Laser lightning: Stanford-led AI crunches crystal physics in milliseconds
Researchers at Stanford, UCLA and SLAC National Accelerator Laboratory built a deep-learning model that simulates how laser pulses behave inside special crystals. It targets second-order nonlinear optics, which matters for the lasers that help generate X-rays at SLAC's LCLS-II. Conventional split-step Fourier simulations account for about 95 percent of processing time in the full laser simulation studied. The team's LSTM neural network, described in Advanced Photonics, cut average run time to a few milliseconds per simulation on GPUs, without substantially compromising accuracy.
Why it matters: Faster laser simulations could speed the design and tuning of lasers that help generate X-rays at SLAC's LCLS-II, benefiting researchers who rely on them, since the AI model cuts run time to milliseconds without substantially losing accuracy.
Comments
No comments yet.