High-entropy alloy and intermetallic design by Super-Cell Random Approximates (SCRAPs)

Structural models of high-entropy alloys with specific short-range order were developed through a highly accelerated hybrid Cuckoo Search – an evolutionary algorithm that combines Levi flights for global searches with Monte Carlo for local searches [Nat. Comput. Sci. 1, 54 (2021)]. We extend SCRAPs (ver. 2.0) to rapidly create specific (h k l)-oriented bulk, interface, or surface models by utilizing an efficient algorithm by Hermann [Surf. Rev. Lett. 4, 1063 (1997)] as the starting unit. As an example, we find an optimal solution from 10⁷³ unique configurations in ~13 seconds that also exhibits strong scaling when run on parallel architectures. These SCRAPs 2.0 representative structural solutions can then be used within density-functional theory to design properties of high-entropy alloy and intermetallic surfaces, such as to design catalytic surfaces for hydrogen-evolution reactions. Finally, we demonstrate that the integration of an Integer Linear Programming Library (LPSolve) adds functionality and improves efficiency by avoiding searches when a solution in unfeasible and setting system-specific variability to prevent Monte Carlo from stagnating.