Comprehensive Mapping of Tracer Diffusivities Across Composition Space in Ternary NiAlTi and Quinary NiCoFeAlTi High-Entropy Alloy Using Diffusion Couple Experiments and Physics Informed Neural Network Inversion

A comprehensive experimental and physics informed neural network numerical inverse diffusion analysis is conducted in technologically important NiAlTi ternary and NiCoFeAlTi quinary solid solutions for estimating and extracting composition dependent diffusion coefficients. A systematic variation of tracer, intrinsic and interdiffusion coefficients with composition could be estimated in the ternary solid solution. Following, the possibility of producing Al Ti constant PB diffusion profiles keeping constant Ni, Co, Fe in the quinary system is demonstrated. The estimation of diffusion coefficients of all the elements at the Kirkendall marker plane of a single diffusion couple profile is elaborated. PINN optimisation parameters are established using self and impurity diffusion coefficients in Ni and tracer diffusion coefficients at the Kirkendall marker plane. The reliability of optimized parameters is validated by comparing with the interdiffusion coefficients estimated from binary NiTi, NiAl and PB diffusion profiles, indicating extendibility to even lower order systems.