CaltechTHESIS
  A Caltech Library Service

GaAs optoelectronic integrated circuits for optical neural network applications

Citation

Lin, Steven H. (1992) GaAs optoelectronic integrated circuits for optical neural network applications. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/tqm4-en29. https://resolver.caltech.edu/CaltechETD:etd-08072007-133843

Abstract

Optoelectronic integrated circuits (OEIC's) have emerged as a viable method in the implementation of optical neurons required for a neural network. This is due to the increased capability in both the material and the device engineering in GaAs technology, which has proliferated incredibly fast during the last decade. In this thesis, two different approaches to monolithically integrate various electronic and optical devices are explored for the implementation of optical neurons. The first approach utilizes the technology from double heterojunction bipolar transistor for its potentially high current gain and its structural compatibility with optical devices. In achieving the current gain required for optical neurons, modeling of the base leakage current, effect of surface passivation and diffusion characteristics is performed for Zn-diffused bipolar transistors. The second approach employs metal semiconductor field-effect transistors as the driver for the optical devices. It is found that, by properly designing the circuit, high optical gain, low electrical power dissipation and low optical switching energy thresholding devices can be accomplished in this approach with large input-output isolation. Such performance is required if large arrays of optoelectronic neurons are to be inserted into a neural network to perform tasks that make neural computation a unique approach in solving a certain class of problems. In this thesis, an optical gain of 80 is demonstrated along with an electrical power dissipation of 1.6 mW and an optical switching energy of 10 pJ. These results generate high promises and optimism for the realization of a physical neural computer in the near future.

Item Type:Thesis (Dissertation (Ph.D.))
Degree Grantor:California Institute of Technology
Division:Engineering and Applied Science
Major Option:Electrical Engineering
Thesis Availability:Public (worldwide access)
Research Advisor(s):
  • Psaltis, Demetri
Thesis Committee:
  • Psaltis, Demetri (chair)
  • Yariv, Amnon
  • Rutledge, David B.
  • Nicolet, Marc-Aurele
Defense Date:24 September 1991
Record Number:CaltechETD:etd-08072007-133843
Persistent URL:https://resolver.caltech.edu/CaltechETD:etd-08072007-133843
DOI:10.7907/tqm4-en29
Default Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:3039
Collection:CaltechTHESIS
Deposited By: Imported from ETD-db
Deposited On:14 Aug 2007
Last Modified:19 Apr 2021 22:39

Thesis Files

[img]
Preview
PDF (Lin_sh_1992.pdf) - Final Version
See Usage Policy.

10MB

Repository Staff Only: item control page