Challenging the electrochemical properties of Nb2CO2/Si3N4 heterojunction by first principles combined with machine learning
摘要
Two-dimensional ceramics-based materials exhibit fascinating performance in the electrochemical field in recent years. In the present contribution, the electronic performance of the Nb2CO2/Si3N4 heterojunction as an anode material are systematically investigated by the first principles calculations. The structural stability of the Nb₂CO₂/Si₃N₄ heterostructure is verified by a negative binding energy, indicating favorable energetic stability. Moreover, the heterostructure demonstrates excellent electrochemical kinetics, with ultra-low diffusion energy barriers of 0.231 eV for Li ion and 0.316 eV for Na ion. With the increase of Li and Na ions concentrations, the average adsorption energy of the heterostructure is -0.12 eV and -0.53 eV, respectively. The theoretical specific capacity reaches 429.73 mAh/g for Li and 330.46 mAh/g for Na, surpassing that of conventional graphite anodes. Furthermore, the average open-circuit voltage (OCV) at maximum loading is 0.41 V and 0.35 V, which is within the reasonable expected range, thus ensuring the safety of the electrode material. These results show that Nb2CO2/Si3N4 has great potential as anode materials for metal ion batteries. The gradient boosting regression (GBR) algorithm with R2 of 0.91 is the most accurate for predicting adsorption energy by using the method of machine learning, followed by random forest (RF) with R2 of 0.897 and support vector regression (SVR) with R2 of 0.848. This high accuracy confirms the robustness of the descriptors, providing a valuable framework for the accelerated design of high-performance anode materials.