Co-free and low strain cathode materials for sodium-ion batteries: Machine learning-based materials discovery

Abstract

Sodium-ion batteries (SIBs) are promising alternatives to lithium-ion batteries (LIBs) owing to their cost-effectiveness and similar intercalation mechanisms. Layered transition metal oxides (LTMOs) are the promising cathode candidates for SIBs owing to their high voltages and ease of synthesis. However, O3-type LTMOs undergo structural deformation and performance degradation during de-/intercalation. Owing to the limitations of single-element compounds, doping them with various elements can enhance their stability and performance. In this study, machine learning (ML) algorithms are used to predict the structural stability of O3-type materials without phase transitions to facilitate efficient material selection. ML classification models assess the phase stability of cathodes in the pristine and desodiated states. Data sampling and feature engineering enhanced the accuracy of the pristine model from 0.886 to 0.962 and desodiated model from 0.642 to 0.954. Furthermore, a novel database in the form of (NaxNi0.5MyMzO2)-M-a-O-b (0.5 <= x <= 1, y + z = 0.5; M = transition metal) was constructed using density functional theory (DFT) calculations. Out of 1,451 LTMOs candidates, we present 128 cathode candidates that satisfy the following conditions: (1) O3 phase is maintained during dis-/charging processes, (2) average voltage >= 3 V, (3) theoretical capacity >= 200 mAh/g, and (4) -5 % <= volume change <= 5 %. Among them, 125 materials showed the possibility of stable Co-free cathodes, and 13 materials exhibited -0.5 % < volume change < 0.5 %. This study suggests optimal LTMO candidates that satisfy both the high energy density and electrochemical stability and provides a reliable battery material screening platform.