Advancing adaptive optics towards machine learning for laser-driven experiments
Shaping the field of a driving laser offers a powerful route to controlling highly nonlinear processes, such as high harmonic generation (HHG), and lies at the heart of coherent control techniques. Adaptive optics elements, such as deformable mirrors (DMs) or spatial light modulators (SLMs), together with optimization routines are often used to control the properties of laser fields. However, the large parameter space associated with these devices often makes conventional optimization methods slow to converge.
Future applications increasingly demand the rapid generation and characterization of laser beams with complex spatial and temporal profiles. As a result, there has been growing interest in applying machine learning techniques to adaptive optics [1].
In this talk, I will first discuss experiments to control the brightness of a HHG source using spatially shaped pulses [2]. I will then present our recent work combining adaptive optics with machine learning approaches, including neural networks and vision transformers, for image-based wavefront sensing [3].
[1] Guo et al. “Adaptive optics based on machine learning: a review” Opto-Electron Adv, 5, 200082 (2022)
[2] Treacher et al. “Increasing the brightness of harmonic XUV radiation with spatially-tailored driver beams”. J. Opt. 23, 015502 (2021)
[3] O’Rourke and O’Keeffe. “Dual-plane wavefront sensing using a vision transformer” Opt. Express, 34, 6456, (2026)