Machine-learning optimization of laser-pulse shapes and target structures for direct-drive implosions

The 23^rd Academic Exchange Seminar Between Shanghai Jiao Tong University and Osaka University

Fuyuan Wu

( School of Physics and Astronomy, Shanghai Jiao Tong University )

ABSTRACT

Double-cone ignition (DCI) scheme is a fast-ignition scheme based on laser direct-drive configuration. DCI uses a relatively small drive energy in a direct-drive conical implosion to form a high-density colliding plasma, which is then be heated by magnetic field guided fast electron beams generated by PW laser pulses.

Genetic algorithm is an optimization algorithm which imitates the evolution process of population in biology. It has the advantages of high parallel efficiency and strong versatility. However, there is a risk of falling into local optimum in the optimization process. Random forest is a machine learning algorithm in data science based on decision trees. It can be used for big data classification, regression, prediction of complex physical system and feature importance analysis of independent variables. In this report, we propose an automated approach for designing laser-pulse shapes and target structures for the DCI scheme based on the machine-learning approach. This machine-learning design approach has been successfully demonstrated in the experimental campaign in 2020 conducted on the SG-II upgrade laser facility with a drive energy up to 12 kJ from 8 laser beams