Although the X-ray absorption fine-structure (XAFS) technique has contributed enormously to our understanding of materials and phenomena, the ability to spatially resolve and identify local atomic and electronic structure on the micrometre and/or nanometre length scales with high sensitivity (for example at a sub-parts per billion level) is crucial to unravel macroscopic responses in heterogeneous materials. Magnetic domains, extremely diluted microstructures, nanoparticles and/or embedded agglomerates in amorphous systems are just a few examples in which ultrasmall features mediate changes, modifying their average properties and macroscopic behaviours. Accordingly, the addition of spatial resolving power to XAFS-related methods represents a key step to study local structural transformations influenced by heterogeneities, surfaces, grain boundaries, interfaces etc. in a wide range of disciplines from biomedicine to nanotechnology, chemistry and surface and environmental sciences.

Superior brilliant and low-emittance synchrotron sources (diffraction-limited storage rings; Eriksson et al., 2014), advanced nanofocusing optics (Ice et al., 2011; Sakdinawat & Attwood, 2010) and enhanced detection schemes (Ryan et al., 2010), as well as new imaging and improved microscopy approaches (Ade & Stoll, 2009), are key factors which have made it possible for spatially resolved XAFS to become an established technique. The combination of large penetration depths, high throughput and excellent tunability in terms of the probe (energy, spot size, photon flux, time structure, polarization etc.) with in situ or operando exploration of short-range ordering has also been an unprecedented asset at micrometre and nanometre lengths (Smit et al., 2008). Deeply buried polymeric microstructures (Takao et al., 2015), nano­scale magnetic phenomena (Nolting et al., 2000; Wang et al., 2015) and dynamics of catalytic solids and related nano­materials (Andrews & Weckhuysen, 2013) under industrially and environmentally relevant conditions are just a few scientific examples (Buurmans & Weckhuysen, 2012).

Complementing spatially averaged tools and electron-microscopy methods (for example electron energy loss spectroscopy and extended electron energy loss fine structure) that operate under ultrahigh vacuum with ultrathin sample thickness and greater beam damage, micro/nano-XAFS offers unique advantages for 2D/3D imaging of local atomic environments with lower detection limits (especially for Z > 20). It also produces maps of mixed-valence states or depth chemical analysis (Lüh et al., 2012) without sectioning large samples that might even be subject to realistic environments (Li, Meyer et al., 2015). However, in developing approaches for spatially resolved XAFS it has proven to be difficult to technically combine a fixed focal spot position, chromaticity of the focusing lens, angular stability of the double-crystal monochromator (DCM), fast energy scanning speeds, incident beam intensity monitoring, vibration-free operation, large k-space information and low radiation dose (Ziegler et al., 2008; Bertsch & Hunter, 2001). In addition, data collection may be affected locally by the highly focused X-ray beam due to sample instability (for example photoreduction) or local environment changes (for example temperature). Finally, the data analysis of spatially resolved XAFS may also become more complex because of mixing of crystallographic phases or structural disorder, large surface-to-volume ratios, nanoscale effects etc. (Kuzmin & Chaboy, 2014). Advanced simulation techniques are sometimes necessary to describe, for example, the structure and dynamics of nanoparticles, as well as to take into account the influence of nanosized effects (charge-transfer, vacancies etc.). The use of complementary techniques [for example transmission electron microscopy (TEM) and X-ray diffraction (XRD)] to extract reliable structural information might be required to properly process the data.

Several spatially resolved XAFS layouts are currently in operation (Holt et al., 2013; Karim et al., 2015). They can be classified as a function of the energy of the X-ray beam (soft or hard X-rays), the specific setup (surface- or bulk-sensitive), the nature of the sample (magnetic materials, organic or biological samples) and/or the signal that is detected (photoelectrons, characteristic X-rays or transmitted beam). For example, there are multiple schemes: scanning transmission X-ray microscopes (Watts & Ade, 2012), scanning X-ray fluorescence microscopes (Martínez-Criado et al., 2012), near-edge X-ray absorption fine-structure microscopes (Ade & Stoll, 2009), photoemission electron microscopes (He et al., 2011; Gu et al., 2014; Karim et al., 2015) and full-field transmission X-ray microscopes (Streubel et al., 2015; Meirer et al., 2011). However, the most common methods can be grouped into three classes: (i) full-field transmission X-ray microscopy (TXM), (ii) energy-dispersive EXAFS microscopy (EDXAM) and (iii) scanning X-ray microscopy (SXM).

