Few-fs resolution of a photoactive protein traversing a conical intersection
Summary¶
Hosseinizadeh et al. use serial femtosecond X-ray crystallography and machine-learning–assisted reconstruction to obtain time-resolved structural information on photoactive yellow protein (PYP) as it evolves through a conical intersection after photoexcitation (Nature 598, 2021; DOI 10.1038/s41586-021-04050-9). The advance is experimental: sub-100 fs time steps and near-atomic spatial resolution for a thousands-of-atoms protein system, mapping nuclear motion to nonadiabatic return at a conical seam—peripheral to standard ReaxFF (ground-state reactive MD) workflows in this knowledge base but relevant as a resolution benchmark for ultrafast structure.
Methods¶
Literature scope and protocol (non-MD, non-DFT production in the vanDuinWiki sense). The Nature workflow combines serial femtosecond X-ray crystallography with data-driven reconstruction to extract ultrafast structural trajectories of photoactive yellow protein after optical excitation. In practical terms, the protocol depends on high-temporal-resolution pump-probe timing, careful handling of diffraction snapshots that are sparse/noisy at each delay point, and a machine-learning-assisted inversion/reconstruction layer that turns those snapshots into time-ordered structural coordinates.
The article then interprets those reconstructed geometries with conical-intersection concepts (seam crossing, branching-space-relevant motions) to connect structural evolution with nonadiabatic relaxation. This is why the paper is methodologically important even for a mostly ReaxFF-focused corpus: it provides an experimentally anchored map of ultrafast structural change at a resolution class that many simulation workflows use as an aspirational benchmark.
Blueprint mapping for this wiki: - MD application block: N/A (no classical production MD protocol reported as the study core). - Force-field training block: N/A. - Static QM-only block: partial/interpretive only; the main contribution is experimental reconstruction and analysis rather than a standalone DFT campaign. - Review/non-simulation framing: applicable; methods should be read directly from the Nature Methods + Extended Data sections for full instrument and reconstruction specifics.
Molecular dynamics / ReaxFF: N/A — not the subject of this publication.
Force-field training / ReaxFF: N/A.
Static QM in silico for production MD: N/A; QM-level interpretation of photophysics may be cited in Nature in relation to electronic structure, but the contribution here is structural dynamics from experiment + reconstruction.
Findings¶
Outcomes & mechanisms (authored): The central claim is that the photoactive protein trajectory can be followed with few-fs temporal granularity through the region associated with a conical intersection, giving experimentally informed access to nuclear motions tied to ultrafast internal conversion. The paper frames this as direct structural evidence for dynamics that are often inferred indirectly.
Comparisons & sensitivity: The reconstruction is evaluated across time slices and structural coordinates rather than through force-field benchmarking. Sensitivity is therefore tied to data quality, inversion choices, and interpretation of trajectory features in the conical-intersection framework, not to a classical-potential hyperparameter scan.
Design relevance for this KB: Even though this is not a ReaxFF paper, it anchors a useful cross-domain message: experimental ultrafast structural datasets can constrain or challenge simulation narratives about excited-state relaxation pathways. In that sense, it is a benchmark-style reference for what "resolved dynamics" can look like when instrumentation and analysis pipelines are pushed aggressively.
Limitations & outlook (as authored): Reconstruction quality is model- and data-dependent; uncertainty handling, inversion assumptions, and derived coordinate interpretation remain important caveats. The article's own discussion and Extended Data should be treated as the authority for quantitative confidence intervals and edge cases.
Corpus / KB honesty: vanDuinWiki is ReaxFF-centric; links from PYP work to reaxff-family are thematic only unless the reader adds nonadiabatic / excited-state simulation (surface hopping, TD-DFT MD, etc.) consistent with the phenomenon. Definitive numbers from pdf_path and publisher ED.
Limitations¶
The wiki page is intentionally high-level and does not substitute for the full Nature Methods, Extended Data, and supplemental protocol details (timing calibration, reconstruction settings, and uncertainty analysis). Readers needing exact numerical settings or reproducibility specifics should use the version-of-record publication directly.
Scope limitation for vanDuinWiki users: this entry is adjacent to, not inside, the core ReaxFF workflow family. It should not be cited as evidence for classical force-field parametrization quality, but it can be cited as an experimental reference point for ultrafast structural dynamics and conical-intersection interpretation.
Relevance to group¶
Peripheral: experimental and ML-inversion limits on large-molecule nonadiabatic dynamics, complementary to simulation-centered corpus entries.
Citations and evidence anchors¶
Nature 598 (2021), DOI 10.1038/s41586-021-04050-9 — main text, figures, Extended Data.