Catalysis is the branch of science aimed at (i) increasing the rate of a chemical reaction A + B → C (typically by finding a new path involving new intermediates characterized by a lower activation energy) and (ii) improving the selectivity towards the desired product C (usually, A + B may yield different products C, D, E … with different relative ratios; Chorkendorff & Niemantsverdriet, 2007; Norskov et al., 2014). This goal is achieved by inserting a substance called a `catalyst' into the reaction environment. Usually the catalyst undergoes multiple chemical transformations, but at the end of the cycle, unlike the reactants (A, B), it remains unchanged. Hence, the catalyst is not consumed and it can continue converting reactants A + B into C indefinitely. This chapter will describe how X-ray absorption spectroscopy (XAS) and X-ray emission spectroscopy (XES) can contribute to understanding the structure of the catalytic site and its modification throughout the whole catalytic cycle at the molecular level.

In most cases, catalysts do not exhibit long-range order. This is obviously the case for homogeneous catalysis (Tromp, 2015), where the reactants, products and catalysts are in the liquid phase (Chadwick et al., 2011; although much less common, gas–gas or solid–solid cases are also included in this definition). For heterogeneous catalysis (Bordiga et al., 2013), in which the reaction is confined to the interface between the catalyst (a solid) and the reactants either in the liquid or the gaseous phase, the lack of long-range order still holds in most cases. In fact, in order to maximize such an interface, most of the solid supports are high surface-area materials (10²–10³ m² g⁻¹) that exhibit poor crystallinity [for example γ-Al₂O₃ (Muddada et al., 2011), activated carbon (Furimsky, 2008; Lazzarini et al., 2016; Pellegrini et al., 2009; Serp & Figueiredo, 2009) and polymers (Bekturov & Kudaibergenov, 2002; Groppo et al., 2010)] or amorphous metal oxides (for example silica; Groppo, Lamberti et al., 2005). This implies that X-ray diffraction (XRD) usually cannot be used to determine the structure of the catalytically active centres, which are typically represented by isolated transition-metal ions (single-site catalysts) or dispersed metal nanoparticles. This makes extended X-ray absorption fine structure (EXAFS) the most suitable characterization technique to understand the structure of the catalytic sites (Bonino et al., 2015; Bordiga et al., 2013; Evans, 1997; Frenkel, 2012; Iwasawa et al., 2017; Lamberti & van Bokhoven, 2016; Newton, 2008; Prins & Koningsberger, 1988), while X-ray absorption near-edge structure (XANES; Fernández-García, 2002; Guda et al., 2015, 2016; van Bokhoven & Lamberti, 2014) and XES (de Groot, 2001; Glatzel & Bergmann, 2005; Singh et al., 2010) are informative on their electronic structure. Exceptions to this statement are made for three main classes of crystalline microporous catalysts, in which XRD has played an important role: zeolites, which are widely used industrially (Agostini et al., 2010; Milanesio et al., 2003), metal–organic frameworks (MOFs), which are of potential interest in the near future (Bordiga et al., 2010; Borfecchia et al., 2017; Butova et al., 2016; Corma et al., 2010), and, partially, metal nanoparticles, which exhibit poor crystallinity. Nonetheless, for crystalline or partially crystalline catalysts, a better understanding of the material was generally achieved when XRD data were complemented with an XAS study (Castillejos-López et al., 2017; Clausen et al., 1998; Martin et al., 2016; Valenzano et al., 2011).

The great benefit of the application of XAS and XES to catalytic studies is related to the high penetration depth of hard X-rays, which allows measurements on catalysts under operation conditions, i.e. in the presence of reactants and products from the gas or liquid phases. This requires specific experimental setups, as described in Agostini et al. (2023). Due to space limitations, hereafter we will focus on applications of XAS/XES in heterogeneous catalysis, reporting representative examples of single-site catalysts and metal nanoparticle-based catalysts; this selection is based on the experience of the authors and is far from being exhaustive.

