PROGRAMMABLE 2D NONLINEAR PHOTONIC WAVEGUIDES FOR ANALOG COMPUTATION USING LIGHT PROPAGATION
Abstract
Our overarching objective in this proposal is to develop a nonlinear-optical device that is capable of performing analog computation. We have designed our proposed device in a way that allows us to harness the benefits of optics over electronics for computation. There is a clear path towards achieving superior computational-performance-per-Watt using our type of device versus conventional, digital computers for tasks that are suited to analog processing. We seek to demonstrate the applicability of our proposed nonlinear-optical computing device for machine learning. There are several fundamental differences between photonics and electronics that can be leveraged to build a photonic processor that has speed and/or energy efficiency advantages over electronic processors. We have designed a photonic device (and a surrounding system for using it) that in particular makes use of i.) the large bandwidth of optics (>1 THz); ii.) large spatial parallelism (it is possible to manipulate 10-million spatial modes); iii.) the nearly dissipationless propogation of light, where the light performs computation as it moves through a medium. Our proposed device is a 2D waveguide whose refractive index can be programmed as a function of space, and which has an induced second-order nonlinear susceptibility that is also programmed as a function of space. Both properties are programmed via the same physical mechanism: a photoconductor that allows the voltage over the waveguide to be modified by impinging control light from above. Our objectives concern demonstrating a functioning device that exhibits the desired programmability, and using the device to perform analog computation.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 20, 2023
- Source ID
- FA95502210378
Entities
People
- Peter L. McMahon
Organizations
- Air Force Office of Scientific Research
- Cornell University
- United States Air Force