TXM produces a chemical mapping by the rapid acquisition of a set of images recorded by scanning across an absorption edge. Usually, a large and incoherent X-ray beam is focused onto the sample with a condenser lens, which defines the field of view. According to the X-ray energy and focusing optics, the spot size can vary from tens of micrometres to a few micrometres in diameter. A Fresnel zone plate often collects the X-rays that are scattered and transmitted through the sample, which must have an appropriate thickness and uniformity. Thus, the objective lens, which determines the spatial resolution, forms a real-space intensity image on a CCD camera. When properly aligned, the image stack has an XAFS spectrum for each pixel.

EDXAM, on the other hand, is based on the diffraction of polychromatic X-rays from a large horizontal divergence source by a polychromator crystal (Katayama et al., 2015). With a large energy bandwidth, the beam is focused on the sample and then diverges towards a position-sensitive detector, which correlates position with energy. Thus, EDXAM acquires a full X-ray absorption fine-structure (EXAFS) transmission spectrum per pixel in a single shot, which makes the technique especially useful for time-resolved studies (Pascarelli et al., 2006). Although still not fast enough to obtain large fields of view (dwell time of >1 s per pixel), the focused beam (with a spot size of typically a few micrometres) is very stable because no movement of optical elements is required to collect the spectral data.

Finally, SXM acquires images on a pixel-by-pixel basis. The beam is focused to a very small spot size by the nanofocusing optics (preferably Kirkpatrick–Baez mirrors in a grazing-incidence configuration). This spot size ranges from hundreds to tens of nanometres and usually determines the resolution of the microscope. The sample is raster-scanned through the X-ray beam, while the intensity of the characteristic X-rays, photoelectrons and/or transmitted light is recorded at each sample position. When the transmitted X-ray intensity is collected using a point detector, this instrument is called a scanning transmission X-ray microscope. Based on the images generated by a contrast mechanism (for example X-ray fluorescence; XRF), spatially resolved XAFS spectra can be collected across the sample at individual points to extract information on valence states, symmetry, electronic states, coordination number, interatomic distances etc. (Segura-Ruiz et al., 2011). Alternatively, by acquiring images at different X-ray photon energies, full spectral information collected at every pixel can be used to image and distinguish between specific chemical species (Martínez-Criado et al., 2005). Micro-spectroscopy and spectro-microscopy are also terminologies that are commonly used to describe spatially resolved XAFS analysis. In micro-spectroscopy, a spectrum is collected with a micro/nanobeam for a specific volume element or region of interest defined by an optical, phase or chemical contrast image or an elemental map generated via XRF. In spectro-microscopy, on the other hand, full spectra are collected at each pixel and are then used to produce an image based on spatially variable chemical differences in the sample.

Although the acquisition time per image is considerably slower compared with TXM, SXM requires a lower radiation dose and has more flexibility in terms of field of view, sample environment (electrochemical or humidity-controlled cells, magnetic and electric fields, cryostats, micro-heaters, diamond anvil cells) and detection schemes (photoelectrons, XRF, visible light, transmitted X-rays, linear and circular magnetic dichroism). It also has the potential for ultrahigh-resolution (∼1 nm) imaging with ptychography, although this imaging technique is still under considerable development (Shapiro et al., 2014; Wise et al., 2016).

