ Meanwhile, the Kβ′ intensity of various Mn oxides was found to be closely related to the nominal spin state. The sensitivity of XES to spin states stems from the exchange interactions between unpaired np and final state valence electrons, with significant differences observed as the nominal spin of the metal decreases (e.g., lower Kβ′ intensity and reduced splitting between Kβ′ and Kβ_1,3).^197,198 Debeer et al. observed the differences in the Kβ XES mainline for a series of nominally high spin d⁵ Fe(III) compounds (Fig. 12a).¹⁹³ While the splitting between Kβ mainline features is largely controlled by 3p–3d exchange integration, some researchers have speculated that it should also be subject to metal–ligand covalent modulation.^186,199,200 For this reason, the analytical expression of Kβ mainline splitting of the dⁿ system defines the splitting of Kβ_1,3–Kβ′ into p–d exchange integral, which is reduced because of covalency. This proves that the Kβ XES mainline is also a sensitive probe for covalency of metal complexes.²⁰¹ However, it has also been shown that covalence can greatly influence the shape and energy of the Kβ XES mainline, and it is possible to obtain spin state information from the Kβ mainline. In spite of this, in the in situ experiment, XES can still meet the needs to distinguish the oxidation state and spin state. Herranz et al. did data acquisition for the in situ XES experiments in minutes by combining energy dispersion settings and high incident X-ray flux.²⁰² By fitting the Kβ_1,3-line component to the spectral shape of Fe^II-phenanthroline and fitting the Kβ′ peak to Gaussian function (inset in Fig. 12b),^203,204 they showed that the relative area of the Kβ′ feature relative to the total Kβ area and spin state of the reference compound showed a clear linear correlation (Fig. 12b). The spin state changes experienced by the catalyst under conditions relevant to polymer electrolyte fuel cell applications were evaluated by in situ XES in N₂-saturated 0.5 M H₂SO₄. At 0.2 V vs. RHE, the relative area of Kβ′′ is smaller as compared to that measured at OCV (0.8 V vs. RHE), indicating a decrease in the average spin state of the catalyst. Lowering the potential from OCV/0.9 V vs. RHE to 0.2 V vs. RHE resulted in a decrease in the mean value of relative Kβ′′ area from 0.8 to 0.55, and this change was found to be reversible upon switching the potential to high values (Fig. 12c).^205,206

Fig. 12 (a) Kβ mainline spectra of a series of ferric compounds. Copyright 2014, American Chemical Society.¹⁹³ (b) Relation of the relative area of the Kβ′ peak with regard to the overall Kβ line and the spin state of the reference compounds (filled black dots), along with the corresponding linear fit. The inset shows the Kβ mainline XES spectrum recorded on DW21. (c) Comparison of the in situ Kβ mainline XES spectra recorded on DW21 in N₂-saturated 0.5 M H₂SO₄ at open circuit (OCV) or 0.2 V vs. RHE. Copyright 2021, Wiley.²⁰²

4.2 Identifying ligand atoms

XES overcomes some limitations of XAS; most notable is its ability to distinguish C and N from O, highlighting the significance of XES to determine the coordination structure of SACs. Valence-to-core (VtC) XES detects occupied valence orbitals by detecting electron transitions from ligand–ligand np → metal 1s (Kβ_2,5) and ligand → metal 1s (Kβ′′) (Fig. 13a).^207–209 As the electronegativity of the ligand increases, the energy of VtC Kβ′′ decreases, enabling precise identification of the ligand atom (Fig. 11c).^210–212 Organometallic clusters and metalloenzymes, similar to homogeneous catalysts with a single metal atomic center, have a metal-centered coordination structure similar to SACs, and therefore their analysis of the VtC XES region is also applicable to SACs.^213,214 For example, DeBeer et al. compared the energy of the 7100.2 eV satellite feature in the iron–molybdenum cofactor (FeMoco) with that of the O 2s → Fe 1s and N 2s → 1s transition observed in Fe₂O₃ and Fe₃N, indicating that this feature came from a ligand 5 eV and 8 eV lower than O and N, respectively. Therefore, the signal is not a contribution of N or O 2s and can be most likely assigned to a C 2s → Fe 1s transition.²¹⁵ In another work, comparing the spectra of Fe complexes with different coordinations, the ionization energy of C, N, and O 2s ligand was determined to be ∼7098 eV, ∼7094 eV and ∼7088 eV, respectively (Fig. 13b). Ding et al. confirmed the binding of Sn to oxygen atoms in single Sn atoms anchored on a carbon support (Sn₁–O₃G) via XES.²¹⁶ In Fig. 13c, a peak with a similar energy position is present in SnO₂ and Sn₁–O₃(OH)G (∼29173.9 and 29174.6 eV), which can be attributed to the corresponding Sn–O transition in SnO₂ and Sn₁–O₃(OH)G, respectively. The additional peak at the lower energy side (29166.4 eV) can be assigned to the protonation state of the Sn–O transition (Sn–OH) in Sn₁–O₃(OH)G. In order to probe the action of the Cu-chabazite catalyst during denitration reaction, Lomachenko et al. performed high energy resolution in situ VtC XES by selectively detecting fluorescence channels to resolve the ambiguity of the relevant ligand types and clearly identify two different mechanisms for the atomic-scale behavior of the Cu active site.²¹⁷ Monitoring the energy position of the weak Kβ′′ satellite revealed a major O connection in the high temperature range of 300 and 400 °C (Fig. 13d), whereas at 150 °C, the Kβ′′ satellite shifted significantly to a higher energy, between the energy positions of pure O and N ligands, which clearly indicated a significant increase in the proportion of N-ligated Cu species in the catalyst. Notably, the overall line shape of Kβ_2,5 is highly sensitive to the local coordination geometry and symmetry, with a narrow and strong peak at ∼8976 eV for Kβ_2,5 and a clear evolution of the line shape toward that of Cu(I)(NH₃)₂ at 150 °C (Fig. 13e). Similarly, Ding et al. found the increased peak width of asymmetrically coordinated Co-single-atom on carbon nitride (AC-Co₁/PCN_KOH) Kβ_2,5 XES as compared to symmetrically coordinated Co-single-atom on carbon nitride (SC-Co₁/PCN) (likely comprised of two emission bands), suggesting that there might be two different Co atom-bound ligand structures in AC-Co₁/PCN_KOH. The Co K-edge XANES peak of AC-Co₁/PCN_KOH at 7710 eV is significantly higher than that of SC-Co₁/PCN, indicating its obviously asymmetric structure.⁸⁶ Moreover, Sun et al. changed the Co³⁺/Co²⁺ ratio in the catalyst by grinding, and VtC XES showed a significant increase in the Kβ_2,5 peak intensity of ZIF-8-derived Co-SACs, thus demonstrating the increased Co³⁺ content. This work clearly reflected the correlation between the Kβ_2,5 region in XES and the ligand type in the catalyst.²¹⁸

Fig. 13 (a) Orbital interaction between the metal and ligand. Copyright 2013, American Chemical Society.²¹⁹ (b) Calculated spectra from optimized [Fe₆C(CO)₁₆]²⁻, [Fe₆N(CO)₁₆]⁻, and [Fe₆O(CO)₁₆] structures. Copyright 2011, American Chemical Society.²¹² (c) Normalized Kβ_2,5/Kβ′′ regions of various samples. Representative fits to the VtC spectra. In all panels: data (black), fit (red), background (grey), Kβ′′ (marked with *) and Kβ_2,5 peaks. Copyright 2023, Springer Nature.²¹⁶ (d) Operando XES spectra collected during SCR at temperatures of 150, 300, and 400 °C. (e) XES spectra of selected Cu reference states formed inside the pores of Cu-CHA, representative of Z-Cu(II), m-Cu(II) and m-Cu(I) species. Copyright 2016, American Chemical Society.²¹⁷ (f) and (g) Calculated XES spectra of (f) (^iPrPDI)FeN₂ and (g) (^iPrPDI)Fe(N₂)₂ compounds showing deconvolutions based on ligand identity and metal contributions. Copyright 2012, American Chemical Society.²²⁰ (h) VtC XES spectra of end-on Ni–(NO¹⁻) (green), end-on Ni–(NO²⁻) (blue), and side-on Ni–(NO²⁻) (red) and the calculated VtC XES spectra (A: NO σ; B: nacnac; C: NO σ*; and D: NO 2p). Copyright 2016, American Chemical Society.²⁰⁷

4.3 Probing coordination number and intermediate adsorption mode

The intensity of the Kβ_2,5 region in XES increases with increasing ligand coordination,^220,221 and some Kβ_2,5 features shift to lower energies and the energy separation between 2s σ and σ* decreases as the activation of bond increases (Fig. 11c).²²² Chirik et al. found that the peak intensity of (^iPrPDI)Fe(N₂)₂ was approximately twice that of (^iPrPDI)FeN₂ both experimentally and theoretically, suggesting that this feature is sensitive to the N₂ coordination number (Fig. 13f and g).²²⁰ Stieber et al. reported a new finding on the sensitivity of XES to the coordination patterns of small molecules.²⁰⁷ In Fig. 13h, it is found that end-on Ni–(NO¹⁻) has a distinct shoulder peak C while side-on Ni–(NO²⁻) has the most intensive peak D in Kβ_2,5, which are sensitive to both the NO oxidation state and coordination mode. In the Kβ′′ region, the side-on mode shows a feature peak A, which can be used to distinguish the end-on and side-on coordination mode for NO. Moreover, the NO coordination modes can be further distinguished by the Ni−N−O bond angle and N–O and Ni–N bond distance. As the Ni–N–O bond angle decreases, the overlap with metal orbital increases, resulting in enhanced feature peaks of the NO σ bonding orbital, while the intensity of feature peaks of NO σ* decreases with decreasing Ni–N–O bond angles. Both the β-diketiminate ligand (nacnac) and NO 2p shift to higher energies upon elongating the N–O bond, and the intensity of the NO 2p peak decreases significantly. Upon increasing the Ni–N bond length and decreasing the Ni 4s feature in the molecular orbital, the nacnac peak shifts to higher energy, while the intensities of other peaks decrease. These findings extend the application of XES to characterize small molecule bond activation as well as probe evolution of intermediate adsorption/desorption during chemical reactions.

In summary, the lack of key information on the precise structure of SACs poses a great challenge to understand the origin of the different catalytic activities of “similar” SACs. XES can accurately distinguish ligand species in the immediate vicinity, especially for ligands with close scattering ability such as C, N, and O. XES is also sensitive to the valence state, coordination number and molecular adsorption mode, offering an important technique to study the reaction mechanism and probe the structure–activity relationship. Meanwhile, it is worth noting that the average signal problem for synchrotron radiation spectroscopy is inevitable, and thus the structural information of SACs obtained by XES still has uncertainties. To better analyze the complex heterogeneous structure of SACs, it is impossible to solely rely on a single technique. Therefore, we need to combine characteristics of multiple characterization techniques and theoretical calculations to effectively analyze the structure of SACs.

5. Integrated multi-spectral technique (IMST)

In recent years, significant progress has been made in overcoming the technical limitations that previously hindered the application of in situ/operando experiments to observe SACs under reaction conditions. However, information obtained solely from hard-XAS, sXAS and XES is still insufficient to conduct an in-depth study of the intrinsic nature of SACs. Therefore, there is a pressing need to develop new generations of characterization techniques. To explore SACs beyond steady-state or in situ measurement using a single technique, one promising approach is to apply a combination of spectroscopic techniques. For instance, resonant inelastic X-ray scattering (RIXS) (including resonant X-ray emission spectroscopy (RXES)), which offers a higher dimensional data analysis than standard XAS and XES, can give more detailed information about electronic structure. Additionally, integrated multi-spectral technique that integrate multiple spectral techniques with energy dispersion XAS (ED-XAS) hold great promises to enable multi-angle and omni-directional temporal/spatial synchronous transient characterization of SACs. These advancements will lay a solid research foundation for future large-scale development of high-performance SACs through rational design and controllable tuning methods.

5.1 RIXS/RXES and its application in SACs

The biggest advantage of RIXS/RXES is the combination of XAS and sXAS/XES; it is able to provide more hierarchical and higher dimensional data than traditional XAS and XES and has garnered great attention for its effectiveness in investigating the electronic structure of 3d transition metal complexes, offering the possibility for a more comprehensive study of the electronic structure and coordination environment.³⁴ In addition to the discussed SACs, mention is also made of metal complex molecules with similar metal center structural patterns to SACs. Recently, numerous articles have reported the application of molecular complexes in RIXS/RXES research, and their related spectroscopic analysis methods are more mature and advanced; this undoubtedly provides valuable insights for studying SACs using RIXS/RXES. In addition, it has become a widely recognized strategy to use these homogeneous molecules as active sites for the synthesis of SACs. Furthermore, combined with the same charge-transfer multiple-state model, we can accurately reproduce the experimental RIXS planes and integrated spectra.^124,223 Leveraging the versatility of the RIXS test environment, in situ/operando methods can enable real-time exploration of structural evolution and surface intermediate adsorption/desorption over catalysts. Unfortunately, RIXS also has significant limitations.²²⁴ (1) Due to the complexity of RIXS spectra involving transitions between multiple energy levels, accurate interpretation and understanding of the data require a deep understanding of relevant theoretical models and the use of appropriate algorithms. (2) RIXS can only be applied to samples containing high atomic number elements or heavy elements, thus limiting its applicability for research on low atomic number or light element samples. Although in situ/operando RIXS/RXES still has many limitations in SAC studies at present, it has been proven to be highly effective in the research field of homogeneous catalysis, and we believe that in situ/operando RIXS/RXES will provide valuable insights into the field of single atom catalysis in the near future.²²⁵ In this section, we primarily discuss the spectroscopic characteristics of RIXS and RXES as well as their advantages as compared to single XAS/XES.

XES and XAS are directly related to each other; both are sensitive to the local electronic structure and bonding configuration of the absorbing atoms.²²⁶ There are many examples in which the information from XES is complementary to or even more detailed than that from XANES.^172,227,228Fig. 14 shows that in the XAS process, after absorbing X-rays with incident energy Ω = hν₁, the ground state electrons are excited to the intermediate state, leaving the holes in the core level. Filling the core–hole with an electron from one of the upper orbitals emits a photon of an electron or X-ray fluorescence, and the fluorescence energy of the photon is measured as ω = hν₂. The difference between the incident energy and the fluorescence energy is the remaining energy in the system, i.e., the final state energy or energy transfer (Ω − ω). These energy diagrams are usually converted to contour plots (RIXS planes).^34,229,230

Fig. 14 Summary diagram showing the SAC structure information that can be obtained by RIXS/RXES. Simplified energy scheme for absorption of a photon with incident energy Ω and emission of a photon with emission energy ω. The energy transfer is the difference between the energy of the absorbed and the emitted photon. In XAS, the incident energy Ω is scanned, while XES analyzes the emitted energy ω. The difference energy (Ω − ω) is the energy transfer or the final state energy.

