Static Photocrystallography Principles

• Static photocrystallography is the study of long-lived, photoinduced states in crystals, where excited states are accumulated and analyzed post-irradiation.
• The technique distinguishes weakly populated metastable structures from the dominant dark state using occupancy refinement and mixed-state reconstruction methods.
• It has practical applications in analyzing spin crossover phenomena, strain gradients, and lattice distortions with tools such as Bragg ptychography.

Static photocrystallography is the crystallographic study of long-lived excited or metastable structural states that can be accumulated under illumination and then interrogated with conventional diffraction data collection after irradiation has stopped, rather than direct ultrafast pump–probe timing. In this regime, the measured intensities may represent a mixture of dark and photoinduced states, so the central problem is to distinguish an average structure from weakly populated structural variants. Recent work places this problem in three closely connected settings: cryogenically trapped photoinduced spin crossover measured by single-crystal X-ray diffraction, mixed-state inverse formulations that reconstruct an incoherent structural ensemble rather than a single averaged crystal, and Bragg ptychographic imaging that remains stable for large lattice distortions in micro-crystals (Basuroy et al., 3 Aug 2025, Gladyshev et al., 2023, Li et al., 12 Mar 2026).

1. Definition, scope, and relation to time-resolved methods

Static photocrystallography is most powerful when a photoinduced state is long-lived enough at low temperature to survive after irradiation has stopped, so that a crystal can be illuminated to build up a photostationary or metastable population and then interrogated crystallographically with full reciprocal-space coverage. That regime is distinct from ultrafast pump–probe crystallography. The former addresses long-lived excited or metastable structural states; the latter addresses direct temporal trajectories. The distinction is not merely instrumental. It determines the inverse problem itself, because static measurements often encode a population mixture rather than a single structure (Basuroy et al., 3 Aug 2025).

In conventional photocrystallographic refinement, that mixed-state problem is typically handled as a small number of discrete structures with refined occupancies, constrained against Bragg intensities and often difference-density maps. The conceptual challenge is that a weakly populated metastable photoinduced state is superimposed on the ground state, and the average structure can conceal the physically relevant minority component. This same challenge appears outside optical excitation, for example when diffraction records thermally fluctuating crystals as an incoherent mixture of coexisting structural states rather than as a single time-averaged object (Gladyshev et al., 2023).

This scope also clarifies what static photocrystallography does not do. It does not, by itself, recover a real-time movie with known temporal ordering. A plausible implication is that its most natural interpretive objects are trapped populations, occupancy-weighted structural ensembles, and spatially heterogeneous metastable states, rather than uniquely time-labeled pathways.

2. Structural signatures and state assignment

A representative static photocrystallographic target is the photoinduced spin crossover transition in a [2x2] Fe(II) metallogrid. In the low-temperature dark state used in the reported synchrotron experiment, the crystal is described as 2HS-2LS; after UV irradiation at low temperature, a partial photoinduced conversion toward a 3HS-1LS state was observed. The structural reporter used throughout is the average Fe–N bond distance, written as $\langle \mathrm{Fe-N} \rangle$, because in octahedral Fe(II) the HS state typically has longer Fe–N bonds than the LS state and a more distorted coordination sphere. The interpretation adopted there is that an LS$\rightarrow$HS conversion lengthens $\langle \mathrm{Fe-N} \rangle$ by about $0.2$ Å per fully converted center, although the observed LS/HS separation in that study was closer to $\sim 0.1$ Å, and the authors explicitly cautioned that conversion estimates based on the canonical $0.2$ Å value are only indicative (Basuroy et al., 3 Aug 2025).

The same experiment also used nonbonded Fe$\cdots$Fe distances and octahedral distortion parameters from OctaDist as supplementary spin-state markers. The supporting information lists $D_{\rm mean}$, $\zeta$, $\Delta$, $\rightarrow$0, and $\rightarrow$1 for each Fe site. At the LS-designated site A in one representative crystal, $\rightarrow$2 changed from $\rightarrow$3 to $\rightarrow$4 Å after irradiation, while at the HS-designated site B it changed from $\rightarrow$5 to $\rightarrow$6 Å. In the strongest-conversion dataset, $\rightarrow$7 increased from $\rightarrow$8 to $\rightarrow$9 Å and $\langle \mathrm{Fe-N} \rangle$0 decreased from $\langle \mathrm{Fe-N} \rangle$1 to $\langle \mathrm{Fe-N} \rangle$2 Å. The interpretation offered is a cage distortion accompanying spin-state redistribution, with the site-A Fe atoms moving apart and the site-B Fe atoms coming slightly closer.

These readouts show both the power and the limits of static photocrystallography. On the one hand, local metric shifts, distortion parameters, and reciprocal-space completeness can reveal metastable structural trapping that remains invisible in a purely dark-state refinement. On the other hand, when the light-state population is modest and the bond-length changes are comparable to standard uncertainties, structural assignment depends strongly on model choice, uncertainty treatment, and the distinction between explicit occupancy refinement and inference from local metric changes.

