Uniform design and deep learning based liquid lens optimization strategy toward improving dynamic optical performance...

Area of Science:

• Optics and Photonics
• Computational Engineering
• Artificial Intelligence

Background:

• Liquid lenses offer tunable focal lengths but face challenges in optimizing optical performance and minimizing driving force.
• Existing optimization methods often result in localized solutions, failing to achieve global performance improvements.

Purpose of the Study:

• To develop an efficient optimization strategy for liquid lenses by combining uniform design and deep learning.
• To achieve simultaneous improvements in dynamic optical performance and reductions in driving force.
• To obtain a globally optimized liquid lens design with superior aberration control and image quality.

Main Methods:

• A plano-convex liquid lens membrane was designed with optimized contour functions and central thickness.
• Uniform design selected representative parameter combinations, with performance data simulated using MATLAB controlling COMSOL and ZEMAX.
• A four-layer deep learning neural network was trained to predict performance based on parameter combinations.

Main Results:

• The trained deep neural network demonstrated effective performance prediction for all parameter combinations.
• The globally optimized design significantly reduced spherical and coma aberrations across the focal length tuning range.
• Compared to conventional and locally optimized designs, the new method achieved substantial reductions in driving force and improved modulation transfer function (MTF) curves.

Conclusions:

• The proposed hybrid optimization strategy effectively achieves global optimization for liquid lenses.
• This approach leads to superior optical performance, reduced driving force, and enhanced image quality.
• The method provides a robust framework for designing advanced liquid lens systems.