Researchers from Saint Petersburg State University and Samara State Technical University have developed a new method for analyzing crystal structures. By combining topological analysis with artificial intelligence, the team can identify chemical patterns and particle configurations that are difficult to detect using standard search parameters.
The process utilizes the ToposPro software to map atomic neighborhoods through Voronoi polyhedra. To handle the vast volume of data generated, researchers integrated a large language model to identify statistical regularities within the Kembirdge Structural Database, which contains over 1.4 million entries.
The team tested the approach on 1,553 contacts across 473 crystals involving hypervalent chlorine, bromine, and iodine salts. The analysis revealed five distinct spatial arrangements, including previously overlooked perpendicular contacts where particles approach the iodine atom from above or below the C-I-C plane.
This methodology allows for the discovery of hidden interactions without requiring researchers to pre-define distance or angle constraints. The study, supported by the Russian Science Foundation, was published in the journal Coordination Chemistry Reviews, and the team plans to expand the method to other chemical classes.
