Physics-driven machine learning model on temperature and time-dependent deformation in lithium metal and its finite element implementation

Precise understanding on the temperature and time-dependent deformation in lithium-metal anode is of compelling need for durable service of Li-based batteries. Due to both temporal and spatial intertwined thermal agitations and the scarcity of experiments, faithful deformation map of Li-metal covering a broad range of service condition is still lacking. Here we design a physicsdriven machine learning (PD-ML) algorithm to map the temperature, stress and rate-dependent deformation in Li-metal. We demonstrate that the PD-ML model, fed with limited experimental results, can predict the mechanical response of Li-metal in a wide span of temperature and deformation rate, and help to realize a deformation map of Li-metal with high fidelity. A finite element (FE) procedure based on the PD-ML constitutive model is then developed. The integration of PD-ML with FE procedure inherits the power of FE analysis and the accuracy originated from PD-ML in describing temperature, stress and rate-dependent mechanical response of Limetal. The method introduced here paves a new way for constitutive modelling to capture the complex deformation in solids involving multi-field and multiscale mechanics.