| Abstract Scope |
Polymer derived transition metal carbonitrides have attracted significant attention for ultrahigh temperature applications due to their tremendous thermal stability, oxidation resistance, and structural tunability. However, the fundamental mechanisms involved during polymer to ceramic conversion remain unexplored and difficult to predict through experiments. Atomistic simulations can provide critical data about atomic interactions and new species formation. In this study, moment tensor potential based machine learning interatomic potential was developed using ab initio molecular dynamics dataset of tetrakis(dimethylamido)zirconium(IV), tetrakis(dimethylamido)-hafnium(IV) and their 1:1 mixture at a temperature range of 300 to 3000 K. The trained MTP model exhibited near density functional theory accuracy. Additionally, the model successfully captured bond evolution, atomic diffusion and structural transformation during pyrolysis while reducing the computational cost in comparison to first-principles method.
Keywords: Machine learning, Moment tensor potential, Interatomic potential, Pyrolysis, Transition metal carbonitride |