Time-Resolved X-ray Resonant Magnetic Scattering

• Time-resolved XRMS is a pump–probe technique that measures magnetic scattering by decomposing the signal into charge and magnetic components with element-specific resonance.
• It employs reciprocal-space selectivity to resolve features such as domain periodicity, wall chirality, and multilayer depth profiles with high temporal resolution.
• Experimental implementations range from synchrotrons and free-electron lasers to tabletop setups, accessing dynamic regimes from nanoseconds to femtoseconds.

Time-resolved X-ray resonant magnetic scattering (XRMS) is a pump–probe family of resonant X-ray methods in which magnetic scattering is measured as a function of delay time at selected reciprocal-space coordinates. In the published literature, the term encompasses reflection-geometry resonant magnetic scattering from thin films, resonant magnetic diffuse small-angle scattering from domain networks, diffraction and reflectometry ferromagnetic resonance, and resonant diffuse scattering around magnetic Bragg points. Across these implementations, the defining features are resonant enhancement at an element-specific absorption edge, momentum selectivity, and magnetic contrast controlled by polarization, geometry, or both, enabling direct access to magnetization precession, domain periodicity, domain-wall chirality, internal domain-wall modes, depth-dependent multilayer dynamics, and momentum-resolved magnetic fluctuations [1009.3389][2501.11506][2301.03256][2007.08583][2606.02154][2602.13113].

1. Scattering formalism and measured observables

A common starting point is the decomposition of the resonant scattering amplitude into structural and magnetic terms,
[
f=f_0+f_m,
]
with (f_0) the charge/structural amplitude and (f_m) the magnetic amplitude. In chiral wall-resolved XRMS, the measured intensities for left- and right-circular probe helicities are denoted (I^{CL}) and (I^{CR}), and are combined as
[
I_{+}=I^{CL}+I^{CR}, \qquad I_{-}=I^{CL}-I^{CR}.
]
To leading order,
[
I_{+}\propto |f_0|^2+|f_m|^2, \qquad I_{-}\propto \Im!\left(f_0f_m\right).
]
In that formulation, the helicity-summed channel (I_+) carries the non-dichroic response of the magnetic texture, whereas the dichroic channel (I_-) isolates charge–magnetic interference and, in the reflection geometry used for chiral Néel walls, is primarily sensitive to the in-plane magnetization inside the wall core [2606.02154].

For labyrinthine stripe domains, reciprocal-space analysis is central. Orientation disorder in real space produces a ring of resonant magnetic scattering rather than discrete Bragg spots. Measuring at the characteristic ring wavevector selects the dominant Fourier component of the domain pattern, so changes in domain contrast, domain periodicity, wall width, or wall angle appear directly as changes in the ring intensity, radius, azimuthal modulation, or peak shape. In transmission diffuse scattering from maze domains, the integrated intensity is used as a magnetization proxy,
[
M(\tau)\propto \sqrt{\int I(q,\tau)\,dq},
]
and the characteristic real-space period is related to the peak position by
[
d\approx \frac{2\pi}{q^*}.
]
This reduction makes it possible to track demagnetization and domain-network reorganization separately [2501.11506].

Time-resolved circular dichroism in XRMS adds a normalized wall-sensitive observable. In work on chiral walls in ([\mathrm{Pt}|\mathrm{Co}|\mathrm{Al}]_5), the decisive metric is the asymmetry ratio
[
\frac{CL-CR}{CL+CR},
]
which remains constant if the wall magnetization follows the same dynamics as the neighboring domains, but changes if the internal wall texture evolves independently [2007.08583].

At still broader scope, resonant diffuse scattering around antiferromagnetic Bragg points probes equal-time spin-spin correlations rather than only the long-range order parameter. In CuO, the diffuse magnetic signal is written as
[
I_\mathrm{RDS}(\mathbf{q}, t) \propto \sum_{\alpha \alpha^\prime} [\hat{\epsilon}i \times \hat{\epsilon}_o]^*\alpha [\hat{\epsilon}i \times \hat{\epsilon}_o]{\alpha^\prime} S_{\alpha \alpha^\prime}(\mathbf{q}, t),
]
with (S_{\alpha\alpha'}(\mathbf q,t)) the spin-spin correlation function. In leading order this quantity tracks magnon occupation numbers weighted by polarization and structure factors, so diffuse resonant magnetic scattering becomes a momentum-resolved probe of nonequilibrium spin fluctuations [2602.13113].

