Polarization-Encoded Diffractive Network for Optical Transformations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Diffractive optical networks using isotropic materials are insensitive to different polarization states, making it impossible to perform polarization multiplexed all-optical computation of different transformations.
Innovation Solution
Incorporating a non-trainable, pre-determined array of linear polarizers at 0°, 45°, 90°, and 135° within the diffractive network to act as polarization seeds, allowing the trainable isotropic diffractive layers to perform different linear transformations through input/output polarization multiplexing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If isotropic diffractive materials are used in the diffractive network, then the device complexity is reduced and ease of manufacture is improved, but the ability to perform polarization multiplexed all-optical computation is lost
Solution Approach 1:
The patent introduces polarizer arrays as intermediary elements within the diffractive network. These polarizers act as mediators that couple the isotropic diffractive layers, enabling polarization-dependent transformations. The polarizer arrays convert the polarization-insensitive diffraction process into polarization-sensitive computation by selectively transmitting or blocking specific polarization states between layers, thus resolving the contradiction between using simple isotropic materials and achieving polarization multiplexing capability.
2Productivity
If polarization multiplexing is implemented to perform multiple transformations, then the computing capacity is increased, but the device complexity increases due to the need for polarizer arrays
Solution Approach 1:
The patent implements a universal diffractive network architecture where the same isotropic diffractive layers and polarizer arrays can perform multiple different linear transformations by simply changing the input polarization state. The system is designed to be multi-functional, where a single physical device can compute different transformations (encoded in different polarization states) without requiring separate dedicated networks for each transformation, thus increasing computing capacity while managing device complexity.
Solution Approach 2:
The patent utilizes parameter changes in the polarization state of light to enable multiple transformations through a single diffractive network. By encoding different transformations into different polarization parameters (horizontal, vertical, diagonal, anti-diagonal), the system achieves multiple computing functions from a single device configuration. The polarizer arrays are trained to respond to these parameter changes, allowing the same physical structure to perform different computational tasks based on the input polarization parameter.
3Ease of manufacture
If standard isotropic diffractive layers are used without polarizers, then the fabrication is simplified, but the transformations cannot be encoded into different polarization states
Solution Approach 1:
The patent segments the diffractive network into distinct functional components: isotropic diffractive layers and polarizer arrays. The isotropic layers are fabricated using standard techniques without polarization-sensitive materials, maintaining fabrication simplicity. The polarizer arrays are separately introduced and positioned between the diffractive layers. This segmentation allows each component to be optimized independently - the diffractive layers for easy fabrication and the polarizers for polarization control - while working together to achieve accurate transformation encoding.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables the successful approximation of multiple arbitrary linear transformations with negligible error by encoding transformations into different input/output polarization combinations, overcoming the insensitivity of isotropic diffractive materials to polarization states.
Implementation Method 1
a diffractive network is first trained using deep learning and error-backpropagation methods implemented in a digital computer, after which the resulting transmissive layers are fabricated to form a physical network that computes based on the diffraction of the input light through these spatially-engineered transmissive layers
Implementation Method 2
Incorporating a non-trainable, pre-determined array of linear polarizers at 0°, 45°, 90°, and 135° within the diffractive network to act as polarization seeds, allowing the trainable isotropic diffractive layers to perform different linear transformations through input/output polarization multiplexing
Data Source
AI summary
A polarization multiplexed diffractive processor is disclosed that all-optically performs multiple, arbitrarily-selected transformations (e.g., linear) through a single diffractive network trained using deep learning. In this framework, an array of pre-selected linear polarizers is positioned between trainable transmissive diffractive materials that are isotropic, and different target linear transformations (complex-valued) are uniquely assigned to different combinations of input/output polarization states. The transmission layers of this polarization multiplexed diffractive network are trained and optimized via deep learning and error-backpropagation by using thousands of examples of the input/output fields corresponding to each one of the complex-valued linear transformations assigned to different input/output polarization combinations. This polarization-multiplexed all-optical diffractive processor can find various applications in optical computing and polarization-based machine vision tasks.