In the RIXS plane, the incident energy is located in the X-axis and the energy transfer (Ω − ω) is located in the Y-axis (if the emission energy is located in the Y-axis, it is called RXES).^231,232 The higher dimensionality data of RIXS/RXES provide more information than standard XAS, and the experimental spectra can include the C, N, and O K-edges in the soft X-ray range (1s2p Kα-RIXS) as well as d–d interactions, ligand fields and charge transfer effects in transition metals (2p3d Lα-RIXS etc.).^172,233 In this plane, the X-axis represents the incident energy and the Y-axis represents either the energy transfer or the X-ray emission energy. If energy transfer is used in 1s2p RIXS, the Y-axis is related to the L-edge, while the X-axis represents the K-edge. The lifetime broadening of the L-edge and K-edge induces vertical and horizontal Lorentzian broadening, respectively.

The RIXS/RXES technique exhibits exceptional beam penetration capabilities in experimental settings, enabling investigations at length scales ranging from hundreds of nanometers to micrometers. This facilitates the detection of intrinsic properties of materials and provides valuable positional selection information, primarily relying on hard X-rays without the need for a vacuum environment. For example, the Kα-RIXS plane of CuO around the Cu K-edge can be obtained by recording and integrating the RIXS spectrum using DCM scanning to adjust the incident energy to the range of the XANES region. HERFD–XANES was obtained by integrating the distribution along a constant emission energy (CEE) line, corresponding to the maximum intensity of the full profile plane (8046.3 eV), with a width scale of 0.3 eV.^186,234 The fine features of CuO are clearly prominent in the HERFD–XANES spectrum as compared to the conventional XANES spectrum, which is attributed to the fact that the pre-edge peak (8978 eV) and the shoulder peak (8987 eV) of the dipole forbidden jump (1s → 3d) are greatly amplified in HERFD–XANES.^235,236 In addition, the fine features above 9010 eV are clearly visible due to the multiple scattering that occurs in HERFD–XANES. Thus, RIXS has a greater degree of freedom with respect to external environmental facilities and becomes an extension of traditional XAS, holding great promise for providing more precise structural information of SACs.

Metal complexes have a metal-centered structural pattern similar to SACs, and researchers have used them to prepare SACs. Solomon et al. applied RIXS to obtain metal–ligand covalency in 3d orbitals of transition metal complexes and collected RIXS data for LS iron complexes.²³⁷ As shown in the Kα RIXS spectra of ferrous tacn (Fe^II-tacn) (Fig. 15a), the final states of Kα₁ and Kα₂ emission correspond to the L₃ (J = 3/2) and L₂ (J = 1/2) edges of the L-edge XAS, and there is a split of ∼12 eV between them, which is due to the 2p spin–orbit coupling of the final states. Each resonance has a lifetime spreading in the direction of incident energy caused by the finite lifetime of the 1s hole (∼1.2 eV) and a lifetime spreading in the direction of energy transfer caused by the finite lifetime of the 2p hole (∼0.4 eV for L₃ and ∼0.8 eV for L₂).¹⁸⁶ The Kα RIXS of Fe^II-tacn in the final state is split because of 2p spin–orbit coupling into two high-intensity regions with K pre-edge absorption resonances centered at an incident energy of 7111.1 eV. The pre-edge peak has two separate maxima with energy transfer at the L₃-edge of 706.8 and 707.6 eV. A very weak resonance is also present at 709.7 eV at the same incident energy. Multiple cuts through the RIXS plane allow visualization of high-dimensional RIXS spectral details, which can be directly compared to the L-edge XAS. The main cutting methods are constant emission energy (CEE) cut (diagonally across the plane to obtain high-energy resolution HERFD XANES) and constant incident energy (CIE) cut (taking advantage of the high resolution in the energy transfer direction to produce “L-edge” energy transfer spectra). Integration of the CIE spectra along the incident energy spectra can include the contributions of all 2p⁵3dⁿ⁺¹ final states.^238–241 The complexity of RIXS could be illustrated by the presence of unresolved shoulders in the K-edge XAS spectra at 7112.0 eV, 7114.2 and 7115.7 eV, as seen in the XANES spectra obtained from the diagonal CEE of maximum values of the RIXS leading edge (7111.1 eV and 706.8 eV) (Fig. 15b).^242–244 The shoulder at 7112.0 eV is the contribution of the large lifetime spread of the second emission maximum (707.6 eV) in the direction of incident energy. For a single dominant K pre-edge jump, the CIE cut should access almost all of the 2p⁵3d⁷ states, hence a high-resolution L-edge-like spectrum is obtained from CIE at the resonance maximum (7111.1 eV). All 2p⁵3d⁷ final states resulting from a given 2p → 1s intermediate state are shown in Fig. 15c, and the energy difference between these states leads to a complex emission distribution and therefore to significant differences with respect to their counterparts in the L-edge XAS spectra. Compared to L-edge XAS, the width of the L₃ and L₂-edges of the Fe^II-tacn CIE is larger, with a double-peaked structure in the main features of the L₃-edge, significant intensity between the edges, and the appearance of a broad resonance near 714 eV. The increased width in the CIE cut can be explained by the difference in the selection rules between the single-photon L-edge XAS and two-photon RIXS processes. The latter can reach additional final states and the energy difference between these states leads to a spectral broadening in the direction of energy transfer. In the L-edge XAS of Fe^II-tacn, the 2p → 3d excitation involves an electric dipole selection rule and visits only the T_1u symmetric final state. In Kα RIXS of Fe^II-tacn, the initial (A_1g symmetry) 1s → 3d excitation involves an electric quadrupole selection rule (transformed to t_2g and e_g in O_h) to form the E_g intermediate state (Fig. 15d). The T_1u symmetric final state has a 2p hole and an electron in the 3d (e_g) orbital, which are coupled to form T_1u and T_2u symmetric final states. Therefore, the subsequent Kα emission is allowed for the electric dipole (T_1u), which accesses the T_1u and T_2u symmetric final states. Neglecting spin–orbit coupling, one component of the T_1u final state wavefunction can be written as the product of 2p_z electrons and 3d_z² electrons, while one component of the T_2u final state wavefunction can be written as the product of 2p_z electrons and 3d_x²−y² electrons. Due to the difference in the d orbital shape relative to 2p_z, the T_1u state experiences a greater electron repulsion than the T_2u state, leading to a higher energy and thus additional intensity in the CIE cut as compared to the L-edge spectrum. To further investigate Kα RIXS bonding in iron compounds, researchers analyzed the RIXS data using the same charge-transfer multiple-state model for the L-edge XAS data.^223,231,232 The charge transfer multiple-state model can accurately reproduce the experimental RIXS planes, CEE cuts, and CIE cuts, suggesting that simultaneous fitting of RIXS and L-edge XAS data can provide insights into the Fe center bonding. For example, the energy difference between the T_1u and T_2u final states can be used to determine the repulsive strength of the 3d–3d electron repulsion and thus the covalency of the atomic centers.²³⁷ Significant features were found in the examples of RIXS measurements performed on SACs. Fig. 15e shows the RIXS planes measured by Frenkel and co-workers for Pt single atoms loaded on CeO₂ (Pt SAC) and Pt nanoparticles loaded on carbon (Pt NPs/C), as well as control samples of Pt foil and PtO₂. The main feature that can be observed is the transition that occurs between occupied and unoccupied Pt 5d (d–d) states.²⁴⁵ Due to the strong background of the elastic peak, no d-band features below 2 eV can be observed in the energy transfer, which is thought to be an effect of the substrate (CeO₂ and carbon). Nevertheless, significant differences were observed in the RIXS signal shape of the Pt SAC as compared to Pt NPs/C. For Pt NPs/C and Pt foil, the d-band tail along the diagonal of the RIXS plane corresponds to the incident energy scan at a fixed emission energy. The emission energy correlates with the transition between the occupied Pt 5d and Pt 2p energy levels. In contrast, a similar feature was not observed in Pt SAC and PtO₂; instead, the Pt 5d–O 2p jump occurs at higher energy transfers due to the fact that Pt is bonded to the O atoms.²⁴⁶

Fig. 15 (a) Experimental RIXS planes of ferrous tacn [Fe^II(tacn)₂]Br₂. In the ferrous tacn spectrum, white lines show constant incident energy (CIE) and constant emission energy (CEE) cuts. The second row shows expanded pre-edge regions where the maximum of the color scale corresponds to the peak of the pre-edge. The third row shows VBCI modeling results for the corresponding pre-edge region. (b) Spectra in the K pre-edge region for ferrous tacn. Experimental K pre-edge spectra (blue) are compared to the spectra calculated from the VBCI multiplet model (light blue). Experimental CEE cuts through the maximum of the pre-edge resonance (red) are compared to the corresponding cuts through the modeled RIXS plane (gray). (c) Comparison of the RIXS CIE cut from the maximum of the pre-edge resonance (dark red) to the L-edge XAS spectrum (dark blue) of ferrous tacn. The results from the VBCI multiplet model are shown below the experimental results for both CIE cut (light red) and L-edge XAS (light blue). Separate contributions to the RIXS CIE spectra from final states with T_1u (Γ⁴⁻) and T_2u (Γ⁵⁻) symmetry are shown in purple and gray, respectively. (d) Relevant symmetry selection rules in O_h symmetry for Kα RIXS and L-edge XAS from the A_1g (Γ⁺₁) ground state in LS ferrous complexes. Copyright 2013, American Chemical Society.²³⁷ (e) RIXS planes measured across the Pt L₃ edge for Pt foil, PtO₂ standards, Pt single atoms on CeO₂ and Pt NPs on carbon. Copyright 2019, American Chemical Society.²⁴⁵ Reproduced with permission from ref. 247. Copyright 2014, Wiley.

In recent years, RIXS/RXES has received great attention for probing the electronic structure of 3d transition metal complexes, which offers the possibility to study the exact electronic structure and coordination environment of SACs. Benefitting from the high degree of freedom of the RIXS/RXES test environment, the in situ/operando RIXS/RXES method allows a simultaneous real-time study of catalyst structure evolution and adsorption/desorption of reaction intermediates.²⁴⁸ Unfortunately, no in situ/operando RIXS/RXES experiments have been performed to study single atom catalysis yet. But thankfully, there are some reported in situ/operando RIXS/RXES studies of homogeneous catalysis, which shall provide valuable insights for characterizing SACs under reaction conditions.^225,249 For example, Glatzel et al. tracked the changes in the electronic structure, d-band center, and magnetic and catalytic properties induced by CO adsorption on Pt via operando RIXS.²⁵⁰ Stampfl et al. investigated the electronic structure change of Au during warming reduction of Au₂O₃ by time-resolved RIXS to better understand the reaction mechanism of Au(III) reduction. Further combined with a genetic algorithm, it became possible to determine the electronic structure of sub-stable Au₂O intermediates.²⁵¹ Yamashita et al. performed a Pt–L_III RIXS study of heterometallic Ln–Pt complexes to investigate the unique effect of hidden Ln–Pt interactions on the luminescence of Tb–Pt molecules.²⁵²

5.2 Integration of ED-XAS and multi-spectroscopy for dynamics investigation

Today, the development of novel catalysts has gone beyond a basic understanding of the catalyst structure. Information on how the catalyst structure changes under reaction conditions is crucial to probe the reaction mechanism, which in-turn can guide the design and development of better catalysts. However, a catalytic reaction typically occurs on the surface of the catalyst and the limited lifetime of the intermediates associated with surface catalysis poses a great challenge to track all elementary steps in a chemical reaction.²⁵³ The dynamic evolution of reaction intermediates cannot be observed using the current vibration spectroscopy. Conventional EXAFS can in principle provide element-specific local electronic/structural information, but it does not have enough temporal resolution. Several emerging variants based on different ideas and technical developments allow this X-ray based spectroscopy to be applied with high temporal resolution. Quick scanning EXAFS and ED-XAS techniques are essential for study of fast irreversible processes, offering the possibility of a high spectral repetition rate from 1 Hz up to 40 kHz. However, both methods have inherent limitations to study catalysts under reaction conditions, and the ability to very quickly get the desired X-ray bandwidth is only the first step to obtain statistically valid and reliable EXAFS information. Nevertheless, they have been progressively refined and successfully applied to study catalytic reactions.^254,255

Quick EXAFS is based on optimization of the mechanical motion of the X-ray monochromator used to sequentially scan the range of X-ray energies required to obtain EXAFS spectra.^256–258 The culmination of techniques currently available for this approach is based on the external center cam-driven channel cutting system developed by Frahm and coworkers. This arrangement allows repeated scans over the required bandwidth of EXAFS at frequencies of up to about 80 Hz.²⁵⁹ Chen et al. developed operando second-resolved X-ray absorption spectroscopy to reveal the chemical state evolution of the working catalyst. Combining a potential switching approach with second time resolution, the Cu model catalyst was found to achieve a stable chemical state of half Cu(0)-half Cu(I) during the electrochemical CO₂RR, which provides an understanding of the underlying chemical state of the Cu catalyst towards achieving selective CO₂RR.^260,261

ED-EXAFS is based on a completely different concept, which obtains temporal resolution by eliminating the need for any mechanical motion in the X-ray optics.^40,262 As shown in Fig. 16a, an elliptical curved crystal is used to focus and disperse the quasi-parallel and multi-colored X-ray beams. The energy-dispersed beams from the crystal converge on a focal point at the sample location; the beams passing through the sample then diverges toward the position sensitive detector (PSD). Therefore, a “bent” polychromator is used to apply the desired X-ray bandwidth instantaneously at the focus point of the sample position. With no mechanical motion involved, dispersive EXAFS experiments can be very fast, very stable, and inherently highly spatially resolved.^263–265 ED-EXAFS has several distinct advantages over traditional EXAFS. (1) Since the entire spectrum is measured simultaneously by PSD, it can be easily applied to time-resolved studies. (2) Only a small amount of sample is needed in the focal point for testing. (3) Mechanical scanning with monochromators is not required. As a result, high stability requirements for XAS measurement, such as XMCD and high pressure (DAC) XAS, can be achieved. (4) The beam does not move vertically when scanning the required energy range, thus eliminating the effects of uneven sample thickness.²⁶⁶ However, while ED-XAS has many advantages, there are also some limitations.^267,268 (1) In practical operation, highly pure and stable samples are required, and the instruments and environmental conditions have high demands. (2) Due to the large difference in characteristic energy levels between different elements, special design and adjustment of instrument parameters are needed for each element. (3) During measurement, the effects of sample surface morphology, crystal structure, and temperature on the results must be taken into account. (4) The strong polychromatic beam of ED-XAS has a high energy density, which may result in nonlinear absorption on the surface or inside the material and generate side effects such as local temperature rise. This is inevitable, therefore, when designing an experimental protocol, it is necessary to reasonably select appropriate power and wavelength ranges and take corresponding measures to protect samples and reduce their risk of damage. (5) In data processing, in order to extract useful information and eliminate interference signals, it is necessary to use complex calculation methods. For example, when analyzing elemental composition, statistical methods such as linear regression or principal component analysis (PCA) can be employed. On the other hand, the study of crystal structures necessitates frequency domain analysis methods like Fourier transform or wavelet transforms are needed. Moreover, by applying the corresponding algorithms (including but not limited to the least squares method, support vector machines, neural networks, and so on), it is feasible to convert raw data into understandable and meaningful results through practical operations and calculations.