In both the SXM and TXM approaches the choice of the X-ray energy range plays a key role in the resolution limits, depth of foci, sample thicknesses and available sample environments (Winarski et al., 2012). The instruments are often grouped into hard (energy > 4 keV) and soft (energy < 4 keV) X-ray instruments. With depths of focus of tens of micrometres and lateral resolutions down to 25–50 nm (for example Kirk­patrick–Baez mirrors), the former can operate in air with focal lengths of centimetres, accommodating multiple sample environments (Martínez-Criado et al., 2013). On the other hand, the latter operate in vacuum or helium conditions and feature depths of focus below 10 µm, better lateral resolutions of about 12–40 nm and focal lengths in the millimetre range. Several comprehensive reviews have described the principles, capabilities and applications of these micro/nano-XAFS-related techniques (Suzuki & Terada, 2016; Fischer et al., 2010; Ade, 2002). A collection of papers also discuss progress in the field in areas such as focusing optics, data-collection modes, evolution and impacts on various research topics (Bertsch & Hunter, 2001; Hitchcock et al., 2005; Grunwaldt et al., 2005; Pascarelli et al., 2006; Sakdinawat & Attwood, 2010; Ice et al., 2011). Generally, X-ray spot sizes from micrometres down to tens of nanometres are produced by three types of focusing devices: refractive, reflective and diffractive optics. The choice depends on the X-ray wavelength range, the experimental setup (TXM or SXM) and the source characteristics. For details and up-to-date references on focusing methods and diffraction-limited optics for spectroscopy, the reader is referred to Simionovici et al. (2022). Finally, the EDXAM approach, in addition to the other spatially resolved XAFS schemes, was introduced by Newton and Dent in the book In-situ Characterization of Heterogeneous Catalysts (Newton & Dent, 2013).

To date, little information has been published concerning the experimental difficulties of conducting spatially resolved XAFS. Some major considerations should be carefully taken into account to obtain a highly stable operation and vibration-free design (Tucoulou et al., 2008; Kaznatcheev et al., 2013; Sung et al., 2015).

Concerning the X-ray source, combining the requirements for high spatial resolution, high photon flux and a large energy range calls for an undulator source (De Andrade et al., 2011). Since the focus size is roughly proportional to the square root of the emittance, operation of a micro/nano-XAFS beamline with a highly stable (less than microradians) and small beam is crucial in addition to full tunability over the chosen energy range (Weckert, 2015). To fulfil the conditions for EDXAM, on the other hand, the requirement to produce a sufficiently large energy bandpass after diffraction by the polychromator also demands an insertion device coupled with a horizontally focusing mirror to enhance the divergence to the required value. An incoming horizontal divergence of about 1 mrad is necessary at the level of the polychromator (Pascarelli et al., 2006).

The high heat load produced by these insertion devices must be effectively minimized on the optical elements. As is usual, unwanted high-energy photons at low energies can be suppressed by a mirror, whereas unwanted low-energy photons at high energies can be suppressed by absorption filters. Optimization of the positions, cooling systems and quality of all optical components (for example slope errors) is the key technical challenge (Siewert et al., 2014). When a long-length beamline design (100–200 m) is adopted to obtain large source-size demagnifications (Schroer et al., 2010; Somogyi et al., 2015; Martínez-Criado et al., 2016), extreme stability of the fixed-exit DCM becomes imperative. Thus, for routine nano-XAFS operations the DCM needs cutting-edge capabilities not only in terms of fixed exit (angular stability ≪ 1 µrad) but also of repeatability, thermal stability, speed and long-term stability (Yoder et al., 2011; Kristiansen et al., 2015). Microradian angular fluctuations can induce movements of the effective point source, resulting in deviations of the focused beam (Tucoulou et al., 2008). Furthermore, the limited beam acceptance of common nanofocusing optics also implies that the angular fluctuations can cause movements of the incident beam that turn into intensity fluctuations in the focused beam. Thus, an overillumination approach of the focusing optics can help to avoid large intensity changes. To further optimize the intensity variation of the incoming beam, the change in energy (i.e. the DCM) should be followed by the undulator gap in the scanned energy range, which can become very complicated due to speed-synchronization issues. Moreover, monitoring of the intensity before the sample at the same time as the transmitted beam or XRF signal also involves serious difficulties for the normalization of absorption spectra. Different strategies might be adopted: sequential data acquisition, collection of the XRF intensity from an element homogeneously distributed in the sample, scanning a slit over the beam fan after the polychromator etc.

Concerning the nanofocusing optics, the influence of thermal distortions, roughness (for example superpolished, <1 Å), fabrication slope errors (∼0.1 µrad r.m.s.) and figure errors (∼1 nm peak to valley) on the focusing properties and energy resolution is critical (Ziegler et al., 2008; Barrett et al., 2011). Efficient and high-quality achromatic 2D focusing lenses are essential to achieve large numerical aperture optics that simultaneously provide a large flux and a fixed and small focus (Matsuyama et al., 2006). The use of achromatic Kirkpatrick–Baez mirrors, for example, offers the possibility of measuring spatially resolved XAFS spectra of several elements in the very same sample spot without realignment of the experimental setup (Segura-Ruiz et al., 2011).

