Ammonium Organosulfonates as Machine Learning-Driven “Crystallization Cocktails” for the Structure Determination of Liquid Molecules

Abstract

A concept of “crystallization cocktails” for crystal structure determination of small (liquid) molecules is introduced, offering a simple one-step procedure of one-pot mixing commercial sulfonic acids (such as p-toluenesulfonic and 1,5-naphtalenedisulfonic) and geometrically diverse amines (such as p-phenylenediamine, 4,4′-oxidianiline, and 4,4′-diamino-3,3′-dichlorodiphenylmethane) in different combinations to cocrystallize with a chosen liquid compound by simple solvent evaporation. These “crystallization cocktails” may adapt to different guests owing to “labile” intermolecular interactions between sulfonate and ammonium ions of diverse geometries and flexibility, as demonstrated by a successful cocrystallization of three model phenol derivatives (2,4-dimethyl-, 2-isopropyl-, and 2-ethylphenol); one of them being characterized by X-ray diffraction for the first time. The guest-to-host ratio in the obtained cocrystals, three of which featured the same ammonium organosulfonate but different phenol guests, was found to depend on the size of the individual components of the “crystallization cocktails”. To get insight into the preferable formation of a cocrystal by some of these “cocktails”─as common (descriptive) crystallographic tools failed to provide a definitive answer─a machine learning (ML) algorithm was developed, which allowed limiting the list of their components to a few most suitable ones for a successful cocrystallization with liquid phenols. “Crystallization cocktails” that can be “mixed to order”─and boosted by the ML─pave the way toward routine structure determination of liquids or other poorly crystallizing substances by X-ray diffraction, streamlining the search for new chemical compounds and identification of “old” ones in natural products, waste and natural waters, and living organisms.