Development of electrolyte descriptors for predicting cycling performance of electrochemical cells
||Development of electrolyte descriptors for predicting cycling performance of electrochemical cells|
||NAISS Small Compute|
||Giovanni Volpe <email@example.com>|
||2023-03-20 – 2024-04-01|
A common problem for battery cell developers is predicting which electrolyte
compositions will give rise to good battery performance. The optimization of electrolyte compositions in batteries is critical for achieving high performance and stability. In this project, we propose to use neural networks to model the relationship between the chemical structure of electrolytes and their performance characteristics. The built model captures important atomic and bonding properties. We will train the network on a set of experimental measurements of electrolyte performance for various compositions, to learn a network to map the representation of an electrolyte to its performance. This work can potentially lead to significant advances in the design and optimization of electrolytes for batteries, ultimately resulting in improved battery performance and new opportunities for developing next-generation energy storage technologies.
Our training data will be generated by molecular dynamics simulations of a selection of electrolytes using a density-functional tight-binding approach. Such simulations provide detailed insight into the dynamics of electrolytes, including for example the diffusive behavior of cations and their solvation shells. The hypothesis is that these details carry the information necessary to predict battery performance when run through a neural network.