Ultra-Soft Resonant Inelastic X-ray Scattering

• Ultra-soft RIXS is a resonant scattering technique performed at low-photon energies, enabling precise measurement of neutral excitations and element-specific information.
• It leverages the small momentum transfer regime to examine complex low-energy electronic dynamics such as charge transfer, crystal field effects, and exciton-phonon interactions.
• Recent instrumentation advances have achieved energy resolutions as low as 12–40 meV, significantly boosting its capability to resolve subtle electronic features.

Ultra-soft resonant inelastic X-ray scattering is best understood as RIXS performed at the low-photon-energy end of the soft-X-ray domain, although the label is used inconsistently in the literature: one study applies it to transition-metal M-edge RIXS in the tens-of-eV regime, whereas another uses it for B $K$-edge measurements near $187$ eV; closely related work on C, B, N, O, lanthanide $M$, and transition-metal $L$ edges is usually described simply as soft X-ray RIXS (Wray et al., 2016, Lomeli et al., 12 Sep 2025, Dashwood et al., 2021, Amorese et al., 2018). In all of these usages, the method is a photon-in/photon-out, element-specific energy-loss spectroscopy in which the incident energy is tuned to a core-level resonance and the scattered-photon energy loss is used to identify neutral excitations. At these low photon energies, ultra-soft RIXS combines resonance enhancement, orbital and chemical selectivity, and access to the small-$q$ regime, while retaining the distinctive interpretive complications of a second-order resonant process (Braicovich et al., 2014, Nicolaou et al., 8 Jan 2025, Lomeli et al., 12 Sep 2025).

1. Scope, terminology, and operating regime

The practical meaning of “ultra-soft” is usage-dependent rather than standardized. In the transition-metal context it can denote M-edge RIXS, where the experiments exploit overlapping $M_3$ and $M_2$ resonances and very high energy resolution on low-energy electronic excitations (Wray et al., 2016). In a different usage, the term describes very low-energy soft-X-ray work at the B $K$ edge, where the incident energy is tuned near a pre-edge feature around $187$ eV and the photon momentum is small enough that $|\mathbf k|\approx 0.095~\text{\AA}^{-1}$, making ultra-low-$187$0 metallic charge dynamics experimentally accessible (Lomeli et al., 12 Sep 2025). Related low-$187$1 soft-edge work spans the C $187$2 edge near $187$3 eV in graphite, the B and N $187$4 edges near $187$5 and $187$6 eV in $187$7-BN, the O $187$8 edge around $187$9 eV, and lanthanide $M$0 edges around $M$1 eV (Dashwood et al., 2021, Nicolaou et al., 8 Jan 2025, Singh et al., 2020, Amorese et al., 2018).

Instrumentation papers make the same boundary visible from the hardware side. PEAXIS at BESSY II provides linearly polarized light from $M$2 eV to $M$3 eV and documents RIXS operation from $M$4 eV to $M$5 eV, explicitly noting that this lower end approaches what is reasonably called the ultra-soft regime in practical synchrotron work (Schulz et al., 2019). The Heisenberg-RIXS instrument at the European XFEL is framed as a soft-X-ray spectrometer for $M$6–$M$7 eV in concept and $M$8–$M$9 eV with its installed gratings, so it covers the low soft-X-ray regime but not the extreme-EUV range below that (Schlappa et al., 2024). This suggests that “ultra-soft RIXS” is less a sharply delimited spectral band than a family of low-energy resonant scattering implementations that exploit the unusually small photon momentum and edge selectivity available below the canonical transition-metal $L$0-edge range.

2. Scattering formalism, kinematics, and selectivity

The basic kinematics are standard. In RIXS and NRIXS the momentum transfer is

$L$1

and the energy loss is

$L$2

with energy conservation $L$3 (Nicolaou et al., 8 Jan 2025). The corresponding dynamical structure factor used for comparison with nonresonant probes is

$L$4

but the resonant cross section is governed instead by the Kramers-Heisenberg expression, which depends on the intermediate-state manifold, its lifetime broadening, and resonance-specific matrix elements (Nicolaou et al., 8 Jan 2025).