3. Experimental implementation at synchrotron beamlines

The reported beamline implementation at P11, PETRA III, DESY provides a concrete static photocrystallography workflow. Monochromatic hard X-rays of wavelength $\langle \mathrm{Fe-N} \rangle$3 Å, corresponding to $\langle \mathrm{Fe-N} \rangle$4 keV, were used in $\langle \mathrm{Fe-N} \rangle$5-scan mode. The main experimental temperatures were $\langle \mathrm{Fe-N} \rangle$6 K and $\langle \mathrm{Fe-N} \rangle$7 K. Photoexcitation was provided by a HORIBA DeltaDiode DD-375L pulsed laser diode, a class 3B source, with nominal specifications of center wavelength $\langle \mathrm{Fe-N} \rangle$8 nm, spectral FWHM $\langle \mathrm{Fe-N} \rangle$9 nm, typical pulse width $0.2$0 ps, maximum pulse width $0.2$1 ps, peak power $0.2$2 mW, average power $0.2$3 mW, and maximum repetition rate $0.2$4 MHz. A custom compound-lens focusing optic was attached directly to the diode head, and the combined setup of the diode head plus focusing optics had a longest dimension of not more than $0.2$5 cm. The focused laser spot could be reduced to about $0.2$6 $0.2$7m. To align the pump beam reproducibly, the laser module was mounted on a dedicated three-dimensional translation stage, and alignment was first carried out using highly fluorescent pyrene-derivative crystals of size roughly $0.2$8–$0.2$9 $\sim 0.1$0m before mounting the actual Fe(II) metallogrid crystals (Basuroy et al., 3 Aug 2025).

The irradiation protocol was deliberately simple. Each crystal was irradiated for a total of $\sim 0.1$1 minutes, implemented as $\sim 0.1$2 minutes from each of four directions separated by $\sim 0.1$3. After irradiation was stopped, a $\sim 0.1$4 $\sim 0.1$5-scan dataset took only $\sim 0.1$6 s to collect, although practical delays added about $\sim 0.1$7–$\sim 0.1$8 min before post-irradiation collection could start. Additional datasets were collected after further delays of $\sim 0.1$9 min, $0.2$0 min, $0.2$1 min, and $0.2$2 min. The use of $0.2$3–$0.2$4 different crystals of the same compound reflects a common practical constraint in static photocrystallography, namely the need to balance irradiation history, crystal quality, and low-temperature survival over multiple data collections.

Temperature control was decisive. At $0.2$5 K, measurable structural changes consistent with photoinduced population of the metastable 3HS-1LS state were observed. At $0.2$6 K, no significant changes in $0.2$7, Fe$0.2$8Fe distances, or angular distortion parameters were detected after irradiation. In the same study, the inferred conversion fraction was modest, about $0.2$9–$\cdots$0, and the structural changes remained measurable hours after irradiation at $\cdots$1 K. This defines a canonical static-photocrystallography regime: cryogenic trapping, post-irradiation data collection, and structural comparison across dark and illuminated states without sub-nanosecond temporal gating.

4. Mixed-state reconstruction and the ensemble view of illuminated crystals

A central conceptual development relevant to static photocrystallography is the reformulation of the inverse problem from “one structure fits all data” to “the data arise from an incoherent mixture of structures.” In electron ptychography, the conventional thin-sample model writes the exit wave at scan position $\cdots$2 as

$\cdots$3

with diffraction intensity

$\cdots$4

The reconstruction is posed as minimization of

$\cdots$5

with gradient-descent updates of unknown parameters. For thicker specimens, the same work used a multislice forward model and then generalized partial-coherence treatment on the object side by representing the crystal as an ensemble of transmission functions $\cdots$6 rather than a single $\cdots$7. The master equation is an incoherent sum over object and probe modes:

$\cdots$8

The reconstructed states are explicitly interpreted not as a real-time movie, but as states of an incoherent structural ensemble (Gladyshev et al., 2023).

That formulation is directly suggestive for static photocrystallography. A weakly populated metastable photoinduced state superimposed on the ground state is also an incoherent or occupancy-weighted mixture in the measurement. The mixed-object ptychographic concept therefore resembles occupancy refinement, but in a richer representation: instead of refining a few atomic coordinates and occupancies in reciprocal space under a crystallographic model, it reconstructs a set of complex-valued projected transmission functions in real space, from which displacements can then be inferred.

The silicon grain-boundary demonstration is methodologically important because it showed that reconstructing the object as an ensemble of $\cdots$9 different states allowed observation of atomic movements in the range of $D_{\rm mean}$0–$D_{\rm mean}$1 Å, in agreement with expectation. At the same time, it also exposed underconstraint. The different object states could not all be initialized identically, full-dataset rather than mini-batch gradients were required for physically meaningful decomposition, and mixed-probe overparameterization worsened results for that dataset. For static photocrystallography, this suggests that mixed-state recovery is plausible but not automatically unique: physically interpretable decomposition may require strong global constraints, carefully chosen state number, and a forward model that genuinely matches the source of heterogeneity.

