Integrated Photonic Technologies for Attojoule Optoelectronics and Neuromorphic Information Processing
Speaker: Prof. Volker J. Sorger
Affiliation: George Washington University
Abstract: Photonic technologies are at the forefront of the ongoing 4th industrial revolution of digitalization supporting applications such as virtual reality, autonomous vehicles, and electronic warfare. The development of integrated photonics in recent years enabled functional devices and circuits through miniaturization. However, fundamental challenges such as the weak light-matter integration have limited silicon and III-V-based devices to millimeter-scale footprints demanding about a million photons-per-bit. Overcoming these challenges, in the first part of this talk I will show how nanoscale photonics together with heterogeneous integration of emerging materials into foundry-based photonic chips leads to the exciting topic of attojoule optoelectronics. Here I will discuss our recent devices demonstrating 100aJ/bit modulators, graphene photodetectors, and exotic epsilon-near-zero modes empowering record-efficient phase shifters for applications in data-com, LiDAR, and photonic neural networks (NN).
With Moore’s law and Dennard scaling now being limited by fundamental physics, optical information processing has reemerged for special-purpose compute systems such as NNs or RF-signal & image filtering. Here unique opportunities exist, for example, given by signal parallelism (e.g. multiplexing) or avoidance of capacitive wire loading, thus opening prospects for distributed non-van Neumann architectures. In the second part of this talk, I will share our latest examples of photonic chip-based processors to include a) a feed-forward fully-connected NN, b) a convolutional NN processor where convolutions are expressed as multiplications in the Fourier domain enabling 1 PMAC/s throughputs at nanosecond-short delays for real-time processing in automated target recognition and ranging, c) a photonic residue arithmetic adder, d) mirror symmetry perception via coincidence detection of spiking NNs, and e) mesh-based reconfigurable photonic computing for solving PDEs non-iteratively.
In summary, integrated (nano)photonics connects the worlds of electronics and optics, thus enabling new classes of a) attojoule optoelectronics, and b) neuromorphic circuits by employing the distinctive properties of light for parallel processing and order O(1) operations.
Biography: Volker J. Sorger is an associate professor in the Department of Electrical and Computer Engineering and the director of the Integrated Nanophotonics lab at the George Washington University. He received his Ph.D. from the University of California Berkeley and MS from UT Austin. His research focuses on integrated photonics and plasmonics, and analog information processing such as programmable photonic circuits and neuromorphic computing. His work was recognized by the Emil Wolf prize from the Optical Society of America, the AFOSR Young Investigator (YIP) award, the Hegarty Innovation Prize, the National Academy of Sciences paper-of-the-year award, the MRS Gold medal, and both the Early Career and Outstanding Research awards at GWU. Prof. Sorger is the editor-in-chief of the Nanophotonics and the OSA division chair for Optoelectronics-and-Photonics. He serves at the boards of OSA and SPIE, and is a senior member of IEEE, OSA & SPIE.
For more information, contact Prof. Aydogan Ozcan (ozcan@ucla.edu)
Date/Time:
Date(s) - Mar 20, 2019
3:00 pm - 4:00 pm
Location:
E-IV Tesla Room #53-125
420 Westwood Plaza - 5th Flr., Los Angeles CA 90095