Stabilisation of transverse mode purity in a radially polarised Ho:YAG laser using machine learning

Radially polarised solid-state lasers offer attractive improvements in materials processing applications, but selection and stabilisation of the appropriate radially polarised mode is much more challenging than for the fundamental mode. Here, we demonstrate automated stabilisation of a radially polarised Ho:YAG laser by utilising laser mode analysis computed from a convolutional neural network. The neural network predicts the transverse modal content from single plane intensity images with high accuracy on timescales of a few milliseconds, permitting real-time self-adjustment of the laser cavity. Radially polarised emission has been maintained across a 30 W range of pump power, with the stabilisation of other arbitrary laser modes using the same neural network also demonstrated.