Fig. 16 (a) Principle of an ED-XAS spectrometer. The incident angle θ of the X-ray beam varies continuously across the crystal surface (θ_α to θ_γ) and this results in different X-ray energies being reflected from different points on the crystal surface. (b) A photograph showing an energy dispersive X-ray absorption beamline station based on the tapered undulator source on the beamline BL05U of Shanghai Synchrotron Radiation Facility (SSRF). (c) The Cu foil K-edge XAS performed in the D-line (BL05U) at SSRF. (d) A schematic diagram displaying IMST.

Fig. 16b shows a digital photograph of ED-XAS taken on the beamline BL05U from SSRF, whose spot size can be focused to 5 × 20 μm² (FWHM), and the time resolution of ED-XAS can reach below 25 μs. Intrinsically, high temporal resolution about 100 ps can be achieved using pump–probe techniques in the single bunch mode, and continuous fast transients can be taken for the purpose of relaxation experiments.²⁶⁹Fig. 16c displays a K-edge XAS spectrum of Cu foil recorded on the dynamics line (D-Line) of SSRF. Supported by ultra-fast time-resolved ED-EXAFS, high-frequency spectral collection during dynamic processes can be achieved, and the high-degree of freedom of the experimental environment permits many possible reaction conditions, which shall allow in-depth exploration of reaction kinetic relationships, such as time-resolved acquisition of the geometric and electronic structure information of the catalyst under reaction conditions,²⁷⁰ transient kinetic analysis of reaction processes through introducing isotopes, etc. Meanwhile, it is worth noting that ultra-fast data acquisition of ED-XAS can only be carried out in the transmission mode, and thus for sample with low-metal-concentration, it is necessary to carefully control the thickness and uniformity of the sample.²⁷¹ More importantly, the construction and innovation of new generation synchrotron radiation sources (such as the Swiss Light Source) and the related technologies have improved the stability and accuracy of instruments, effectively reducing the impact of noise sources such as mechanical vibrations and electromagnetic interferences on experimental results during instrument operation.^272,273 For example, (1) introducing an adaptive feedback control system in beamline design allows real-time monitoring and correction of beam current deviations, thereby reducing associated errors. (2) Using high-sensitivity detectors, precise data processing algorithms, and other means can improve signal acquisition efficiency while simultaneously reducing background noise without compromising data quality. Undoubtedly, these developments provide a foundation for the application of ED-XAS to low-metal-concentration catalysts to achieve time-resolved data collection.²⁷⁴ Additionally, the beamline can even support in situ experiments under extreme conditions, which provides a broad experimental platform for observing structure changes of catalysts in complex systems from different perspectives. Notably, inspired by the advanced nature of ED-XAS, to gain insights into the local structure dynamics of catalysts during reaction, other characterization techniques such as synchrotron radiation infrared spectroscopy (IR) can be coupled with ED-XAS.^275,276 As shown in Fig. 16d, line 1 is the ED-XAS optical path, and the line 2 IR optical path (as well as the line 3 other optical path) can be transferred and focused to the same spot as the ED-XAS optical path, so as to achieve simultaneous detection of relevant reaction intermediates, atomic structures and electronic structures during dynamic processes at high time/space-resolved scales and to observe changes in structures of substances in complex systems from different dimensions.^277,278

When characterizing catalysts in in situ experiments, the combination of complementary techniques offers a robust tool to achieve a comprehensive and in-depth understanding of the system.²⁷⁹ In particular, the combination of XAS with diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) and mass spectrometry (MS) is powerful because XAS describes the major elements in the catalytic system (selective absorption edges), DRIFTS monitors surface adsorption, and MS identifies and quantifies the products. ED-XAS shows great advantages in this combination. Agostini et al. reported a new reactor design and an optimized experimental setup to perform time-resolved experiments on heterogeneous catalysts under working conditions (Fig. 17a).^280–282 One of the main features of the setup is its ability to work at high temperatures and pressures with small volume in the reaction cell. The synchronization of an X-ray detector, an IR spectrometer, a mass spectrometer, a switching valve and other devices (e.g., UV-vis shutter) ensures the combination and correlation of information collected from XAS, DRIFTS, MS, etc. Catalytic tests in the presence and absence of X-rays confirm the reliability and accuracy of the kinetic data obtained in this new cell setup. The time-resolved capability was demonstrated by tracking the evolution of the Rh–Rh scattering path in EXAFS, the IR band associated with Rh⁰–CO (2025 cm⁻¹) and the signal at m/z = 44 corresponding to the generated CO₂ (Fig. 17b–d). Newton et al. demonstrated simultaneous combination of time-resolved (energy dispersive) EXAFS, DRIFTS and MS for an in situ and time-resolved study of Rh/Al₂O₃ and Pd/Al₂O₃ catalysts during CO/NO redox at 573 K.²⁸³ This approach applies transmission-mode-based and highly time-resolved XAS to low-concentration catalysts (0.3 wt% Rh/Al₂O₃), and it is shown that this approach can provide a new experimental window through which it is possible to recover highly time-resolved structure-kinetic information from previously intractable systems. On this basis, the spontaneous oscillatory behavior of the 5 wt% Rh/Al₂O₃ catalyst oxidized by O₂ under stoichiometric and net oxidation conditions was investigated by parallel application of time-resolved XAFS, DRIFTS and mass spectrometry.²⁸⁴ Skoglundh et al. analyzed electronic states, surface coverage of reaction intermediates/products and catalytic activity/selectivity of a 4% Pt/Al₂O₃ catalyst during methane oxidation under transient inlet conditions via combined techniques.²⁸⁵ The formation of inhomogeneous surface oxides on Pt microcrystals during cyclic gas feed between net oxidation and net reduction conditions increased methane oxidation activity and facilitated dissociative methane adsorption (Fig. 17e and f). By matching Pd K-edge XAS measurements over Pd nanoparticles to CO coverage from DRIFTS, Wells et al. found that most Pd catalysts underwent a dramatic transition from CO-poisoning catalysts to highly active oxidized Pd at multiple locations in a fixed bed reactor.²⁸⁶ Wu et al. used a combination of XAS, DRIFTS and MS to study the catalytic mechanism during integrated carbon capture and utilization and found that both adsorption and catalytic sites underwent cyclic regeneration for subsequent integrated carbon capture and utilization.²⁸⁷ Additionally, although the above examples support the use of ED-XAS in the SAC study, the sample requirements for SACs in the transmission mode are very strict. Even if a sample meets the requirements for ED-XAS characterization, there may still be situations that it is not possible to fully utilize the active sites, especially in electrocatalytic reaction due to limitation of mass transfer at the gas–solid–liquid interface. Therefore, most of the current reported studies focus on thermal catalytic systems and it is still difficult to apply ED-XAS to electrocatalytic systems.

Fig. 17 (a) A schematic illustration showing the design of the catalytic cell. (b) k²χ(k) XAS spectra collected on a Rh catalyst in CO (red curve) and NO (blue curve) environments. (c) ROI analysis performed using PyMca on EXAFS, DRIFTS and MS data. (d) Evolution of DRIFTS spectra. Copyright 2019, International Union of Crystallography.²⁸⁰ (e) Evolution of XANES Pt L₃-edge spectra and (f) IR bands in the interval of 1500–4000 cm⁻¹ over a 4 wt% Pt/Al₂O₃ catalyst exposed to 1000 ppm CH₄ in He while periodically switching O₂ concentration between 0 (60 s) and 1.5 vol% (60 s) at 280 °C. Copyright 2019, Elsevier.²⁸⁶

SACs are some of the ideal choices for future commercial catalysts. Identification of reactive sites in SACs and tracking their dynamic evolution under reaction conditions shall provide deep insights for the future design and development of high-performance SACs. Time-resolved in situ observations can be performed by spectroscopic techniques such as ED-XAS and DRIFTS to explore the adsorption of reaction intermediates on the atomic surface and the changes in the electronic structure of reactive sites. The coupled application of these characterization techniques will greatly improve the sensitivity of the experiment as an integrated body-sensitive probe. Nevertheless, the simultaneous use of multiple spectral techniques inevitably faces a series of challenges such as sample structure requirements (the structure needs to meet the test conditions of different spectra at the same time), development of in situ reactors and the design of beam paths, catalyst damage caused by intense light beams, and limitations imposed by reaction systems, and so on. With the development of next generation synchrotron radiation techniques, multiple spectroscopic characterization methods with their own advantages (including sXAS/XES/RIXS) will eventually form a collection to achieve multi-angle and full-range transient exploration of SACs synchronized in time and space, which will provide new insights into the structures of SACs and their applications in various fields.

6. Machine learning and spectral analysis by X-ray based synchrotron radiation spectroscopy

To more accurately identify catalytically active sites and track their evolution under reaction conditions, a promising approach is to establish correlations between experimental measurements and theoretical simulations, such as ab initio simulations and DFT calculations.^288,289 However, for nanoscale catalysts (especially SACs) with unique structures and/or significant disorders, neither the forward model nor the fitting method can reliably produce sufficient results in the active state. Machine learning methods trained on large DFT simulation groups enable the utilization of machine learning models to predict properties of novel structures and quantify spectral data without the need to explicitly execute time-consuming simulations. With robust computing power, machine learning methods can detect subtle changes in experimental data (previously “hidden” spectral features), rapidly analyze experimental spectra through supervised machine learning algorithms, and track catalyst structure changes in real-time during reactions.^290–292 These advancements are expected to revolutionize traditional computational analysis and enable more precise acquisition of structural information. Next, we comprehensively analyze intricate characteristics present in both experimental and theoretical spectral data, subsequently further discussing the cutting-edge capabilities of machine learning in in situ/operando experiments as well as its potential extensions to other spectroscopic techniques, particularly considering application feasibility and development prospect of IMST.

Synchrotron radiation X-ray-based spectral techniques are essentially methods of ensemble averaging, and the heterogeneity of atomic distribution in SACs often affects their quantitative analysis of metal centers and the corresponding coordination environments. Due to the absence of metal–metal bonding and the heterogeneity of multiple binding geometries, some of the spectral fine features of SACs are often inconspicuous and imperceptible. Traditional forward modeling calculation and the fitting analysis method heavily rely on high structural order, which pose great challenges for SACs with unique structures and/or disorders, as well as lack scalability.²⁹³ The goal of machine learning (ML) models is to generate accurate predictions by analyzing underlying patterns and the corresponding relationships in data.^294–296 Each new sample requires the development of a new classification equation (such as the artificial neural network method, etc.). By assuming that there exists a unique relationship between the structure, electronic properties, and composition of a given catalyst and its XAS spectrum, observing numerous examples and training and analyzing large amounts of data allow ML to extract features and discover hidden patterns within the data.^295,297 Currently, the most pressing need is the heterogeneity source separation of SAC spectra. Under the conditions of no labels, reliable structural characteristic regions can be identified and divided from a large amount of experimental data by adding additional constraints. This allows for obtaining mixed matrices of relevant structural information and discrete quantities of pure source signals. Frenkel et al. employed PCA, Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS), and K-means clustering^298,299 methods to assist analysis of in situ XANES data, obtaining quantitative structural information on the coordination environment of Co SACs. The combination of PCA^300,301 and MCR-ALS^{291,302–304} helps to reduce the dimensionality of the data matrix and explain a series of spectra using a small (2–4) concentration distribution of pure species. Furthermore, neural networks were used to aid in XANES analysis for identifying local structural descriptors of known species, such as a CO group attached to the Co SAC (Fig. 19a and b).³⁰⁵ Subsequently, various methods (such as logistic regression,³⁰⁶ support vector machines,^307–309 random forests,^310,311 artificial neural networks,^312,313etc.) were applied to perform supervised classification of the XAS spectra (Fig. 18a). By utilizing multi-label classification algorithms, it becomes possible to identify multiple sources that contribute to the average spectrum and train classifiers to map feature vectors to output label vectors indicating the presence or absence of certain categories. After completing the classification step, the inversion method will be used to map the spectra of each category onto specific atomic structures. This approach combines unsupervised and supervised machine learning methods to directly classify structural models for identification, enabling the deciphering of the spectral features of SACs and even unraveling the contributions of different metal sites coexisting in working SACs.^292,300,314 Cuenya et al. employed unsupervised machine learning methods, such as PCA combined with transformation matrix techniques, to determine the number of different coexisting nickel species along with their corresponding kinetic profiles and XANES spectra (Fig. 19c). Subsequently, a supervised machine learning-based XANES fitting program was utilized to derive the atomic structure of each identified Ni species.^315,316 Finally, reverse Monte Carlo (RMC)^317–319 simulations were conducted to validate the predicted structures obtained from processing the respective EXAFS spectra, taking into account structural disorder in the local environment surrounding the identified Ni species (Fig. 19a and b).

Fig. 18 Schematics showing the two most common approaches in dealing with heterogeneous X-ray based synchrotron radiation spectral data. (a) Approach 1: classification is used to identify the structural model, which can be refined further using neural networks. (b) Approach 2: embedding-based structural similarity in the low-dimensional structure database to find candidate species and their distributions. Adapted with permission from ref. 320 and 321.

