- Research Article
- 10.1149/ma2025-023481mtgabs
Machine Learning Enhanced Phase Field Modeling of Battery Materials
- Nov 24, 2025
- ECS Meeting Abstracts
- Edwin García
Historically, the properties of cathode materials for rechargeable lithium-ion batteries (LIBs) have been investigated by predicting the properties by starting from atomistics, or by self-consistently averaging the effect of an individual representative material feature and incorporating it into coarse grained porous electrode theory descriptions that capture the time-dependent electronic and ionic transport. Such approaches are based simplified representations of grains, pores, or inclusions, and averaging them into models that aim to understand or engineer the macroscopic properties of the device or component at hand. However, if effects due to phase transformation kinetics, or the impact of defects such as vacancies or interstitials, dislocations, or grain boundaries are to be engineered, currently used averaging approaches will not capture the underlying multiphysical and microstructural richness of behavior of the device. In general, the modern development of advanced technology for LIBs demands: 1) the formulation of a methodology that provides an accurate description of the materials that integrate these devices at each length scale; 2) the systematic and mindful coarse-graining of lower length scale methodologies into higher length scale descriptions, and 3) the establishment of meaningful databases that enable the development of insight to understand and then engineer advanced, reliable, next generation devices. Here, by defining a thermodynamically consistent representation of materials that spatially resolves the multiphysical fields that results from formally considering microstructural features such as grain size, crystallographic texture, grain boundaries, particle size, and porosity, the time-dependent behavior is analyzed in materials for lithium-ion battery applications. Progress towards integrating the physical contributions of each individual phase and its processing-induced spatial distribution into advanced models is presented. The effect on the performance and degradation is analyzed. Reaches and limitations of well-known and emerging theories will be reviewed and compared, and efforts to accelerate to the limit of real time performance and degradation computations will be presented in an effort to explore the space of what is physically possible.
Read more