Time-Resolved Resonant X-ray Emission Spectroscopy

• Time-resolved resonant X-ray emission spectroscopy is a pump–probe method that uses core excitations to prepare resonant intermediate states and monitor evolving electronic structures.
• It achieves element specificity by measuring both incident and emitted photon energies, providing insights into occupied and unoccupied states in complex systems.
• Advanced experimental architectures, including soft/hard X-ray setups and stochastic as well as analyzer-based methods, enable detailed studies of molecular dynamics, surface electron behavior, and correlated phenomena.

Time-resolved resonant X-ray emission spectroscopy is a pump–probe, resonant photon-in/photon-out spectroscopy in which a core excitation prepares a resonant intermediate state and the system emits an X-ray photon while evolving; the final photon energy reflects the evolving electronic structure. In practice, the term spans resonantly excited X-ray emission spectroscopy, resonant X-ray emission spectroscopy maps recorded as a function of incident and emitted photon energy, and time-resolved resonant inelastic X-ray scattering when the emitted spectrum is analyzed in energy loss and, in solids, momentum space. Across molecular photodynamics, adsorbate electron dynamics, correlated solids, silicon membranes, and solid-density plasmas, the method is valued for element specificity, sensitivity to occupied and unoccupied states, and compatibility with femtosecond pump–probe operation (Freibert et al., 2024, Mitrano et al., 2020, Schreck et al., 2022, Huang et al., 14 Aug 2025).

1. Process, nomenclature, and scope

In resonant X-ray emission spectroscopy, the incident x-ray energy $E_i$ is swept across and above an absorption edge while the emitted x-ray energy $E_e$ is recorded, so the measured quantity is effectively the emission intensity as a function of incident energy and energy transfer. At the U $L_{\mathrm{III}}$ edge, for example, the process involves the $2p_{3/2}$ threshold and the $L_{\alpha1}$ emission line from $3d_{5/2}\rightarrow 2p_{3/2}$ decay, and the resulting RXES signal is bulk sensitive because the measurement depth was greater than $1.9\ \mu\mathrm{m}$ in URu$_2$Si$_2$ (Booth et al., 2016).

In resonant inelastic X-ray scattering, an incident X-ray photon is tuned to a characteristic absorption edge, excites a core electron into the valence shell or above it, creates a short-lived intermediate state with a core hole, and then decays when another electron fills the core hole and a scattered photon is emitted. Time-resolved RIXS extends this process into a pump–probe geometry in which an ultrafast optical pump prepares a nonequilibrium state and a delayed resonant X-ray probe records the evolving energy-loss and momentum-resolved spectrum. The relation to resonant X-ray emission spectroscopy is direct: RXES is the energy-resolved emission process near resonance, while RIXS and trRIXS make the inelasticity and momentum transfer central observables (Mitrano et al., 2020).

For transient molecular spectroscopy, the same second-order process can be viewed from the emission side. A fully time-dependent formulation for pyrazine states explicitly that the formalism also naturally describes time-resolved resonant X-ray emission spectroscopy because the emission step is treated as part of the same time-dependent second-order process; in that sense, the signal expression is a general time-domain resonant photon-in/photon-out formalism (Freibert et al., 2024).

A recurring point of nomenclature is that not every time-resolved X-ray emission measurement is a continuously tuned resonant inelastic scattering experiment. In the C/Ni(100) adsorbate study, the measurements include ordinary C K-edge XES and also a specifically resonantly excited XES case at $282.5\ \mathrm{eV}$ to probe states just below the Fermi edge more selectively. This distinction matters because resonant tuning changes intermediate-state selectivity even when the observable is still the emitted spectrum (Schreck et al., 2022).

2. Dynamical formulations beyond steady-state Kramers–Heisenberg–Dirac theory

Steady-state RIXS is often handled by the Kramers–Heisenberg–Dirac formula, which assumes a stationary initial state and frequency-domain scattering amplitudes. For time-resolved RIXS and time-resolved resonant X-ray emission spectroscopy, this becomes insufficient once an optical pump prepares a nonstationary molecular wavepacket and an ultrashort intense X-ray probe interrogates it. The pyrazine study therefore develops a fully time-dependent theory in which nuclear motion, state-to-state population transfer, and core-excited-state dynamics are treated explicitly in time (Freibert et al., 2024).