Fig. 19 (a) A schematic to represent the application of neural networks–XANES to SACs. (b) Comparison between the experimental XANES spectrum of Co–cyclam–CO (black) and the theoretical XANES spectra. The spectra with colour from green to yellow are simulated XANES spectra with two descriptors changing (d_C and d_O). Copyright 2021, Royal Society of Chemistry.³⁰⁵ (c) A set of possible deformations applied to the structural models used for the construction of the SML training data set and XANES fitting. (d) Points in a structural parameter space obtained using the adaptive sampling employed to establish the [small mu, Greek, circumflex](E; p) (a mathematical model descriptor, which is a function of energy E and a set of structural parameters p) interpolating functions for model 1 depicted. (e) Calculated spectra are obtained from the structural parameters corresponding to the points. Representative calculated spectra of models 2 and 3. Copyright 2023, American Chemical Society.³¹⁴

On the other hand, spectral embedding is a low-dimensional latent space representation (also known as dense representation) of spectra learned through self-supervised learning (Fig. 18b).³²⁰ Embedding is a useful feature of spectra that can be used for various tasks such as classification, regression, data visualization, etc. This vector space representation of embedding also allows for different measures of similarity.³²² It clusters input data into different groups, pulling similar data points close together and pushing dissimilar data points further apart. Additionally, decoders or generative models can be used to learn the probability distribution of latent candidate structures, serving as a direct pathway to ‘invert’ spectra back to structures.³²³ Seidler et al. improved the sensitivity to finer details of stable chemical substances by gradually reducing the constraints of unsupervised machine learning algorithms, transitioning from PCA to variational autoencoders (VAE)^324–326 and then to t-distributed stochastic neighbor embedding (t-SNE).^327,328 When embedding spectral collections into two dimensions, t-SNE not only distinguishes between oxidation states and general sulfur bonding environments, but also identifies the aromaticity of bond-free radical groups with 87% accuracy, revealing more intricate details about their internal electronic structures. This enables visualization of clustering in reduced-dimensional space and facilitates inference of general features encoded in chemical descriptors for XANES and VtC-XES.

For ED-XAS and IMST, the characteristic of simultaneous collection of various spectroscopic data in a short period of time leads to a large amount of experimental data. Subsequent analysis of the data requires a significant amount of time and effort. Machine learning algorithms can not only be implemented in real-time data analysis pipelines in synchrotron beamlines to provide immediate feedback for experimental setups but also optimize experimental parameters such as X-ray energy, sample positioning, and data acquisition time through continuous monitoring and analysis of evolving spectra.³²⁹ This algorithmic approach maximizes information extraction while minimizing beam damage by ensuring effective data collection, thereby improving overall measurement quality. Most importantly, these algorithms can automatically extract relevant spectroscopic features and reduce dimensionality to simplify data interpretation and analysis, enabling real-time precise quantification of catalyst structure evolution along with the corresponding changes in reaction activity patterns.^330–332

Overall, the application of machine learning to X-ray based synchrotron radiation characterization techniques has revolutionized the analysis and interpretation of complex spectroscopic data.^333,334 These methods have accelerated research progress, leading to discoveries in various fields, such as materials science, catalysis, energy storage, and environmental science. The combination of synchrotron radiation with machine learning has the potential to unlock further insights into the fundamental properties of matter and drive future technological advancements.^335,336

7. Conclusions and outlook

Synchrotron radiation techniques are now widely used for characterizing SACs. Currently, the lack of precise structural information of SACs poses a great challenge for understanding the origin of the very different activities among similar SACs. In this review, the research progress of synchrotron radiation techniques for the characterization of SACs and the application of in situ/operando synchrotron radiation techniques in the field of SACs were systematically summarized. Techniques such as XAS, XES, RIXS, and RXES play a crucial role in achieving accurate measurements of valence electrons and tracking reaction on catalyst surfaces, which is important for understanding the source of activity at the electronic level under in situ/operando conditions.^40,281,337 Therefore, the correlation between the fine spectral characteristics of XAS (including hard XAS and sXAS) and XES and the geometric and electronic structure information of SACs is also emphatically discussed. On the other hand, the applications of these characterization techniques in SAC studies still face many challenges; especially it is almost not possible to prepare SACs that exhibit only a single type of coordination environment on a substrate; most of them possess a range of different coordination environments. This complexity makes accurate interpretation of data challenging. Among the various synchrotron radiation techniques, each provides complementary information to the other (by using EXAFS, the distance between the central atom and its coordinating atoms can be obtained). However, there is a great controversy in distinguishing the types of coordinating atoms. On the other hand, analyzing the Kβ′′ peak of XES can effectively differentiate these coordinating atoms, thus helping to obtain an overall picture involved in the reaction process. Additionally, based on ED-XAS, various spectroscopic techniques can be coupled together to achieve multi-angle and all-round transient characterization of SACs in time and space. Although ED-XAS is considered difficult to be applied to the SAC study due to the need for data collection in the transmission mode and the low metal content of SACs, it can still be applied to some reaction systems. We believe that the main challenge lies in the design of customized in situ reactors and beam paths, and this technique is more susceptible to external factors (such as mechanical vibrations and electromagnetic interferences). Therefore, the construction of new generation synchrotron light sources (such as the Swiss Light Source) and the development of related technologies are crucial.

Notably, with the development of reliable spectral ab initio simulation, machine learning and DFT modeling methods, quantitative spectral analysis of SACs has become a key factor to understand the accurate structural information of SACs and the chemical reactions taking place over SACs.^2,338,339 For these synchrotron radiation based spectroscopic techniques, it is worth noting that traditional data fitting is only applicable to certain uniform and precisely synthesized catalysts, and there is controversy when it comes to more complex systems like SACs. In order to better authenticate the effective active sites in SACs and identify their structural characteristics, the qualitative rationalization of experimentally observed trends using DFT models is largely required. Thus, spectral calculation and DFT modelling have become increasingly sophisticated, enabling accurate analysis and authentication of the fine structure in the spectra of SACs. On this basis, we discussed and summarized the fine characteristic signatures of the spectra in experiments and theoretical calculations. In situ/operando experiments offer the possibility to examine the dynamic changes in catalysts under reaction conditions. However, under realistic in situ/operando experimental conditions, it is often impossible to obtain high-quality spectral data over SACs. To this end, the sensitivity of machine learning methods to subtle variations in experimental data can be exploited to discover “hidden” information in relation to the structure and/or activity of SACs.^{292,326,340,341} Thanks to the rapid (in a second or less) analysis of spectra by pre-trained neural networks, “dynamic” data analysis becomes possible, allowing real-time analysis of spectral data during data collection and consequently automatic control of reaction systems towards achieving improved catalytic activity, selectivity and stability. Additionally, machine learning can be extended to interpret other spectra, which is certainly very applicable to IMST.³⁴²

In summary, all forms of characterization techniques, whether used alone or in combination, are very powerful probes for a variety of chemical systems. However, it is worth noting that although X-ray based synchrotron radiation spectroscopy can provide information about the structure of SACs, the complexity and microscopic features of SACs require careful consideration of potential sources of errors when conducting quantitative analysis. Moreover, the heterogeneity of bonding geometries in SACs can further complicate and obscure certain spectral fine features, which may contribute to the differences in activity among similar SACs. Therefore, it is crucial to develop high-resolution spectral experimental platforms and reliable spectroscopic modeling techniques, as well as deep machine learning methods. Based on the current trend of in-depth research trying to understand the structure–activity relationship in catalysis, this process will undoubtedly provide an important opportunity to greatly advance the development and expansion of high-precision, micro-scale spectroscopic characterization techniques and spectral analysis methods. Through combination of technologies with temporal and spatial resolution, the experimental environment can become more realistic. Furthermore, integrating real-time data supervision through artificial intelligence will enable us to describe them more accurately, further accelerating resolution of scientific problems in catalysis-related fields and ushering in a new era of catalysis. In the future, it is reasonable to expect that these types of time/space scales themselves will become routine procedures, providing new approaches for solving important physical and chemical problems across various fields and disciplines. Through this system, we will be able to achieve multi-angle in situ observation from growth to reaction processes for establishing SACs, understanding localized structure and electronic information throughout their entire “life” cycles from synthesis to deactivation, and improving the theoretical foundation for macroscopical preparation and large-scale application of SACs.

Data availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of interest

There are no conflicts of interest to declare.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (no. 22075195, 22102207, 22475145 and U1932203), the Shanghai Municipal Science and Technology Major Project, the Photon Science Center for Carbon Neutrality, the City University of Hong Kong startup fund (9020003), the ITF-RTH-Global STEM Professorship (9446006), the JC STEM Lab of Advanced CO₂ Upcycling, and the Youth Innovation Promotion Association of the Chinese Academy of Sciences (2023303).

