Chapter 3.6. Energy-scanning and energy-dispersive spectrometers for XAFS

In this mode the Bragg angle is changed in a step-by-step manner through the full angular range. At each step, the values of the incident and transmitted intensity I and I₀ are recorded. To cover an extended X-ray absorption fine-structure (EXAFS) spectrum, the Bragg angle is scanned over several degrees in angular steps, the amplitude of which generally varies as a function of energy. An efficient scan uses steps of several electronvolts to cover the pre-edge region (roughly ∼100 to ∼10 eV before the onset of absorption) and the smallest angular step (a function of the required energy resolution and typically a fraction of an electronvolt) to cover the energy interval close to the absorption edge. The EXAFS region is generally scanned using constant δk steps, where k is the photoelectron wavevector. A typical Cu K-edge EXAFS spectrum up to k = 20 Å⁻¹ covers an energy range of the order of 1600 eV, with a minimum step at the edge of ∼0.5 eV and a total number of 700 steps. In step-by-step mode, the dwell time Δt per step includes the acquisition time δt and the sum of all other deadtimes (motor movement, vibration settling and detector deadtime). Synchrotron EXAFS spectra were recorded in the 1970s at SPEAR, Stanford, USA using a Si(220) channel-cut monochromator operated in step-by step mode (Kincaid & Eisenberger, 1975). The flux on the sample was ∼10⁸ photons eV⁻¹ s⁻¹, leading to a signal-to-noise ratio (S/N) of ∼10⁴ with an integration time of 1 s per point. These presently remain typical acquisition times for concentrated samples measured in transmission mode in step-by-step mode, leading to a total duration of an EXAFS spectrum of ∼10–20 min.

Today, flux on a sample varies from 10⁹ to 10¹³ photons s⁻¹ and the S/N on bulk and moderately concentrated samples is almost never limited by the photon flux, but rather by mechanical instabilities of the monochromator as it is rotated from one position to another in the step-by-step scan. In the continuous-scan mode, the motors of the monochromator are moved in a continuous way while acquisition is performed on-the-fly. The first EXAFS spectrum ever measured at a synchrotron was recorded in continuous-scan mode at SPEAR in 1974. Using a Si(220) monochromator crystal rotating at 1.2 deg s⁻¹ and a strip-chart recorder, a back-and-forth scan through the Cr K edge was recorded on a sample of stainless steel (Lytle, 2015). A similar approach was adopted and further developed in the 1980s (Frahm, 1988). Today, the angular speed of monochromators in continuous mode can be varied, typically from 0.001 to a few deg s⁻¹, depending on the working energy and the required time resolution. The values of I, I₀ and motor positions are read out simultaneously at a high frequency. Assuming an incoming flux of 10⁹ photons s⁻¹ (typical of bending-magnet XAS beamlines) and a transmitted flux of 10⁸ photons s⁻¹ (i.e. 10⁶ photons over 10 ms), a total acquisition time of ∼10 s is required for a full XAS spectrum of 1000 points with an effective S/N of about 10³. The continuous-scan mode was initially developed for time-resolved applications and was rapidly implemented at many synchrotrons (Lagarde et al., 1989; Murphy et al., 1995; Frahm et al., 1995; Als-Nielsen et al., 1995; Solé et al., 1999; Prestipino et al., 2011; Nonaka et al., 2012). In recent years, it has become more and more popular at synchrotron XAS beamlines as a default operation mode as it leads to more efficient use of beamtime.

To match the timescale of kinetics of many chemical reactions, and to allow X-ray absorption near-edge structure (XANES) tomography, which requires the collection of several thousand scans, the time resolution can be pushed to below 1 s per spectrum by adapting both monochromator mechanics and data collection. Oscillating monochromators are based on a cam-driven tilt table in combination with a small-gap channel-cut crystal (Frahm et al., 2004). In the oscillatory scan mode, the monochromator is oscillated continuously without any interruption around a central Bragg angle in a sinusoidal movement. Fast angular encoders are used to track the energy calibration. Typical oscillation speeds used are ∼10 Hz, leading to the recording of about 20 full XAS spectra in 1 s. With full ∼1000 eV spectra in 50 ms, a time response in the microsecond range is required for a spectral resolution of the order of 1 eV. This requires high-speed, low-noise current amplifiers (Müller et al., 2013). Assuming an incoming flux of 10¹³ photons s⁻¹ (typical of undulator XAS beamlines) and a transmitted flux of 10¹² photons s⁻¹, each point of the spectrum records ∼5 × 10⁷ photons over ∼50 µs with an effective S/N of about 7 × 10³. Oscillating monochromators have now been implemented on several synchrotron beamlines (Stötzel et al., 2010; Fonda et al., 2012; Müller et al., 2016).

In `movie' mode, the ED spectrometer acts as a fast camera that records the time evolution of the XAS signal throughout the dynamic process. The time resolution is given by the maximum repetition rate of the detector, which accounts for the minimum exposure time and the deadtime between exposures. Phenomena on timescales of a few milliseconds have been investigated with repetition rates as fast as a few tens of microseconds. This mode has been applied to study reaction mechanisms and intermediates (Bal et al., 2006; Kong et al., 2012), structural and chemical phase transitions induced by rapid heating (Marini et al., 2014; Kantor et al., 2014) and, more recently, the kinetics of pressure-induced phase transitions (Dewaele et al., 2016). For a more detailed review, see Mathon, Kantor et al. (2015).

