Explainable Deep Learning-Guided Screening of LiNO₃-Based Molten-Salt Phase Change Materials for Lithium-Ion Battery...

Area of Science:

• Materials Science
• Electrochemistry
• Spectroscopy

Background:

• Lithium nitrate-based phase change materials (PCMs) are crucial for mitigating thermal runaway in lithium-ion batteries (LIBs).
• Selecting optimal PCM formulations requires efficient screening methods due to numerous possibilities.

Purpose of the Study:

• To develop and validate an interpretable deep learning framework for rapid PCM characterization.
• To establish a correlation between spectral features and thermal properties (onset temperature, latent heat).

Main Methods:

• Characterization of 12 LiNO3-based PCMs using laser-induced plasma spectroscopy (LIPS).
• Application of a self-attention-enhanced 1D convolutional neural network (1D CNN) for sample discrimination.
• Comparison with conventional methods like PCA-SVM.

Main Results:

• The deep learning approach demonstrated superior performance and interpretability over traditional methods.
• High-performing PCMs showed lower onset temperatures, higher latent heat, and distinct K/Na spectral intensity ratios.
• Na-normalized K emission intensity was identified as a key spectral probe.

Conclusions:

• A rapid spectroscopic method using Na-normalized K emission intensity can effectively screen optimal LiNO3-based PCMs.
• This approach offers a fast, complementary alternative to traditional thermal analysis for PCM selection.
• Enhanced battery safety through optimized PCM selection is achievable.