Research

The design of soft materials is being revolutionized through the application of computational modeling and in-depth mechanism studies across multiple scales. By using computational tools combined with advanced algorithms and machine learning techniques, this approach provides insights into materials from molecular to macroscopic levels, predicting how they react under different conditions. This is particularly useful in developing new materials tailored for specific applications, such as cost-effective drugs or drug delivery vehicles in biomedical engineering, and flexible, biocompatible, biodegradable materials for electronics. The focus on molecular-level research is crucial, uncovering the fundamental behaviors for more innovative and efficient designs. This combination of modeling and investigation is leading to significant advancements in soft materials, offering new prospects in various scientific and industrial fields. 

Refractory metals, known for their high melting points, corrosion resistance, and exceptional thermal conductivity, face limitations with traditional BCC structure interatomic potentials. These limitations are evident in phonon dispersion curves, screw dislocation core structures, and point defect formation energies. ML interatomic potentials, designed to emulate DFT results, offer improved accuracy and efficiency in MD simulations but require extensive DFT data for parameterization. To address this, we introduce a physics-informed Generalized EAM (GEAM) potential. The primary advantages of the GEAM potential are: (a) its use of basis functions, providing flexibility for various atomic environments and easing free parameter optimization; (b) employing linear regression to reduce free parameters, allowing for effective training with smaller datasets compared to ML potentials; and (c) a nonlinear three-body term that adds higher-order interactions without excessively increasing coefficients.