Monolithic GaAs VLSI optoelectronic neuron arrays

This thesis focuses on the design and fabrication of GaAs monolithic optoelectronic integrated circuits (OEIC's) for use in optical neural networks. The basic circuit in a neuron array consists of GaAs MESFET (Metal-Semiconductor Field Effect Transistor) circuits and optical input/output (I/O)...

Full description

Bibliographic Details
Main Author: Luo, Jiafu
Format: Others
Language:en
Published: 1996
Online Access:https://thesis.library.caltech.edu/1560/1/Luo_J_1996.pdf
Luo, Jiafu (1996) Monolithic GaAs VLSI optoelectronic neuron arrays. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/F7DG-YM82. https://resolver.caltech.edu/CaltechETD:etd-05012006-124513 <https://resolver.caltech.edu/CaltechETD:etd-05012006-124513>
id ndltd-CALTECH-oai-thesis.library.caltech.edu-1560
record_format oai_dc
spelling ndltd-CALTECH-oai-thesis.library.caltech.edu-15602021-03-18T05:01:43Z https://thesis.library.caltech.edu/1560/ Monolithic GaAs VLSI optoelectronic neuron arrays Luo, Jiafu This thesis focuses on the design and fabrication of GaAs monolithic optoelectronic integrated circuits (OEIC's) for use in optical neural networks. The basic circuit in a neuron array consists of GaAs MESFET (Metal-Semiconductor Field Effect Transistor) circuits and optical input/output (I/O) devices. By implementing the I/O process optically, we can greatly increase the neuron density for a 2-dimensional array and thus achieve highly parallel computation. Because of the high loss involved in optical interconnections, high density neuron arrays require high gain photodetectors and high efficiency output devices. With responsivities up to 10(4)A/W and structure compatibility with MESFET circuits, optical FET detectors (OPFET's) are an excellent choice as photodetectors. Several techniques have been investigated in order to fabricate high efficiency LED's (light-emitting-diodes) at low current levels. Low power consumption neurons based on OPFET's and GaAs/A1GaAs double-Zn-diffusion double-heterojunction LED's are fabricated using in-house facilities. Industrial foundries provide the most convenient answer to the challenge of fabricating high density 2-D neuron arrays. Two approaches will be described. The first approach utilizes the FET-SEED (self-electrooptic effect device) process from AT&amp;T Bell labs. It provides monolithically integrated circuits with optical I/O devices and depletion mode FET's. In the other approach, GaAs/AlGaAs multiple quantum well modulators are grown on MOSIS GaAs MESFET circuits by MBE regrowth. It is found that in both approaches, the FET's can be used as high gain photodetectors even though the mechanisms are different, thus making it possible to achieve low power consumption high density neuron arrays. Various kinds of complex optoelectronic circuits can be fabricated through these two approaches. 1996 Thesis NonPeerReviewed application/pdf en other https://thesis.library.caltech.edu/1560/1/Luo_J_1996.pdf Luo, Jiafu (1996) Monolithic GaAs VLSI optoelectronic neuron arrays. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/F7DG-YM82. https://resolver.caltech.edu/CaltechETD:etd-05012006-124513 <https://resolver.caltech.edu/CaltechETD:etd-05012006-124513> https://resolver.caltech.edu/CaltechETD:etd-05012006-124513 CaltechETD:etd-05012006-124513 10.7907/F7DG-YM82
collection NDLTD
language en
format Others
sources NDLTD
description This thesis focuses on the design and fabrication of GaAs monolithic optoelectronic integrated circuits (OEIC's) for use in optical neural networks. The basic circuit in a neuron array consists of GaAs MESFET (Metal-Semiconductor Field Effect Transistor) circuits and optical input/output (I/O) devices. By implementing the I/O process optically, we can greatly increase the neuron density for a 2-dimensional array and thus achieve highly parallel computation. Because of the high loss involved in optical interconnections, high density neuron arrays require high gain photodetectors and high efficiency output devices. With responsivities up to 10(4)A/W and structure compatibility with MESFET circuits, optical FET detectors (OPFET's) are an excellent choice as photodetectors. Several techniques have been investigated in order to fabricate high efficiency LED's (light-emitting-diodes) at low current levels. Low power consumption neurons based on OPFET's and GaAs/A1GaAs double-Zn-diffusion double-heterojunction LED's are fabricated using in-house facilities. Industrial foundries provide the most convenient answer to the challenge of fabricating high density 2-D neuron arrays. Two approaches will be described. The first approach utilizes the FET-SEED (self-electrooptic effect device) process from AT&amp;T Bell labs. It provides monolithically integrated circuits with optical I/O devices and depletion mode FET's. In the other approach, GaAs/AlGaAs multiple quantum well modulators are grown on MOSIS GaAs MESFET circuits by MBE regrowth. It is found that in both approaches, the FET's can be used as high gain photodetectors even though the mechanisms are different, thus making it possible to achieve low power consumption high density neuron arrays. Various kinds of complex optoelectronic circuits can be fabricated through these two approaches.
author Luo, Jiafu
spellingShingle Luo, Jiafu
Monolithic GaAs VLSI optoelectronic neuron arrays
author_facet Luo, Jiafu
author_sort Luo, Jiafu
title Monolithic GaAs VLSI optoelectronic neuron arrays
title_short Monolithic GaAs VLSI optoelectronic neuron arrays
title_full Monolithic GaAs VLSI optoelectronic neuron arrays
title_fullStr Monolithic GaAs VLSI optoelectronic neuron arrays
title_full_unstemmed Monolithic GaAs VLSI optoelectronic neuron arrays
title_sort monolithic gaas vlsi optoelectronic neuron arrays
publishDate 1996
url https://thesis.library.caltech.edu/1560/1/Luo_J_1996.pdf
Luo, Jiafu (1996) Monolithic GaAs VLSI optoelectronic neuron arrays. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/F7DG-YM82. https://resolver.caltech.edu/CaltechETD:etd-05012006-124513 <https://resolver.caltech.edu/CaltechETD:etd-05012006-124513>
work_keys_str_mv AT luojiafu monolithicgaasvlsioptoelectronicneuronarrays
_version_ 1719383851721031680