Machine learning helps to facilitate jittering of high power lasers

Researchers at the Department of Energy's Lawrence Berkeley National Laboratory (Berkeley Lab) have made breakthroughs in laser technology by stabilizing high-power lasers using machine learning (ML).

The advance, led by Berkeley Lab's Accelerator Technology & Applied Physics (ATAP) and engineering departments, promises to accelerate advances in physics, medicine and energy. Researchers report their work in journals High Power Laser Science and Engineering.

Jitter prediction

High power lasers have become an essential tool for both scientific research and industry. One exciting application of these lasers is the Laser Plasma Accelerator (LPA). This allows particles to accelerate to high energy over short distances.

LPA provides more compact and cost-effective particle corridors and new light sources, allowing for the search of matter at the atomic and molecular scale. High power lasers also support advances in inertial fusion, which promises rich and reliable energy.

https://www.youtube.com/watch?v=i4ijtsjahzy

However, variations in beam pointing, known as “jitter,” caused by mechanical vibrations, hinder laser performance and prevent advances in these applications.

“Laser pointing errors are particularly problematic because they cause instability in the electron beam from which the LPA is generated,” said Dan Wang, a research scientist at ATAP's Berkeley Accelerator Controls & Instrumentation program and one of the paper's lead authors.

However, traditional laser control systems “has struggled to accommodate rapid changes in laser positions. “This creates shot-to-shot errors that have negative impact on the experiment,” according to Anthony Gonsalves, staff scientist and associate director of the experiment at ATAP's Bella Center.

To overcome this limitation, the team turned to machine learning.

Unlike traditional control systems after laser pointing errors occur, “Our method predicts jitter, adjusts the laser's optical components in real time, rapidly improving shot-to-shot stabilization and more accurate beam pointing.”

“Pilot” beam and real-time adjustment

To test the effectiveness of this method, researchers adopted a low-power, altitude, “pilot” laser beam as the proxy for the high-power, low-recovery rate main beam of Bella Petawatt Laser, a major LPA research facility.

“The pilot beam is fired much more frequently than the main beam, allowing you to map the movement of the beam caused by the vibrations of the mirror,” Goncalves said. “This information can be used to predict where the beam will be when the high-power pulse arrives. Since you know the pointing errors in advance, you can adjust the mirror to correct these errors.”

They fed this position data to an ML-enabled control system and adjusted the beam pointing using a correction mirror. After testing performance, the system reduced jitter by 65% ​​in the X direction of the beam and 47% in the Y direction.

“We will be using field programmable gate arrays, electronic control circuitry that provides advanced timing and synchronization to enhance methods that allow for faster, more accurate real-time corrections,” Wang said. “This is expected to improve the stabilization of the shot-to-shot laser as testing is planned with Verapetawatt lasers in full power and a wide range of applications.”