Time-Resolved X-Ray Bragg Diffraction

• Time-resolved X-ray Bragg diffraction is a method that uses time-dependent Bragg reflections to probe transient changes in lattice spacing, strain, and electronic order.
• It integrates diverse X-ray sources—from synchrotron pulse shaping to XFEL single-pulse readouts—with advanced detector technologies to capture ultrafast structural dynamics.
• Advanced analysis techniques, including phase retrieval and quantum-field formalisms, enable precise inversion and modeling of evolving material properties.

Time-resolved X-ray Bragg diffraction is the use of Bragg reflections as explicitly time-dependent observables after a controlled perturbation, most commonly in pump-probe form, to track transient changes in lattice spacing, strain, domain structure, magnetic or electronic order, and nonstationary quantum states. In contemporary practice it spans synchrotron experiments with picosecond pulse shaping, laboratory measurements based on high-harmonic generation or laser-plasma sources, and XFEL implementations capable of single-pulse structural readout, while corresponding theoretical descriptions range from kinematical peak-shift analysis to quantum-field-based correlation-function formalisms (Sander et al., 2018, Jarecki et al., 2024, Afshari et al., 2019, Bortel et al., 2023, Yuan et al., 30 Apr 2026).

1. Diffraction observables and measurement principles

The central geometric condition remains Bragg’s law,

$2d\sin\theta = n\lambda,$

with time dependence entering through the transient evolution of the lattice spacing $d$, the scattering geometry, or the underlying structure factor. In energy-dispersive implementations, the out-of-plane momentum transfer is written as

$$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$

so that time-dependent peak shifts can be transformed directly into reciprocal-space trajectories (Jarecki et al., 2024).

The experimentally monitored quantities are not restricted to peak position. Time-resolved Bragg diffraction routinely follows changes in full width at half maximum, integrated intensity, reciprocal-space-map anisotropy, and, in phase-retrieval settings, complex real-space profiles. For ferroelectric thin films, peak broadening is expressed through the normalized width

$\tilde w=\frac{w(t)-w(0)}{w(0)},$

while the relative $c$-axis change is

$\eta=\frac{\Delta c(t)}{c(0)},$

allowing structural disorder and piezoelectric strain to be separated operationally (Kwamen et al., 2018). In soft-X-ray superlattice diffraction, the strain is extracted from reciprocal-space shifts through

$$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$

which makes Bragg-peak motion itself the primary dynamical observable (Jarecki et al., 2024).

Asymmetric reflection pairs provide an additional layer of tensor sensitivity. In BiFeO$_3$, the $(\pm h01)_c$ reflections respond differently to off-diagonal strain components because ferroelectric symmetry breaking lifts the equivalence that would hold in a cubic crystal. From the measured shifts of $d^*_{h01}$ and $d$0, both longitudinal and shear strain can be extracted simultaneously, making time-resolved Bragg diffraction sensitive not only to dilation but also to ultrafast unit-cell distortion (Juvé et al., 2020).

2. Temporal resolution strategies and source classes

A defining constraint in storage-ring Bragg diffraction is the intrinsic synchrotron pulse duration, stated as typically $d$1–$d$2 ps. The picosecond Bragg switch, or PicoSwitch, addresses this by using photoacoustically induced transient Bragg matching in a multilayer thin film, so that only a narrow temporal slice of a longer synchrotron pulse is diffracted. In the demonstrated ESRF experiment, a $d$3 ps incident pulse was shortened to about $d$4 ps FWHM, with peak on-state diffraction efficiency $d$5, switching contrast $d$6, area loss factor $d$7, and total contrast $d$8; the switching window is governed by $d$9 (Sander et al., 2018).

Laboratory femtosecond operation is realized differently. In the water-window soft-X-ray experiment, a $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$0 kHz mid-infrared OPCPA system drives helium high-harmonic generation to produce broadband pulses from $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$1 to $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$2 eV with probe duration $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$3 fs FWHM, while the effective temporal resolution is about $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$4 fs because it is limited by pulse cross-correlation and geometry. The key methodological decision is to avoid monochromatization and use energy-dispersive detection, thereby preserving photon flux from the broadband HHG source (Jarecki et al., 2024).

A separate laboratory route uses short-pulse laser-plasma emission. In the Ti K$$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$5 setup, a sub-$$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$6 fs Ti:Sapphire CPA laser generates $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$7 keV radiation from a Ti wire target, and a toroidally bent Ge crystal performs collection, monochromatization, and focusing. The system delivers $$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$8 Ti-K$$Q_z = \frac{4\pi E_{\text{ph}}}{h c_0}\sin\theta, \qquad \vec Q=\vec k_{\text{out}}-\vec k_{\text{in}}=n\vec G,$$9 photons per pulse at $\tilde w=\frac{w(t)-w(0)}{w(0)},$0 Hz with relative bandwidth $\tilde w=\frac{w(t)-w(0)}{w(0)},$1, and the time resolution is $\tilde w=\frac{w(t)-w(0)}{w(0)},$2–$\tilde w=\frac{w(t)-w(0)}{w(0)},$3 ps, limited by pump-probe geometry rather than by the X-ray pulse duration (Afshari et al., 2019).