5. Bragg ptychography and three-dimensional imaging of distorted metastable states

Three-dimensional Bragg ptychography extends the static-photocrystallographic problem into the regime of spatially heterogeneous lattice distortion in finite crystals. In Bragg coherent imaging, the reconstructed crystalline object is written

$D_{\rm mean}$2

where $D_{\rm mean}$3 is the amplitude and the phase is the Bragg-projected displacement,

$D_{\rm mean}$4

From this, one obtains the displacement component $D_{\rm mean}$5 and, through spatial derivatives of phase, strain and lattice tilt. In the reported comparison between 3DBP and BCDI, 3DBP used a focused beam from a Fresnel zone plate a few millimeters downstream of focus, producing a divergent structured beam about $D_{\rm mean}$6 in diameter, and a raster scan of $D_{\rm mean}$7 positions with $D_{\rm mean}$8 nm step size at each rocking angle. The central result is that 3DBP tolerates lattice distortions more than six times larger than BCDI: BCDI remained satisfactory only up to approximately $D_{\rm mean}$9, whereas 3DBP remained satisfactory up to approximately $\zeta$0 (Li et al., 12 Mar 2026).

The highly distorted Au micro-crystal in that study illustrates why this matters. The 3DBP reconstruction resolved one part of the crystal with approximately constant phase and another with phase increasing linearly along $\zeta$1, with a measured phase gradient of $\zeta$2 rad per retrieved pixel or $\zeta$3 rad/nm over $\zeta$4m, total phase range $\zeta$5, and extracted lattice tilt of about $\zeta$6. BCDI failed to reconstruct this state, although diffraction simulated from the 3DBP reconstruction agreed with the measured BCDI data. On a weakly distorted crystal, both methods worked, but 3DBP yielded smoother amplitude and phase fields with reduced short-length-scale artifacts.

For static photocrystallography, the relevance is direct. Long-lived illuminated crystals can exhibit localized excited-state expansion or contraction, strain gradients near an optical penetration front, lattice tilts between differently illuminated sub-domains, or partial transformation of only one region of a crystal. These are precisely the kinds of states that may be too distorted or too inhomogeneous for standard BCDI but still static on the acquisition timescale. A plausible implication is that 3DBP can serve as a route to image frozen or long-lived photoinduced states when conventional coherent Bragg imaging stagnates.

6. Limitations, ambiguities, and open directions

The present literature also delineates the limits of static photocrystallography. In the synchrotron Fe(II) metallogrid study, the authors openly described the work as a proof-of-concept demonstration rather than a final, fully quantified photoconversion study. Results were “sometimes quite varied” across datasets; different crystals were often used before and after irradiation; photocrystallographic refinements were simplified, with anisotropic refinement only for Fe and N atoms; unresolved positional disorder had not yet been modeled by partial occupancies; the inferred photoconversion fractions depended on an assumed $\zeta$7 Å full-conversion bond expansion even though the measured LS/HS separation was closer to $\zeta$8 Å; and some observed bond-length changes were close to experimental uncertainty. The paper also contains an inconsistency in the laser repetition-rate description: one section says $\zeta$9 kHz, whereas the irradiation protocol states “80 MHz rep-rate” and Table 2 reaches $\Delta$0 kHz (Basuroy et al., 3 Aug 2025).

The mixed-object ptychography work defines a different set of limits. It was demonstrated on realistically simulated, not experimental, data, and at infinite dose. The decomposition depends on the chosen number of states, the recovered states are not guaranteed to correspond one-to-one to physically distinct phonon modes or real temporal snapshots, and the method is computationally expensive because each diffraction prediction may require multiple multislice propagations across object and probe modes. On a single NVIDIA Tesla V100 GPU, $\Delta$1 iterations of pure-object/pure-probe reconstruction took $\Delta$2 hours, whereas $\Delta$3 iterations of mixed-object reconstruction with $\Delta$4 transmission-function states took $\Delta$5 hours. Convergence was also fragile: mixed-probe overparameterization worsened results, and mixed-object reconstruction failed when mini-batches rather than the full diffraction set were used to form updates (Gladyshev et al., 2023).

Bragg ptychography imposes yet another set of conditions. Acquisition is slower and more data-heavy than BCDI; the metastable state must remain static throughout the full scan; the method is underdetermined along the beam direction and therefore used thickness-support regularization; only the displacement component projected onto a chosen $\Delta$6 is recovered from a single reflection; and the paper did not explicitly analyze dynamical diffraction effects. Even so, it demonstrated that overlap redundancy can stabilize inversion in regimes where strong phase gradients and multi-$\Delta$7 phase spans defeat standard coherent Bragg imaging, including successful reconstruction for phase ranges up to at least $\Delta$8 and an additional supplementary example at $\Delta$9 (Li et al., 12 Mar 2026).

Taken together, these studies define static photocrystallography as a structurally rich but model-sensitive field. Its mature form is not merely “dark minus light” diffraction, but a family of mixed-state and spatially resolved inverse problems in which the decisive question is whether measured intensities are being interpreted as a single averaged lattice or as a distribution of coexisting structural states. This suggests that future progress will depend on better occupancy-aware refinement, stronger treatment of heterogeneity, and imaging strategies that remain robust when the illuminated crystal is only partially transformed, strongly distorted, or both.