Gaussian processes for choosing laser parameters for driven, dissipative Rydberg aggregates

TL;DR

This paper demonstrates how Gaussian processes can efficiently predict optimal laser parameters to achieve thermalization in Rydberg atom aggregates, reducing computational effort in parameter searches for quantum simulations.

Contribution

The study introduces the application of Gaussian processes to predict thermalization performance in Rydberg aggregates, enabling efficient parameter optimization for larger systems.

Findings

Gaussian processes accurately predict thermalization with only 1000 simulations.

Effective laser parameters for thermalization are identified and shown to be robust.

Different thermalization dynamics are characterized using the predictive model.

Abstract

To facilitate quantum simulation of open quantum systems at finite temperatures, an important ingredient is to achieve thermalization on a given time-scale. We consider a Rydberg aggregate (an arrangement of Rydberg atoms that interact via long-range interactions) embedded in a laser-driven atomic environment. For the smallest aggregate (two atoms), suitable laser parameters can be found by brute force scanning of the four tunable laser parameters. For more atoms, however, such parameter scans are too computationally costly. Here we apply Gaussian processes to predict the thermalization performance as a function of the laser parameters for two-atom and four-atom aggregates. These predictions perform remarkably well using just 1000 simulations, demonstrating the utility of Gaussian processes in an atomic physics setting. Using this approach, we find and present effective laser parameters for generating thermalization, the robustness of these parameters to variation, as well as different thermalization dynamics.