Electrooptic Materials Requirements for Optical Information Processing and Computing
- PDF / 848,551 Bytes
- 12 Pages / 420.48 x 639 pts Page_size
- 105 Downloads / 264 Views
ELECTROOPTIC MATERIALS REQUIREMENTS FOR OPTICAL INFORMATION PROCESSING AND COMPUTING Praveen Asthana, Edward Herbulock, Zaheed Karim, Chris Kyriakakis, Gregory P. Nordin, and Armand R. Tanguay, Jr. Optical Materials and Devices Laboratory, Center for Photonic Technology, and National Center for Integrated Photonic Technology, Departments of Electrical Engineering and Materials Science, University of Southern California, 520 Seaver Science Center, University Park, MC-0483, Los Angeles, CA 90089-0483.
ABSTRACT Photorefractive materials comprise an important category of electrooptic materials, and are important constituent elements in a wide range of devices designed specifically for use in optical information processing and computing systems. Critical issues affecting the development and applicability of photorefractive materials are examined from the perspective of photonic neural network implementations that incorporate photorefractive volume holographic interconnections.
INTRODUCTION Electrooptic materials are those (primarily single crystal) materials that exhibit a change in the local index of refraction in response to an applied or internally developed electric field. Such materials are of considerable importance for a wide range of active optical components, including one- and two-dimensional spatial light modulators, volume holographic optical elements, and threshold arrays, that in turn form the device repertoire for advanced optical information processing and computing systems [1]. Applications for such systems include pattern recognition, target classification, spectrum and texture analysis, machine vision, radar ambiguity function generation, synthetic aperture radar (SAR) image formation, and photonic neural network implementation. In this paper, we examine a number of the materials requirements and desirable characteristics that apply to photosensitive electrooptic (or so-called photorefractive)materials, as utilized, for example, in volume holographic optical elements (VHOEs). Dynamically reprogrammable VHOEs are essential for applications requiring highly multiplexed and densely packed interconnections, such as those envisioned for neural network implementations in which many individual switching devices (neuron units) are interconnected to (in some cases) very large numbers of other neuron units [2). The approach taken herein is to first describe a specific application of photorefractive materials in an optical information processing/computing system in order to illustrate the implied functional and materials requirements. For this purpose, we have chosen to describe a novel approach to the photonic implementation of neural networks. In this context, we describe several critical issues facing the continued development and successful establishment of photorefractive materials and devices as an interconnection technology base.
PHOTONIC IMPLEMENTATIONS OF NEURAL NETWORKS The principle advantage of photonic implementations of neural networks is the potential for multiplexing a very large number of ind
Data Loading...