The nanopositioning required for 2D/3D chemical imaging also demands sample (piezo)stages with very high thermal/vibration stability, resolution, repeatability, low angular errors and ultrahigh precision (Zhu et al., 2015; Schroer & Falkenberg, 2014; Nazaretski et al., 2013). For tomographic acquisitions, in addition, a high-accuracy spindle (angular resolution of ∼10⁻³, eccentricity of less than micrometres, wobble of less than microradians) must be anticipated. Online metrology can also be crucial to reach ultrahigh-precision positioning (error compensation using laser interferometers/capacitive sensors to measure online angular errors and introduce online compensation, vibration-control techniques etc.). To speed up data collection and to reduce possible beam-induced sample damage, the development of fast-scanning strategies (degrees per second) is highly desirable, such as, as mentioned, quick energy scans involving continuous movement of the DCM with parallel tuning of the undulator gap (Müller et al., 2016).

The detectors are also one of the keystones of the micro/nano-XAFS experiment itself (Heald, 2015). Depending on the detection energy range, several options are commercially available for XRF analysis (Letard et al., 2006). Multi-element silicon drift detectors (SDDs) are commonly used to quantitatively measure XRF signals from samples at energies from 2 keV to ∼25 keV, at which the detector efficiency falls to ∼20% due to the limited silicon X-ray absorption; high-energy measurements are preferentially obtained with a germanium diode array. These systems reach spectral energy resolutions down to about 140 eV FWHM at the Mn Kα line (5.9 keV) but featured a limited count rate (up to 4 × 10⁶ counts per second), involving fairly long dwell times per point/pixel or point/energy (i.e. 0.1–1 s). Faster performance energy-resolving detectors for XRF-based XAS acquisitions with a larger solid angle, a higher count rate and higher efficiency are under development. The Maia-384 massively parallel detector is a good example (Ryan et al., 2010). The system consists of an annular array of 384 silicon diode detectors positioned in a back-scattering geometry with respect to the incident X-ray micro/nanobeam; this geometry ensures that a large solid angle is subtended (approximately 1.3 sr). An approach based on an event-mode data collection with real-time processing capabilities allows the recording of maps with a minimum dwell time down to ∼0.05 ms per pixel and a total count-rate capacity greater than 10⁶ s⁻¹ while avoiding readout overheads. A real-time full spectra elemental deconvolution can be obtained through an integrated algorithm based on a matrix-transform method called dynamic analysis (Ryan et al., 2010). However, the Maia detectors suffer from higher scatter and backgrounds, difficulty in positioning visible optics, poor low-energy performance and an energy resolution that is not comparable to SDDs, as well as working-distance limitations.

EDXAM acquisitions in transmission mode, on the other hand, are presently based on a 2D CCD camera exploited in a 1D fashion (Labiche et al., 2007; Salvini et al., 2005). The general requirements include a pixel size of <50 µm for energy-resolution purposes, a pixel array of >2048 to span the required energy range, a large dynamic range (10⁴–10⁵), a fast readout time (<1 µs) and very low noise. For time-resolved experiments on the millisecond time scale, a strip detector based on germanium technology is a good choice. An XSTRIP for high energies has been developed specifically for EDXAM on third-generation synchrotrons (Headspith et al., 2007). The system collects data at high energies with good efficiency (>90% absorption at 40 keV), spatial resolution and radiation-damage tolerance. A key element of this system is the 1024-element germanium microstrip sensor, fabricated using amorphous germanium (a-Ge) contact technology, featuring a strip pitch of 50 µm. This sensor coupled to the STFC-designed X2CHIP ASIC readout electronics provides an excellent linearity (better than 0.1%) and ultrafast readout performance (100 kHz frame rate).

