Physics-informed CNN convolution for emulating scattering in quantum-dot color conversion displays.

Area of Science:

• Optics and Photonics
• Computational Physics
• Materials Science

Background:

• Quantum-dot color conversion films (QDCFs) are crucial for advanced display technologies.
• Accurate simulation of optical scattering in QDCFs is essential for optimizing device performance.
• Existing simulation methods can be computationally intensive.

Purpose of the Study:

• To develop a physics-informed convolutional framework for efficient optical scattering simulation in QDCFs.
• To leverage convolutional neural network (CNN) principles for modeling light propagation.
• To provide an interpretable and accurate tool for QDCF display design.

Main Methods:

• Computed scattering intensity distributions using LightTools (LT) and discretized them into grids.
• Constructed convolution kernels representing spatial energy redistribution.
• Applied iterative convolution with LT-derived kernels to input images.
• Evaluated simulation accuracy using metrics like SSIM, standard deviation, and MSE.

Main Results:

• LT-derived kernels closely matched full LT simulation scattering patterns.
• The proposed framework achieved high accuracy (e.g., SSIM of 0.95 with a 5x5 kernel).
• Larger kernels (9x9) required fewer iterations for optimal performance and accuracy.

Conclusions:

• The physics-informed convolutional framework offers an efficient and interpretable method for simulating optical scattering in QDCFs.
• The approach demonstrates robustness and accuracy, validated by edge analysis and structural similarity metrics.
• This tool aids in predicting scattering behavior for QDCF-based display systems under varying conditions.