Voltage Mining for (De)lithiation-Stabilized Cathodes and a Machine Learning Model for Li-Ion Cathode Voltage

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Lithium-metal anodes drive interest in Li-free cathode discovery.
• Commercial lithium-ion battery cathodes are stable in discharged states, unlike many Li-free candidates found in charged states.

Purpose of the Study:

• To analyze voltage distributions for Li-free and commercial cathode materials.
• To establish design principles for high-voltage cathode discovery.
• To develop and validate a machine learning model for cathode voltage prediction.

Main Methods:

• Calculated cathode voltage data for 5577 charged-state and 2423 discharged-state unique structure pairs.
• Analysis of voltage distributions based on redox pairs and anion types.
• Training a machine learning model using chemical formulas for voltage prediction.

Main Results:

• High-voltage cathodes favor later Period 4 transition metals and electronegative anions (e.g., fluorine, polyanions).
• Charged-state cathodes generally exhibit lower voltages than lithiated counterparts, with deviations linked to anion distribution.
• The developed machine learning model achieved state-of-the-art performance compared to Roost and CrabNet.

Conclusions:

• Key design principles for high-voltage Li-free cathodes are identified.
• Machine learning offers a powerful tool for predicting cathode voltages and accelerating materials discovery.
• Understanding anion distribution is crucial for predicting cathode voltage behavior.