In FMR-driven resonant X-ray methods, the time dependence is commonly fitted as
[
S(t)=X\sin(2\pi \nu t)+Y\cos(2\pi \nu t),
]
from which amplitude and phase are extracted as
[
A=\sqrt{X^2+Y^2}, \qquad \psi = 2\arctan\left(\frac{Y}{A+X}\right).
]
This quadrature form is used in XFMR and extends naturally to diffraction and reflectometry ferromagnetic resonance, which are scattering-based dynamic probes in reciprocal space [2301.03256].

2. Experimental architectures and geometries

Published implementations span synchrotrons, free-electron lasers, and laboratory sources. A reflection-geometry tr-XRMS setup at BESSY II used the ALICE diffractometer on the UE52 beamline, with a magnetic field pulse as pump and delayed single-bunch synchrotron X-rays as probe. The single-bunch conditions were a 50 ps photon pulse width, 800 ns pulse separation, and 1.25 MHz repetition rate; the minimum electronic delay step was 10 ps, and the accessible scan range extended from about 100 ps to a few ns. The geometry probed the magnetization component collinear with the X-ray beam and was explicitly designed for thin films, multilayers, laterally structured samples, and temperature-dependent studies in reflection [1009.3389].

Free-electron-laser implementations pushed the method into the femtosecond regime. At the DiProI endstation of FERMI, one reflection-geometry pump–probe XRMS experiment used a 790 nm, (\sim 60) fs infrared pump and circularly polarized 22.75 nm, (\sim 60) fs FEL pulses at the Fe (M_{2,3}) edge, with incidence angle (\sim 45^\circ), delays up to 1 ns, 500 shots averaged per delay, and absorbed pump fluence from (1.5) to (4.5\ \mathrm{mJ\,cm^{-2}}) [2606.02154]. Another FERMI study of circular-dichroic XRMS on chiral walls used a 100 fs, 780 nm pump, a 60 fs XUV probe at the Co (M) edge near 60 eV, overall time resolution of about 120 fs, 45° incidence, and 50 Hz repetition rate [2007.08583].

Transmission-based resonant magnetic diffuse scattering has also been realized outside large-scale FEL facilities. A laboratory soft-X-ray pump–probe scattering beamline based on a laser-driven plasma source demonstrated resonant magnetic diffuse SAXS from a Fe/Gd maze-domain texture with 9 ps temporal resolution. The probe covered the Fe (L_{2,3}) and Gd (M_{4,5}) edges, the geometry was normal-incidence transmission, the source repetition rate was 100 Hz, the transported bandwidth was (<10) eV, the photon flux at the sample was (10^6)–(10^7\ \mathrm{ph/s/eV}), and delay scans reached 2 ns [2501.11506].

Tabletop high-harmonic generation extended tr-XRMS to rare-earth resonances. At the Tb (N) edge near 155 eV, a laboratory HHG diffractometer used a 1550 nm, 80 fs optical pump and a 5-eV-wide selected harmonic probe focused onto a (\mathrm{Co}{0.88}\mathrm{Tb}{0.12}) film, with far-field diffraction recorded on a CCD camera. The source covered 100–220 eV and achieved (>2\times 10^9) photons/s/1% bandwidth, enabling the first tabletop tr-XRMS experiment at the Tb (N) edge [1910.14263].

A separate scattering-based branch is GHz stroboscopic DFMR and RFMR. In these methods the synchrotron master clock near 500 MHz is phase-locked to the microwave drive, accessible frequencies lie in the 1–10 GHz range, the programmable delay resolution is about 0.5 ps, and the effective time resolution is set mainly by 35–70 ps X-ray pulse widths. DFMR detects the modulation of diffracted intensity at selected reciprocal-space points, whereas RFMR detects the modulation of resonant reflectivity as a function of
[
Q_z=\frac{4\pi}{\lambda}\sin\vartheta.
]
These are direct dynamic extensions of resonant magnetic diffraction and resonant magnetic reflectometry [2301.03256].

