Solid polymer electrolytes (SPEs) are considered promising building blocks for next-generation lithium-ion batteries. However, the low ionic conductivity of current SPEs prevents their incorporation into real-world applications. This motivates the development of novel, highly conductive SPE materials, which will require a thorough understanding of the ionic transport mechanisms in order to provide clear guidelines for materials design. Here we develop a coarse-grained molecular dynamics (CGMD) model, which captures the ion motion and the polymer conformational evolution during the transport process, at a much lower computational cost than fully atomistic simulations. From the CGMD simulations, the ionic conductivity is calculated by the Green-Kubo relation, and the transference number is estimated by analyzing the mean-square-displacement curve of each species. The inputs of the CG model form a continuous parameter space, consisting of intuitive descriptors of the polymer-salt system, such as particle size and intermolecular interaction strength. This high-dimensional space is explored to map the molecular level information to the electrical conductance. An efficient machine learning (ML) assisted exploration protocol is established, where automated CGMD simulations are performed iteratively, with the searching direction determined by a Bayesian optimization (BO) algorithm. Adopting this ML-CGMD approach, we obtain detailed descriptions of the correlations between the CG level material properties and the electrolyte performance. Many of the trends discovered by the model can be directly validated by either fully atomistic simulations or experiments. In addition, the model predicts the optimal conductivity and the associated properties of the anion, the secondary site, and the backbone chain, respectively. This research sheds light on the mechanisms of ionic transport in polymers, and provides useful insights as well as a new tool for designing tailored SPE materials.<br/>

Machine learning-assisted coarse-grained molecular dynamics for designing highly conductive polymer electrolytes

2019 Spring National Meeting

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