LED-pump-X-ray-multiprobe crystallography for sub-second timescales

The study addresses how to monitor solid-state chemical and biological processes in real time across millisecond-to-minute timescales using single-crystal X-ray diffraction. Traditional photocrystallography and time-resolved methods have focused on microsecond and faster dynamics (often using lasers and Laue diffraction), which pose challenges including uneven illumination, photo-bleaching, crystal heating, high X-ray flux requirements, and difficulty solving unknown intermediates from Laue data. There is a methodological gap for slower processes (milliseconds to minutes) relevant to phenomena like protein folding, ligand binding, phase transitions, and radiation damage. The authors propose an LED-pump-X-ray-multiprobe SCXRD technique that enables sub-second time resolution, higher excited-state populations, minimized damage, and complete 3D datasets to unambiguously determine structures and follow kinetics.

State-of-the-art time-resolved macromolecular crystallography uses Laue diffraction at synchrotrons (picoseconds) and XFELs (femtoseconds), but structure solution is challenging and often requires prior models, hindering identification of unknown intermediates. Molecular crystallography typically employs monochromatic X-rays due to flux and analysis constraints, with a few Laue studies reaching microsecond lifetimes. Existing pump–probe experiments are typically synchronized to pulsed lasers with nanosecond–femtosecond pulses, imposing constraints on timing and requiring high X-ray flux between pulses; they risk photo-bleaching, uneven illumination, and sample heating. Prior work on millisecond timescales has used gated detectors to monitor selected Bragg reflections, insufficient for full structure solution. Recent advances in synchrotron brightness and gated photon-counting detectors motivate new approaches that can capture complete datasets with sub-second resolution.

• Model system: [Pd(Bu4dien)(NO2)][BPh4] (1; Bu4dien = N,N,N',N'-tetrabutyldiethylenetriamine; BPh4 = tetraphenylborate). Under 400 nm light, GS nitro-(η1-NO2) converts to ES endo-nitrito-(η1-ONO); reverse is thermally activated and strongly temperature-dependent, enabling matching of ES lifetime to measurement timescale.
• Pump-multiprobe strategy: Use a bespoke pulsed LED array as pump and an electronically gated Pilatus 300K detector as probe. Define cycles of duration t_cyc = t_exc + t_dec with LEDs on for excitation (t_exc) and off for decay (t_dec). During each cycle, rotate the crystal back-and-forth through the same φ-range Δφ multiple times; gate the detector to record separate diffraction images for multiple time windows within each cycle. Acquisition time per image t_acq (set by gate length and rotation speed) defines temporal averaging of each 3D snapshot. Repeat cycles while incrementing φ to build a 180° scan, yielding a series of time-stamped 3D datasets in a single experiment (<120 min total).
• Tested sequences: Five t_cyc from 14 to 170 s with corresponding t_acq from 0.4 to 8.0 s; example parameters include (t_cyc, t_exc, t_dec, Δφ, t_acq): (170 s, 55 s, 115 s, 1.6°, 8.0 s); (108 s, 35 s, 73 s, 8.0°, 4.0 s); (35 s, 14 s, 21 s, 3.2°, 1.6 s); (22 s, 8 s, 14 s, 1.6°, 0.8 s); (14 s, 5 s, 9 s, 0.8°, 0.4 s).
• Experimental design and simulations: Numerical simulations parameterized by conventional SCXRD excitation/decay kinetics and steady-state ES populations guided selection of t_cyc, t_exc, t_dec, and temperature (260–290 K suggested) to maximize ES/GS ratio at the target timescales.
• Hardware and excitation source: 3D-printed part-spherical LED holder with up to 25 T-1 (3 mm) LEDs; used 18 Bivar UV3TZ-400-30 LEDs (λ ≈ 400 nm). LED sphere mounted below the sample; crystal-to-detector distance 9.57 cm; uniform illumination aided by rotation about φ. Measured power at crystal: 23 mW. LED rise/fall times negligible on experiment timescale; pulse separations down to ~200 µs achievable. Timing synchronization implemented via beamline control software.
• Beam and damage control: Pre-tests established X-ray-induced excitation; experiments used 5% beam intensity to limit baseline X-ray excitation to ~3%. LED over-irradiation (~4 h) leads to surface damage and photobleaching; thus a fresh crystal was used per run.
• Data collection and processing: XPS four-circle diffractometer, Pilatus 300K detector, Oxford Cryostream 7 for temperature control. Automated, on-the-fly sorting by time, indexing and integration via DIALS through xia2, scaling/absorption via AIMLESS, structure solution with SHELXT and refinement with SHELXL. ES populations α(t) refined with a standard disorder model (SHELX PART). Photo-difference maps computed between fixed GS coordinates and ES data to visualize electron-density changes.
• Kinetic modeling: Johnson–Mehl–Avrami–Kolmogorov model applied to excitation (GS ↔ ES) and decay (ES → GS), with Avrami exponents fixed to unity; simulation tool fitted pump–probe data to extract rate constants.
• Timepix3 proof-of-concept: Single-module Timepix3 hybrid active pixel detector (25 ns timestamp accuracy; continuous readout of ToA/ToT) used to follow a single Bragg reflection (−2 1 0) identified from SCXRD as sensitive to ES population. Photon events integrated into 50 ms bins; timing synchronized to pump cycles. Demonstrated qualitative tracking of ES dynamics and photobleaching across cycles; quantitative kinetics inferior to full SCXRD but paves way for larger-area Tristan 1M/10M Timepix3 detectors to collect complete datasets without fixed t_acq.