Supported metal nanoparticle systems were among the very first catalysts to be investigated using the XAFS technique by Lytle himself in collaboration with researchers from Exxon (Sinfelt et al., 1978; Via et al., 1979). Particle shape and size are the most relevant parameters in determining the adsorption and reactivity properties of metal nanoparticles; therefore, a huge effort has been devoted to developing experimental tools that are able to provide statistically significant information on these two activity-determining properties. Atoms at the surface of nanoparticles exhibit a lower coordination number than atoms in the bulk. As a consequence, by experimentally determining the coordination numbers of the different shells with EXAFS, it is possible to estimate the average particle size assuming a certain particle shape (Agostini et al., 2009; Agostini, Piovano et al., 2014; Bezemer et al., 2006; Frenkel et al., 2001; Karim et al., 2009; Miller et al., 2006; Mostafa et al., 2010; Newton, 2008; Witkowska et al., 2007; see Fig. 6a). In this regard, the main strengths of EXAFS include (i) the ability to probe all metal atoms crossed by the beam (overcoming the limitation of XRPD, which only detects particles that are sufficiently large to give Bragg diffraction) and (ii) the large number of particles probed [overcoming the intrinsic statistical weakness of transmission electron microscopy (TEM) studies]. However, only average values can be obtained by EXAFS and therefore data analysis should be performed carefully and the EXAFS results should be compared with those obtained from other independent techniques such as XRPD, TEM, scanning TEM (STEM), small-angle X-ray scattering (SAXS), chemisorption, total X-ray or neutron scattering (Agostini, Lamberti et al., 2014; Agostini et al., 2009; Agostini, Piovano et al., 2014; Briggs et al., 2015; Cuenya et al., 2010; Deganello et al., 2006; Groppo, Agostini et al., 2015; Groppo et al., 2012, 2014; Lazzarini et al., 2017; Pellegrini et al., 2011) or supported by theoretical modelling (Bugaev et al., 2017; Chill et al., 2015).

Figure 6 (a) Comparison of the average coordination numbers of the first four shells for clusters with different shapes, cuboctahedron (CO), truncated cuboctahedron (TCO) and icosahedron (ICO), shown as solid, dashed and dotted lines, respectively. The data are plotted in terms of order m (bottom abscissa axis) and of the particle diameter in the case of palladium nanoparticles (top abscissa axis) with cell parameter 3.889 Å and first-shell Pd–Pd distance 2.75 Å. Reproduced with permission from Agostini, Piovano et al. (2014); copyright 2014 American Chemical Society. (b) Determination of the model cluster sizes (shown for two different types of clusters) using the H:Pd ratio obtained from the measured Pd–Pd distances. The (111) model fits the data (shown by red arrows) for both the H:Pd ratio at 186 K and the particle size range, while the (001) model does not. Reproduced with permission from Wang et al. (2015); copyright 2015 American Chemical Society. (c) Experimental Pd K-edge XANES spectra (left ordinate axis) and Δ-XANES (right ordinate axis) collected at 100°C on palladium nanoparticles in vacuum (black), in 100 mbar H2 (violet) and in 1000 mbar acetylene (orange). (d) As in (c) for a mixture of 650 mbar H2 and 350 mbar acetylene (red) and in 600 mbar H2 (blue). (e) Theoretical Pd K-edge Δ-XANES spectra for the PdHx phase in the 0 ≤ x ≤ 0.5 stoichiometry interval. (f) As in (e) for the PdCy phase in the 0 ≤ y ≤ 0.2 stoichiometry interval. Reproduced with permission from Bugaev et al. (2017) with permission from Elsevier.

Through an accurate and unambiguous determination of the coordination number for the first four shells, interpretation of the EXAFS data collected on metal nanoparticles allows determination of both the particle size and shape (Agostini et al., 2009; Frenkel et al., 2011; see Fig. 6a). This however holds for ideal cases only, i.e. for low-temperature data collections, narrow particle size and shape distributions, and symmetric bond-length distributions. The analysis becomes delicate when the catalysts are investigated under reaction conditions at high temperature and high reactant pressure. Further complications arise when the particles undergo modifications along the reaction such as changes in (i) size (sintering or disaggregation), (ii) shape or (iii) chemical composition (metal/oxide, metal/hydride or metal/carbide transformation). Hereafter, we will briefly discuss the case study of palladium-based catalysts, the active phase of which is subjected to chemical changes under reaction conditions.

