Logo image
Back
Computational design of refractory high entropy alloys with optimized mechanical properties
Dissertation

Computational design of refractory high entropy alloys with optimized mechanical properties

Christopher A. Tandoc
Doctor of Philosophy (Ph.D.), Drexel University
Dec 2025
DOI:
https://doi.org/10.17918/00011452
pdf
Tandoc_Christopher_20253.58 MB
PDF Embargoed Access, Embargo ends: 30 Jun 2028

Abstract

Refractory high-entropy alloys (RHEAs) have emerged as promising materials for structural applications in extreme environments such as gas turbines for aircraft propulsion and power generation. Their exceptional thermal stability and strength retention at high temperatures make them attractive for replacing conventional superalloys. RHEAs are composed of multiple refractory metallic elements in near-equiatomic proportions, rather than a single dominant base element. This design strategy dramatically expands the compositional design space, resulting in an effectively infinite number of possible alloy combinations that are impractical to explore experimentally. First-principles, Density functional theory (DFT) calculations present a powerful approach to calculate atomic-level properties that govern alloy behavior, but its computational cost remains non-trivial, especially when evaluating large combinatorial spaces. To address this challenge, the present work develops a suite of computational tools that combine first-principles calculations with physics-informed machine learning models to enable efficient, interpretable prediction of key crystalline defect and structural properties from alloy composition. These tools accelerate the design of ductile and strong RHEAs by enabling rapid exploration of the large compositional design space. First, a bond-based model is introduced to predict lattice parameters. Accurate prediction of lattice parameter is important to support the design of alloys with controlled lattice mismatch at phase interfaces. The model uses DFT-calculated bond lengths from binary B2 intermetallics crystals to estimate the average of atomic bond lengths in multicomponent alloys. This approach captures nonlinear compositional trends arising from charge transfer and local chemical environment effects, outperforming the current standard approach, Vegard's law, on a DFT-predicted lattice parameter dataset for 292 alloy compositions. The predicted lattice parameters and bond-length statistics also serve as inputs for later models of lattice distortion and defect energetics. Second, a statistical framework is developed to quantify lattice distortion in RHEAs based on structural and electronic descriptors. Lattice distortion is the key driver of solid solution strengthening in RHEAs and invaluable for estimating an alloy's yield strength. Using DFT-calculated root-mean-squared atomic displacements as a target, the model links lattice distortion with atomic bond lengths, and d-band filling. The resulting distortion metric correlates with yield strength, providing a physically meaningful way to evaluate local structural disorder and its impact on mechanical behavior. Third, to address solid-solution softening and promote sustained strain hardening, machine-learning surrogate models are constructed to predict unstable stacking fault energies ([gamma]_[USF]) for dislocation glide on the {112} and {123} slip systems. Support vector regression models trained on DFT data accurately reproduce [gamma]_[USF] across a wide compositional space using physics-informed features such as elastic moduli, valence electron concentration, bond-length statistics, and electronic descriptors. By comparing [gamma]_[USF] across intersecting slip planes, a new descriptor--the H-parameter--is proposed to quantify the energetic proximity of {112} and {123} slip systems by taking the ratio [gamma]_[USF]^[112] / [gamma]_[USF]^[123]. This metric correlates strongly with experimental measurements of work-hardening capacity and necking strain, suggesting that slip multiplicity, enabled by similar fault energies across multiple planes, promotes dislocation interactions that sustain plastic deformation. Together, these models provide a unified, physically interpretable framework for linking alloy chemistry to local structure, defect energetics, and macroscopic deformation behavior in body-centered-cubic (bcc) RHEAs. This dissertation establishes a foundation for data-driven, mechanism-informed alloy design in chemically complex bcc systems, with broad implications for high-temperature materials engineering.

Metrics

1 Record Views

Details

Logo image