A deep learning approach to predict thermophysical properties of metastable liquid Ti-Ni-Cr-Al alloy

The physical properties of liquid alloy are crucial for many science fields. However, acquiring these remains challenging. By means the deep neural network (DNN), here we presented a learning interatomic potential Ti–Ni–Cr–Al system. Meanwhile, thermophysical were experimentally measured by electrostatic levitation and electromagnetic technologies. DNN predicted this system accurately in terms both atomic structures properties, results agreement with ab initio molecular dynamics calculation experimental values. A further study on local structure carried out Voronoi polyhedron analysis showed that cluster exhibited tendency to transform into high-coordinated decrease temperature, indicating enhancement stability. This eventually contributed linear increase density surface tension, exponential variation viscosity diffusion coefficient rise undercooling.