Machine-Learning-Driven High-Throughput Screening for High-Energy Density and Stable NASICON Cathodes.

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Sodium-ion batteries (SIBs) face limitations in specific energy and volume distortion.
• Sodium super ionic conductor (NASICON) materials offer structural stability and high operating voltage for SIBs.

Purpose of the Study:

• To develop a computational screening platform for discovering novel NASICON cathode materials.
• To identify NASICON structures with enhanced electrochemical performance and stability for SIB applications.

Main Methods:

• Utilized density functional theory (DFT) calculations and machine learning (ML) for high-throughput screening.
• Generated a training database from existing electrode data and constructed a test set of 3126 NASICON structures with various dopants.
• Employed ML surrogate models and ML interatomic potentials (MLIP) for material property prediction and stability assessment.

Main Results:

• Screened 3126 potential NASICON structures, identifying 796 materials meeting criteria for formation energy, energy above hull, volume change, and theoretical capacity.
• Selected thermodynamically stable doped NASICON configurations using MLIP.
• Identified 50 candidate NASICON materials with an average voltage of ≥3.5 V after DFT calculations.

Conclusions:

• The developed DFT and ML-based platform efficiently accelerates the discovery of optimal NASICON cathode materials for SIBs.
• The identified 50 candidate materials represent promising avenues for next-generation sodium-ion battery development.
• This approach significantly reduces the resources required for exploring advanced battery materials.