Viewing high entropy alloys through glasses: Linkages between solid solution and glass phases in multicomponent alloys

Abstract

High entropy alloys (HEAs) represent highly promising multicomponent crystals that form concentrated solid solutions (CSSs) and may violate traditional thermodynamic rules of mixing, ultimately leading to excellent physical properties. For a deeper understanding, we investigate seven CSSs, including Co-Cr-Ni-Fe-Mn elements, at experimentally relevant compositions and conditions, through molecular simulations, and we use 1-1 comparisons to corresponding glass state characteristics, attained through rapid cooling protocols. We determine the behavior of various structural features, including the configurational entropy for a set of CSSs in their crystalline and vitreous states numerically. We employ swap Monte Carlo (MC) simulations, in combination with the reversible scaling method, to efficiently compute the configurational entropy (${S}_{\text{conf}}$), and show that the entropic rule of mixing is not always adequate for predicting alloy formation. We study the stability and formability of crystalline solid solutions, as well as glasses, while following the thermodynamics of ${S}_{\text{conf}}$. An apparent entropic similarity between CSSs and corresponding glasses leads us to use a Kauzmann-like ansatz, relating the CSSs at ${S}_{\text{conf}}\ensuremath{\rightarrow}0$ with the emergence of a CSS order-disorder transition, at temperature ${T}_{OD}$. In the context of glasses, a comparison between kinetic and thermodynamic fragilities allows the association of sluggish diffusion onset to a drop in ${S}_{\text{conf}}$ at ${T}_{K}$. Analogously, we classify CSSs as ``strong'' or ``fragile'' in the sense of their ability to migrate across CSS crystal configurations at high temperatures, distinguishing its formability. We argue that the magnitude of ${T}_{OD}$ may be an excellent predictor of CSS single-phase stability, which appears to scale with well-known HEA predictors, in particular we notice that VEC and ${T}_{OD}$ have in relation to the others a significantly large Pearson correlation coefficient, much larger than most other observables (except $\mathrm{\ensuremath{\Delta}}{H}_{\text{mix}}$).