All-Optical Transformations Performed Using Diffractive Materials
Abstract
University of California Los Angeles will conduct basic research on the analyses of complex-valued all-optical linear transformations, passively performed by diffractive materials between an input and output field-of-view, without the need for an external power source, except for the illumination light. These all-optical, passive processors will be composed of transmissive diffractive layers, where each feature (neuron) of a given layer has an engineered complex-valued transmission coefficient, connecting itself to the successive diffractive layers by modulating the phase and/or amplitude of a local spherical wave, following the Huygens-Fresnel principle of light diffraction. The individual features/neurons of a diffractive layer will each be at the scale of half a wavelength and will be designed to collectively control the modes of light propagation within a diffractive material volume. By using an analytical inverse design strategy as well as deep learning-based optimization and error-backpropagation methods, a diffractive material system that is composed of K diffractive layers with a total of N neurons will be spatially-engineered to approximate an arbitrary complex valued linear transformation (defined by A) between the input and output fields-of-view. After the numerical design phase, the resulting diffractive layers will be fabricated and physically assembled together to all-optically perform an approximate transformation (A?) between a complex-valued input field and the output field, through light-matter interaction within the spatially-engineered volume of the diffractive material. In our analyses, we will also compare the all-optical transformation errors and diffraction efficiencies that can be achieved using data-free inverse designs and data-driven designs of diffractive materials built to perform a desired transformation. These all-optical transformations will act on both the phase and amplitude information of the input, computing the output at the speed of light propagation within the diffractive material and will not use of any external computing power, except for the illumination source.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 21, 2022
- Source ID
- FA95502110324XX0
Entities
People
- Aydoğan Özcan
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California, Los Angeles