References

1. V. R. Stamenkovic, D. Strmcnik, P. P. Lopes and N. M. Markovic, Nat. Mater., 2017, 16, 57–69.
2. Z. W. Seh, J. Kibsgaard, C. F. Dickens, I. Chorkendorff, J. K. Nørskov and T. F. Jaramillo, Science, 2017, 355, eaad4998.
3. Y. Wang, J. Mao, X. Meng, L. Yu, D. Deng and X. Bao, Chem. Rev., 2018, 119, 1806–1854.
4. G. Giannakakis, M. Flytzani-Stephanopoulos and E. C. H. Sykes, Acc. Chem. Res., 2018, 52, 237–247.
5. A. Wang, J. Li and T. Zhang, Nat. Rev. Chem., 2018, 2, 65–81.
6. L. Liu and A. Corma, Chem. Rev., 2018, 118, 4981–5079.
7. S. Wang, A. Y. Borisevich, S. N. Rashkeev, M. V. Glazoff, K. Sohlberg, S. J. Pennycook and S. T. Pantelides, Nat. Mater., 2004, 3, 143–146.
8. J. M. Thomas, R. Raja and D. W. Lewis, Angew. Chem., Int. Ed., 2005, 44, 6456–6482.
9. S. Cao, F. F. Tao, Y. Tang, Y. Li and J. Yu, Chem. Soc. Rev., 2016, 45, 4747–4765.
10. X.-F. Yang, A. Wang, B. Qiao, J. Li, J. Liu and T. Zhang, Acc. Chem. Res., 2013, 46, 1740–1748.
11. L. Zhang, M. Zhou, A. Wang and T. Zhang, Chem. Rev., 2020, 120, 683–733.
12. H. Wang and J. Lu, Chin. J. Chem., 2020, 38, 1422–1444.
13. P. Gotico, B. Boitrel, R. Guillot, M. Sircoglou, A. Quaranta, Z. Halime, W. Leibl and A. Aukauloo, Angew. Chem., Int. Ed., 2019, 58, 4504–4509.
14. G. W. K. Moore, S. E. L. Howell, M. Brady, X. Xu and K. McNeil, Nat. Commun., 2021, 12, 1.
15. X. Cui, W. Li, P. Ryabchuk, K. Junge and M. Beller, Nat. Catal., 2018, 1, 385–397.
16. X. Ren, S. Liu, H. Li, J. Ding, L. Liu, Z. Kuang, L. Li, H. Yang, F. Bai and Y. Huang, Sci. China: Chem., 2020, 63, 1727–1733.
17. S. Liu, H. B. Yang, S. F. Hung, J. Ding, W. Cai, L. Liu, J. Gao, X. Li, X. Ren and Z. Kuang, Angew. Chem., Int. Ed., 2020, 59, 798–803.
18. K. Asakura, H. Nagahiro, N. Ichikuni and Y. Iwasawa, Appl. Catal., A, 1999, 188, 313–324.
19. Q. Fu, H. Saltsburg and M. Flytzani-Stephanopoulos, Science, 2003, 301, 935–938.
20. X. Zhang, H. Shi and B.-Q. Xu, Angew. Chem., Int. Ed., 2005, 44, 7132–7135.
21. S. F. J. Hackett, R. M. Brydson, M. H. Gass, I. Harvey, A. D. Newman, K. Wilson and A. F. Lee, Angew. Chem., Int. Ed., 2007, 46, 8593–8596.
22. B. Qiao, A. Wang, X. Yang, L. F. Allard, Z. Jiang, Y. Cui, J. Liu, J. Li and T. Zhang, Nat. Chem., 2011, 3, 634–641.
23. J. Ding, J. Huang, Q. Zhang, Z. Wei, Q. He, Z. Chen, Y. Liu, X. Su and Y. Zhai, Catal. Sci. Technol., 2022, 12, 2416–2419.
24. Z. Zheng, Y. Yue, H. Zhuo, Q. Liu and Y. Huang, Catal. Sci. Technol., 2024, 14, 43–56.
25. S. A. Cook and A. S. Borovik, Acc. Chem. Res., 2015, 48, 2407–2414.
26. C. Tang, L. Chen, H. Li, L. Li, Y. Jiao, Y. Zheng, H. Xu, K. Davey and S.-Z. Qiao, J. Am. Chem. Soc., 2021, 143, 7819–7827.
27. B. B. Sarma, F. Maurer, D. E. Doronkin and J.-D. Grunwaldt, Chem. Rev., 2023, 123, 379–444.
28. M. Che and J. C. Védrine, Characterization of solid materials and heterogeneous catalysts: From structure to surface reactivity, John Wiley & Sons, 2012.
29. G. E. Ice, J. D. Budai and J. W. Pang, Science, 2011, 334, 1234–1239.
30. Y. Zhu, H. Inada, K. Nakamura and J. Wall, Nat. Mater., 2009, 8, 808–812.
31. X. Ye, J. Ma, W. Yu, X. Pan, C. Yang, C. Wang, Q. Liu and Y. Huang, J. Energy Chem., 2022, 67, 184–192.
32. Y. Iwasawa, K. Asakura and M. Tada, XAFS techniques for catalysts, nanomaterials, and surfaces, Springer, 2017.
33. H. F. J. Van't Blik, J. B. A. D. Van Zon, T. Huizinga, J. C. Vis, D. C. Koningsberger and R. Prins, J. Am. Chem. Soc., 1985, 107, 3139–3147.
34. F. de Groot, Chem. Rev., 2001, 101, 1779–1808.
35. G. E. Cutsail III and S. DeBeer, ACS Catal., 2022, 12, 5864–5886.
36. J. Singh, C. Lamberti and J. A. van Bokhoven, Chem. Soc. Rev., 2010, 39, 4754–4766.
37. L. J. Ament, M. Van Veenendaal, T. P. Devereaux, J. P. Hill and J. Van Den Brink, Rev. Mod. Phys., 2011, 83, 705.
38. C. Jia, K. Wohlfeld, Y. Wang, B. Moritz and T. P. Devereaux, Phys. Rev. X, 2016, 6, 021020.
39. S. Bordiga, E. Groppo, G. Agostini, J. A. van Bokhoven and C. Lamberti, Chem. Rev., 2013, 113, 1736–1850.
40. M. A. Newton and W. van Beek, Chem. Soc. Rev., 2010, 39, 4845–4863.
41. M. Xiao, Z. Xing, Z. Jin, C. Liu, J. Ge, J. Zhu, Y. Wang, X. Zhao and Z. Chen, Adv. Mater., 2020, 32, 2004900.
42. C. H. Wu, C. Liu, D. Su, H. L. Xin, H.-T. Fang, B. Eren, S. Zhang, C. B. Murray and M. B. Salmeron, Nat. Catal., 2019, 2, 78–85.
43. A. Bergmann and B. Roldan Cuenya, ACS Catal., 2019, 9, 10020–10043.
44. Y. Zhu, T.-R. Kuo, Y.-H. Li, M.-Y. Qi, G. Chen, J. Wang, Y.-J. Xu and H. M. Chen, Energy Environ. Sci., 2021, 14, 1928–1958.
45. F. Lin, Y. Liu, X. Yu, L. Cheng, A. Singer, O. G. Shpyrko, H. L. Xin, N. Tamura, C. Tian and T.-C. Weng, Chem. Rev., 2017, 117, 13123–13186.
46. S. P. Cramer, X-ray spectroscopy with synchrotron radiation, 2020 DOI:10.1007/978-3-030-28551-7.
47. D. C. Koningsberger and D. E. Ramaker, Handb. Asymmetric Heterog. Catal., 2008, 774–803.
48. M. Newville, Rev. Mineral. Geochem., 2014, 78, 33–74.
49. J. G. Chen, Surf. Sci. Rep., 1997, 30, 1–152.
50. S. Nehzati, N. V. Dolgova, A. K. James, J. J. H. Cotelesage, D. Sokaras, T. Kroll, G. N. George and I. J. Pickering, Anal. Chem., 2021, 93, 9235–9243.
51. Q. He, J. Ding, H.-J. Tsai, Y. Liu, M. Wei, Q. Zhang, Z. Wei, Z. Chen, J. Huang and S.-F. Hung, J. Colloid Interface Sci., 2023, 651, 18–26.
52. L. Liu, C. Mao, H. Fu, X. Qu and S. Zheng, ACS Appl. Mater. Interfaces, 2023, 15, 16654–16663.
53. M. Bauer, Phys. Chem. Chem. Phys., 2014, 16, 13827–13837.
54. L. Liu and A. Corma, Nat. Catal., 2021, 4, 453–456.
55. J. Finzel, K. M. Sanroman Gutierrez, A. S. Hoffman, J. Resasco, P. Christopher and S. R. Bare, ACS Catal., 2023, 13, 6462–6473.
56. K. Feng, H. Zhang, J. Gao, J. Xu, Y. Dong, Z. Kang and J. Zhong, Appl. Phys. Lett., 2020, 116, 191903.
57. D. C. Koningsberger and R. Prins, X-Ray Absorption: Principles, Applications, Techniques of EXAFS, SEXAFS and XANES, Wiley-Interscience, 1988.
58. H. Fei, J. Dong, C. Wan, Z. Zhao, X. Xu, Z. Lin, Y. Wang, H. Liu, K. Zang and J. Luo, Adv. Mater., 2018, 30, 1802146.
59. Z. Wei, Y. Liu, J. Ding, Q. He, Q. Zhang and Y. Zhai, Chin. J. Chem., 2023, 41, 3553–3559.
60. J. Ding, F. Li, J. Zhang, Q. Zhang, Y. Liu, W. Wang, W. Liu, B. Wang, J. Cai, X. Su, H. B. Yang, X. Yang, Y. Huang, Y. Zhai and B. Liu, J. Am. Chem. Soc., 2023, 145(21), 11829–11836.
61. J. Qi, J. Finzel, H. Robatjazi, M. Xu, A. S. Hoffman, S. R. Bare, X. Pan and P. Christopher, J. Am. Chem. Soc., 2020, 142, 14178–14189.
62. H. Fei, J. Dong, Y. Feng, C. S. Allen, C. Wan, B. Volosskiy, M. Li, Z. Zhao, Y. Wang, H. Sun, P. An, W. Chen, Z. Guo, C. Lee, D. Chen, I. Shakir, M. Liu, T. Hu, Y. Li, A. I. Kirkland, X. Duan and Y. Huang, Nat. Catal., 2018, 1, 63–72.
63. J. Gao, H. b Yang, X. Huang, S.-F. Hung, W. Cai, C. Jia, S. Miao, H. M. Chen, X. Yang, Y. Huang, T. Zhang and B. Liu, Chem, 2020, 6, 658–674.
64. S. Wei, A. Li, J.-C. Liu, Z. Li, W. Chen, Y. Gong, Q. Zhang, W.-C. Cheong, Y. Wang, L. Zheng, H. Xiao, C. Chen, D. Wang, Q. Peng, L. Gu, X. Han, J. Li and Y. Li, Nat. Nanotechnol., 2018, 13, 856–861.
65. Z.-Y. Wu, M. Karamad, X. Yong, Q. Huang, D. A. Cullen, P. Zhu, C. Xia, Q. Xiao, M. Shakouri, F.-Y. Chen, J. Y. Kim, Y. Xia, K. Heck, Y. Hu, M. S. Wong, Q. Li, I. Gates, S. Siahrostami and H. Wang, Nat. Commun., 2021, 12, 2870.
66. J. Huang, Q. Zhang, J. Ding and Y. Zhai, Mater. Rep. Energy, 2022, 100141.
67. H. Wang, J.-X. Liu, L. F. Allard, S. Lee, J. Liu, H. Li, J. Wang, J. Wang, S. H. Oh, W. Li, M. Flytzani-Stephanopoulos, M. Shen, B. R. Goldsmith and M. Yang, Nat. Commun., 2019, 10, 3808.
68. J. Ding, F. Li, J. Zhang, H. Qi, Z. Wei, C. Su, H. B. Yang, Y. Zhai and B. Liu, Adv. Mater., 2024, 36, 2306480.
69. X. Ren, J. Zhao, X. Li, J. Shao, B. Pan, A. Salamé, E. Boutin, T. Groizard, S. Wang and J. Ding, Nat. Commun., 2023, 14, 3401.
70. W. Zhang, S. Liu, Y. Yang, H. Qi, S. Xi, Y. Wei, J. Ding, Z. J. Wang, Q. Li and B. Liu, Angew. Chem., Int. Ed., 2023, 62, e202219241.
71. L. Liu, P. Zhou, X. Su, Y. Liu, Y. Sun, H. Yang, H. Fu, X. Qu, S. Liu and S. Zheng, Sens. Actuators, B, 2022, 351, 130983.
72. R. Rana, F. D. Vila, A. R. Kulkarni and S. R. Bare, ACS Catal., 2022, 12, 13813–13830.
73. S. D. Kelly, D. Hesterberg and B. Ravel, Methods of Soil Analysis Part 5—Mineralogical Methods, 2008, pp. 387–463 DOI:10.2136/sssabookser5.5.c14.
74. Y. Chen, R. Rana, T. Sours, F. D. Vila, S. Cao, T. Blum, J. Hong, A. S. Hoffman, C.-Y. Fang, Z. Huang, C. Shang, C. Wang, J. Zeng, M. Chi, C. X. Kronawitter, S. R. Bare, B. C. Gates and A. R. Kulkarni, J. Am. Chem. Soc., 2021, 143, 20144–20156.
75. A. Zitolo, N. Ranjbar-Sahraie, T. Mineva, J. Li, Q. Jia, S. Stamatin, G. F. Harrington, S. M. Lyth, P. Krtil, S. Mukerjee, E. Fonda and F. Jaouen, Nat. Commun., 2017, 8, 957.
76. S. Sun, N. Jiang and D. Xia, J. Phys. Chem. C, 2011, 115, 9511–9517.
77. Y. Cai, J. Fu, Y. Zhou, Y.-C. Chang, Q. Min, J.-J. Zhu, Y. Lin and W. Zhu, Nat. Commun., 2021, 12, 586.
78. H. Shang, X. Zhou, J. Dong, A. Li, X. Zhao, Q. Liu, Y. Lin, J. Pei, Z. Li, Z. Jiang, D. Zhou, L. Zheng, Y. Wang, J. Zhou, Z. Yang, R. Cao, R. Sarangi, T. Sun, X. Yang, X. Zheng, W. Yan, Z. Zhuang, J. Li, W. Chen, D. Wang, J. Zhang and Y. Li, Nat. Commun., 2020, 11, 3049.
79. D. Chen, L.-H. Zhang, J. Du, H. Wang, J. Guo, J. Zhan, F. Li and F. Yu, Angew. Chem., Int. Ed., 2021, 60, 24022–24027.
80. H. Fei, J. Dong, D. Chen, T. Hu, X. Duan, I. Shakir, Y. Huang and X. Duan, Chem. Soc. Rev., 2019, 48, 5207–5241.
81. J. J. Rehr and R. C. Albers, Rev. Mod. Phys., 2000, 72, 621.
82. H. Funke, A. Scheinost and M. Chukalina, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 094110.
83. M. Munoz, P. Argoul and F. Farges, Am. Mineral., 2003, 88, 694–700.
84. R. Sarangi, Coord. Chem. Rev., 2013, 257, 459–472.
85. K. Cząstka, A. A. Oughli, O. Rüdiger and S. DeBeer, Faraday Discuss., 2022, 234, 214–231.
86. J. Ding, Z. Teng, X. Su, K. Kato, Y. Liu, T. Xiao, W. Liu, L. Liu, Q. Zhang and X. Ren, Chem, 2023, 9, 1017–1035.
87. Y. Deng, J. Zhao, S. Wang, R. Chen, J. Ding, H.-J. Tsai, W.-J. Zeng, S.-F. Hung, W. Xu and J. Wang, J. Am. Chem. Soc., 2023, 145, 7242–7251.
88. Y. Pan, Y. Chen, K. Wu, Z. Chen, S. Liu, X. Cao, W.-C. Cheong, T. Meng, J. Luo, L. Zheng, C. Liu, D. Wang, Q. Peng, J. Li and C. Chen, Nat. Commun., 2019, 10, 4290.
89. H. B. Yang, S.-F. Hung, S. Liu, K. Yuan, S. Miao, L. Zhang, X. Huang, H.-Y. Wang, W. Cai, R. Chen, J. Gao, X. Yang, W. Chen, Y. Huang, H. M. Chen, C. M. Li, T. Zhang and B. Liu, Nat. Energy, 2018, 3, 140–147.
90. L. A. Avakyan, A. S. Manukyan, A. A. Mirzakhanyan, E. G. Sharoyan, Y. V. Zubavichus, A. L. Trigub, N. A. Kolpacheva and L. A. Bugaev, Opt. Spectrosc., 2013, 114, 347–352.
91. G. J. Colpas, M. J. Maroney, C. Bagyinka, M. Kumar, W. S. Willis, S. L. Suib, P. K. Mascharak and N. Baidya, Inorg. Chem., 1991, 30, 920–928.
92. C. D. Douglas, A. V. Dias and D. B. Zamble, Dalton Trans., 2012, 41, 7876–7878.
