GeCSn Lasers for Silicon-Based Processors, IR Photonics, and Machine Learning
Abstract
This project seeks to grow and fabricate GeCSn lasers that are compatible with conventional electronic chip fabrication and operate at the high temperatures of modern processors. Direct bandgap devices built with these materials would enable new silicon photonics such as extreme bandwidth data communications within and between million-core processors, real time machine learning using optical accelerators, and mass produced sensors for swarms. GeCSn offers a new route to strong and efficient lasing even at high temperatures due to its strongly direct bandgap, strong confinement in deep quantum wells, and separate confinement heterostructures to maximize optical gain. We recently developed growth techniques that eliminate unwanted carbon-carbon bonds and other defects that previously hindered dilute Ge carbides. The gas precursors we are developing are expected to be transferrable to industry fabrication tools. The anticipated outcomes of this project are to demonstrate lasers grown on silicon that operate in continuous wave at 100 C under electrical power, for nanofabricated photonic integrated circuits. The techniques developed here would be suitable for direct growth at the post-metal stage of CMOS fabrication, allowing thousands of lasers to be integrated on chip, while avoiding common alignment and material problems that frequently limit yield. Chips built with these lasers would avoid the memory bottleneck where many processors wait for access to shared data. On-chip lasers could also enable tensor computations such as multiply-and-accumulate to be performed in a single clock cycle, reducing the size, weight, and power (SWaP) of artificial intelligence architectures by an order of magnitude, for real time decision making and low-cost autonomous swarms. They would offer near-instantaneous image processing and recognition for high frame rate hyperspectral and LiDAR target discrimination and tracking.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 06, 2024
- Source ID
- FA95502310458
Entities
People
- Mark Wistey
Organizations
- Air Force Office of Scientific Research
- Texas State University
- United States Air Force