Supported palladium catalysts are among the most used systems for the hydrogenation of hydrocarbons, such as alkynes and alkenes. Under reaction conditions, palladium nanoparticles may undergo phase changes to hydride and carbide phases, the nature of which affects the catalytic properties (Teschner et al., 2008, 2010). Therefore, determining hydride and carbide formation during a catalytic process becomes an important problem that is also relevant to industry. Being the subject of numerous theoretical (Shabaev et al., 2010; Zhdanov & Kasemo, 2008) and experimental (Bugaev et al., 2014; Davis et al., 1989; Flanagan & Oates, 1991; Lewis, 1982; Mitsui et al., 2003; Shegai & Langhammer, 2011; Soldatov et al., 1993; Tew et al., 2009) studies, palladium hydride is among the most-studied metal hydrides.

The in situ formation of the PdH_x phase in palladium nanoparticles supported on θ-Al₂O₃ has been investigated by in situ EXAFS in the 186–483 K range (Wang et al., 2015). The experiments were performed in both hydrogen and helium atmospheres in order to disentangle thermal effects from modifications due to hydride formation in the first-shell EXAFS signal. By comparing the two data sets, the authors were able to isolate the effects of both the finite particle size and hydrogen absorption on the Pd—Pd bond length, to obtain hydrogen-intake values and to estimate the particle size, which agrees with independent ex situ STEM analysis. In more detail, they constructed two families of models of truncated hemispherical f.c.c. cuboctahedra of increasing size, with the support surface having a (001) or (111) orientation. The values of the H:Pd ratios were calculated for each cluster model (see Fig. 6b). Calculations demonstrate that the estimated H:Pd ratio will be 0.52 for the (111)-oriented clusters and 0.65 for the (001)-oriented clusters of ∼2.1 nm in size. The latter value is far beyond the experimental value of 0.54 ± 0.02 at 186 K. Hence, the (111)-oriented clusters better match the experimental data. As a result, the particle size of 2.1 ± 0.4 nm obtained by STEM is in agreement with the particle size inferred from the experimentally determined H:Pd ratio of 0.54 ± 0.02. Subsequently, the experimental H:Pd ratio determined by expansion of the bond length was used to estimate the particle size by occupancy model calculation. It should be noted that this method is only effective for particle sizes below 3 nm because the H:Pd ratio varies strongly with size in this size range.

In contrast to the hydride, the structure and properties of the carbide phase are still under discussion (McCaulley, 1993; Teschner et al., 2008; Tew et al., 2011). The formation of both the hydride and carbide phases is accompanied by an expansion of the palladium lattice, which can be followed by EXAFS, as discussed above, or by XRPD (Suleiman et al., 2003; Vogel et al., 2010). However, both techniques are scarcely sensitive to light atoms, such as carbon and especially hydrogen, due to their low scattering amplitudes compared with palladium. Thus, PdH_x and PdC_y phases are only indirectly observed by EXAFS and XRPD via Pd–Pd distance elongation or lattice expansion. During hydrogenation reactions, both H₂ and alkynes (or alkenes) are fed to the catalyst so that PdH_x, PdC_y and PdH_xC_y phases could potentially be formed. Since all of them result in an elongation of the Pd–Pd distance, it will be very difficult to discriminate among the different phases using EXAFS or XRPD. Conversely, the formation of palladium hydride or carbide directly affects the shape of the Pd K-edge XANES spectra due to mixing of unoccupied states of hydrogen or carbon and palladium. As the mixing is different in the two cases, the perturbation of the XANES spectra upon hydride formation is different from that induced by carbide (Fig. 6c), making XANES the technique of choice to distinguish the two cases under operando conditions (Bugaev et al., 2017). In particular, XANES was able to prove that on feeding the palladium catalyst at 100°C with 350 mbar H₂ and 650 mbar acetylene the active phase is a PdH_x phase (see Fig. 6d).