At longer timescales, temporal resolution and flux are traded differently. Hadamard-encoded timing schemes use patterned probe pulse sequences generated by the WaveGate solid-state pulse picker. Instead of probing a single delay per excitation, multiple time windows are encoded in one acquisition and then decoded through

$\tilde w=\frac{w(t)-w(0)}{w(0)},$4

The reported demonstrations covered $\tilde w=\frac{w(t)-w(0)}{w(0)},$5 ms windows in the millisecond regime and $\tilde w=\frac{w(t)-w(0)}{w(0)},$6 ns windows in the microsecond regime, with switching as fast as $\tilde w=\frac{w(t)-w(0)}{w(0)},$7 ns (Schmidt et al., 13 May 2025).

XFEL and attosecond-oriented schemes extend the same logic to the opposite extreme. A single $\tilde w=\frac{w(t)-w(0)}{w(0)},$8 fs XFEL pulse at $\tilde w=\frac{w(t)-w(0)}{w(0)},$9 keV was used to collect hundreds of Bragg reflections in parallel through Kossel line patterns, enabling structure solution from one shot (Bortel et al., 2023). Convergent-beam attosecond X-ray crystallography proposes that dispersive high-NA optics can encode time directly into Bragg streaks; in one experimental geometry discussed there, the accessible span is $c$0 fs with a minimum per-pixel temporal resolution of $c$1 as and an average of $c$2 fs (Chapman et al., 2024).

3. Beamline geometries, optics, and detectors

The canonical geometry remains pump-probe, but the implementation varies strongly with source class and observable. In the PicoSwitch configuration at ESRF ID09, two synchronized optical pulses derived from the same laser are used: one excites the switch and defines the temporal slice of the X-ray probe, and the other excites the sample. Because both optical beams originate from the same oscillator, the pump-probe jitter is negligible. The switch is placed as a Bragg optic in the beamline, and its acceptance angle is several times larger than the focused X-ray beam divergence, so the downstream beam properties are not distorted (Sander et al., 2018).

Energy-dispersive soft-X-ray diffraction uses a different architecture. The reported setup combines a $c$3 goniometer with a rotatable spectrometer containing a variable-line-spacing grating and an in-vacuum CCD camera, with about $c$4 eV energy resolution at $c$5 eV. Pump and probe are quasi-collinear with a $c$6 angle between them, and the detector records the diffracted spectrum across a broad photon-energy range rather than at a single monochromatic setting. This allows a large slice of reciprocal space to be captured in each exposure (Jarecki et al., 2024).

In the laser-plasma instrument, the critical optical element is the toroidally bent Ge(100) crystal used in its (400) reflection. The Rowland-circle condition,

$c$7

is used for $c$8 imaging of source and sample, while the apparent rocking-curve width is decomposed as

$c$9

Because the beam is convergent, the full rocking curve can be recorded in one shot rather than by mechanical scanning (Afshari et al., 2019).

Detector technology is similarly heterogeneous. Time-correlated single-photon counting with gated detection was used in ferroelectric thin-film measurements to obtain time-resolved rocking curves in the sub-microsecond regime, combining a photomultiplier, a sub-nanosecond scintillator, and a PicoHarp 300 with static reciprocal-space maps from a PILATUS 100K detector (Kwamen et al., 2018). In BiFeO$\eta=\frac{\Delta c(t)}{c(0)},$0, a two-dimensional XPAD3.2 detector recorded full reciprocal-space maps of asymmetric Bragg peaks, which is essential when the strain tensor is inferred from peak trajectories rather than from scalar intensity changes (Juvé et al., 2020).

Fast pulse segregation at synchrotrons requires detector-side timing. An analog integrating pixel array detector with in-pixel storage achieved around $\eta=\frac{\Delta c(t)}{c(0)},$1 ns temporal resolution, sufficient to isolate bunch trains with $\eta=\frac{\Delta c(t)}{c(0)},$2 ns separation at CHESS and compatible with bunch spacing at APS and ESRF. Its eight in-pixel analog storage elements enable rapid sequential frame capture, and the device was explicitly identified as appropriate for time-resolved Bragg spot single-crystal experiments (Koerner et al., 2010). At the single-shot structural frontier, a $\eta=\frac{\Delta c(t)}{c(0)},$3 megapixel Jungfrau detector recorded Kossel line patterns from a single XFEL pulse in forward transmission geometry (Bortel et al., 2023).

