Machine Learning-Driven Optimization of Burst Femtosecond Laser Processing for High-Performance Anti-Reflective...

Area of Science:

• Materials Science
• Optical Engineering
• Biomedicine

Background:

• Femtosecond laser processing enables micro/nanostructure fabrication for diverse applications.
• Current methods rely on inefficient trial-and-error, hindering high-performance manufacturing.

Purpose of the Study:

• To develop an efficient machine learning (ML) strategy for real-time prediction and optimization of femtosecond laser fabrication.
• To achieve high-performance anti-reflective microstructures with improved efficiency.

Main Methods:

• Trained a multilayer perceptron model using finite-difference time-domain simulation data.
• Established a nonlinear mapping between microstructural parameters and transmittance.
• Integrated the ML model into the fabrication system for rapid prediction (0.004 s).

Main Results:

• Achieved efficient manufacturing of large-area (12 × 12 mm²) anti-reflective microhole arrays on MgF₂ windows.
• Fabricated structures at a rate of 10,000 holes per second with 2 μm periodicity.
• Obtained an average transmittance of 99.03% (3-5 μm range) with broad angular stability (0-50°).

Conclusions:

• ML-assisted femtosecond laser processing significantly enhances fabrication efficiency and microstructure performance.
• The developed anti-reflective MgF₂ windows demonstrate excellent infrared imaging capabilities.
• This approach facilitates the deployment of advanced optical components in demanding environments.