Control and stabilization of spatial mode quality in a radially polarized solid-state laser using machine learning (Conference Presentation)

The automated selection and stabilization of the transverse mode of a radially polarized Ho:YAG laser is reported. A convolutional neural network (CNN) was developed to analyze the modal composition of the laser output in real-time. Calculated error signals from the CNN are compared to the desired mode, allowing a PID control algorithm to dynamically optimize the position of an intracavity lens and therefore maintain desired modal content over pump power changes. This CNN based diagnostic system provides a fast method for selection and stabilization of transverse modes in order to advance radially polarized sources for applications such as laser processing.