Sandia Labs FY22 Laboratory Directed Research & Development Annual Report

NEUROMORPHIC INFORMATION PROCESSING BY OPTICAL MEDIA OFFERS ENERGY EFFICIENCY AND INCREASED SPEED.

Identifying features in a scene from emitted or reflected light is at the core of many applications such as remote sensing, autonomous driving, and biological imaging. Usually, this is accomplished by converting the incoming photons to electrons using a camera, and then analyzing the digital images using algorithms running on computers. This approach requires significant energy and time. In this LDRD, Sandia explored a radically different approach where a passive material can

directly process the incoming light and act as a neuromorphic classifier at the speed of light. This demonstrates that such systems can be as good as electronic neural networks, but with much lower energy consumption and faster speeds. While this work focused on optical fields, the approach can be generalized to other fields (e.g., temperature, strain), opening a rich area of research that requires contributions from many scientific disciplines. This work resulted in two key

publications published in Optics Express and ACS Photonics . The PI facilitated collaborations with Sandia Securing Top Academic Research Talent at Historically Black Colleges and Universitites partner Norfolk State University, where five students supported work focused on optical learning. (PI: Francois Leonard)

Side by side comparison of electronic (left) and optical (right) neuromorphic recognition processes.

“GAME CHANGER” FOR OPTICAL DIAGNOSTICS USED IN BIOLOGY AND CHEMISTRY.

Experimental setup for massively entangled multimode squeezed light quantum imaging.

81

LABORATORY DIRECTED RESEARCH & DEVELOPMENT