Machine learning prediction of perovskite sensors for monitoring the gas in lithium-ion battery

Room-temperature gas-sensitive materials are urgently needed for lithium-ion battery monitoring to ensure the safety of battery. In this work, we proposed a strategy for predicting gas-sensitive materials to sense gas in lithium-ion batteries by the combination of machine learning and ab initio calculations. Copper acetylacetonate functionalized perovskite CsPbBr3, an excellent room-temperature gas-sensitive material, was chosen as an example to demonstrate the correctness and extensibility of the strategy. It is found that the variation of adsorption characteristic parameters of different gases determines the uniqueness of the electrical response behavior by calculations. However, it is difficult to obtain the correlation between characteristic data and gassensitive performance directly. Therefore, using machine learning, combining multiple features with algorithmic voting classifiers can achieve a prediction accuracy of 85.71%. The results show that the features of adsorption energy, band structure and state of density play a major role in the prediction, making a great influence on carrier transport and further affecting the gas-sensitive performance. This work provides a theoretical framework and a new perspective for the follow-up work.