Machine Learning Improves High-power Laser Experiments
Commercial fusion energy plants and advanced compact radiation sources are increasingly reliant on high-intensity, high-repetition-rate lasers firing numerous times per second. However, the rapid pace at which these lasers operate presents challenges for human operators in responding to changes effectively.
To address this issue, an international team of scientists from Lawrence Livermore National Laboratory (LLNL), Fraunhofer Institute for Laser Technology (ILT), and the Extreme Light Infrastructure (ELI ERIC) collaborated on an experiment to optimize a high-intensity, high-repetition-rate laser using machine learning technology.
Lead researcher Matthew Hill from LLNL highlighted the objective of achieving reliable diagnosis of laser-accelerated ions and electrons from solid targets under high-intensity and repetition rates. The team utilized a machine-learning optimization algorithm to provide rapid feedback to the laser front end, enabling them to maximize the total ion yield of the system.
The researchers leveraged a closed-loop machine learning code developed by LLNL’s Cognitive Simulation team to train on laser-target interaction data, allowing real-time adjustments to the laser pulse shape during the experiment. The machine learning-based optimizer received data generated throughout the experiment, enabling dynamic pulse shape modifications on the fly.
Operating at a firing rate of every 5 seconds with laser intensities exceeding 3×10^21 W/cm², the laser achieved impressive results over 120 shots before the copper target foil needed replacement. The team monitored diagnostics for damage and evaluated debris accumulation from vaporized targets. Throughout three weeks of experiments at ELI, with laser firing up to 500 shots daily, the researchers maintained a focus on applying machine learning to the high-rate-laser experiment.
Working at the ELI Beamlines Facility in the Czech Republic, the team capitalized on the advanced High-Repetition-Rate Advanced Petawatt Laser System (L3-HAPLS) to generate protons in the ELIMAIA laser-plasma ion accelerator. The collaboration between LLNL, ILT, and ELI aimed to simplify experiment elements, such as utilizing a basic copper foil target, while exploring the application of machine learning algorithms in high-repetition-rate laser experiments.
By firing over 4,000 shots during the campaign, the team achieved statistical analysis of results and demonstrated optimization of ion yield beyond the baseline performance. The experimental physicists embraced the spectator aspect of using machine learning, as they observed real-time data and anticipated the optimizer’s actions, marking a significant departure from traditional manual intervention experiments.
The successful execution of this complex experiment underscored the quality and reliability of the L3-HAPLS laser system. ELI’s Bedrich Rus praised the laser’s exceptional performance in providing a stable, high-repetition-rate beam for generating secondary sources like electrons, ions, and x-rays.
LLNL’s partnership in advancing the HAPLS laser with ELI Beamlines showcased the robustness and precision of the laser system, following its installation in the Czech Republic in 2017. This experiment, awarded through a competitive worldwide call for proposals, highlighted the collaborative efforts of teams from LLNL, ILT, and ELI, supported by state-of-the-art diagnostics and instruments.
The dedicated preparation and thorough implementation of the experiment demonstrated the deep commitment of the research team to exploring innovative applications of machine learning in laser physics. With a focus on optimizing laser performance and maximizing ion yield, the successful campaign at ELI Beamlines heralded a new era in high-intensity, high-repetition-rate laser experiments driven by advanced computing techniques and international collaboration.