Viscosities of over 2,500 small organic molecules were curated from scientific literature, publications, and online databases to enable data-driven approaches that could rapidly predict viscosities. Quantitative Structure–Property Relationships (QSPR) consisting of traditional machine learning algorithms and graph convolution neural networks were developed to predict viscosities in the range of 0.01-100 cP with considerable accuracy. To further improve model accuracy, classical molecular dynamics simulations were used to extract physically relevant descriptors and encoded to machine learning models for molecules at 25°C. Experimental temperatures were then included as an external descriptor for machine learning models, resulting in generalized temperature-dependent viscosity models. The data curation of a large viscosity dataset and the development of accurate machine learning models enables the high-throughput screening of small organic molecules for the design of new materials.

Viscosity prediction of small molecules using machine learning and physics-based models

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.