3. Reciprocal-space selectivity in domains, walls, multilayers, and antiferromagnets

The major strength of time-resolved XRMS is that the measured quantity is momentum selective rather than purely spatially averaged. In domain-textured films, the reciprocal-space signature itself encodes the morphology. Labyrinthine stripe domains produce a circular diffuse ring because the domain periodicity is well defined while the in-plane orientation is disordered. The ring radius gives the characteristic domain wavevector, the ring width reflects the distribution of periodicities and correlations, and azimuthal modulation of dichroic contrast distinguishes wall type and chirality [2501.11506][2606.02154].

For chiral Néel walls, dichroic XRMS resolves the wall-specific in-plane magnetization component. In the Fe-edge reflection geometry used for (\mathrm{Ir/CoFeB/MgO}) and (\mathrm{W/CoFeB/MgO}) multilayers, static dichroic scattering maxima at (90^\circ) and (270^\circ) around the ring were consistent with Néel walls of unique chirality, whereas maxima shifted to (135^\circ) and (315^\circ) indicated mixed Néel/Bloch character. In the corresponding Co-edge circular-dichroic XRMS study, the equilibrium sign pattern of (CL-CR) around the ring identified clockwise homochiral Néel walls. In both cases the reciprocal-space dichroic pattern encoded an internal wall degree of freedom that is not equivalent to the average out-of-plane domain magnetization [2606.02154][2007.08583].

In stripe-domain ferrimagnets the same reciprocal-space logic appears in a grating form. At the Tb (N) edge, alternating oppositely magnetized stripes in (\mathrm{Co}{0.88}\mathrm{Tb}{0.12}) acted as a magnetic diffraction grating, giving (+1)st-order diffraction peaks whose intensity tracked magnetic contrast and whose momentum transfer tracked the stripe periodicity. This was explicitly interpreted as simultaneous access to demagnetization and domain expansion through the same diffraction dataset [1910.14263].

In multilayers, specular resonant reflectivity provides depth sensitivity rather than in-plane periodicity. RFMR on ([\mathrm{CoFeB}/\mathrm{MgO}/\mathrm{Ta}]_4) used the (Q_z) dependence of the dynamic reflected intensity to infer a (5^\circ) phase lag between four magnetic layers. The point is not lateral imaging but interference-sensitive depth discrimination within a buried stack [2301.03256].

In antiferromagnets the distinction between Bragg and diffuse sectors is especially sharp. In CuO, the resonant magnetic Bragg reflection ((\tfrac12\,0\,-\tfrac12)) measured the long-range AFM order parameter, while resonant diffuse scattering around that point measured momentum-resolved magnetic fluctuations. The diffuse upturn was larger for (\pi) than (\sigma) polarization, appeared only once ultrafast demagnetization was triggered, and was separated from phonon diffuse scattering by complementary non-resonant measurements. This established that time-resolved resonant magnetic scattering can access the fluctuation sector in (\mathbf q)-space, not only the static order parameter [2602.13113].

4. Dynamical regimes accessible to time-resolved XRMS

One well-established regime is field-driven precession. In reflection-geometry tr-XRMS at BESSY II, a 10 ns field pulse launched damped free precession in a 25 nm Py layer on a Cr stripline. The reflected resonant intensity showed a step at the pulse edge and damped oscillations at both leading and trailing edges; the observed frequencies were in the low GHz regime and the oscillations persisted for a few nanoseconds. Measurements at both the Fe (L_3) and Ni (L_3) edges demonstrated element-resolved precessional dynamics [1009.3389].

A second regime is ultrafast demagnetization coupled to reciprocal-space reorganization of a domain network. In the laboratory Fe/Gd resonant diffuse-scattering experiment, the resonant SAXS intensity dropped rapidly after optical pumping, and the corresponding magnetization proxy fell to (M\approx 0.56) for the Fe-edge example at (25\ \mathrm{mJ/cm^2}), remaining nearly constant until roughly 1 ns before remagnetization began. Simultaneously, the first-order peak position first shifted to larger (q), implying smaller real-space periodicity, and later reversed toward smaller (q), indicating an increase in periodicity. The authors emphasized that this sign reversal of the transient peak shift had not been seen previously in that context [2501.11506].