93. Q. Jia, N. Ramaswamy, H. Hafiz, U. Tylus, K. Strickland, G. Wu, B. Barbiellini, A. Bansil, E. F. Holby, P. Zelenay and S. Mukerjee, ACS Nano, 2015, 9, 12496–12505.
94. G. Rossi, F. d'Acapito, L. Amidani, F. Boscherini and M. Pedio, Phys. Chem. Chem. Phys., 2016, 18, 23686–23694.
95. F. Calle-Vallejo, J. I. Martínez and J. Rossmeisl, Phys. Chem. Chem. Phys., 2011, 13, 15639–15643.
96. J. Qin, H. Liu, P. Zou, R. Zhang, C. Wang and H. L. Xin, J. Am. Chem. Soc., 2022, 144, 2197–2207.
97. A. Zitolo, V. Goellner, V. Armel, M.-T. Sougrati, T. Mineva, L. Stievano, E. Fonda and F. Jaouen, Nat. Mater., 2015, 14, 937–942.
98. B. Sheng, D. Cao, C. Liu, S. Chen and L. Song, J. Phys. Chem. Lett., 2021, 12, 11543–11554.
99. D. Cao, W. Xu, S. Chen, C. Liu, B. Sheng, P. Song, O. A. Moses, L. Song and S. Wei, Adv. Mater., 2023, 35, 2205346.
100. X. Su, Z. Jiang, J. Zhou, H. Liu, D. Zhou, H. Shang, X. Ni, Z. Peng, F. Yang and W. Chen, Nat. Commun., 2022, 13, 1322.
101. L. Zhang, X. Yang, Q. Yuan, Z. Wei, J. Ding, T. Chu, C. Rong, Q. Zhang, Z. Ye and F.-Z. Xuan, Nat. Commun., 2023, 14, 8311.
102. X. Li, Y. Zeng, C.-W. Tung, Y.-R. Lu, S. Baskaran, S.-F. Hung, S. Wang, C.-Q. Xu, J. Wang and T.-S. Chan, ACS Catal., 2021, 11, 7292–7301.
103. X. Li, C.-S. Cao, S.-F. Hung, Y.-R. Lu, W. Cai, A. I. Rykov, S. Miao, S. Xi, H. Yang and Z. Hu, Chem, 2020, 6, 3440–3454.
104. L. DeRita, J. Resasco, S. Dai, A. Boubnov, H. V. Thang, A. S. Hoffman, I. Ro, G. W. Graham, S. R. Bare, G. Pacchioni, X. Pan and P. Christopher, Nat. Mater., 2019, 18, 746–751.
105. Y. Lu, J. Wang, L. Yu, L. Kovarik, X. Zhang, A. S. Hoffman, A. Gallo, S. R. Bare, D. Sokaras, T. Kroll, V. Dagle, H. Xin and A. M. Karim, Nat. Catal., 2019, 2, 149–156.
106. S. Liu, H. B. Yang, S.-F. Hung, J. Ding, W. Cai, L. Liu, J. Gao, X. Li, X. Ren, Z. Kuang, Y. Huang, T. Zhang and B. Liu, Angew. Chem., Int. Ed., 2020, 59, 798–803.
107. X. Guo, G. Fang, G. Li, H. Ma, H. Fan, L. Yu, C. Ma, X. Wu, D. Deng, M. Wei, D. Tan, R. Si, S. Zhang, J. Li, L. Sun, Z. Tang, X. Pan and X. Bao, Science, 2014, 344, 616–619.
108. J. Gu, C.-S. Hsu, L. Bai, H. M. Chen and X. Hu, Science, 2019, 364, 1091–1094.
109. L. Cao, Q. Luo, W. Liu, Y. Lin, X. Liu, Y. Cao, W. Zhang, Y. Wu, J. Yang, T. Yao and S. Wei, Nat. Catal., 2019, 2, 134–141.
110. J. Gu, C.-S. Hsu, L. Bai, H. M. Chen and X. Hu, Science, 2019, 364, 1091–1094.
111. D. Karapinar, N. T. Huan, N. Ranjbar Sahraie, J. Li, D. Wakerley, N. Touati, S. Zanna, D. Taverna, L. H. Galvão Tizei, A. Zitolo, F. Jaouen, V. Mougel and M. Fontecave, Angew. Chem., Int. Ed., 2019, 58, 15098–15103.
112. G. Xing, M. Tong, P. Yu, L. Wang, G. Zhang, C. Tian and H. Fu, Angew. Chem., Int. Ed., 2022, 61, e202211098.
113. J. Yang, H. Qi, A. Li, X. Liu, X. Yang, S. Zhang, Q. Zhao, Q. Jiang, Y. Su, L. Zhang, J.-F. Li, Z.-Q. Tian, W. Liu, A. Wang and T. Zhang, J. Am. Chem. Soc., 2022, 144, 12062–12071.
114. P. B. Thompson, B. N. Nguyen, R. Nicholls, R. A. Bourne, J. B. Brazier, K. R. Lovelock, S. D. Brown, D. Wermeille, O. Bikondoa and C. A. Lucas, J. Synchrotron Radiat., 2015, 22, 1426–1439.
115. F. Tao, W. F. Schneider and P. V. Kamat, Heterogeneous Catalysis at Nanoscale for Energy Applications, John Wiley & Sons, 2014 DOI:10.1002/9781118843468.
116. C.-L. Dong and L. Vayssieres, Chem. – Eur. J., 2018, 24, 18356–18373.
117. P. Chen, N. Zhang, S. Wang, T. Zhou, Y. Tong, C. Ao, W. Yan, L. Zhang, W. Chu, C. Wu and Y. Xie, Proc. Natl. Acad. Sci. U. S. A., 2019, 116, 6635–6640.
118. Y. Tong, P. Chen, T. Zhou, K. Xu, W. Chu, C. Wu and Y. Xie, Angew. Chem., Int. Ed., 2017, 56, 7121–7125.
119. H. B. Yang, J. Miao, S.-F. Hung, J. Chen, H. B. Tao, X. Wang, L. Zhang, R. Chen, J. Gao, H. M. Chen, L. Dai and B. Liu, Sci. Adv., 2016, 2, e1501122.
120. L. Wei, L. Wen, T. Yang and N. Zhang, Energy Fuels, 2015, 29, 5088–5094.
121. J. Shi, Y. Wei, D. Zhou, L. Zhang, X. Yang, Z. Miao, H. Qi, S. Zhang, A. Li, X. Liu, W. Yan, Z. Jiang, A. Wang and T. Zhang, ACS Catal., 2022, 12, 7760–7772.
122. F. Frati, M. O. J. Y. Hunault and F. M. F. de Groot, Chem. Rev., 2020, 120, 4056–4110.
123. D. Attwood, Soft x-rays and extreme ultraviolet radiation: principles and applications, Cambridge University Press, 2000. https://ilsf.ipm.ac.ir/News/2014-03-03BeamlineOpWrkshp/SXR_EUV_excerps.pdf.
124. M. L. Baker, M. W. Mara, J. J. Yan, K. O. Hodgson, B. Hedman and E. I. Solomon, Coord. Chem. Rev., 2017, 345, 182–208.
125. D. Coster and R. D. L. Kronig, Physica, 1935, 2, 13–24.
126. H. Wang, P. Ge, C. Riordan, S. Brooker, C. Woomer, T. Collins, C. Melendres, O. Graudejus, N. Bartlett and S. Cramer, J. Phys. Chem. B, 1998, 102, 8343–8346.
127. H. Wang, D. S. Patil, W. Gu, L. Jacquamet, S. Friedrich, T. Funk and S. P. Cramer, J. Electron Spectrosc. Relat. Phenom., 2001, 114–116, 855–863.
128. J. Van der Zwaan, S. Albracht, R. Fontijn and E. Slater, FEBS Lett., 1985, 179, 271–277.
129. J. J. Moura, M. Teixeira and I. Moura, Pure Appl. Chem., 1989, 61, 915–921.
130. H. Wang, S. M. Butorin, A. T. Young and J. Guo, J. Phys. Chem. C, 2013, 117, 24767–24772.
131. P. Olalde-Velasco, J. Jiménez-Mier, J. Denlinger and W.-L. Yang, Phys. Rev. B: Condens. Matter Mater. Phys., 2013, 87, 245136.
132. R. K. Hocking, E. C. Wasinger, F. M. de Groot, K. O. Hodgson, B. Hedman and E. I. Solomon, J. Am. Chem. Soc., 2006, 128, 10442–10451.
133. R. K. Hocking, E. C. Wasinger, Y.-L. Yan, F. M. Degroot, F. A. Walker, K. O. Hodgson, B. Hedman and E. I. Solomon, J. Am. Chem. Soc., 2007, 129, 113–125.
134. J. Zhou, Y. Hu, Y.-C. Chang, Z. Hu, Y.-C. Huang, Y. Fan, H.-J. Lin, C.-W. Pao, C.-L. Dong and J.-F. Lee, ACS Catal., 2022, 12, 3138–3148.
135. B. T. Thole, G. Van Der Laan and P. H. Butler, Chem. Phys. Lett., 1988, 149, 295–299.
136. P. S. Miedema, M. M. van Schooneveld, R. Bogerd, T. C. R. Rocha, M. Hävecker, A. Knop-Gericke and F. M. F. de Groot, J. Phys. Chem. C, 2011, 115, 25422–25428.
137. M. Xiao, J. Zhu, L. Ma, Z. Jin, J. Ge, X. Deng, Y. Hou, Q. He, J. Li, Q. Jia, S. Mukerjee, R. Yang, Z. Jiang, D. Su, C. Liu and W. Xing, ACS Catal., 2018, 8, 2824–2832.
138. L. Liu, T. Xiao, H. Fu, Z. Chen, X. Qu and S. Zheng, Appl. Catal., B, 2023, 323, 122181.
139. Z. Chen, H. Niu, J. Ding, H. Liu, P. H. Chen, Y. H. Lu, Y. R. Lu, W. Zuo, L. Han and Y. Guo, Angew. Chem., 2021, 133, 25608–25614.
140. J. Ding, Z. Wei, F. Li, J. Zhang, Q. Zhang, J. Zhou, W. Wang, Y. Liu, Z. Zhang and X. Su, Nat. Commun., 2023, 14, 6550.
141. A. Tanaka and T. Jo, J. Phys. Soc. Jpn., 1994, 63, 2788–2807.
142. T. Kroll, R. Kraus, R. Schönfelder, V. Y. Aristov, O. V. Molodtsova, P. Hoffmann and M. Knupfer, J. Chem. Phys., 2012, 137(5), 054306.
143. S. K. Beaumont, Phys. Chem. Chem. Phys., 2020, 22, 18747–18756.
144. L. Xi, M. Schellenberger, R. F. Praeg, D. Gao, D. Drevon, P. Plate, P. Bogdanoff, R. van de Krol and K. M. Lange, ACS Appl. Energy Mater., 2019, 2, 4126–4134.
145. R. Toyoshima and H. Kondoh, J. Phys.: Condens. Matter, 2015, 27, 083003.
146. P. T. Kristiansen, T. Rocha, A. Knop-Gericke, J. Guo and L. Duda, Rev. Sci. Instrum., 2013, 84(11), 113107.
147. A. Yamaguchi, N. Akamatsu, S. Saegusa, R. Nakamura, Y. Utsumi, M. Kato, I. Yagi, T. Ishihara and M. Oura, RSC Adv., 2022, 12, 10425–10430.
148. J. D. Smith, C. D. Cappa, K. R. Wilson, B. M. Messer, R. C. Cohen and R. J. Saykally, Science, 2004, 306, 851–853.
149. P. Wernet, D. Nordlund, U. Bergmann, M. Cavalleri, M. Odelius, H. Ogasawara, L. Å. Näslund, T. K. Hirsch, L. Ojamäe, P. Glatzel, L. G. M. Pettersson and A. Nilsson, Science, 2004, 304, 995–999.
150. J. W. Smith and R. J. Saykally, Chem. Rev., 2017, 117, 13909–13934.
151. V. Pfeifer, T. E. Jones, J. J. Velasco Vélez, C. Massué, M. T. Greiner, R. Arrigo, D. Teschner, F. Girgsdies, M. Scherzer, J. Allan, M. Hashagen, G. Weinberg, S. Piccinin, M. Hävecker, A. Knop-Gericke and R. Schlögl, Phys. Chem. Chem. Phys., 2016, 18, 2292–2296.
152. V. Pfeifer, T. E. Jones, J. J. Velasco Vélez, R. Arrigo, S. Piccinin, M. Hävecker, A. Knop-Gericke and R. Schlögl, Chem. Sci., 2017, 8, 2143–2149.
153. C. Kolczewski, R. Puttner, O. Plashkevych, H. Agren, V. Staemmler, M. Martins, G. Snell, A. S. Schlachter, M. Sant'Anna, G. Kaindl and L. G. M. Pettersson, J. Chem. Phys., 2001, 115, 6426–6437.
154. M. Nagasaka, H. Yuzawa and N. Kosugi, Anal. Sci., 2020, 36, 95–99.
155. H. Bin Yang, C.-Q. Xu, S. Baskaran, Y.-R. Lu, C. Gu, W. Liu, J. Ding, J. Zhang, Q. Wang, W. Chen, J. Li, Y. Huang, T. Zhang and B. Liu, EES Catal., 2023, 1, 774–783.
156. C. Yan, H. Li, Y. Ye, H. Wu, F. Cai, R. Si, J. Xiao, S. Miao, S. Xie, F. Yang, Y. Li, G. Wang and X. Bao, Energy Environ. Sci., 2018, 11, 1204–1210.
157. M. F. Tesch, S. A. Bonke, T. E. Jones, M. N. Shaker, J. Xiao, K. Skorupska, R. Mom, J. Melder, P. Kurz, A. Knop-Gericke, R. Schlögl, R. K. Hocking and A. N. Simonov, Angew. Chem., Int. Ed., 2019, 58, 3426–3432.
158. H.-T. Lien, S.-T. Chang, P.-T. Chen, D. P. Wong, Y.-C. Chang, Y.-R. Lu, C.-L. Dong, C.-H. Wang, K.-H. Chen and L.-C. Chen, Nat. Commun., 2020, 11, 4233.
159. F. M. F. de Groot, M. Grioni, J. C. Fuggle, J. Ghijsen, G. A. Sawatzky and H. Petersen, Phys. Rev. B: Condens. Matter Mater. Phys., 1989, 40, 5715–5723.
160. S.-T. Chang, C.-H. Wang, H.-Y. Du, H.-C. Hsu, C.-M. Kang, C.-C. Chen, J. C. Wu, S.-C. Yen, W.-F. Huang and L.-C. Chen, Energy Environ. Sci., 2012, 5, 5305–5314.
161. J.-K. Chang, M.-T. Lee and W.-T. Tsai, J. Power Sources, 2007, 166, 590–594.
162. Y.-C. Huang, W. Chen, Z. Xiao, Z. Hu, Y.-R. Lu, J.-L. Chen, C.-L. Chen, H.-J. Lin, C.-T. Chen, K. T. Arul, S. Wang, C.-L. Dong and W.-C. Chou, J. Phys. Chem. Lett., 2022, 13, 8386–8396.
163. L. A. J. Garvie and A. J. Craven, Phys. Chem. Miner., 1994, 21, 191–206.
164. Q. Wang, Y. Xiao, S. Yang, Y. Zhang, L. Wu, H. Pan, D. Rao, T. Chen, Z. Sun, G. Wang, J. Zhu, J. Zeng, S. Wei and X. Zheng, Nano Lett., 2022, 22, 10216–10223.
165. S. M. Butorin, K. O. Kvashnina, J. R. Vegelius, D. Meyer and D. K. Shuh, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, 8093–8097.
166. S.-Y. Chen, Y.-H. Lu, T.-W. Huang, D.-C. Yan and C.-L. Dong, J. Phys. Chem. C, 2010, 114, 19576–19581.
167. M. Münzenberg, W. Felsch and P. Schaaf, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 014427.
168. Y.-C. Lin, P.-Y. Teng, P.-W. Chiu and K. Suenaga, Phys. Rev. Lett., 2015, 115, 206803.
169. F. Zhao, Q. Gong, B. Traynor, D. Zhang, J. Li, H. Ye, F. Chen, N. Han, Y. Wang, X. Sun and Y. Li, Nano Res., 2016, 9, 3162–3170.
170. H. C. Choi, S. Y. Bae, J. Park, K. Seo, C. Kim, B. Kim, H. J. Song and H. J. Shin, Appl. Phys. Lett., 2004, 85, 5742–5744.
171. M. Katsikini, F. Pinakidou, E. C. Paloura, E. Wendler, W. Wesch and R. Manzke, J. Phys.: Conf. Ser., 2009, 190, 012065.
172. P. Glatzel, T.-C. Weng, K. Kvashnina, J. Swarbrick, M. Sikora, E. Gallo, N. Smolentsev and R. A. Mori, J. Electron Spectrosc. Relat. Phenom., 2013, 188, 17–25.
