The use of soft X-rays brings its own inconveniences and complexities to an X-ray absorption fine-structure (XAFS) experiment. Firstly, air absorption at low photon energies is severe, and the sample and associated detector apparatus are therefore typically encased in an ultrahigh-vacuum (UHV) chamber. In addition, as shown in Fig. 1, the transmission of soft X-rays through beryllium (and other window materials that are typically used to isolate detectors from the sample environment; Henke et al., 1993; Perkins et al., 1990) falls off rapidly below 1000 eV, so the beamline and experimental chamber often share vacuum with the storage ring and the typical safety issues associated with it. These considerations are certainly true for the low-Z elements of the second row of the periodic table and their K-shell energies, boron (188 eV), carbon (284 eV), nitrogen (410 eV), oxygen (543 eV) and fluorine (697 eV), whereas the higher (or `tender') energy K edges of the third-row elements phosphorus (2149 eV), sulfur (2472 eV) and above may be studied with beryllium windows and a helium beam path, although both contribute to the slope of the background; i.e. increasing flux and transmission with increasing energy. In addition, the lower soft-energy edges can only be accessed with a grating monochromator, whereas the higher tender-energy edges may be accessed with a Si(111) double-crystal monochromator that typically has 2140 eV as its lower cutoff energy. [The Si K edge (1839 eV), which is also in the tender-energy range, is typically accessed with a double-crystal monochromator operating with InSb(111) crystals (Hussain et al., 1986; Ohta et al., 1986), while the Al K edge (1559 eV) must be assessed with either a grating monochromator or quartz or beryl crystals in a double-crystal monochromator.]

Figure 1 X-ray transmission as a function of photon energy through polypropylene (polyp.), aluminium, boron nitride, silicon nitride and beryllium. The thicknesses for each correspond to the thicknesses typically used in detector windows. Note the large dips in transmission at the absorption-edge energies of the constituent atoms, which can lead to `colouring' of the transmitted X-rays.

In addition to the fundamental importance of the soft-energy K edges, experimental work studying the low-energy L edges of the fourth-row transition metals scandium through copper, with L-edge energies in the range of approximately 400–1100 eV, is also routinely performed to address both their linear and circular-dichroic signals. These investigations provide information on the orientation of the atoms relative to the polarization axis of the synchrotron beam (linear dichroism) in geometries that differ from cubic symmetry and information on electron spin and angular momentum (circular dichroism) through the different quantum-mechanical couplings of the dipole matrix element to the synchrotron-beam polarization (Thole et al., 1992). In addition to being able to isolate non-dipole (quadrupole) transitions from dipole transitions in cubic crystals (Brouder, 1990), linear dichroism is useful for studying antiferromagnetism and circular dichroism is useful for studying ferromagnetism (Thole et al., 1985, 1992; van der Laan & Thole, 1991; Stöhr et al., 1999).

In a typical UHV experiment, the samples are either powders that have been pressed into pellets, supported single crystals or thin films deposited on substrates. Samples are introduced into the UHV chamber from the laboratory atmosphere via a sample load lock or, as in the early days of traditional surface science, a single-crystal sample can be installed on the sample manipulator and `baked' together with the experimental chamber. An example of a modern chamber optimized for rapid sample introduction and low-energy XAFS is shown in Fig. 2. It consists of a long-travel sample manipulator, an upper sample-introduction preparatory chamber and a lower UHV analysis chamber. The manipulator is equipped with either a multi-sample bar for studying a large number of samples concurrently that have been prepared ex situ or a bolt-on, single-sample stage that has both heating and liquid-nitrogen cooling capabilities. The top preparatory chamber houses an ion gun and a variety of gas lecture bottles for ion sputtering and gas dosing. The lower analysis or `main' chamber houses a hemispherical electron analyzer for X-ray photoelectron spectroscopy (XPS) measurements and energy-discriminated partial-electron yield XAFS measurements. The I_o section that houses in-vacuum high-precision horizontal and vertical slits for defining the incident X-ray beam and a grid/channeltron/drain current I_o monitor is directly coupled to the main chamber. There is also a total electron-yield channeltron detector for total electron-yield detection and an energy resolving seven-element germanium fluorescence detector, both of which are housed in the analysis chamber and are dedicated to XAFS measurements. The analysis chamber is also fitted with reverse-view low-energy electron-diffraction (LEED) optics for sample surface-structure characterization and a low-energy electron `flood' gun to mitigate charging of insulating samples. The manipulator has the capability to either electrically bias the sample or collect the sample drain current directly.

Figure 2 Typical ultrahigh-vacuum chamber designed for NEXAFS and HAXPES measurements. The chamber is equipped with a hemispherical multi-channel electron analyzer (left), a seven-element germanium detector with polymer window (back left) and a long-travel sample manipulator and pass-through fast load–lock system with its own dedicated turbomolecular pump (top). A large ion pump with a liquid nitrogen-cooled titanium sublimation pump is on the lower right and services the main chamber.

In order for windowless operation to the synchrotron, the chamber must be maintained at UHV or at least be differentially pumped. UHV is typically associated with a base pressure better than 10⁻⁷ Pa or 10⁻⁹ Torr. UHV is a necessity for surface-science experiments, where the surface and its interaction with a chosen adsorbate is the focus of the experiment. For example, assuming a unity sticking coefficient at a base pressure of 10⁻⁶ Torr, one monolayer of contaminant will condense on a clean surface in 1 s (this is the definition of the Langmuir unit). Consequently, even for the best attainable pressures in the low 10⁻¹¹ Torr range, a reactive surface will become contaminated in a short 8 h shift at a synchrotron. The unfettered transport of electrons through vacuum also requires UHV conditions, hence the mandated low operating base pressure of the synchrotron itself, although significant progress has been made towards operando/ambient pressure conditions (see, for example, Liu & Bluhm, 2016).

