Ab initio Design of Alloy Materials

Alloys present a great challenge for computational materials design, as their intrinsic chemical disorder requires advanced ab initio techniques capable of accurately capturing both chemical and magnetic complexity.
This thesis develops ab initio–based methodologies for predicting the mechanical properties of alloys by combining density functional theory (DFT), Green’s function-based methods, and machine-learned interatomic potentials to achieve high accuracy with computational efficiency.
The first major contribution is the development of a computational framework for efficient and robust compositional screening and Pareto-optimal design of alloys through the integration of DFT workflows with Bayesian multi-objective optimization. Two primary modeling workflows were established: (1) a strengthening model, based on linear elastic constants and atomic misfit volumes obtained from DFT within the coherent potential approximation (CPA); and (2) a ductility model, based on stacking-fault and surface energies predicted using actively learned interatomic potentials with ab initio accuracy.
This framework was applied to predict solid-solution strengthening in face-centered cubic (fcc) 3d transition-metal alloys and to investigate the temperature dependence of yield strength in these systems. Then, using a surrogate-assisted genetic algorithm, Pareto-optimal body-centered cubic (bcc) refractory compositionally complex alloys were found using the combined two-objective workflow (strength and ductility).
The second major contribution involves a direct ab initio investigation of dislocation–impurity interactions, addressing cases where model-based approaches are insufficient. This includes the development of tools for constructing dislocation geometries, visualizing Nye tensors and differential displacement maps, and quantifying impurity–dislocation interaction energies for various 3d elements (V, Cr, Mn, Cu, Ni, and Co) in and around 1/2<111> screw dislocations in alpha-Fe under both ferromagnetic and paramagnetic conditions. Furthermore, groundwork is laid for computing the quantum-mechanical microscopic stress tensor through an interpolation scheme for solving the Poisson equation using screened spherical waves.