Bayesian optimization of laser-plasma accelerators assisted by reduced physical models

TL;DR

This paper presents a Bayesian optimization approach that leverages reduced physical models to efficiently optimize laser-plasma accelerators, significantly reducing computational costs while maintaining high accuracy.

Contribution

It introduces a multitask Bayesian optimization method that combines high-fidelity and reduced models, enabling faster optimization of laser-plasma accelerators.

Findings

Achieved an order-of-magnitude speedup in optimization process.

Demonstrated effective integration of reduced models with high-fidelity simulations.

Enabled cost-effective exploration of large parameter spaces.

Abstract

Particle-in-cell simulations are among the most essential tools for the modeling and optimization of laser-plasma accelerators, since they reproduce the physics from first principles. However, the high computational cost associated with them can severely limit the scope of parameter and design optimization studies. Here, we show that a multitask Bayesian optimization algorithm can be used to mitigate the need for such high-fidelity simulations by incorporating information from inexpensive evaluations of reduced physical models. In a proof-of-principle study, where a high-fidelity optimization with FBPIC is assisted by reduced-model simulations with Wake-T, the algorithm demonstrates an order-of-magnitude speedup. This opens a path for the cost-effective optimization of laser-plasma accelerators in large parameter spaces, an important step towards fulfilling the high beam quality requirements of future applications.