Towards Full Ab Initio Simulations of Mechanical Response of Ti-N Binary System under Extremely Operational Conditions Via Neural Network Potential

In the process of high temperature service, the mechanical properties of cutting tools decrease sharply due to the peeling of the protective coating. However, the mechanism of such coating failure remains obscure due to complicated interaction between atomic structure and extreme temperature and stress. This non-equilibrium nature demands both large system size and accurate description from atomic scale, raising challenges for all existing methods. Here, we developed a neural network potential (NNP) of Ti-N binary systems based on density functional theory (DFT) datasets to achieve quantum-accurate large-scale atomic simulation. Compared with empirical interatomic potential based on embedded-atom-method (EAM), the developed NNP predict accurate lattice constant, phonon properties and mechanical properties for a wide range of thermodynamic conditions. Moreover, for the first time, we present the atomic evolution of the fracture behavior of large-scale TiN systems coupled with temperature and stress conditions. Our study validates that an extension of TiN beyond its tensile yield point leads to interatomic brittle fracture. Such simulation of coating fracture and cutting behavior based on large-scale atoms can shed a new light on understanding the microstructure and mechanical properties of coating tools under extreme operating conditions.