The underlying wavepacket dynamics are propagated with multiconfiguration time-dependent Hartree in multilayer form. The nuclear wavefunction is written as $_2$1 with time-dependent single-particle functions, $_2$2 and equations of motion derived from the Dirac–Frenkel variational principle. For pyrazine, the propagation uses a 24-dimensional vibronic coupling model with a diabatic Hamiltonian separated into valence and core manifolds, $_2$3 where valence and core manifolds are decoupled from each other for non-adiabatic transitions because their energy gap is large, while non-adiabatic coupling inside each manifold is retained (Freibert et al., 2024).

The total Hamiltonian is $_2$4 with dipole interaction $_2$5 The pump is treated as effectively $E_e$0-like, whereas the probe is modeled explicitly as a Gaussian-shaped coherent X-ray pulse with temporal FWHM $E_e$1 in the pyrazine calculations. The central dynamical object is the Raman wavefunction, $_2$6 which collects the evolving valence-state nuclear wavepacket, propagation into the core manifold, finite core-hole lifetime through $E_e$2, and probe detuning and timing through $E_e$3 (Freibert et al., 2024).

Two differences from the standard KHD picture are emphasized. First, the initial state is not stationary but a time-evolving pumped wavepacket. Second, the probe is not infinitely long or monochromatic but has finite duration and explicit temporal gating. The resulting signal depends on instantaneous nuclear geometry and population distribution rather than on a single equilibrium initial state. The pyrazine analysis further concludes that transient RIXS tracks both valence and core dynamics, that nuclear wavepacket motion broadens and distorts spectral bands, and that symmetry breaking in the core-excited manifold can generate anti-Stokes emission and otherwise weak transitions (Freibert et al., 2024).

3. Experimental architectures and signal reconstruction

One experimental architecture uses direct soft X-ray pump–probe spectroscopy with conventional detection channels. For carbon adsorbed on Ni(100), the experiment at FERMI employed a 400 nm optical pump with 3.1 eV photon energy, pulse duration 90–100 fs, pulse energy about 152 $E_e$4J, and spot size $E_e$5m. The probe consisted of soft X-ray FEL pulses in the carbon K-edge range, pulse duration 45–50 fs, and spot size $E_e$6m, giving an overall temporal resolution of about 100–110 fs. XAS was recorded in fluorescence yield, and XES was measured with an energy-dispersive soft X-ray spectrometer with overall spectral resolution about 0.3 eV (Schreck et al., 2022).

A second architecture replaces monochromatic scanning by stochastic spectroscopy. In the silicon $E_e$7-edge experiment, broadband stochastic EUV pulses were measured upstream with PRESTO and downstream with WEST, and the sample response was reconstructed from the covariance of shot-to-shot spectral fluctuations. The method uses singular value decomposition and Tikhonov regularization to stabilize the ill-posed inverse problem

$E_e$8

solving instead

$E_e$9

In the reconstructed 2D response map, diagonal terms correspond to transmission or XAS, while off-diagonal vertical terms correspond to XES. The reported energy resolution is a few tens of meV, and the time-resolved implementation used a 390 nm optical pump, 90 fs pulse duration, about 50,000 FEL shots per delay, and an FEL pulse duration of about $L_{\mathrm{III}}$0 (Angelis et al., 2023).

A third architecture is hard X-ray trRIXS with analyzer-based spectroscopy. At the Bernina beamline of SwissFEL, a compact Johann-type spectrometer with Rowland radius $L_{\mathrm{III}}$1 and up to 3 crystal analyzers was commissioned at the Ir $L_{\mathrm{III}}$2 edge of $L_{\mathrm{III}}$3-Li$L_{\mathrm{III}}$4IrO$L_{\mathrm{III}}$5. The resonant probe energy was $L_{\mathrm{III}}$6, the demonstrated elastic peak width was $L_{\mathrm{III}}$7, the pump wavelength was 400 nm with 100 fs FWHM, and the reported temporal resolution was around 50 fs. The setup combines energy, momentum, and temporal resolution and is explicitly framed as a time-resolved, excitation-selective version of resonant X-ray emission with energy-loss sensitivity (Chen et al., 2023).