4. Signal theory, inversion, and dynamical diffraction

The interpretation of time-resolved Bragg diffraction depends on whether the measured signal is dominated by coherent inter-molecular scattering at reciprocal-lattice positions or by single-particle diffuse contributions. A unified formulation separates the signal into a two-molecule term,

$\eta=\frac{\Delta c(t)}{c(0)},$4

and a one-molecule term,

$\eta=\frac{\Delta c(t)}{c(0)},$5

For traditional Bragg diffraction, $\eta=\frac{\Delta c(t)}{c(0)},$6 dominates and scales as $\eta=\frac{\Delta c(t)}{c(0)},$7; in nonstationary states it contains cross-terms between ground- and excited-state amplitudes, giving heterodyne interference at Bragg peaks. In weak excitation, the ground-state scattering acts as an in situ local oscillator for the excited-state signal (Bennett et al., 2016).

A recurrent interpretive issue is whether ultrafast diffraction measures an instantaneous charge density. In the QED treatment of off-resonant single-molecule scattering, the coherent contribution is proportional to the modulus square of a charge-density amplitude, but the incoherent contribution involves a two-point correlation function of the charge-density operator. The stationary elastic limit reduces to genuine diffraction; outside that limit, especially for single molecules and inelastic channels, the signal is not simply the modulus square of a time-dependent density (Bennett et al., 2014).

Finite crystals introduce another level of complexity through dynamical diffraction. Numerical solutions of modified Takagi-Taupin equations show that refraction, absorption, and multiple scattering can generate strong artifacts in Bragg coherent X-ray diffractive imaging when the crystal size becomes comparable to or exceeds the extinction length. The practical criterion is that the kinematical approximation is valid, apart from refraction and absorption, when the relevant crystal dimension $\eta=\frac{\Delta c(t)}{c(0)},$8 is smaller than the extinction length $\eta=\frac{\Delta c(t)}{c(0)},$9; once $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$0, dynamical coupling must be modeled explicitly. An analytical correction factor $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$1 can remove refraction- and absorption-induced phase artifacts after reconstruction (Shabalin et al., 2017).

Time-resolved Bragg diffraction also increasingly relies on inversion rather than direct peak fitting. In one-dimensional diffraction imaging, the measured intensity is

$$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$2

so phase retrieval is required to reconstruct the depth profile $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$3. Because one-dimensional phase retrieval is nonunique, time-spliced diffraction imaging uses temporal continuity as an added constraint, aligning neighboring reconstructions with the metric

$$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$4

This turns a sequence of rocking curves into a self-consistent movie of depth-dependent order-parameter evolution (Beyerlein, 2017).

At the most general level, the unified 2026 quantum-field-based formalism expresses time-resolved X-ray diffraction and ultrafast electron diffraction within the same framework: the measured signal is a two-time, two-point correlation function of the target weighted by the first-order coherence function of the probe beam. In that description, density-current and current-current couplings can be incorporated consistently, making explicit which dynamical features are common to TR-XRD and TR-UED and which arise from probe-specific interaction mechanisms (Yuan et al., 30 Apr 2026).

5. Structural dynamics across material systems

A major use of time-resolved Bragg diffraction is the direct readout of coherent acoustic dynamics. The PicoSwitch experiment demonstrated this in nanometer-thin epitaxial films by shaping a long synchrotron pulse into a $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$5 ps probe and using it to monitor propagating sound waves. The same work established that the short pulse could be characterized quantitatively from diffraction data, including pulse duration, efficiency, and switching contrast, and argued that the method can deliver up to $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$6 photons/sec in high-repetition-rate synchrotron experiments (Sander et al., 2018).

In a soft-X-ray water-window experiment on a Mo/Si superlattice composed of $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$7 double layers of $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$8 nm Mo and $$\varepsilon(t)=\frac{\Delta d(t)}{d_0}\approx-\frac{\Delta Q_z(t)}{Q_z^0},$$9 nm amorphous Si, the first-order superlattice Bragg peak shifted to lower $_3$0 after optical excitation, indicating transient lattice expansion. The strain dynamics were extracted by reciprocal-space transformation and compared with udkm1Dsim simulations. The measured shift exceeded the average strain and instead matched the strain averaged over the X-ray information depth of about $_3$1 nm at $_3$2 eV, demonstrating the surface sensitivity of the method (Jarecki et al., 2024).

Ferroelectric switching provides a distinct application in which Bragg peak widths and intensities encode domain nucleation and disorder rather than only elastic strain. In a $_3$3 nm PZT capacitor, sub-coercive pulses of $_3$4 to $_3$5 V caused the normalized Bragg width to increase slowly and saturate after about $_3$6s, reaching a $_3$7 increase, while the normalized peak intensity decreased by up to $_3$8. These changes persisted after the field was removed, evidencing a remanent disordered domain state that could be erased by an appropriate pulse sequence. For above-coercive pulses of $_3$9 or $(\pm h01)_c$0 V, the width first increased and then decreased again as domains grew and merged (Kwamen et al., 2018).

