Finding the global minimum of the potential-energy surface is a fundamental problem in material science, with applications ranging from protein folding to cluster physics—or more generally, to systems where the number of (meta)stable configurations becomes prohibitively large. Classical annealing (CA) is a time-honored, robust technique that explores the configurational phase space via thermal fluctuations. However, in recent decades, quantum annealing (QA) has emerged as a promising alternative, relying instead on quantum fluctuations and often outperforming CA in multi-minima systems with large energy barriers [1,2].
We propose a novel QA implementation using path-integral molecular dynamics, a method well-suited for sampling quantum nuclear densities. This offers an alternative to the path-integral Monte Carlo approaches used in previous works [1,3]. After validating our method on standard benchmarks such as Lennard-Jones clusters, we apply it—together with machine-learning potentials—to problems of broad relevance in material science.
[1] G. E. Santoro et al., J. Phys. A 39, R393 (2006).
[2] L. Stella et al., Phys. Rev. B 72, 014303 (2005).
[3] T. Gregor et al., Chem. Phys. Lett. 412, 125 (2005)
Laboratory for Materials Simulations (LMS)