173. M. Kubin, J. Kern, M. Guo, E. Källman, R. Mitzner, V. K. Yachandra, M. Lundberg, J. Yano and P. Wernet, Phys. Chem. Chem. Phys., 2018, 20, 16817–16827.
174. G. Vankó, A. Bordage, P. Glatzel, E. Gallo, M. Rovezzi, W. Gawelda, A. Galler, C. Bressler, G. Doumy, A. M. March, E. P. Kanter, L. Young, S. H. Southworth, S. E. Canton, J. Uhlig, G. Smolentsev, V. Sundström, K. Haldrup, T. B. van Driel, M. M. Nielsen, K. S. Kjaer and H. T. Lemke, J. Electron Spectrosc. Relat. Phenom., 2013, 188, 166–171.
175. S. Mitchell, E. Vorobyeva and J. Pérez-Ramírez, Angew. Chem., Int. Ed., 2018, 57, 15316–15329.
176. Z. Chen, E. Vorobyeva, S. Mitchell, E. Fako, M. A. Ortuño, N. López, S. M. Collins, P. A. Midgley, S. Richard, G. Vilé and J. Pérez-Ramírez, Nat. Nanotechnol., 2018, 13, 702–707.
177. H. Zhang, W. Tian, X. Duan, H. Sun, S. Liu and S. Wang, Adv. Mater., 2020, 32, 1904037.
178. Z. Németh, J. Szlachetko, É. G. Bajnóczi and G. Vankó, Rev. Sci. Instrum., 2016, 87, 103105.
179. A.-P. Honkanen, S. Ollikkala, T. Ahopelto, A.-J. Kallio, M. Blomberg and S. Huotari, Rev. Sci. Instrum., 2019, 90, 033107.
180. P. Zimmermann, S. Peredkov, P. M. Abdala, S. DeBeer, M. Tromp, C. Müller and J. A. van Bokhoven, Coord. Chem. Rev., 2020, 423, 213466.
181. A. Meisel, G. Leonhardt and R. Szargan, X-ray spectra and chemical binding, Springer, 1989. https://link.springer.com/book/9783642822643.
182. M. Van Bay, N. K. Hien, P. T. Quy, P. C. Nam, D. U. Van and D. T. Quang, Vietnam J. Chem., 2019, 57, 389–400.
183. N. Lee, T. Petrenko, U. Bergmann, F. Neese and S. DeBeer, J. Am. Chem. Soc., 2010, 132, 9715–9727.
184. J. K. Kowalska, A. W. Hahn, A. Albers, C. E. Schiewer, R. Bjornsson, F. A. Lima, F. Meyer and S. DeBeer, Inorg. Chem., 2016, 55, 4485–4497.
185. T. Fransson, R. Chatterjee, F. D. Fuller, S. Gul, C. Weninger, D. Sokaras, T. Kroll, R. Alonso-Mori, U. Bergmann and J. Kern, Biochemistry, 2018, 57, 4629–4637.
186. P. Glatzel and U. Bergmann, Coord. Chem. Rev., 2005, 249, 65–95.
187. U. Bergmann, P. Glatzel, F. de Groot and S. Cramer, J. Am. Chem. Soc., 1999, 121, 4926–4927.
188. K. Tsutsumi, H. Nakamori and K. Ichikawa, Phys. Rev. B: Condens. Matter Mater. Phys., 1976, 13, 929.
189. P. Glatzel, J. Yano, U. Bergmann, H. Visser, J. H. Robblee, W. Gu, F. M. de Groot, S. P. Cramer and V. K. Yachandra, J. Phys. Chem. Solids, 2005, 66, 2163–2167.
190. M. Rovezzi and P. Glatzel, Semicond. Sci. Technol., 2014, 29, 023002.
191. J. Kawai, M. Takami and C. Satoko, Phys. Rev. Lett., 1990, 65, 2193.
192. J. Kern, R. Alonso-Mori, R. Tran, J. Hattne, R. J. Gildea, N. Echols, C. Glöckner, J. Hellmich, H. Laksmono and R. G. Sierra, Science, 2013, 340, 491–495.
193. C. J. Pollock, M. U. Delgado-Jaime, M. Atanasov, F. Neese and S. DeBeer, J. Am. Chem. Soc., 2014, 136, 9453–9463.
194. R. Alonso-Mori, J. Kern, R. J. Gildea, D. Sokaras, T.-C. Weng, B. Lassalle-Kaiser, R. Tran, J. Hattne, H. Laksmono, J. Hellmich, C. Glöckner, N. Echols, R. G. Sierra, D. W. Schafer, J. Sellberg, C. Kenney, R. Herbst, J. Pines, P. Hart, S. Herrmann, R. W. Grosse-Kunstleve, M. J. Latimer, A. R. Fry, M. M. Messerschmidt, A. Miahnahri, M. M. Seibert, P. H. Zwart, W. E. White, P. D. Adams, M. J. Bogan, S. Boutet, G. J. Williams, A. Zouni, J. Messinger, P. Glatzel, N. K. Sauter, V. K. Yachandra, J. Yano and U. Bergmann, Proc. Natl. Acad. Sci. U. S. A., 2012, 109, 19103–19107.
195. S. Limandri, J. Robledo and G. Tirao, Spectrochim. Acta, Part B, 2018, 144, 29–37.
196. U. Bergmann and P. Glatzel, Photosynth. Res., 2009, 102, 255–266.
197. G. Vankó, T. Neisius, G. Molnar, F. Renz, S. Karpati, A. Shukla and F. M. De Groot, J. Phys. Chem. B, 2006, 110, 11647–11653.
198. S. Lafuerza, A. Carlantuono, M. Retegan and P. Glatzel, Inorg. Chem., 2020, 59, 12518–12535.
199. X. Wang, M. M. Grush, A. G. Froeschner and S. P. Cramer, J. Synchrotron Radiat., 1997, 4, 236–242.
200. J.-P. Rueff and A. Shukla, Rev. Mod. Phys., 2010, 82, 847.
201. J. Badro, J.-P. Rueff, G. Vankó, G. Monaco, G. Fiquet and F. Guyot, Science, 2004, 305, 383–386.
202. V. A. Saveleva, K. Ebner, L. Ni, G. Smolentsev, D. Klose, A. Zitolo, E. Marelli, J. Li, M. Medarde and O. V. Safonova, Angew. Chem., Int. Ed., 2021, 60, 11707–11712.
203. G. Vankó and F. M. de Groot, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 75, 177101.
204. S. Mebs, B. Braun, R. Kositzki, C. Limberg and M. Haumann, Inorg. Chem., 2015, 54, 11606–11624.
205. Q. Jia, N. Ramaswamy, U. Tylus, K. Strickland, J. Li, A. Serov, K. Artyushkova, P. Atanassov, J. Anibal, C. Gumeci, S. C. Barton, M.-T. Sougrati, F. Jaouen, B. Halevi and S. Mukerjee, Nano Energy, 2016, 29, 65–82.
206. J. Li, S. Ghoshal, W. Liang, M.-T. Sougrati, F. Jaouen, B. Halevi, S. McKinney, G. McCool, C. Ma, X. Yuan, Z.-F. Ma, S. Mukerjee and Q. Jia, Energy Environ. Sci., 2016, 9, 2418–2432.
207. P. N. Phu, C. E. Gutierrez, S. Kundu, D. Sokaras, T. Kroll, T. H. Warren and S. C. E. Stieber, Inorg. Chem., 2020, 60, 736–744.
208. M. A. Beckwith, M. Roemelt, M.-N. Collomb, C. DuBoc, T.-C. Weng, U. Bergmann, P. Glatzel, F. Neese and S. DeBeer, Inorg. Chem., 2011, 50, 8397–8409.
209. B. Lassalle-Kaiser, T. T. Boron III, V. Krewald, J. Kern, M. A. Beckwith, M. U. Delgado-Jaime, H. Schroeder, R. Alonso-Mori, D. Nordlund and T.-C. Weng, Inorg. Chem., 2013, 52, 12915–12922.
210. G. Smolentsev, A. V. Soldatov, J. Messinger, K. Merz, T. Weyhermüller, U. Bergmann, Y. Pushkar, J. Yano, V. K. Yachandra and P. Glatzel, J. Am. Chem. Soc., 2009, 131, 13161–13167.
211. S. G. Eeckhout, O. V. Safonova, G. Smolentsev, M. Biasioli, V. A. Safonov, L. N. Vykhodtseva, M. Sikora and P. Glatzel, J. Anal. At. Spectrom., 2009, 24, 215–223.
212. M. U. Delgado-Jaime, B. R. Dible, K. P. Chiang, W. W. Brennessel, U. Bergmann, P. L. Holland and S. DeBeer, Inorg. Chem., 2011, 50, 10709–10717.
213. Y. Pushkar, X. Long, P. Glatzel, G. W. Brudvig, G. C. Dismukes, T. J. Collins, V. K. Yachandra, J. Yano and U. Bergmann, Angew. Chem., Int. Ed., 2010, 49, 800–803.
214. J. C. Swarbrick, Y. Kvashnin, K. Schulte, K. Seenivasan, C. Lamberti and P. Glatzel, Inorg. Chem., 2010, 49, 8323–8332.
215. K. M. Lancaster, M. Roemelt, P. Ettenhuber, Y. Hu, M. W. Ribbe, F. Neese, U. Bergmann and S. DeBeer, Science, 2011, 334, 974–977.
216. J. Ding, H. Bin Yang, X.-L. Ma, S. Liu, W. Liu, Q. Mao, Y. Huang, J. Li, T. Zhang and B. Liu, Nat. Energy, 2023, 1–9.
217. K. A. Lomachenko, E. Borfecchia, C. Negri, G. Berlier, C. Lamberti, P. Beato, H. Falsig and S. Bordiga, J. Am. Chem. Soc., 2016, 138, 12025–12028.
218. Z. Chen, G. Zhang, Q. Hu, Y. Zheng, S. Cao, G. Chen, C. Li, T. Boyko, N. Chen and W. Chen, Small Struct., 2022, 3, 2200031.
219. C. J. Pollock, K. Grubel, P. L. Holland and S. DeBeer, J. Am. Chem. Soc., 2013, 135, 11803–11808.
220. S. C. E. Stieber, C. Milsmann, J. M. Hoyt, Z. R. Turner, K. D. Finkelstein, K. Wieghardt, S. DeBeer and P. J. Chirik, Inorg. Chem., 2012, 51, 3770–3785.
221. T. T. Lu, T. C. Weng and W. F. Liaw, Angew. Chem., 2014, 126, 11746–11750.
222. G. E. Cutsail III, N. L. Gagnon, A. D. Spaeth, W. B. Tolman and S. DeBeer, Angew. Chem., Int. Ed., 2019, 58, 9114–9119.
223. T. Kroll, R. G. Hadt, S. A. Wilson, M. Lundberg, J. J. Yan, T.-C. Weng, D. Sokaras, R. Alonso-Mori, D. Casa and M. H. Upton, J. Am. Chem. Soc., 2014, 136, 18087–18099.
224. M. Lundberg and P. Wernet, Synchrotron Light Sources and Free-Electron Lasers: Accelerator Physics, Instrumentation and Science Applications, Springer, 2020, pp. 2315–2366.
225. J. Forsberg, L.-C. Duda, A. Olsson, T. Schmitt, J. Andersson, J. Nordgren, J. Hedberg, C. Leygraf, T. Aastrup and D. Wallinder, Rev. Sci. Instrum., 2007, 78, 083110.
226. F. M. F. de Groot, M. W. Haverkort, H. Elnaggar, A. Juhin, K.-J. Zhou and P. Glatzel, Nat. Rev. Methods Primers, 2024, 4, 45.
227. C. Milne, T. Penfold and M. Chergui, Coord. Chem. Rev., 2014, 277, 44–68.
228. C. D. Rankine and T. J. Penfold, J. Phys. Chem. A, 2021, 125, 4276–4293.
229. A. Kotani and S. Shin, Rev. Mod. Phys., 2001, 73, 203–246.
230. P. Glatzel and U. Bergmann, Coord. Chem. Rev., 2005, 249, 65–95.
231. F. Gel'mukhanov and H. Ågren, Phys. Rep., 1999, 312, 87–330.
232. A. Kotani and S. Shin, Rev. Mod. Phys., 2001, 73, 203.
233. P. Eisenberger, P. Platzman and H. Winick, Phys. Rev. Lett., 1976, 36, 623.
234. M. L. Baker, M. W. Mara, J. J. Yan, K. O. Hodgson, B. Hedman and E. I. Solomon, Coord. Chem. Rev., 2017, 345, 182–208.
235. E. Domashevskaya, V. Gorbachev, V. Terekhov, V. Kashkarov, E. Panfilova and A. Shchukarev, J. Electron Spectrosc. Relat. Phenom., 2001, 114, 901–908.
236. D. G. Strawn and L. L. Baker, Environ. Sci. Technol., 2008, 42, 37–42.
237. M. Lundberg, T. Kroll, S. DeBeer, U. Bergmann, S. A. Wilson, P. Glatzel, D. Nordlund, B. Hedman, K. O. Hodgson and E. I. Solomon, J. Am. Chem. Soc., 2013, 135, 17121–17134.
238. K. Hämäläinen, D. Siddons, J. Hastings and L. Berman, Phys. Rev. Lett., 1991, 67, 2850.
239. H. Hayashi, M. Kawata, R. Takeda, A. Sato, Y. Udagawa, N. Kawamura and S. Nanao, J. Phys. Chem. Solids, 2005, 66, 2168–2172.
240. P. Harrison and A. Valavanis, Quantum wells, wires and dots: theoretical and computational physics of semiconductor nanostructures, John Wiley & Sons, 2016. https://homes.nano.aau.dk/kp/nano6-2009/quantum%20wells%20wires%20and%20dots.pdf.
241. J. J. Kas, J. J. Rehr, J. A. Soininen and P. Glatzel, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 83, 235114.
242. P. Loeffen, R. Pettifer, S. Müllender, M. Van Veenendaal, J. Röhler and D. Sivia, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 14877.
243. P. Carra, M. Fabrizio and B. Thole, Phys. Rev. Lett., 1995, 74, 3700.
244. A. Kotani, J. Synchrotron Radiat., 2001, 8, 110–114.
245. M. Kottwitz, Y. Li, R. M. Palomino, Z. Liu, G. Wang, Q. Wu, J. Huang, J. Timoshenko, S. D. Senanayake and M. Balasubramanian, ACS Catal., 2019, 9, 8738–8748.
246. I. Jarrige, K. Ishii, D. Matsumura, Y. Nishihata, M. Yoshida, H. Kishi, M. Taniguchi, M. Uenishi, H. Tanaka and H. Kasai, ACS Catal., 2015, 5, 1112–1118.
247. M. W. Small, J. J. Kas, K. O. Kvashnina, J. J. Rehr, R. G. Nuzzo, M. Tromp and A. I. Frenkel, ChemPhysChem, 2014, 15, 1569–1572.
248. B. Mei, S. Gu, X. Du, Z. Li, H. Cao, F. Song, Y. Huang and Z. Jiang, X-Ray Spectrom., 2020, 49, 251–259.
249. T. Ishihara, T. Tokushima, Y. Horikawa, M. Kato and I. Yagi, Rev. Sci. Instrum., 2017, 88, 104101.
250. P. Glatzel, J. Singh, K. O. Kvashnina and J. A. van Bokhoven, J. Am. Chem. Soc., 2010, 132, 2555–2557.
251. J. Szlachetko, J. Sá, M. Nachtegaal, U. Hartfelder, J.-C. Dousse, J. Hoszowska, D. L. Abreu Fernandes, H. Shi and C. Stampfl, J. Phys. Chem. Lett., 2014, 5, 80–84.
252. T. Yoshida, A. Shabana, D. C. Izuogu, K. Fuku, T. Sato, H. Zhang, Y. Yamamoto, J. Kamata, H. Ohmagari and M. Hasegawa, J. Phys. Chem. C, 2022, 126, 7973–7979.
