RINSE: A reciprocal-space fingerprint for comparing crystal structures
- 1Department of Chemistry
- 2Durham University
Abstract
Comparing crystal structures, whether to identify polymorphs, follow a phase transition, or search a database for a known motif, requires a descriptor that encodes the chemistry of a crystal while remaining invariant to the arbitrary choices made during structure determination or descriptor construction. RINSE (Reciprocal-space INvariant Spectral Embedding) is such a fingerprint, computed from a crystal’s diffraction pattern. By projecting the intensity-weighted reciprocal lattice onto a combined radial and spherical-harmonic basis, RINSE produces a compact power spectrum that is invariant to crystal orientation and to the choice of unit-cell representation, in tens of milliseconds per structure. The descriptor is inspired by the Smooth Overlap of Atomic Positions (SOAP) kernel but maps the full periodic structure rather than a local atomic neighbourhood. A locality-sensitive hash derived from principal-component analysis of the descriptor space enables rapid approximate retrieval of structurally similar crystals, and provides a convenient pronounceable identifier for a related group of structures. The RINSE code is freely available as an open-source Python package.
Related
- The CPOSS Database of Computed Crystal Structures
- Polymorphism, Crystal Habit and Intermolecular Contact Topology in Benzothiazine Derivatives: A Comparative Hirshfeld Surface Analysis of Seven Crystal Structures
- Investigation of the crystal structures and electronic features for trivalent praseodymium doped KY(WO4)2 crystal: a theoretical study
- Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures
- CrystalGRW: generative modeling of crystal structures with targeted crystallographic properties via geodesic random walks
BibTeX
@misc{fellowes2026rinse,
title = {{RINSE: A reciprocal-space fingerprint for comparing crystal structures}},
author = {Thomas Fellowes and John S. O. Evans and Aurora J. Cruz-Cabeza},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15008615/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008615/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - RINSE: A reciprocal-space fingerprint for comparing crystal structures
AU - Thomas Fellowes
AU - John S. O. Evans
AU - Aurora J. Cruz-Cabeza
PY - 2026
DA - 2026/09/10
PB - ChemRxiv
DO - 10.26434/chemrxiv.15008615/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008615/v1
AB - Comparing crystal structures, whether to identify polymorphs, follow a phase transition, or search a database for a known motif, requires a descriptor that encodes the chemistry of a crystal while remaining invariant to the arbitrary choices made during structure determination or descriptor construction. RINSE (Reciprocal-space INvariant Spectral Embedding) is such a fingerprint, computed from a crystal’s diffraction pattern. By projecting the intensity-weighted reciprocal lattice onto a combined radial and spherical-harmonic basis, RINSE produces a compact power spectrum that is invariant to crystal orientation and to the choice of unit-cell representation, in tens of milliseconds per structure. The descriptor is inspired by the Smooth Overlap of Atomic Positions (SOAP) kernel but maps the full periodic structure rather than a local atomic neighbourhood. A locality-sensitive hash derived from principal-component analysis of the descriptor space enables rapid approximate retrieval of structurally similar crystals, and provides a convenient pronounceable identifier for a related group of structures. The RINSE code is freely available as an open-source Python package.
ER -