A third regime is wall-specific ultrafast dynamics. In Fe-edge time-resolved XRMS on chiral Néel walls embedded in a disordered stripe texture, the helicity-summed channel (I_+) showed ultrafast demagnetization, recovery, and in some cases a weak oscillatory contribution assigned to laser-launched coherent surface phonons. By contrast, the dichroic channel (I_-) exhibited a pronounced damped oscillatory response over the first few hundred picoseconds, with frequency decreasing as pump fluence increased. The wall-sensitive signal was modeled through
[
I_{-}(t) \propto M_s(t)\,\Delta(t)\,\cos\phi(t)\,\mathrm{sech}!\left(\frac{\pi Q_M \Delta(t)}{2}\right),
]
so the strong dichroic oscillation was interpreted as an internal domain-wall mode dominated by oscillations of the wall angle (\phi(t)), with additional sensitivity to wall width (\Delta(t)). The observed frequency softening was attributed to pump-induced changes in effective anisotropy and saturation magnetization that reduce the restoring field of the wall mode [2606.02154].

An earlier Co-edge CD-XRMS study resolved an even earlier wall-texture transient. After 780 nm pumping, both (CL+CR) and (CL-CR) decreased, but the normalized asymmetry ratio showed a (\sim 15\%) dip at about 0.7 ps and remained below its initial value up to around 2 ps. Modeling required a transient reduction of the wall’s effective Néel chirality, described as a mixed Bloch–Néel–Bloch profile with stronger demagnetization inside the wall than in the neighboring domains. This established that time-resolved dichroic XRMS can follow ultrafast evolution of the internal spin texture of a wall, not just domain periodicity or average magnetization [2007.08583].

Tabletop HHG-based tr-XRMS at the Tb (N) edge accessed both demagnetization and wall-mediated domain expansion in a rare-earth ferrimagnet. For pump fluence (8\ \mathrm{mJ/cm^2}), the diffraction-peak intensity was suppressed by about 70%, corresponding to demagnetization up to about 50% because (M\propto\sqrt{I_{\mathrm{diff}}}). The momentum transfer decreased by about 3% in roughly 10 ps, indicating domain expansion perpendicular to the stripe direction, and the inferred domain-expansion velocity was (\sim 750\ \mathrm{m/s}) [1910.14263].

At the antiferromagnetic end of the field, time- and momentum-resolved resonant diffuse scattering in CuO revealed a hierarchical sequence: a (\sim 70) fs drop of magnetic Bragg intensity, broad nonthermal magnon generation, picosecond magnon quasi-thermalization, and (\sim 7) ns recovery governed by momentum-selective magnon–phonon scattering. The essential advance was that the diffuse resonant magnetic channel tracked where the spin entropy flowed in reciprocal space after the order parameter collapsed [2602.13113].

5. Sources, sensitivity, time scales, and analysis constraints

Time-resolved XRMS is unusually sensitive to source characteristics. Synchrotron implementations provide mature reciprocal-space control but are typically pulse-width limited in the tens-of-picoseconds range. The 2010 BESSY II reflection setup stated explicitly that processes faster than 50 ps were not accessible because the single-bunch photon pulse width was 50 ps [1009.3389]. The XFMR/DFMR/RFMR review similarly places typical synchrotron pulse widths at (\sim 35) ps at Diamond and BESSY and (\sim 70) ps at ALS, even though delay-line steps can be much smaller [2301.03256].

Free-electron lasers enable sub-ps and fs-resolved scattering. FERMI-based experiments on chiral walls used (\sim 60) fs XUV pulses and accessed the sub-ps to ps regime directly in reciprocal space [2606.02154][2007.08583]. SwissFEL-based resonant diffuse scattering in CuO reached (\sim 60) fs FWHM time resolution and made momentum-dependent magnetic fluctuation kinetics experimentally accessible [2602.13113].

Laboratory platforms trade temporal resolution against accessibility. The laser-driven plasma source achieved 9 ps temporal resolution, shot-resolved normalization, and sufficient flux for weak diffuse magnetic scattering, with typical delay scans using (10^4) X-ray pulses or 100 s per time point and a 40-point scan taking at most 60 min. The authors stressed that multidimensional scans over fluence, edge, field, and temperature become feasible in a normal laboratory workflow under these conditions [2501.11506]. The HHG implementation emphasized a different tradeoff: weak XUV magnetic scattering requires high photons per pulse, but low repetition rate is beneficial because it suppresses cumulative sample heating. In that work the laser repetition rate was deliberately reduced from 2 kHz to 500 Hz to avoid thermal damage [1910.14263].

