Design of nanostructure in solid electrolyte interphase for enhancing the mechanical durability of lithium metal anode by deep-learning approach

Abstract

Lithium metal anode (LMA) encounters significant safety challenges due to the growth of Li dendrites. The solid electrolyte interphase (SEI) plays a crucial role in inhibiting dendrites growth. SEI shows a nanostructure consisting of embedded crystalline particles (CP), and the distribution of these CP strongly impacts the mechanical durability and thus Li dendrites growth. Therefore, establishing a correlation between the nanostructure of the SEI and its mechanical durability is essential to design a SEI with optimized security properties. Herein, we present a Convolutional Neural Network (CNN) that has been trained on a high-throughput Finite-Elements Method (FEM) dataset based on the experimentally observed Cryo-TEM image. The CNN model can accurately predict the mechanical failure time (FT) of SEI structures. Furthermore, we employ the Reverse Monte Carlo (RMC) method coupled with CNN model to explore the structures with longer FT, ultimately identifying an optimized structure with uniform arrangement CP. Additionally, Ablation-Classification Activate Map (Ablation-CAM) technique highlights the critical role of CP distribution in failure, as clustering CP can lead to nonuniform current density and uneven Li plating. This work provides design strategies and insights into the failure mechanisms for SEI, offering a potential solution to address the safety concerns of LMA.