A fourth architecture is resonant probing of dense plasmas. In the Cu-wire experiment, a 10 $L_{\mathrm{III}}$8m diameter Cu wire was driven by a 3 J, 30 fs, 10 Hz Ti:sapphire pump laser focused to $L_{\mathrm{III}}$9 FWHM at up to $2p_{3/2}$0, while an XFEL probe at 8.2 keV with $2p_{3/2}$1 duration, $2p_{3/2}$2 FWHM spot size, and $2p_{3/2}$3 bandwidth measured resonant emission with a von Hámos spectrometer and simultaneous X-ray transmission imaging. The timing uncertainty was up to about 200 fs, and the probe delivered sub-picosecond temporal resolution (Huang et al., 14 Aug 2025).

4. Molecular photodynamics and surface electron dynamics

Pyrazine at the nitrogen K-edge is a benchmark case for the fully dynamical description because both the valence manifold and the core-excited manifold exhibit important non-adiabatic and symmetry-breaking dynamics. The model includes all 24 vibrational modes, four relevant valence states, four core-excited states, and a finite core-hole lifetime of about 8 fs. The pump creates population primarily in the bright $2p_{3/2}$4 state, with some direct population of $2p_{3/2}$5; $2p_{3/2}$6 rapidly depopulates to $2p_{3/2}$7 and $2p_{3/2}$8, and $2p_{3/2}$9 oscillations occur on roughly 20 fs timescales. In fs-XANES, a strong bleach near 402.5 eV from $L_{\alpha1}$0 coexists with excited-state absorption near 399.0 eV and 401.8 eV, leading to the choice of 399.0 eV to target dynamics on $L_{\alpha1}$1 and 401.5 eV to probe $L_{\alpha1}$2 (Freibert et al., 2024).

At 399.0 eV, the probe predominantly excites $L_{\alpha1}$3, producing a strong elastic-like $L_{\alpha1}$4 band near zero loss, inelastic features to $L_{\alpha1}$5, $L_{\alpha1}$6, and $L_{\alpha1}$7, and anti-Stokes intensity to $L_{\alpha1}$8 and other states after core-excited-state symmetry breaking. A major result is that the core-excited states $L_{\alpha1}$9 and $3d_{5/2}\rightarrow 2p_{3/2}$0 are vibronically coupled by asymmetric $3d_{5/2}\rightarrow 2p_{3/2}$1 modes, causing ultrafast symmetry breaking and localization within the core-hole lifetime. At 401.5 eV, the probe addresses $3d_{5/2}\rightarrow 2p_{3/2}$2 and $3d_{5/2}\rightarrow 2p_{3/2}$3; early-time signal is dominated by $3d_{5/2}\rightarrow 2p_{3/2}$4, after about 20–30 fs $3d_{5/2}\rightarrow 2p_{3/2}$5 dominates, there are no anti-Stokes features, and the observed band is a superposition because $3d_{5/2}\rightarrow 2p_{3/2}$6 and $3d_{5/2}\rightarrow 2p_{3/2}$7 are vibronically coupled and spectrally overlapping. The study argues that reduced-dimensionality models may miss important symmetry-breaking coordinates (Freibert et al., 2024).

At surfaces, time-resolved X-ray absorption and X-ray emission spectroscopy provide an element-specific view of adsorbate-local electron dynamics. For carbon atoms adsorbed on Ni(100), ultrafast changes in both XAS and XES show clear signatures of the formation of a hot electron-hole pair distribution on the adsorbate within about 100 fs after 400 nm pumping of the substrate. At about 0.3 ps delay, the XAS shows increased intensity below 283 eV and reduced intensity above 283 eV. At about 0.2 ps, the XES main peak near 278.5 eV loses intensity, additional intensity appears in the 275–277 eV region, and additional intensity appears above the Fermi level near 283 eV. A resonantly excited XES spectrum at 282.5 eV shows the same behavior more clearly, including a shift of the main emission from about 279 eV to 277.5 eV plus intensity above the Fermi level (Schreck et al., 2022).

The same experiment separates an early electronic regime from a later thermalized regime. By about 1.2 ps, the earliest XAS changes are diminished and a new intensity increase appears in the 284–285 eV region; by 4.0 ps, the spectrum is nearly the same as at 1.2 ps. The interpretation is that the first ultrafast changes arise from a hot electron-hole pair distribution on the adsorbate, whereas the few-picosecond evolution is consistent with thermalization of the complete C/Ni system and increased vibrational or structural disorder, including C motion away from the original ideal hollow-site geometry toward lower coordination geometries. A 5000 K electronic-temperature DFT model reproduces the early-time XAS and XES changes qualitatively, an 800 K AIMD model reproduces the later XAS high-energy shoulder, but the early XES redistribution far below the Fermi level is not fully captured by simple heating and may involve stronger nonadiabatic coupling, many-body effects, demagnetization-related effects, or redistribution among critical points in the 2D band structure (Schreck et al., 2022).

