Holography optimization based on combining iterative Green's function algorithm and deep learning method.

Area of Science:

• Optics and Photonics
• Computational Physics
• Machine Learning Applications

Background:

• Solving the Helmholtz equation is crucial for wave propagation simulations.
• Holographic reconstruction often requires accurate refractive index modulation.
• Existing methods may lack generalization capabilities for complex optical fields.

Purpose of the Study:

• To develop a novel, generalized approach for refractive index modulation holography.
• To efficiently solve the Helmholtz equation using a hybrid numerical and machine learning method.
• To demonstrate the method's effectiveness across diverse holographic reconstruction scenarios.

Main Methods:

• A non-uniform refractive index convolutional neural network (NRI-CNN) was designed to extract feature vectors.
• An iterative Green's function algorithm (IGFA) was proposed to approximate the Helmholtz equation solution.
• A U-net architecture (ERPU-net) was developed to manage equation residuals and optical field phases.

Main Results:

• The combined NRI-CNN and IGFA approach achieved generalized refractive index modulation holography.
• The abstracted feature vector enhanced the generalization ability of the iterative algorithm.
• Successful holographic reconstructions were demonstrated for Gaussian beams, image data, and turbulent phase alterations.

Conclusions:

• The proposed hybrid physics-informed neural network and iterative algorithm offers a powerful tool for solving the Helmholtz equation.
• This method significantly advances the field of generalized holographic reconstructions.
• The approach shows promise for applications involving complex light field manipulations and simulations.