Predict Thermodynamic, Thermal and Mechanical Properties of High Entropy Alloys using Physics-Informed Machine Learning

High entropy alloys represent a new paradigm shift in materials design. The main challenges in high entropy alloys development are the lack of efficient and reliable tools to predict the composition-processing-microstructure-properties relationships of vastly unexplored compositional space as well as their dependence on temperature, pressure, and other environmental/operating conditions. In this talk we will present our ongoing research effort in predicting thermodynamic, thermal and mechanical properties of HEAs by integrating multi-scale computational modeling with machine learning. Modeling techniques include high throughput CALPHAD calculations, density functional theory methods (DFT), and continuum models. CALPHAD method allows us to predict phase diagram information including phases and their mole fractions and phase transformations. Single crystal elastic constants and average elastic properties, defect energetics (such as surface energy and stable and unstable stacking fault energies), and intrinsic ductility are predicted using DFT methods. To predict coefficient of thermal expansion (CTE) and temperature-dependent elastic constants, we adopt an optimized Debye model by calibrating the Debye temperature and its volume dependence of pure elements by comparing the predicted heat capacity and CTE to the experimental results. Based on these predicted properties from physics-based models and reported experimental results, machine learning modeling are performed to predict thermal and mechanical properties of HEAs. The integration of high-fidelity multiscale modeling with machine learning not only helps decipher complex materials challenges such as high temperature strength and creep behavior but also accelerates the design of cost-competitive high-performance alloys for extreme environment applications.