5. Correlated solids, silicon band dynamics, and finite-momentum observables

In correlated materials, time-resolved resonant X-ray emission is often realized through trRIXS because the emitted spectrum is most informative when resolved in both energy loss and momentum transfer. The SwissFEL study on $3d_{5/2}\rightarrow 2p_{3/2}$8-Li$3d_{5/2}\rightarrow 2p_{3/2}$9IrO$1.9\ \mu\mathrm{m}$0 shows that the transient spectrum, defined as

$1.9\ \mu\mathrm{m}$1

changes mainly in the energy-loss region below 2 eV, especially around 0.4 eV and 1.3–1.6 eV. The strong $1.9\ \mu\mathrm{m}$2 on-site excitation shows no clear transient response. The observed changes are therefore ascribed to modulations of the Ir-to-Ir intersite transition scattering efficiency, associated with transient screening of the on-site Coulomb interaction under 3.1 eV optical pumping and ligand-to-metal charge transfer (Chen et al., 2023).

The broader condensed-matter framework emphasizes that trRIXS is a momentum-resolved ultrafast probe of nonequilibrium collective excitations and finite-$1.9\ \mu\mathrm{m}$3 fluctuation spectra. It is sensitive to spin, charge, and orbital fluctuations, and in the ultrashort core-hole lifetime approximation the dominant equilibrium response is given by dynamical charge or spin structure factors. The full nonequilibrium cross section is more demanding because it involves a four-time correlation function and scales roughly as $1.9\ \mu\mathrm{m}$4 in the number of time steps. Predicted observables include bimagnon features at $1.9\ \mu\mathrm{m}$5 with $1.9\ \mu\mathrm{m}$6, Mott-gap excitations near $1.9\ \mu\mathrm{m}$7, and pump-induced spectral weight transfer from the Mott peak into in-gap excitations and bimagnons (Mitrano et al., 2020).

The silicon $1.9\ \mu\mathrm{m}$8-edge stochastic spectroscopy experiment illustrates a complementary limit in which time-resolved XES and XAS are reconstructed simultaneously from broadband FEL fluctuations. The static transmission profile resolves the two spin-orbit-split resonances at 99.76 eV for the $1.9\ \mu\mathrm{m}$9 edge and 100.45 eV for the $_2$0 edge, while the XES lines appear at approximately 99.35, 99.50, 99.76, 100.10, and 100.45 eV. After photoexcitation with a 390 nm optical pulse at $_2$1, pre-edge transmission increases, post-edge transmission decreases, and the XES line at 99.35 eV increases within the first 10 ps before recovering more slowly with recovery time about $_2$2. The interpretation is that the observations are not consistent with simple thermal or Pauli-blocking-only behavior but are compatible with bond softening and the predicted precursor state of a non-thermal solid-liquid phase transition (Angelis et al., 2023).

A plausible implication is that time-resolved resonant X-ray emission becomes most discriminating when emission changes can be cross-correlated with absorption, momentum dependence, or transfer-matrix structure rather than interpreted as isolated fluorescence intensities. That pattern is explicit in hard X-ray trRIXS, stochastic XES/XAS, and resonant surface XES alike (Mitrano et al., 2020, Angelis et al., 2023, Chen et al., 2023).

6. Dense plasmas, charge-state diagnostics, and temporal segmentation

In hot dense plasmas, time-resolved resonant X-ray emission can act as a direct charge-state population diagnostic. The Cu-wire experiment tuned the XFEL to 8.2 keV, resonant with a K–L bound–bound transition in nitrogen-like Cu$_2$3. When enough Cu$_2$4 ions are present, the XFEL photons are absorbed resonantly, exciting a K-shell electron into a higher L-shell state, and the excited configuration decays mainly radiatively, producing an enhanced resonant X-ray emission signal. Because the resonance is charge-state specific, the emitted intensity tracks the population of Cu$_2$5 (Huang et al., 14 Aug 2025).

