Catalytic Strategy for Chemical Analysis of Volatile Iodine with the Assistance of Machine Learning

Abstract

A strategy of catalytic chemical detection (CCD) with the assistance of a machine learning (ML) approach was proposed and evaluated in this work. In the CCD method, the target analyte acts as the catalyst of the detection reaction rather than traditional reactants. The detection of a typical environmental contaminant-volatile iodine was selected as an example to establish the general routine in designing CCD. One major obstacle lies in the complex of manual selection of detection reaction, especially considering that more than 650,000 related reactions were exhibited in SciFinder database. Traditional workflow is time-consuming and material-consuming; therefore, the ML approach with descriptors directly related to CCD was employed. The reaction of indoles and aromatic aldehydes to bis(indolyl)methanes was screened out with the ML approach. After preliminary experiments, the screened reaction for iodine detection achieved desirable sensitivity, specificity, and recognizability simultaneously. The fabricated sensor devices were practicable for portable detection in real gas samples with a low concentration. This work provides a practical example of chemical analysis based on catalytic strategy and exemplifies the powerful application for the ML method in chemistry through the introduction of original descriptors.