In multiferroic BiFeO$(\pm h01)_c$1, scrutiny of the asymmetric $(\pm h01)_c$2 and $(\pm h01)_c$3 peaks showed that longitudinal and shear strains do not evolve identically. The longitudinal strain peaked about $(\pm h01)_c$4 ps after excitation, the shear strain about $(\pm h01)_c$5 ps, and both reached a quasi-steady plateau after about $(\pm h01)_c$6 ps. The measured amplitude ratio $(\pm h01)_c$7 and the paper’s comparison with pure-mechanism estimates led to the conclusion that both thermal and non-thermal processes contribute to acoustic-phonon photogeneration, while the temporal offset was attributed to the interplay of quasi-longitudinal and quasi-transverse acoustic modes (Juvé et al., 2020).

Time-resolved Bragg diffraction can also be extended to depth-resolved order-parameter imaging. In resonant soft-X-ray diffraction from a $(\pm h01)_c$8 nm NdNiO$(\pm h01)_c$9 film, time-spliced phase retrieval revealed that mid-infrared excitation of the substrate launched a demagnetization front from the interface: after $d^*_{h01}$0 ps the interface was demagnetized, the front propagated into the film and stagnated at about $d^*_{h01}$1 nm after about $d^*_{h01}$2 ps, and the extracted velocity was about twice the speed of sound in NdNiO$d^*_{h01}$3 (about $d^*_{h01}$4 m/s). Near-infrared excitation, by contrast, produced a more uniform reduction of antiferromagnetic order (Beyerlein, 2017).

6. Limitations, misconceptions, and emerging directions

The principal experimental limitation is that temporal sharpening and reciprocal-space selectivity usually cost photons. For the PicoSwitch, the overall photon efficiency in the reported experiment was about $d^*_{h01}$5, and although larger bandwidth increases flux nearly linearly, it reduces switching contrast because the broader energy acceptance is less well matched to the transient Bragg shift. The same device does not isolate single bunches from the storage-ring train, so additional choppers or suitable detectors are required, and sustained high fluence at high repetition rate requires thermal management (Sander et al., 2018).

Energy-dispersive and broadband methods avoid monochromator losses but shift the burden to normalization and interpretation. In soft-X-ray HHG diffraction, spectral overlap can complicate analysis when multiple Bragg peaks fall in the detection range, and absolute intensity observables such as Debye-Waller factors are difficult to quantify without rapid and accurate flux normalization or absolute calibration (Jarecki et al., 2024). In laser-plasma diffraction, the high per-pulse flux is offset by a $d^*_{h01}$6 Hz repetition rate, the photon energy is constrained by the accessible K$d^*_{h01}$7 lines and optics, and the time resolution is geometry-limited rather than pulse-limited (Afshari et al., 2019). At synchrotrons, Hadamard-encoded timing schemes mitigate the flux–resolution trade-off but, in the present demonstrations, operate on microsecond and millisecond rather than picosecond scales (Schmidt et al., 13 May 2025).

On the theoretical side, two interpretive misconceptions recur. The first is that heterodyne terms are generically present in any ultrafast diffraction experiment; the formal analyses instead show that robust heterodyne interference is tied to ordered systems and Bragg peaks, whereas in gases and other disordered phases the corresponding two-molecule contributions collapse except at zero momentum transfer (Bennett et al., 2016). The second is that any time-resolved scattering pattern is a direct image of an evolving charge density; for single-particle or inelastic channels, correlation functions rather than densities are the fundamental observables (Bennett et al., 2014).

The frontier of the field is defined by single-shot completeness and intrinsic time encoding. In a forward-transmission XFEL experiment, about $d^*_{h01}$8 unique structure factors were sufficient to reconstruct GaAs and GaP up to $d^*_{h01}$9 Å resolution from the data collected during a single $d$00 fs pulse, using Kossel line profiles that encode both amplitudes and phases of structure factors (Bortel et al., 2023). Convergent-beam attosecond crystallography pushes this idea further by replacing discrete Bragg spots with time-encoded Bragg streaks generated by dispersive optics such as multilayer Laue lenses, aiming at sub-femtosecond and eventually attosecond crystallography in which arrival time is mapped directly onto detector position (Chapman et al., 2024). In parallel, the unified TR-XRD/TR-UED framework indicates that future diffraction imaging will increasingly target not only lattice motion but also transient charge and current correlations, as illustrated by simulations of laser-driven graphene in which density-current terms can dominate selected diffraction features (Yuan et al., 30 Apr 2026).