• Developed LED-pump-X-ray-multiprobe SCXRD enabling sub-second time resolution and complete 3D datasets within <120 minutes per experiment.
• Achieved high excited-state conversions for [Pd(Bu4dien)(NO2)][BPh4] at optimized temperatures/timing: α_max values of 33.7% (t_acq 8 s), 21.3% (4 s), 10.7% (1.6 s), 11.8% (0.8 s), and 10.4% (0.4 s). Across 12 experiments, refined ES populations were 68–98% (mean 84%) of predicted continuous-illumination steady-state maxima at the experiment temperature, indicating near-theoretical photoconversion.
• Data quality: For the nine datasets at shortest t_acq, R_int = 9.16–9.85% (mean 9.55%); R1 = 6.24–6.41% (mean 6.32%). Wilson plots across time points showed no significant deterioration or B-factor changes at each temperature, indicating limited damage.
• Photo-difference maps visualized time-resolved electron-density changes with 400 ms resolution; shorter t_acq reduced map quality due to flux-limited diffraction intensity but still permitted reliable structure solutions.
• LED pumping improved ES fractions relative to traditional laser pump-probe studies (few percent), likely due to uniform illumination and reduced heating/bleaching, and allowed flexible control of pulse width/intensity.
• Temperature rise during LED illumination was negligible: Photo-Wilson temperature scale factor average 1.069 ± 0.028 across 12 datasets.
• Proof-of-concept Timepix3 measurements of the (−2 1 0) reflection reproduced ES dynamics qualitatively and revealed photobleaching over longer t_exc; continuous readout removes detector dead time and allows post hoc adjustment of t_acq.
• Computational NEB pathway between nitro-(η1-NO2) and endo-nitrito-(η1-ONO) found a shallow local minimum with exo-nitrito geometry requiring only 6.6 kJ mol−1 to continue to endo-nitrito compared to an overall activation energy of ~114 kJ mol−1, suggesting the exo intermediate is not long-lived at near-ambient temperatures.
• Operational constraints identified: Pilatus 300K readout (~3 ms) limits sub-10 ms gating; X-ray-induced excitation managed by attenuating beam to 5% (baseline ~3%).

The method addresses the gap in time-resolved SCXRD for millisecond-to-minute processes by enabling multiple time delays to be captured within each pump cycle, ensuring complete 3D structural information for each time point with minimized cumulative damage and consistent comparability across delays. Using pulsed LEDs provides uniform illumination, tunable pulse widths and intensities, and reduced heating relative to lasers, improving ES populations and data quality. The approach directly visualizes electron-density changes over time, allowing unambiguous identification of transient species and extraction of kinetic/mechanistic information. For the model system, only the known endo-nitrito ES was observed between 260–284 K, consistent with a simple binary GS/ES switch; computed pathways indicate any exo-nitrito intermediate is too short-lived to impact operation at near-ambient temperatures. The proof-of-concept Timepix3 measurements validate reflection-intensity tracking and indicate a path to finer temporal resolution and continuous readout for future full-dataset collection. Overall, the methodology is widely applicable to solid-state and biological processes occurring on millisecond and slower timescales.

This work introduces an LED-pump-X-ray-multiprobe SCXRD technique that rapidly acquires complete, time-resolved 3D structures at sub-second resolution while maximizing photoconversion and minimizing damage. Demonstrated on [Pd(Bu4dien)(NO2)][BPh4], the method achieves near-theoretical ES populations and high-quality datasets, enabling kinetic analysis and electron-density visualization across excitation and decay. The LED-based pump affords flexible, uniform, low-cost illumination and can be adapted to diverse wavelengths and timescales. Proof-of-concept Timepix3 experiments show the promise of continuous, event-based detection for improved time resolution without readout dead time. Future directions include integrating larger-area Timepix3 (Tristan 1M/10M) detectors for complete SCXRD datasets with adjustable post hoc time binning, pairing with methods such as HATRX to enhance SNR and access faster timescales, and broader application to millisecond–minute phenomena in materials and biology.

• Signal-to-noise at shorter t_acq is limited by diffraction intensity and X-ray flux constraints; photo-difference maps show spurious features at fastest timescales. Higher flux risks X-ray-induced excitation.
• Detector readout time of Pilatus 300K (~3 ms) limits sub-10 ms time resolution with gated imaging; event-based detectors are needed for faster regimes.
• Photobleaching and potential LED-induced surface damage over long irradiation periods necessitate using fresh crystals and careful optimization of t_exc and intensity.
• X-ray-induced excitation requires beam attenuation (here to 5%), introducing a baseline ES population (~3%) and limiting usable flux.
• Maximum ES conversion is thermodynamically limited by equilibrium with thermal decay under continuous or quasi-continuous illumination; short high-intensity pulses may not yield sustained non-equilibrium populations in practice.
• Single-reflection Timepix3 measurements provide qualitative but not yet quantitative kinetics comparable to full SCXRD due to limited detector area and analysis constraints.