Wavefront Distortion Correction for the Non-Uniform-Intensity Light Based on Extreme Learning Machine

Haiqi Lin,Shuai Wang,Ping Yang,Xing He

doi:10.1109/jphot.2021.3086005

Abstract

This paper aims to employ a neural network method of extreme learning machine to solve the problem of laser beam cleanup under the dynamic non-uniform-intensity distribution. This neural network is used to fit the corresponding nonlinear relationship between the slope information detected by the wavefront sensor and the recovery voltage of deformable mirror under the dynamic non-uniform-intensity distribution. The simulation results show that the beam quality after correction is improved from 1.7 to 1.3 times diffraction limit. And the residual wavefront after correction for the proposed method has been reduced by 55%, while the direct gradient method is 17%.

Highlights

In recent years, many laser application fields need laser to work on a high output power level while simultaneously keep a relatively high beam quality [1]–[3]
The linear regression relationship between the local slopes measured by wavefront sensor (WFS) and the corresponding recovery voltage of deformable mirror (DM) is no longer well satisfied
It is difficult for the direct gradient method to calculate the corresponding recovery voltage of DM and further improve the beam quality

Summary

Introduction

Many laser application fields need laser to work on a high output power level while simultaneously keep a relatively high beam quality [1]–[3]. As the laser output power increases, the thermally induced the non-uniform-intensity distribution or strong scintillation regime of the near-field laser beam will achieve a level that WFS may not achieve outstanding performance. When the laser works in this state, some sub-aperture

Objectives

Results

Conclusion