Spatially resolved XAFS is an evolving field that has unique attributes and is expanding in a number of new and exciting directions. Many full-field transmission X-ray microscopes, energy-dispersive EXAFS microscopes and scanning X-ray microscopes are available today and additional instruments are planned or under construction at third-generation synchrotron-radiation sources (Pascarelli et al., 2006; Schroer et al., 2010; Somogyi et al., 2015; Martínez-Criado et al., 2016). By pushing the spatial resolution, new research strategies exploiting spatially resolved XAFS are emerging: combined data-collection modalities (for example beam-induced current, X-ray-excited optical luminescence), novel spectroscopic concepts (for example ptychography; Shapiro et al., 2014; Wise et al., 2016) and new imaging and microscopy approaches (microscopes at long beamlines, cryomicroscopy, XANES tomography and resonant XRD). Ptychography, for example, combines point scanning with 2D detection of the coherently scattered X-rays. Using a coherent X-ray beam, a sample is illuminated with overlapping spots, overcoming the phase problem. Thus, extended structures can be imaged with elemental and chemical contrast mechanisms (Shapiro et al., 2014). Another promising nanoscale imaging technique of buried structures with elemental specificity is resonant X-ray diffraction microscopy (Song et al., 2008). Different diffraction patterns are acquired near the absorption edge of a specific element and are directly phased to obtain high-resolution chemical state maps. The ultimate resolution of the microscope is limited only by the X-ray wavelengths and can in principle reach the atomic level. Tomographic methods, as discussed, are also promising for 3D imaging of chemical phase transformation at the nanoscale (Andrews & Weckhuysen, 2013; Simionovici et al., 2008).

Future spatially resolved XAFS investigations can also be anticipated under extreme conditions of pressure (>200 GPa), temperature (2–4000 K) and pulse magnetic fields (30 T) by in situ static or dynamic EDXAM approaches, as well as ultrafast processes in a pump–probe scheme. Further developments will emerge in differential EDXAM for quantitative estimation of femtometre-scale atomic displacements in strain-related phenomena such as magnetostriction (Pettifer et al., 2005) or chemical variations using SXM with atomic detection capabilities and/or internal 3D local information with sub-parts per billion sensitivity under controlled sample environments (pH, operando, magnetic and electric fields).

Efforts are also under way to achieve better spatial resolutions, versatile chemical contrast mechanisms and high-throughput and high-energy spatially resolved XAFS analysis. Forthcoming challenges and highly promising developments comprise combined atomic force microscopy (AFM)/scanning tunnelling microscopy (STM)/SXM methods (for example the nanoXAS instrument; Watts & Ade, 2012; Cummings et al., 2012; Pilet et al., 2012), as well as complementary instrumentation integration within hybrid techniques (for example confocal chemical imaging, scanning electron microscopy (SEM)/TEM/SXM, nano-XAFS-SNOM; Takao et al., 2015; Ade & Stoll, 2009; Larcheri et al., 2008). When coupled with the polarization state of the photons, X-ray magnetic circular dichroism (XMCD)–SXM also characterizes bond orientation and magnetic domain structure.

Future impact on spatially resolved XAFS can be also expected from the even higher photon flux, higher brightness and coherence of the synchrotron radiation at future diffraction-limited storage rings and free-electron lasers (FELs; McNeil & Thompson, 2010; Eriksson et al., 2014). Beyond synchrotron radiation, with pulses of about 200 fs in length, FELs provide a peak brilliance 10¹⁰ higher than the best synchrotron light sources and photon pulses of 10¹² photons per femtosecond. Thus, a step forward would be 4D local structural monitoring capabilities (chemical imaging with space, time and energy resolutions) at future FEL facilities to access ultrafast phenomena (kinetics, photocatalysis, rapid fluctuations and chemical reactions).

However, to allow such breakthroughs key instrumental and methodological improvements need to be fostered in the following areas: (i) nanofocusing optics, and more generally optical elements, that are able to provide stable and reliable beams on the nanometre length scale, (ii) nanocharacterization tools to monitor nanodisplacements of samples and mechanical control and stability (piezo, feedback etc.), (iii) sample preparation and nanolocalization or nanomarking protocols to precisely record the areas of interest and relate them to observations on a larger length scale, (iv) low-noise, large solid angle and ultrafast detectors and electronics (data readout), (v) further cryo-techniques to reduce radiation damage and (vi) in situ measurements under monitored environmental conditions.