Machine Learning-Assisted Exploration and Identification of Aqueous Dispersants in the Vast Diversity of Organic Chemicals

Abstract

Dispersion represents a central processing method in the organization of nanomaterials; however, the strong interparticle interaction represents a significant obstacle to fabricating homogeneous and stable dispersions. While dispersants can greatly assist in overcoming this obstacle, the appropriate type is dependent on such factors as nanomaterial, solvent, experimental conditions, etc., and there is no general guide to assist in the selection from the vast number of possibilities. We report a strategy and successful demonstration of the machine-learning-based "Dispersant Explorer", which surveys and identifies suitable dispersants from open databases. Through the combined use of experimental and molecular descriptors derived from SMILES databases, the model showed exceptional predictive accuracy in surveying about ∼1000 chemical compounds and identifying those that could be applied as dispersants. Furthermore, fabrication of transparent conducting films using the predicted and previously unknown dispersant exhibited the highest sheet resistance and transmittance compared with those of other reported undoped films. This result highlights that, in addition to opening new avenues for novel dispersant discovery, machine learning has a potential to elucidate the chemical structures essential for optimal dispersion performance to assist in the advancement of the complex topic of nanomaterial processing.