Abstract

The task of sampling from a multidimensional finite-temperature classical Boltzmann probability distribution is a central problem in numerical simulations of physics, chemistry, and beyond the traditional boundaries of natural sciences. In this talk, I will introduce a recent algorithm that can be executed on quantum computers, offering a scaling advantage compared to state-of-the-art Metropolis schemes. In practice, we can leverage the fact that the collapses of a wave function are uncorrelated and use them as trial updates to obtain non-local but effective moves in the configuration space. The algorithm was invented in 2021 for continuous systems, where a rigorous justification can be found.[1] Subsequently, it was adapted to spin systems amenable to hardware implementation, where it has been experimentally demonstrated.[2]

[1] Mazzola, PRA , 104, 022431 (2021) [2] Layden, Mazzola et al, arXiv:2203.12497 (2022)