Beam Science with LaserNetUS & Other User Facilities

11:30 am – 1:54 pm, Wednesday March 19 // Session MAR-M45 //Anaheim Convention Center, 263A (Level 2)

Chair:: Felicie Albert, Lawrence Livermore National Laboratory
Topics:
Sponsored by: DLS
DPB

machine learning for digital twin surrogates of high-power laser systems

1:06 pm – 1:18 pm

Presenter: Jack E Hirschman (Stanford University)

Authors: Erfan Abedi (University of California, Los Angeles), Hao Zhang (University of California, Los Angeles), Randy Lemons (SLAC National Accelerator Laboratory), Sergio Carbajo (University of California, Los Angeles)

Machine learning (ML) is driving advances in optics and photonics, promising to revolutionize the design and operation of complex laser systems. Laser system optimization is complex, and ML enhancements rely on extensive datasets. We present a modular start-to-end (S2E) software framework for high-power laser systems [1]. We apply the framework to a complex, high-power, application: the drive photoinjector laser system at SLAC’s LCLS-II, a next-generation high-brightness X-ray free electron laser (XFEL) [2]. The drive laser includes pulse shaping, amplification, and nonlinear optical (NLO) methods. We demonstrate how the individual cascaded processes can be linked end-to-end for full integration into a S2E simulation and generate an ML model that achieves a 250x speed-up over traditional numerical simulations [3], paving the way for generalized digital twins and enhanced synergy between simulation and experiment.

[1] Hirschman, et al. "Design, tuning, and blackbox optimization of laser systems." Opt. Exp. (2024).

[2] Zhang, et al. “The LCLS-II Photoinjector Laser Infrastructure.” HPLSE. (2024).

[3] Hirschman, et al. “A Modeling Framework and LSTM Nonlinear Dynamics Predictor for Accelerating Data Generation in Ultrafast Optics and Laser System Design.” Opt. Open (2023).

PRESENTATIONS (10)