Below the ∼10 µs regime, the time structure of the synchrotron cannot be neglected and it becomes necessary to perform the experiment in pump-and-probe mode using single X-ray bunches. In this mode, the time resolution is defined by the delay between the pump (start) and the probe (stop). The time resolution is then limited either by the duration of the pump or by the duration of the probe. In practice, at synchrotrons, the time resolution is very often limited by the duration of the X-ray bunch, i.e. around 100 ps. The time evolution of a reversible phenomenon can be studied by varying the delay between the pump and the probe over many pump-and-probe cycles (Tromp et al., 2013). However, the real added value of using an ED spectrometer in pump-and-probe mode is when the experiment is nonreversible; for example, when the pump itself is not perfectly reproducible (i.e. the energy delivered by the laser is not constant shot to shot) or when the sample is damaged or destroyed at the end of each pump-and-probe cycle. An example of the latter application is the study of dynamically compressed matter induced by pulsed laser sources, which allows the exploration of extreme states of matter that mimic conditions in the interior of the Earth and other planets. Recent work aimed at probing the electronic and local structure in iron under conditions of pressure and temperature that go well beyond those in the Earth's core, where iron is highly compressed but no longer solid, has shown that third-generation synchrotrons provide sufficient flux to record analyzable XANES and EXAFS using a single 100 ps X-ray pulse (Torchio et al., 2016).

With ES spectrometers the sample is illuminated by monochromatic radiation with a defined energy and propagation direction at each step in the scan. This allows photon-in photon-out spectroscopic techniques, such as X-ray emission spectroscopy (XES), to be applied. There are important advantages for XAFS. One of these is the possibility of increasing the sensitivity to dilute species by detecting photons emitted from element-specific de-excitation processes (i.e. total fluorescence yield). On the other hand, the parallel detection of the full XAS spectrum in ED spectrometers relies on the sample being illuminated by a polychromatic divergent X-ray beam: photon-in photon-out spectroscopic techniques are not applicable. In other words, EDXAS only works in transmission geometry. This has the important drawback of low sensitivity to dilute species. By giving up parallel detection over the full energy range and using a sequential acquisition mode (Pascarelli et al., 1999), energy-dispersive spectrometers have demonstrated the capacity for fast time-resolved XAS studies in fluorescence mode (Nagai et al., 2008).

For high-resolution spectroscopic applications, ES spectrometers have several advantages. By using perfect, flat crystals and an appropriate optical scheme, the energy resolution of the incoming beam can approach the intrinsic resolution (equation 2). On the other hand, ED spectrometers rely on the use of curved crystals, and in the more commonly used Bragg geometry the energy resolution is affected by broadening of the reflectivity curve, especially at high X-ray energies where the penetration depth into the crystal is important.

Another important advantage of photon-in photon-out spectroscopies is the possibility of recording XAS spectra with an energy resolution beyond the core-hole lifetime of the absorbing atom by using high-resolution crystal analysers to analyse the energy of the emitted photons (Hämäläinen et al., 1991). With ED spectrometers, the energy resolution is always limited by core-hole lifetime broadening.

ES spectrometers have to provide a very good repeatability for successive scans. The requirements for precision and repeatability are close to ΔE/E = 1 × 10⁻⁵ and ΔE/E = 1 × 10⁻⁶, respectively. This is not an issue for ED spectrometers as the spectrum is acquired with no moving components. Thermal stability issues may affect both spectrometers.

A critical requirement for XAFS spectrometers is the ability to provide a highly stable X-ray beam on the sample in both energy scale and position. From this point of view, the ED spectrometer has superior properties since the full absorption spectrum is acquired with no moving components, as opposed to one or more motorized movements in the ES spectrometer, leading to an efficient reduction in nonstatistical noise. This feature has been exploited for the measurement of femtometre atomic displacements in magnetostrictive FeCo films (Pettifer et al., 2005), as well as for 3d metal K-edge XMCD studies in the multi-megabar regime (Torchio et al., 2014).

The speed of acquisition of a full EXAFS spectrum is an important parameter. Both kinds of spectrometers are used for time-resolved studies, but the ED spectrometer provides an intrinsically high acquisition speed thanks to the parallel detection scheme, which makes it particularly useful in the investigation of very fast processes. The time required to measure a full spectrum is limited by the detector-readout speed in `movie mode' or by the X-ray pulse length (∼100 ps) in `pump–probe' mode (Torchio et al., 2016). ES spectrometers have seen much development in this field in recent years, and the acquisition time for a full EXAFS spectrum is now approaching the 10 ms timescale (Müller et al., 2015).

Whereas the energy range of application of ES spectrometers is very wide, covering the full X-ray range available at synchrotrons (from roughly 2 to 80 keV), the ED spectrometer has a limitation at low energies because the energy bandwidth of the diffracted polychromatic fan range scales with the cotangent of the Bragg angle (see equation 4): below ∼5 keV the EXAFS k-range is severely limited and only XANES is available.

Both kinds of spectrometer can be used to follow the time evolution of reversible phenomena in pump–probe mode. ES spectrometers are operated in step-by-step mode: at each energy step the pump–probe cycle is repeated N times until a sufficient signal-to-noise ratio is achieved in the measurement of the absorption coefficient at a given energy. Once the full spectrum has been recorded, the time delay δt is changed and the full procedure is repeated (Chen et al., 1999; Oyanagi et al., 2001; Bressler & Chergui, 2004). ED spectrometers have been utilized to follow the time evolution of reversible phenomena (Tromp et al., 2013). However, as mentioned in Section 3.2, the real added value of using an ED spectrometer in pump-and-probe mode is when the experiment is nonreversible (Torchio et al., 2016).