253. G. A. Somorjai and Y. Li, Introduction to surface chemistry and catalysis, John Wiley & Sons, 2010.
254. J. Wang, C.-S. Hsu, T.-S. Wu, T.-S. Chan, N.-T. Suen, J.-F. Lee and H. M. Chen, Nat. Commun., 2023, 14, 6576.
255. A. Fontaine, E. Dartyge, J. P. Itie, A. Jucha, A. Polian, H. Tolentino and G. Tourillon, MRS Proc., 1988, 143, 121.
256. A. J. Dent, Top. Catal., 2002, 18, 27–35.
257. O. Müller, M. Nachtegaal, J. Just, D. Lützenkirchen-Hecht and R. Frahm, J. Synchrotron Radiat., 2016, 23, 260–266.
258. R. Frahm, T. Barbee Jr and W. Warburton, Phys. Rev. B: Condens. Matter Mater. Phys., 1991, 44, 2822.
259. D. LuÈtzenkirchen-Hecht, S. Grundmann and R. Frahm, J. Synchrotron Radiat., 2001, 8, 6–9.
260. S.-C. Lin, C.-C. Chang, S.-Y. Chiu, H.-T. Pai, T.-Y. Liao, C.-S. Hsu, W.-H. Chiang, M.-K. Tsai and H. M. Chen, Nat. Commun., 2020, 11, 3525.
261. C.-S. Hsu, J. Wang, Y.-C. Chu, J.-H. Chen, C.-Y. Chien, K.-H. Lin, L. D. Tsai, H.-C. Chen, Y.-F. Liao and N. Hiraoka, Nat. Commun., 2023, 14, 5245.
262. M. A. Newton, A. J. Dent and J. Evans, Chem. Soc. Rev., 2002, 31, 83–95.
263. T. Matsushita and R. P. Phizackerley, Jpn. J. Appl. Phys., 1981, 20, 2223.
264. S. Pascarelli, O. Mathon, M. Munoz, T. Mairs and J. Susini, J. Synchrotron Radiat., 2006, 13, 351–358.
265. S. Pascarelli, O. Mathon, T. Mairs, I. Kantor, G. Agostini, C. Strohm, S. Pasternak, F. Perrin, G. Berruyer and P. Chappelet, J. Synchrotron Radiat., 2016, 23, 353–368.
266. S. Diaz-Moreno, M. Amboage, M. Basham, R. Boada, N. E. Bricknell, G. Cibin, T. M. Cobb, J. Filik, A. Freeman and K. Geraki, J. Synchrotron Radiat., 2018, 25, 998–1009.
267. O. Mathon, I. Kantor and S. Pascarelli, X-Ray Absorption and X-Ray Emission Spectroscopy, 2016, pp. 185–212 DOI:10.1002/9781118844243.ch8.
268. A. Fontaine, E. Dartyge, J. P. Itie, A. Jucha, A. Polian, H. Tolentino and G. Tourillon, Time-resolved X-ray Absorption Spectroscopy Using an Energy Dispersive Optics: Strengths and Limitations, in Synchrotron Radiation in Chemistry and Biology III, ed. E. Mandelkov, De Gruyter, 1989.
269. M. A. Newton, Catalysts, 2017, 7, 58.
270. Q. Kong, F. Baudelet, J. Han, S. Chagnot, L. Barthe, J. Headspith, R. Goldsbrough, F. E. Picca and O. Spalla, Sci. Rep., 2012, 2, 1018.
271. W. Limphirat, N. Wiriya, S. Tonlublao, S. Chaichoy, P. Pruekthaisong, S. Duandmanee, P. Kamonpha, D. Kaewsuwan, N. Meethong, R. P. Poo-arporn, P. Songsiriritthigul, J. Hormes and Y. Poo-arporn, Radiat. Phys. Chem., 2020, 171, 108750.
272. S. Pascarelli and O. Mathon, Phys. Chem. Chem. Phys., 2010, 12, 5535–5546.
273. D. F. Sanchez, A. S. Simionovici, L. Lemelle, V. Cuartero, O. Mathon, S. Pascarelli, A. Bonnin, R. Shapiro, K. Konhauser, D. Grolimund and P. Bleuet, Sci. Rep., 2017, 7, 16453.
274. P. L. Lee, M. A. Beno, G. Jennings, M. Ramanathan, G. S. Knapp, K. Huang, J. Bai and P. A. Montano, Rev. Sci. Instrum., 1994, 65, 1–6.
275. P. Innocenzi, L. Malfatti, T. Kidchob, S. Costacurta, P. Falcaro, M. Piccinini, A. Marcelli, P. Morini, D. Sali and H. Amenitsch, J. Phys. Chem. C, 2007, 111, 5345–5350.
276. A. Marcelli, D. Hampai, W. Xu, L. Malfatti and P. Innocenzi, Acta Phys. Pol., A, 2009, 115, 489–500.
277. F. Meirer and B. M. Weckhuysen, Nat. Rev. Mater., 2018, 3, 324–340.
278. J. Wang, H.-Y. Tan, M.-Y. Qi, J.-Y. Li, Z.-R. Tang, N.-T. Suen, Y.-J. Xu and H. M. Chen, Chem. Soc. Rev., 2023, 52, 5013–5050.
279. M. A. Newton and A. J. Dent, In situ Characterization of Heterogeneous Catalysts, 2013, pp. 75–119 DOI:10.1002/9781118355923.ch3.
280. G. Agostini, D. Meira, M. Monte, H. Vitoux, A. Iglesias-Juez, M. Fernandez-Garcia, O. Mathon, F. Meunier, G. Berruyer and F. Perrin, J. Synchrotron Radiat., 2018, 25, 1745–1752.
281. M. A. Newton, M. Di Michiel, A. Kubacka and M. Fernández-García, J. Am. Chem. Soc., 2010, 132, 4540–4541.
282. A. Kubacka, A. Martínez-Arias, M. Fernández-García, M. Di Michiel and M. A. Newton, J. Catal., 2010, 270, 275–284.
283. D. Ferri, M. S. Kumar, R. Wirz, A. Eyssler, O. Korsak, P. Hug, A. Weidenkaff and M. A. Newton, Phys. Chem. Chem. Phys., 2010, 12, 5634–5646.
284. S. J. Figueroa and M. A. Newton, J. Catal., 2014, 312, 69–77.
285. E. Becker, P.-A. Carlsson, L. Kylhammar, M. A. Newton and M. Skoglundh, J. Phys. Chem. C, 2011, 115, 944–951.
286. E. K. Dann, E. K. Gibson, C. R. A. Catlow, V. Celorrio, P. Collier, T. Eralp, M. Amboage, C. Hardacre, C. Stere and A. Kroner, J. Catal., 2019, 373, 201–208.
287. H. Sun, C. Wang, S. Sun, A. T. Lopez, Y. Wang, J. Zeng, Z. Liu, Z. Yan, C. M. Parlett and C. Wu, Sep. Purif. Technol., 2022, 298, 121622.
288. M. Benfatto, S. Della Longa, E. Pace, G. Chillemi, C. Padrin, C. R. Natoli and N. Sanna, Comput. Phys. Commun., 2021, 265, 107992.
289. J. Timoshenko, A. Anspoks, A. Cintins, A. Kuzmin, J. Purans and A. I. Frenkel, Phys. Rev. Lett., 2018, 120, 225502.
290. S. Tetef, N. Govind and G. T. Seidler, Phys. Chem. Chem. Phys., 2021, 23, 23586–23601.
291. A. A. Guda, S. A. Guda, K. A. Lomachenko, M. A. Soldatov, I. A. Pankin, A. V. Soldatov, L. Braglia, A. L. Bugaev, A. Martini and M. Signorile, Catal. Today, 2019, 336, 3–21.
292. J. Timoshenko and A. I. Frenkel, ACS Catal., 2019, 9, 10192–10211.
293. F. Huber, S. van der Burg, J. J. van der Hooft and L. Ridder, J. Cheminf., 2021, 13, 84.
294. Z. Ghahramani, Nature, 2015, 521, 452–459.
295. M. I. Jordan and T. M. Mitchell, Science, 2015, 349, 255–260.
296. S. Ekins, A. C. Puhl, K. M. Zorn, T. R. Lane, D. P. Russo, J. J. Klein, A. J. Hickey and A. M. Clark, Nat. Mater., 2019, 18, 435–441.
297. K. A. Brown, S. Brittman, N. Maccaferri, D. Jariwala and U. Celano, Nano Lett., 2019, 20, 2–10.
298. T. Mueller, A. G. Kusne and R. Ramprasad, Rev. Comput. Chem., 2016, 29, 186–273.
299. E. Swann, B. Sun, D. Cleland and A. Barnard, Mol. Simul., 2018, 44, 905–920.
300. A. Martini and E. Borfecchia, Crystals, 2020, 10, 664.
301. M. Lerotic, R. Mak, S. Wirick, F. Meirer and C. Jacobsen, J. Synchrotron Radiat., 2014, 21, 1206–1212.
302. W. H. Cassinelli, L. Martins, A. R. Passos, S. H. Pulcinelli, C. V. Santilli, A. Rochet and V. Briois, Catal. Today, 2014, 229, 114–122.
303. D. K. Pappas, A. Martini, M. Dyballa, K. Kvande, S. Teketel, K. A. Lomachenko, R. Baran, P. Glatzel, B. Arstad and G. Berlier, J. Am. Chem. Soc., 2018, 140, 15270–15278.
304. A. Martini, E. Borfecchia, K. Lomachenko, I. Pankin, C. Negri, G. Berlier, P. Beato, H. Falsig, S. Bordiga and C. Lamberti, Chem. Sci., 2017, 8, 6836–6851.
305. S. Xiang, P. Huang, J. Li, Y. Liu, N. Marcella, P. K. Routh, G. Li and A. I. Frenkel, Phys. Chem. Chem. Phys., 2022, 24, 5116–5124.
306. I. Miyazato, L. Takahashi and K. Takahashi, Mol. Syst. Des. Eng., 2019, 4, 1014–1018.
307. C. Zheng, Machine Learning of Big Materials Data, UC San Diego, 2019. https://escholarship.org/uc/item/89j4z0hf.
308. S. Barocas, M. Hardt and A. Narayanan, Fairness and machine learning: Limitations and opportunities, MIT Press, 2023. https://fairmlbook.org/pdf/fairmlbook.pdf.
309. C. Zhang and X. Fu, Chin. Phys. B, 2023, 32, 126103.
310. S. B. Torrisi, M. R. Carbone, B. A. Rohr, J. H. Montoya, Y. Ha, J. Yano, S. K. Suram and L. Hung, npj Comput. Mater., 2020, 6, 109.
311. O. Trejo, A. L. Dadlani, F. De La Paz, S. Acharya, R. Kravec, D. Nordlund, R. Sarangi, F. B. Prinz, J. Torgersen and N. P. Dasgupta, Chem. Mater., 2019, 31, 8937–8947.
312. A. N. Zaloga, V. V. Stanovov, O. E. Bezrukova, P. S. Dubinin and I. S. Yakimov, Mater. Today Commun., 2020, 25, 101662.
313. C.-H. Liu, Y. Tao, D. Hsu, Q. Du and S. J. Billinge, Acta Crystallogr., Sect. A: Found. Adv., 2019, 75, 633–643.
314. A. Martini, D. Hursán, J. Timoshenko, M. Rüscher, F. Haase, C. Rettenmaier, E. Ortega, A. Etxebarria and B. Roldan Cuenya, J. Am. Chem. Soc., 2023, 145, 17351–17366.
315. A. A. Guda, S. A. Guda, A. Martini, A. N. Kravtsova, A. Algasov, A. Bugaev, S. P. Kubrin, L. V. Guda, P. Šot, J. A. van Bokhoven, C. Copéret and A. V. Soldatov, npj Comput. Mater., 2021, 7, 203.
316. A. Tereshchenko, D. Pashkov, A. Guda, S. Guda, Y. Rusalev and A. Soldatov, Molecules, 2022, 27, 357.
317. J. Timoshenko, A. Anspoks, A. Kalinko and A. Kuzmin, Phys. Status Solidi A, 2015, 212, 265–273.
318. J. Timoshenko, A. Kuzmin and J. Purans, Comput. Phys. Commun., 2012, 183, 1237–1245.
319. J. Timoshenko, A. Kuzmin and J. Purans, J. Phys.: Condens. Matter, 2014, 26, 055401.
320. P. K. Routh, N. Marcella and A. I. Frenkel, J. Phys. Chem. C, 2023, 127, 5653–5662.
321. N. F. de Jonge, K. Mildau, D. Meijer, J. J. Louwen, C. Bueschl, F. Huber and J. J. van der Hooft, Metabolomics, 2022, 18, 103.
322. P. K. Routh, Y. Liu, N. Marcella, B. Kozinsky and A. I. Frenkel, J. Phys. Chem. Lett., 2021, 12, 2086–2094.
323. Y. Liu, A. Halder, S. Seifert, N. Marcella, S. Vajda and A. I. Frenkel, ACS Appl. Mater. Interfaces, 2021, 13, 53363–53374.
324. A. Rocchetto, E. Grant, S. Strelchuk, G. Carleo and S. Severini, npj Quantum Inf., 2018, 4, 28.
325. S. K. Portillo, J. K. Parejko, J. R. Vergara and A. J. Connolly, Astron. J., 2020, 160, 45.
326. J. Timoshenko, D. Lu, Y. Lin and A. I. Frenkel, J. Phys. Chem. Lett., 2017, 8, 5091–5098.
327. T. Ronan, S. Anastasio, Z. Qi, P. H. S. V. Tavares, R. Sloutsky and K. M. Naegle, J. Mach. Learn. Res., 2018, 19, 1–6.
328. L. vd Maaten and G. Hinton, J. Mach. Learn. Res., 2008, 9, 2579–2605.
329. S. Madan, T. Henry, J. Dozier, H. Ho, N. Bhandari, T. Sasaki, F. Durand, H. Pfister and X. Boix, Nat. Mach. Intell., 2022, 4, 146–153.
330. J. Hong, E. Marceau, A. Y. Khodakov, L. Gaberová, A. Griboval-Constant, J.-S. Girardon, C. L. Fontaine and V. Briois, ACS Catal., 2015, 5, 1273–1282.
331. C. Ruckebusch and L. Blanchet, Anal. Chim. Acta, 2013, 765, 28–36.
332. X. Wang, J. C. Hanson, A. I. Frenkel, J.-Y. Kim and J. A. Rodriguez, J. Phys. Chem. B, 2004, 108, 13667–13673.
333. M. M. Bajomo, Y. Ju, J. Zhou, S. Elefterescu, C. Farr, Y. Zhao, O. Neumann, P. Nordlander, A. Patel and N. J. Halas, Proc. Natl. Acad. Sci. U. S. A., 2022, 119, e2211406119.
334. S. Tetef, V. Kashyap, W. M. Holden, A. Velian, N. Govind and G. T. Seidler, J. Phys. Chem. A, 2022, 126, 4862–4872.
335. M. R. Carbone, MRS Bull., 2022, 47, 968–974.
336. L. Ruthotto and E. Haber, GAMM-Mitt., 2021, 44, e202100008.
337. J. Timoshenko and B. Roldan Cuenya, Chem. Rev., 2020, 121, 882–961.
338. M. Sun, A. W. Dougherty, B. Huang, Y. Li and C. H. Yan, Adv. Energy Mater., 2020, 10, 1903949.
339. M. Sun, T. Wu, Y. Xue, A. W. Dougherty, B. Huang, Y. Li and C.-H. Yan, Nano Energy, 2019, 62, 754–763.
340. J. Timoshenko and A. I. Frenkel, ACS Catal., 2019, 9, 10192–10211.
341. J. R. Kitchin, Nat. Catal., 2018, 1, 230–232.
342. S. Gong, K. Yan, T. Xie, Y. Shao-Horn, R. Gomez-Bombarelli, S. Ji and J. C. Grossman, Sci. Adv., 2023, 9, eadi3245.

---