This presentation describes the use of machine learning approaches for predicting viscosity and ionic conductivity of protic ionic liquids (PILs) in water mixtures. Specifically, multiple linear regression with expectation maximization (MLREM) and Bayesian regularized artificial neural network with Laplacian prior (BRANNLP) machine learning models were constructed using previously reported literature data for viscosity and conductivity of PILs containing small amounts of water. While the viscosity training data included 91 data points from paring of 39 cations and 16 anions, the training data for ionic conductivity included 105 data points from 48 cations and 17 anions. The optimal models were used to predict the viscosity and ionic conductivity in water mixtures of ~2000 possible PILs generated through paring of 63 cations and 29 anions. The models were validated by experimental values obtained from 10 different PILs containing different amounts of water. This study showed that machine learning models can be constructed from sparse experimental data to not only give meaningful predictions for properties of quite complicated systems/materials but also computationally extract structure features in correlation with the desired properties.

Predicting viscosity and Ionic Conductivity of protic ionic liquids in water mixtures using machine learning methods

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

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