Decoding Lithium Metal Battery Degradation with Symmetric-Cell Artificial Intelligence Diagnostics (SAID)

While the underlying crosstalk effect between cathode and anode complicates the understanding of lithium metal degradation utilizing asymmetric (full)-cell data, employing a symmetric cell configuration enables isolation of contributions from specific electrodes. In this study, a symmetric-cell artificial intelligence diagnostics (SAID) is devised to decipher lithium metal anode degradation. Leveraging easily accessible, early-cycle lithium | lithium symmetric-cell data, SAID is demonstrated to accurately predict the elbow points (indicators of polarization acceleration) with a test mean absolute percentage error of 13.3%. More importantly, SAID reveals the persistent role of an initial-nucleation-related fingerprint in determining long-term cell polarization, which is validated through experiments and extended to full cells across different electrolytes. This approach, therefore, not only offers valuable insights into battery design but also exhibits great potential in uncovering hidden chemical correlations and advancing the field of energy storage in general.