Bayesian Optimization of Direct-Drive Inertial Confinement Fusion Simulations

Inertial confinement fusion (ICF) is a form of controlled nuclear fusion where deuterium-tritium (DT) fuel is compressed to fusion densities and pressures via ablation of the surface of the DT capsule. In ICF experiments, demonstration of sufficient energy gain remains the key obstacle for commercially viable fusion-energy. ICF experiments are expensive and incredibly complex, leading to only a limited number of experiments being run each year. For this reason, state of the art radiation-hydrodynamics simulations are employed alongside experiments to examine a range of design parameters. This allows a wide variety of design spaces to be explored while minimizing cost and time. One of the most well established approaches to ICF as a possible fusion-energy source is direct-drive ICF. In this project the laser pulse for direct-drive ICF was modeled and optimized. Optimization is an attractive approach to ICF parameter spaces because it allows rapid and intelligent exploration of complex design spaces that are difficult to traverse manually. Specifically, Bayesian optimization was used to optimize the laser pulse.