Computational Simulations and Machine Learning Studies of Battery Materials

計算模擬及機器學習研究電池材料

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date10 Aug 2023

Abstract

The increasing demand for battery diversification has posed multiple challenges to battery performance. The performance of batteries is closely related to the performance of battery materials. Therefore, clarifying the relationship between battery materials and battery performance, and searching for battery materials with specific effects has become the focus of research in the field of batteries. Computational simulation research methods can be used to deeply study the mechanism of battery atoms and molecules, and at the same time, they have the advantage of short time period and low cost for the screening of new materials. At the same time, the rapid development of machine learning methods in recent years has also brought new impetus to the research in the field of battery materials. In this study, we will explore the origin of material performance and search for new materials from three different battery components, namely, battery anode and cathode materials and electrolyte, through computational simulation and machine learning methods. The anode and cathode materials determine the working voltage and energy density of the battery. In this paper, based on the first-principles study of density functional theory (DFT), the effect of defect engineering on the working voltage and energy density of MXene electrodes was investigated. The results showed that by defect engineering to remove some transition metals elements, the working voltage and energy density of MXene as a sodium ion battery electrode can be effectively improved. In addition, through the combination of DFT and machine learning methods, organic electrodes with high working voltage and high energy density were searched. High-throughput calculations of lithiation energies of 7,724 organic molecules were performed, and then a graph neural network was used to fit the relationship between molecular structure and lithiation energy. The electrolyte serves as a bridge linking the positive and negative electrodes. The ionic transport efficiency within it influences the energy density of the battery, and its stability has a significant impact on the overall cycle efficiency of the battery. In this paper, we employed Bayesian optimization and coarse-grained molecular dynamics (MD) to investigate the effect of additive features on the self-diffusion coefficient of lithium ions in aqueous electrolyte. Additionally, we searched for 6000 organic small molecules as aqueous additives to investigate how they affect the structure of the lithium ion solvent shell, and experimentally verified that the stability window of the electrolyte can be widened by reducing the number of water molecules in the lithium ion solvation shell.

    Research areas

  • Battery, Simulation, Machine learning, Molecular dynamics, density functional theory