Ultrafast farfield simulation of non-paraxial computer generated holograms

Area of Science:

• Optics and Photonics
• Computational Electromagnetics
• Machine Learning Applications

Background:

• Simulating large-area diffractive optical elements (DOEs) is computationally intensive, especially when considering non-paraxial propagation and electromagnetic coupling between adjacent subwavelength features.
• Existing rigorous simulation methods struggle with the scale and complexity of modern DOEs, particularly computer-generated holograms (CGHs) with features at the wavelength scale.

Purpose of the Study:

• To develop a novel, efficient, and accurate method for the far-field simulation of large-area DOEs, including CGHs with subwavelength features.
• To leverage machine learning to accelerate the prediction of optical performance based on geometric parameters.
• To provide a simulation tool that maintains high fidelity compared to traditional rigorous methods.

Main Methods:

• A machine learning approach was employed, utilizing a linear regression model trained on data generated from physical simulation methods.
• Geometric parameters of the DOEs were used as input features for the machine learning model.
• The model predicts the far-field optical function, effectively bypassing computationally expensive rigorous simulations for prediction.

Main Results:

• The developed machine learning method enables significantly faster computation of the far-field optical performance of DOEs.
• The simulation results obtained using the machine learning model demonstrate high conformity with those derived from rigorous electromagnetic simulation techniques.
• The method effectively addresses the challenges posed by non-paraxial propagation and coupling effects in large-area DOEs.

Conclusions:

• The novel machine learning-based simulation method offers a computationally efficient alternative for analyzing large-area DOEs and CGHs.
• This approach accelerates the design and optimization cycle for optical elements with complex, subwavelength structures.
• The high accuracy and speed make this method suitable for practical applications in optical engineering and design.