To achieve UHV, a typical stainless-steel vacuum vessel is `baked' to temperatures exceeding 100°C. This baking desorbs impurity gases, typically water and hydrocarbons, that are adsorbed on the walls of the vessel, facilitating their pumping by the numerous pumps found in the system. The system shown in Fig. 2 has several oil-free roughing/backing pumps, a large turbomolecular pump directly attached to the preparation chamber and a large ion pump, together with a large turbomolecular pump and a titanium sublimation/liquid-nitrogen cryogenic gettering pump directly attached to the analysis chamber. Several other forms of cryogenic and gettering pumps are available commercially.

Materials that have significant vapour pressure should not be introduced into the vacuum system, as these materials will significantly degrade the pressure, often irreversibly. Additionally, the materials used for UHV must be thoroughly cleaned and degreased by a judicious choice of chemical solvents in an ultrasonic cleaner and handled with clean gloves in a clean room or glovebox environment so as to not recontaminate them. Samples that need to be heated in situ as part of the experimental design must also be thoroughly outgassed, along with the sample heater/holder, before commencement of the experiment. This is achieved by bringing the sample to, and holding it at, elevated temperature for several hours in order to drive off (or out) water and other unwanted contaminants. This process serves to minimize pressure spikes during the course of the experiment that would otherwise either degrade the base pressure of the chamber or contaminate a clean surface in a surface-science experiment.

In addition to the somewhat standard list of detectors, there have been several recent positive developments in detector design. For example, windowless ion chambers (Samson, 1967) can be directly attached to the beamline exit flange and have been developed for use in the low to tender X-ray energy range (Rothe et al., 2012). The silicon drift detector (SDD; Gatti & Rehak, 1984), which may be fitted with a brazed beryllium window to mitigate helium diffusion when used in a helium atmosphere, has demonstrated significant improvement over traditional energy-dispersive lithium drifted silicon and germanium detectors (SiLi and GeLi). An increase of several orders of magnitude in count rate has been documented without the traditional degradation of resolution (Woicik et al., 2010).

Completely new detector schemes have also been introduced to XAFS, such as single-photon, transition-edge sensors (TES; Andrews et al., 1942). These detectors determine the energy of an absorbed photon by the resulting temperature change of a superconducting thin film, and they produce superior energy resolution (of the order of 1–2 eV), allowing chemical resolution of the fluorescence channel with virtually zero dark current (Uhlig et al., 2015). Additionally, the original high-resolution wavelength-dispersive fluorescence detector based on a Rowland circle curved crystal design (Brennan et al., 1989) has been upgraded with a position-sensitive charge-coupled device (CCD) camera (Rothe et al., 2012). Such high-resolution detector schemes are indispensable for the lowest Z measurements, where an energy resolution better than 50 eV is required to discriminate between the fluorescence energies of adjacent atoms in the periodic table. We should also note that the advent of third- and higher generation synchrotron sources mitigates the relatively poor solid-angle acceptance of such detectors as well as, in general, the diminishing fluorescence to Auger branching ratio that hampers fluorescence yield experiments at lower energies (Krause, 1979).

Another consideration for the collection of high-quality XAFS data in the low photon-energy range is the problem of contamination of the X-ray optics and I_o flux monitor with both carbon and oxygen. Such contamination causes a significant decrease in intensity as well as unwanted structure while scanning over these edges. Imperative to the collection of any high-quality absorption data is accurate normalization of the data to the incident flux, which becomes impossible as a typical I_o flux monitor is rendered nonlinear due to these contaminants. Normalizing to a blank sample, i.e. a sample that either has not been dosed with the adsorbate in question (surface science) or one that has a zero concentration of the element being studied, can eliminate these nonlinearities (Stöhr, 1992); however, in general, this method is not robust for all forms of samples, so contamination of the beamline optics should be kept as low as possible. In many UHV applications a fresh layer of a clean metal that does not contain absorption edges in the energy region of interest (often gold) can be evaporated during the course of the experiment onto the I_o grid to bury the contaminant.

Lastly, it is worth noting the added complexity of harmonics for low-energy XAFS. Harmonics plague all aspects of XAFS data collection, but the problem is significantly more acute at low X-ray energies for several reasons. Firstly, in the soft X-ray region the energy differences between the fundamental and its harmonics are not large enough to be sorted by a single-harmonic rejection mirror operating near its critical angle. Effective order suppression requires multiple mirror systems with enough complexity that they are rarely used. An X-ray filter with the appropriate chemical composition may be of some use for order suppression, but the low-energy fundamental will also be significantly suppressed and the selection can therefore never be complete. Secondly, unlike a double-crystal monochromator as used for the tender and higher X-ray energy ranges that can easily `detune' harmonics by taking advantage of their narrower Darwin widths, a grating monochromator in the soft X-ray region requires special designs or operating modes that are not sufficiently convenient for routine operation (Guo et al., 2015). Thirdly, the density of absorption edges is much larger at low photon energies (E α Z<sup>2</sup>), so spectra are often `contaminated' not only by the different angular momentum-dependent edges of the element of interest but also from different elements in the sample that often include unknown or unwanted impurities. Some success has been achieved utilizing undulators built with a quasi-periodic magnetic structure that shifts the energies of the higher-order undulator radiation from their periodic values, allowing direct discrimination by the monochromator (Hashimoto & Sasaki, 1995). However, this clever scheme comes at a significant cost in intensity.