That distinction is not merely formal. In $L$5-BN, the explicit Kramers-Heisenberg form is used to emphasize that RIXS is a second-order process whose matrix elements cannot, in general, be reduced directly to $L$6 without additional hypotheses (Nicolaou et al., 8 Jan 2025). In metallic MgB$L$7, the same issue becomes central for low-$L$8 charge physics: probes such as EELS or nonresonant IXS measure the screened longitudinal response or loss function,

$L$9

with $q$0, so in a metal the low-energy intra-band continuum is strongly suppressed at small $q$1 by screening and charge conservation. Resonant ultra-soft RIXS does not simply measure that fully screened total density response; it couples instead to local, element- and orbital-selective operators, which is why a low-energy Lindhard-continuum feature can be visible in RIXS while being effectively hidden in the total longitudinal charge response (Lomeli et al., 12 Sep 2025).

The same resonance dependence underlies edge selectivity. In $q$2-BN, the lowest longitudinal exciton is seen as a $q$3 eV Raman feature at the B edge in LH polarization but is not observed as a clear Raman peak at the N edge, because the lowest conduction states have predominantly B character and the N $q$4 excitation is not an efficient route to the same neutral excitation (Nicolaou et al., 8 Jan 2025). In graphite, tuning across the C $q$5-edge $q$6 and $q$7 resonances changes the intermediate-state orbital content and therefore changes the momentum weighting of the exciton-phonon coupling encoded in the multi-phonon spectrum (Dashwood et al., 2021).

3. Resolution frontier and instrument architecture

Ultra-soft and low-energy soft RIXS are fundamentally resolution-limited techniques. A major threshold problem was the inability of older soft-RIXS spectrometers on lanthanides to resolve CEF splittings of only a few tens of meV. This changed with the new generation of spectrometers exemplified by ESRF ID32, where about $q$8 meV resolution at the Ce $q$9 edges, and specifically $M_3$0 meV achievable at the Ce $M_3$1 edge with $M_3$2 eV, made it possible to resolve crystal-electric-field excitations directly in CeRh$M_3$3Si$M_3$4; the main high-resolution data were broadened to $M_3$5 meV, while polarization-resolved measurements were taken at $M_3$6 meV because of polarimetry losses (Amorese et al., 2018).

The most aggressive soft-RIXS resolution figures come from beamlines designed to preserve throughput while narrowing the energy-loss response. The Taiwan Photon Source AGM–AGS beamline implements the energy-compensation principle of grating dispersion and achieved a best energy resolution of $M_3$7 meV at $M_3$8 eV with resolving power $M_3$9 while keeping the incident bandwidth at $M_2$0 eV (Singh et al., 2020). PEAXIS documents a best total resolution of about $M_2$1 meV at about $M_2$2 eV and a measured total energy resolution of $M_2$3 meV at an incident energy of about $M_2$4 eV in the monochromator characterization regime, with continuous spectrometer-arm rotation for momentum-dependent work (Schulz et al., 2019). At the European XFEL, hRIXS is designed so that, with optimized spot size and small-pixel detection, the energy resolution can be better than $M_2$5 meV at any photon energy below $M_2$6 eV; commissioning reported $M_2$7 meV FWHM at Cu $M_2$8, $M_2$9 meV FWHM at Ni $K$0, and the summary states that $K$1 meV at the oxygen $K$2 edge has been achieved (Schlappa et al., 2024).

Polarimetry required a parallel instrumental advance. The first implementation in a high-resolution soft-RIXS spectrometer of a device capable of measuring the degree of linear polarization of the scattered photons while retaining full energy-resolved detection used a graded W/B$K$3C multilayer mirror near the CCD on the AXES spectrometer at ESRF beamline ID08 (Braicovich et al., 2014). The pilot implementation suffered a factor $K$4 loss in efficiency, but the concept established that outgoing-photon linear polarization can be measured simultaneously with energy dispersion, and an advanced version was proposed with overall efficiency up to $K$5 (Braicovich et al., 2014).

Alternative detection architectures seek to decouple soft-RIXS resolution from instrument length and X-ray spot size. Photoelectron spectrometry for analysis of X-rays (PAX) replaces the large grating spectrometer by a converter plus