To quantify the PdH_x and PdC_y stoichiometry, the Rostov-on-Don group performed XANES calculations applying Monte Carlo simulations of H- and C-atom occupancy in the octahedral interstitials of the palladium f.c.c. lattice (Bugaev et al., 2017). For each of the selected concentrations of x or y in PdH_x or PdC_y, they generated 1000 different geometries and averaged the resulting 1000 XANES spectra, which were subsequently used for the quantitative fitting of experimental XANES. Fig. 6(e) shows the evolution of the theoretical difference XANES spectra for the PdH_x phase in the 0 ≤ x ≤ 0.5 range; Fig. 6(f) reports the analogous difference for the PdC_y phase in the 0 ≤ y ≤ 0.2 interval. The fitting procedure was performed using a multidimensional interpolation approach implemented in the FitIt-3 code (Smolentsev & Soldatov, 2006, 2007, 2021). In particular, the resulting stoichiometry of the active phase on the catalyst exposed to 100 and 600 mbar H₂ at 100°C was PdH_0.20±0.05 and PdH_0.30±0.05, respectively. Analogously, the stoichiometry of the carbide phase obtained by exposing the catalyst to 1000 mbar C₂H₂ at 100°C was PdC_0.13±0.05. It is clear that the results reviewed here (Fig. 6) open up new opportunities for in situ XAFS investigations of palladium nanoparticle-based catalysts under reaction conditions where catalyst fragmentation, shape change and/or stoichiometry change may occur.

In this chapter, we have provided the necessary basis to understand the relevant role that XAS (and more recently XES) has played in disclosing the structural changes that are undergone by the active site of a working catalyst at the molecular level. Selected examples include copper zeolites for NH₃-SCR and copper-functionalized polymers, the Phillips polymerization catalyst and supported metal nanoparticles.

We foresee that remarkable advances in the use of hard X-ray spectroscopy techniques in catalysis will be available in the future; the principal advances are listed below.

(i) The further improvement and more extended application of modulation excitation spectroscopy, which will potentially allow the selective detection of minority species that are sensitive to an external stimulus (Chiarello & Ferri, 2015), thus allowing the discrimination of active sites from spectators.

(ii) Experimental setups allowing parallel IR, UV–Vis and Raman data to be collected simultaneously with XAS/XES data are expected to be a key area of technical development (Liu et al., 2017).

(iii) Space-resolved (tomographic) techniques will provide a precise three-dimensional insight into the whole catalytic bed (hosted inside a capillary) and the individual catalyst grains to investigate effects such as the change in the reactants:products ratio occurring along the catalytic bed (Beale et al., 2010; Buurmans & Weckhuysen, 2012; Grunwaldt & Schroer, 2010; Kalirai et al., 2016).

(iv) The realization of new and improved secondary emission spectrometers at various beamlines (Kvashnina & Scheinost, 2016) will allow oxidation state-specific EXAFS (to obtain separate EXAFS signals in samples containing the same element in different oxidation states; de Groot, 2000) and spin-selective EXAFS spectra collection (Glatzel et al., 2002).

(v) Hard X-ray Raman scattering will allow the collection of XAS-like signal from relevant light atoms such as, for exanple, C, O and N under catalytic working conditions (Bergmann et al., 2000; Lamberti et al., 2016; Mino et al., 2013).

(vi) Laser pump/X-ray probe experiments, which to date have mainly been applied in studies related to photoinduced structural dynamics, may be employed to investigate photocatalysts, clarifying the structural and electronic rearrangements of the photocatalytic site just after (visible) photon absorption or other external stimuli (Borfecchia et al., 2013; Chen, 2016; Chergui, 2015). Furthermore, in the immediate future, X-ray free-electron laser (XFEL) sources will revolutionize the physics and chemistry of time-resolved experiments (Gawelda et al., 2016).

(vii) The simulation of XANES and XES spectra will be used more and more frequently to confirm or discard local structures hypothesized from the refinement of EXAFS or diffraction data (Borfecchia et al., 2012; Gallo et al., 2014; Guda et al., 2015; Joly & Grenier, 2016; Regli et al., 2007).

(viii) Joint EXAFS/diffraction anomalous fine-structure (DAFS) studies will allows the exploration of biphasic systems and, for example, the contributions from two co-existing crystalline phases or from co-existing amorphous and crystalline phases to be disentangled (Groppo, Agostini et al., 2015; Groppo et al., 2014).

(ix) As far as metal nanoparticles are concerned, total scattering (PDF) experiments (Bozin et al., 2013) will be able to bridge the gap between EXAFS, dominating the 0–30 Å diameter interval, and XRPD, which is informative in the 80 Å to bulk range (Newton et al., 2012).

(x) X-ray magnetic circular dichroism (X-MCD; Rogalev et al., 2016), coupled with more conventional visible-light MCD and electron paramagnetic resonance (EPR), will provide new insights in the investigation of transition metal-supported catalysts (Roa et al., 2010) and in biocatalysis (Staniland et al., 2007).