Novel optical devices for information processing

Abstract

Optics has the inherent advantages of parallelism and wide bandwidths in processing information. However, the need to interface with electronics creates a bottleneck that eliminates many of these advantages. The proposed research explores novel optical devices and techniques to overcome some of these bottlenecks. To address parallelism issues we take a specific example of a content-addressable memory that can recognize images. Image recognition is an important task that in principle can be done rapidly using the natural parallelism of optics. However in practice, when presented with incomplete or erroneous information, image recognition often fails to give the correct answer. To address this problem we examine a scheme based on free-space interconnects implemented with diffractive optics. For bandwidth issues, we study possible ways to eliminate the electronic conversion bottleneck by exploring all-optical buffer memories and all-optical processing elements. For buffer memories we examine the specific example of slow light delay lines. Although this is currently a popular research topic, there are fundamental issues of the delay-time-bandwidth product that must be solved before slow light delay lines can find practical applications. For all-optical processing we examine the feasibility of constructing circuit elements that operate directly at optical frequencies to perform simple processing tasks. Here we concentrate on the simplest element, a sub-wavelength optical wire, along with a grating coupler to interface with conventional optical elements such as lenses and fibers. Even such a simple element as a wire has numerous potential applications. In conclusion, information processing by all-optical devices are demonstrated with an associative memory using diffractive optics, an all-optical delay line using room temperature slow light in photorefractive crystals, and a subwavelength optical circuit by surface plasmon effects.