Computer simulations of molecular systems with complex free-energy landscapes causing metastability are among the most challenging tasks in computational science. Typical examples are macromolecules such as proteins and glassy materials. Consequently, improvements in simulation methods there hold the promise of important advances in biomedical applications and materials science. In the present work the authors propose a combination of molecular dynamics simulations with a swarm technique that is similar in spirit to a genetic algorithm, but instead of merely a ground-state search, it does provide equilibrium populations over a quasi continuous range of temperatures. This new approach yields a dramatic acceleration of thermalization for complex molecular systems that is on par with the best previously known techniques. In contrast to the latter, however, for a wide variety of systems, a parallelized computer implementation of the new population based simulation method scales extremely well, ranging from graphics processing units to the largest supercomputers. It thus promises to become a tool of choice for powerful molecular simulations on petaflop supercomputers of the present and the exaflop machines of the future.

H. Christiansen, M. Weigel, and W. Janke, Accelerating Molecular Dynamics Simulations with Population Annealing, Phys. Rev. Lett. 122, 060602 (2019)