Microlens Array-Based Beam Profile and Wavefront Sensor With Physical Constraint Learning

Abstract

The beam profile and wavefront characteristics of laser beams are essential for numerous laser applications, including micromachining and microfabrication. However, conventional wavefront sensors, such as the Shack-Hartmann wavefront sensor (SHWS), are limited by reduced accuracy in detecting local distortions and sensitivity to non-uniform beam profiles. Additionally, beam profile information is crucial for such applications. This paper introduces a new methodology that utilizes an SHWS-like structure to overcome these limitations. By employing a physical constraint learning approach, the proposed method simultaneously provides highly accurate wavefront and beam profile data. We first develop a pretrained network using microlens array (MLA) simulation datasets. To implement a practical MLA-based measurement system, this pretrained network is further fine-tuned with datasets modulated by a spatial light modulator in the system setup. Experimental results demonstrate that the proposed network can reconstruct both beam profiles and wavefronts in real-time. Compared to traditional SHWS reconstruction techniques, our approach enhances computation speed by over 100 times, while also providing beam intensity profile information and increasing wavefront sensing accuracy by approximately fivefold.