Deciphering failure paths in lithium metal anodes by electrochemical curve fingerprints

Area of Science:

• Materials Science
• Electrochemistry
• Machine Learning

Background:

• Lithium metal batteries (LMBs) are critical for energy storage, but anode failure mechanisms hinder their widespread adoption.
• Current post-mortem analyses of anode failure lack dynamic insights into failure progression and root causes.

Purpose of the Study:

• To develop a predictive model for lithium metal anode failure mechanisms in LMBs.
• To identify early indicators of distinct failure types using electrochemical fingerprints.
• To deepen the understanding of kinetics and reversibility degradation in LMB anodes.

Main Methods:

• Utilized a machine learning model informed by domain knowledge.
• Analyzed a dataset of over 18,000 cycles and 12 million data points from cells cycled to failure.
• Focused on identifying correlations between initial lithium plating/stripping behavior and subsequent anode changes.

Main Results:

• Accurately predicted LMB failure types using data from the first two cycles (less than 2% of total data).
• Identified key electrochemical fingerprints related to lithium microstructure and its interphase with the electrolyte as critical to degradation.
• Demonstrated that these fingerprints influence the formation of ineffective interphase regions and inactive lithium, impacting transport paths.

Conclusions:

• Early electrochemical fingerprints serve as reliable indicators of distinct anode failure modes in LMBs.
• The developed pre-mortem prediction method offers a practical approach to assess battery reliability and guide electrolyte development.
• The model shows versatility and generalizes well across different electrolyte systems.