Optimal parallelisation strategies for flat histogram Monte Carlo methods
Date:
Contributed talk at the 2025 Condensed Matter and Quantum Materials Conference.
Abstract
Flat histogram methods such as Wang Landau sampling [1] provide a means for high throughput calculation of phase diagrams for atomistic/lattice model systems. Many parallelisation schemes have been proposed to accelerate sampling simulations with varying degrees of complexity [2]. In this study, these different schemes are benchmarked - both in isolation and in combination - to establish best practice. The schemes studied include energy domain decomposition with both static domains and a dynamic domain sizing which we propose. We also assess the benefit of replica exchange and including multiple random walkers per domain to determine which factor has the largest impact on parallel efficiency. The influence of flatness criteria and domain overlap will also be discussed. As illustrative test cases, we implement and apply the aforementioned strategies to a lattice-based model describing the internal energies of the the AlTiVNb and AlTiCrMo refractory high-entropy superalloys, both of which are understood to crystallographically order into a B2 (CsCl) structure with decreasing temperature [3].
References
[1] Wang, Landau, Phys. Rev. Lett. 86, 2050 (2001).
[2] Vogel, Li, Wust, Landau, Phys. Rev. Lett. 110, 210603 (2013).
[3] Woodgate, Naguszewski, Redka, Minar, Quigley, Staunton, arXiv:2503.13235.