The analysis burden is correspondingly high. In wall-resolved studies, interpretation can be indirect because the measured dichroic signal depends simultaneously on (M_s(t)), (\Delta(t)), and wall angle (\phi(t)); the 2026 Fe-edge study noted explicitly that exact separation of these contributions was not complete, that the mode-frequency interpretation relied on a scaling argument rather than a full microscopic theory, and that the damping origin was not identified microscopically [2606.02154]. In reciprocal-space domain-network measurements, early-time peak broadening and late-time peak shifts had to be disentangled to separate inhomogeneous domain rearrangement from simple period changes [2501.11506]. In RFMR and XFMR, fitted phases are relative rather than absolute unless the full timing chain is calibrated, because cable delays, electronics, and beamline timing all contribute to the measured phase [2301.03256].

Several recurring misconceptions are therefore corrected by the literature. Time-resolved XRMS is not restricted to specular reflectivity; it includes diffuse SAXS from domains, diffraction from periodic spin textures, reflectometry from multilayers, and resonant diffuse scattering from magnetic fluctuations [2501.11506][2301.03256][2602.13113]. Nor does it measure only average magnetization: it can resolve domain periodicity, wall chirality, internal wall motion, reciprocal-space phase lags in multilayers, and momentum-resolved magnon populations [2007.08583][2606.02154][2301.03256][2602.13113].

6. Relation to adjacent methods and conceptual boundaries

Several closely related time-resolved resonant X-ray techniques are methodologically important but are not themselves canonical XRMS. A BL07LSU endstation at SPring-8 demonstrated time-resolved XMCD on FePt with under-50-ps-class timing and described a time-resolved resonant soft-X-ray scattering capability, but the reported dynamics were XMCD rather than a completed magnetic scattering dataset. Its direct XRMS contribution is infrastructural: synchronization, detector gating, soft-X-ray resonant operation, and sample geometry for future scattering experiments [1701.03156].

Single-shot off-axis zone-plate streaking at FERMI demonstrated resonant magnetic X-ray absorption with a 1.57 ps encoded time window and percent-level sensitivity in transmission. That work is not XRMS, but it addresses the central limitation of stroboscopic averaging for irreversible dynamics. A plausible implication is that analogous time encoding could become relevant for future single-shot XRMS if momentum and time coordinates can be disentangled on the detector [1811.05917].

Time-resolved STXM-FMR is similarly adjacent rather than identical. Its contribution is the rigorous extraction of a true dynamic magnetic signal from resonant transmission data by membrane normalization, log-ratio conversion,
[
\ln!\left(\frac{I_s^{on}}{I_s^{off}}\right)=\Delta\mu\cdot t,
]
and helicity-reversal checks. The methodological lesson transfers directly to XRMS: resonant contrast must be separated carefully from non-magnetic backgrounds rather than assumed to be magnetic simply because the photon energy is resonant [1905.12497].

At the theoretical boundary, ultrafast resonant scattering from nonstationary electronic states shows that broadband probes can mix elastic and inelastic channels inseparably. That framework is not a magnetic XRMS theory, but it implies that in an attosecond or few-femtosecond limit the measured resonant scattering pattern need not reduce to an instantaneous structure factor. This suggests caution in extrapolating stationary XRMS intuition into the ultrabroadband ultrafast regime [1505.07452].

Taken together, the literature presents time-resolved XRMS as a reciprocal-space-resolved, element-specific, polarization-sensitive probe of magnetic dynamics spanning at least three experimentally distinct regimes: tens-of-picoseconds stroboscopic precession at synchrotrons, femtosecond wall and fluctuation dynamics at FELs, and increasingly capable laboratory scattering platforms in the 9 ps to fs domain. The core technical advance across these forms is not merely faster time stamping of magnetic scattering, but selective access to specific Fourier components and specific magnetization components of complex magnetic textures, including disordered domain networks, chiral walls, ordered spin structures, multilayer depth profiles, and nonequilibrium magnetic fluctuations [1009.3389][2501.11506][1910.14263][2606.02154][2602.13113].