Saturday 01 February 2025
Scientists have made a significant breakthrough in understanding how materials can be designed and optimized for energy storage applications, such as lithium-ion batteries. Researchers at the University of Bremen, Paul Scherrer Institut, and École Polytechnique Fédérale de Lausanne (EPFL) have developed a new approach to calculate the electronic properties of these materials, allowing them to better predict their performance.
The team used a combination of theoretical calculations and machine learning techniques to study the behavior of lithium-manganese-phosphorus-oxide (LMPO) compounds. These compounds are commonly used as cathodes in lithium-ion batteries, but their optimal composition is still unclear. By analyzing the electronic properties of LMPO, researchers can better understand how they interact with lithium ions during charging and discharging.
The study found that the electronic properties of LMPO are highly dependent on the oxidation state of the manganese atoms. The team used a novel approach to calculate these oxidation states by combining density-functional theory (DFT) with extended Hubbard functionals. This allowed them to accurately predict the electronic occupations of the manganese atoms, which are crucial for understanding their behavior during battery operation.
The researchers also developed a machine learning interatomic potential that can be trained on a dataset of atomic configurations generated using first-principles molecular dynamics (FPMD). This potential was used to simulate the behavior of LMPO compounds under different conditions, such as varying lithium concentration and temperature. The simulations showed that the electronic properties of LMPO are highly sensitive to these conditions, which is important for optimizing battery performance.
The findings of this study have significant implications for the design and optimization of energy storage materials. By better understanding how these materials interact with lithium ions during charging and discharging, researchers can develop more efficient and sustainable energy storage solutions. The development of machine learning interatomic potentials also has potential applications in other fields, such as materials science and chemistry.
Overall, this study demonstrates the power of combining theoretical calculations with machine learning techniques to understand complex electronic properties of materials. It highlights the importance of accurately predicting oxidation states and electronic occupations for optimizing material performance, and it opens up new avenues for research into energy storage materials.
Cite this article: “Elucidating Electronic Properties of Energy Storage Materials through Theoretical Calculations and Machine Learning”, The Science Archive, 2025.
Lithium-Ion Batteries, Energy Storage, Materials Science, Machine Learning, Density-Functional Theory, Extended Hubbard Functionals, Interatomic Potential, First-Principles Molecular Dynamics, Electronic Properties, Oxidation States
Reference: Cristiano Malica, Nicola Marzari, “Teaching oxidation states to neural networks” (2024).
