Scientists Use AI To Supercharge Ultrafast Laser Simulations by More Than 250x

Researchers built an LSTM-based AI surrogate for χ² nonlinear optical interactions in ultrafast laser simulations.

In tests on noncollinear sum-frequency generation, it matched the accuracy of conventional split-step Fourier solvers across varied pulse conditions.

The model cut runtime from costly iterative calculations to millisecond-scale inference, delivering more than 250x speedup.

The advance could speed up laser design for UV and X-ray sources by improving feedback on pulse timing and shape.