Study on the mechanical properties of beta silicon nitride based on neural network potential

Silicon nitride is widely utilized in the manufacture of cutting tools, bearings, and engine components, making accurate predictions of its mechanical and thermal behaviors crucial for engineering design. This study significantly advances the modeling of complex atomic interactions by employing machine learning, specifically Neural Network Potential (NNP), to enhance precision over traditional empirical potentials. Through NNP, predictions of stress-strain curves, thermal conductivity, and grain boundary behavior are achieved. The results demonstrate that NNP not only reduces computational time but also enhances the accuracy of material property predictions. This work provides the mechanical and thermal performance of Si3N4 and opens new avenues for further applications of machine learning in advanced ceramic materials.