Year: 2026
Type: article
Abstract: Deep learning has achieved remarkable success in data-rich domains, but applying it to physical systems remains challenging because labelled datasets are difficult to acquire and model predictions must be converted into fast, repeatable hardware interventions. Laser-cavity alignment in solid-state lasers is one such problem that still relies heavily on skilled operators. Here, we demonstrate an autonomous solid-state-laser feedback system in which a programmable liquid-crystal-on-silicon spatial light modulator acts as an intracavity mirror and an image-based deep neural network predicts alignment and curvature corrections from pre-lasing amplified-spontaneous-emission images. In the original optical configuration, the system initiated laser oscillation from unseen non-lasing states after a single feedback update. After manual removal and coarse reinstallation of the modulator, the same trained model recovered laser oscillation from non-lasing starting conditions within two feedback updates in three independent tests, without retraining. These results show that weak pre-lasing spatial emission patterns contain actionable information about the latent cavity state, enabling reproducible, electronically programmable initiation of solid-state laser oscillation with practical tolerance to alignment perturbations introduced by manual reinstallation.
Source: Scientific Reports
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7237358256