The measured temporal behavior is a rise-and-fall profile. Before the laser peak, resonant emission is weak and buried in continuum background; at about 0.5 ps it becomes visible, at about 2.5 ps it reaches a maximum, and it then decays gradually over about 10 ps. The same pump–probe delays show a clear inverse correlation between resonant emission yield and XFEL transmission, indicating maximum resonant absorption when the Cu$_2$6 population is largest. Off-resonance satellite lines on both sides of the main resonant line, including neighboring lines such as 8.16 keV and 8.255 keV, are also enhanced, and their near-symmetry suggests that ionization and recombination rates are comparable and both are within roughly an order of magnitude of the radiative decay rate (Huang et al., 14 Aug 2025).

The interpretation depends on collisional–radiative kinetics rather than on a single temperature. SCFLY modeling includes collisional ionization, radiative recombination, three-body recombination, excitation/de-excitation, Auger ionization, and dielectronic recombination, and predicts that at around 500 eV the population of Cu$_2$7 peaks, as does the opacity at 8.2 keV. Comparison with PIC and MHD simulations reveals a substantial discrepancy: the simulations predict bulk electron temperatures around 1–2.5 keV at about 2.5 ps, which is much higher than the $_2$8 inferred from the resonant charge-state diagnostics. The conclusion is that typical models overestimate plasma heating and ionization under these conditions, highlighting the need for improved modeling of NLTE collisional processes (Huang et al., 14 Aug 2025).

A related lesson from time-resolved X-ray emission line spectroscopy is that temporal averaging can suppress the strongest line response. In the DROXO study of Fe K$_2$9 emission from young stellar objects, time-resolved equivalent widths are often much larger than time-averaged values, and long integrations can hide the true peak line strength. The same work shows that line production is source-dependent and may involve photofluorescence, electron impact ionization, occultation geometry, or emission from significantly ionized iron rather than neutral fluorescence. This suggests that for resonant emission spectroscopy as well, segmentation by activity state or pump–probe delay can be as important as absolute spectral resolution (Stelzer et al., 2010).

7. Limitations, misconceptions, and future directions

A persistent misconception is that time-resolved resonant X-ray emission mainly tracks static state populations. The pyrazine calculations argue instead that transient RIXS tracks both valence and core dynamics, and that core-excited-state vibronic coupling matters as much as valence-state non-adiabatic dynamics. The C/Ni(100) measurements similarly show that the adsorbate is not merely a passive spectator to substrate excitation because its own electronic structure changes immediately, while the silicon study argues that the observed dynamics are not consistent with simple thermal or Pauli-blocking-only behavior (Freibert et al., 2024, Schreck et al., 2022, Angelis et al., 2023).

Another limitation is the tension between temporal resolution, spectral bandwidth, and signal strength. The silicon work explicitly states that Tr-XES/XAS typically requires X-ray pulses with both a narrow bandwidth and sub-picosecond pulse duration, whereas the stochastic method offers an alternative route by exploiting broadband fluctuations. The SwissFEL commissioning study gives a demonstrated overall resolution of about 180 meV, while noting that the use of Si(555) and tighter focusing around 10 $_2$0m could improve the resolution to about 60 meV at the cost of flux. The condensed-matter perspective identifies improved repetition rate, brightness, and energy resolution as central requirements and points to endstations at LCLS-II, European XFEL, FERMI, PAL-XFEL, and SwissFEL, with soft X-ray resolution expected to reach about 0.03 eV (Angelis et al., 2023, Chen et al., 2023, Mitrano et al., 2020).

Theory remains a major bottleneck. Full nonequilibrium trRIXS calculations are expensive because of the four-time correlation structure; plasma modeling remains sensitive to collisional physics and recombination; and reduced-dimensionality molecular models may omit symmetry-breaking coordinates that become active within the core-hole lifetime. The unresolved early XES redistribution in C/Ni(100) and the overestimated heating in Cu plasmas are concrete examples in which existing models capture the broad interpretation but not all selective spectral details (Mitrano et al., 2020, Schreck et al., 2022, Huang et al., 14 Aug 2025, Freibert et al., 2024).

The present trajectory is therefore toward more explicit probe-pulse treatments, stronger integration of XES with XAS or transmission imaging, higher-resolution analyzer spectrometers, and broader use of resonant, element-specific probes to isolate nonequilibrium pathways in molecules, surfaces, correlated solids, and high-energy-density matter. This suggests that the defining contribution of time-resolved resonant X-ray emission spectroscopy is not a single experimental geometry but a unified dynamical framework for following resonant intermediate states, emitted-photon spectra, and the evolving many-body or vibronic system on its natural timescale (Freibert et al., 2024, Chen et al., 2023, Huang et al., 14 Aug 2025).