Extraction of molecular information from experimental data on liquids using manifold learning

Abstract

Listen

Translate article

This study proposes that microscopic information can be extracted from experimental data on liquid substances using a manifold learning technique. Here, 98 liquid substances could be classified based on their functional groups by applying the isomap method, a manifold learning technique, to nine thermal properties of the substances. Moreover, information related to their intermolecular interactions, which could be divided into the contributions of van der Waals and polar interactions, molecular weight, and number of carbon atoms could be extracted by the proposed method. This study could guide future research on the utilization of manifold learning techniques in liquid-substance characterization and development.