Optical Analog Computing and Communications with Configurable Nonlocal Photonics
Abstract
The exponential growth of signal processing and computing demands in modern societies has become a major challenge due to the associated energy demands. Currently these tasks are accomplished by power-hungry electronic systems. However, there is an emerging opportunity for more energy-efficient optics to offload some of the heavy lifting done by electronics, opening the door to a more sustainable future. The potential benefit of using light waves to perform computation and processing has been known since the 1960’s, yet conventional optical computing systems based on Fourier-Optics have generally lacked flexibility, require bulky devices and systems and are typically restricted to Fourier-transform-based processing. Recently, however, conceptually new nanophotonic approaches have been introduced and these have brought exciting opportunities for arbitrary linear processing of data with extremely compact nanophotonic and metamaterial systems. They enable the creation of fully self-configuring optical meshes that can self-adapt to a problem of interest. Once configured, such optical computing systems can, in principle, perform arbitrary analog operations in a massively parallel fashion and with nearly zero power dissipation on incoming optical signals. They also offer a highly desirable ultrathin, planar form-factor and easy integration with electronics. For instance, it has been recently shown that a mesh of Mach-Zehnder interferometers can realize arbitrary linear transforms in optics, going well beyond Fourier transforms, and including direct applications for self-aligning optics, sensor preprocessing, mode conversion, linear quantum circuits, and matrix coprocessors. Most recently, metamaterials and metastructures have also been shown to enable sophisticated mathematical operations on the incident optical wavefront by designing a transfer functions based on the complex nanostructuring of the metamaterial layers, and even solving integral equations. A particularly promising concept is the idea of flexible non-locality - each optical output value is built controllably from inputs over a selectable area or number of inputs, a concept that greatly expands the range of possible mathematical operations. Extensions to nonlinear operations, dynamic reconfigurability, and demanding computational tasks remain to be pursued. The successful implementation of these computing systems demands a substantial new materials development to allow for reconfigurability, and new theoretical approaches, such as full modal and nonlocal descriptions to expose fundamental scaling limits to performance and function. In this program, our team will realize a range of highly-programmable and reconfigurable materials and develop universal nonlocal metasurface-based photonic elements to perform a wide range of analog mathematical operations. We will aim to overcome the current limitations of Fourier optics technologies and perform linear and nonlinear operations with flat optical systems that will be dynamically reconfigurable and offer robust scaling. A multidisciplinary research effort will bring together mathematics, physics, engineering, materials science, quantum technologies and control theory, in order to design, prototype, and analyze the fundamental limitations of novel meta-structures and photonic meshes. There are many open, fundamental questions pertaining to the optimum design for photonic mesh architectures, but it has become eminently clear that the nature and range of mathematical operations greatly depends on the degree of nonlocality of the meshes. To address this point, our team will explore the possibility of creating universal nonlocal elements capable of delivering currently unimaginable levels of nonlocality. These revolutionary systems will be capable of performing near-zero energy computations that go well beyond the capabilities of both electronics and current optical systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 29, 2024
- Source ID
- FA95502310307
Entities
People
- Mark L. Brongersma
Organizations
- Air Force Office of Scientific Research
- Stanford University
- United States Air Force