Optical Computing

F.A.P Tooley, B.S Wherrett, F.A.P Tooley, B.S Wherrett

• Publisher: CRC Press
• Year: 2020
• ISBN:
  9781000156942
• Edition: 1
• Topic: Computer Science
• Subtopic: Optical Data Processing
• Index: Computer Science

Written by ten leading experts in the field, Optical Computing cover topics such as optical bistability, optical interconnects and circuits, photorefractive devices, spatial light modulators, associative memory, and optical computer architectures.

INTRODUCTION

Brian S. Wherrett

Department of Physics
Heriot-Watt University
Edinburgh, Scotland, UK

The uses and prospects for optics in information processing are now wide ranging; from spatial filtering of images by attenuating specific Fourier components through to proposals for general purpose optical computers. A definition and overview of the subject is given by J.W. Goodman in Chapter 2 of this edited text. The purpose of this introduction is to set the various contributions of these Summer School Proceedings in context.

Optical computing, as a research field, can be thought of as starting in 1964 with the first publication of a collection of papers, presented at the Symposium on Optical and Electrooptical Information Processing [1]. It is no coincidence that this meeting was held just four years after the invention of the laser, with the dramatic increase in information processing ability afforded by coherent, intense, narrow bandwidth radiation. Since that time one can point to at least 19 conference proceedings and journal special issues devoted to research publications in the field [2].

It is less than easy to draw from this collation of papers the threads of the subject. Individual topics appear over a few meetings and then disappear from the research literature, often without apparent influence on other topics. The absence until recently of a textbook covering the subject as a whole compounds the confusion of researchers entering the field [3]. It is therefore hoped that this tutorial level text will help establish the structure and bounds of the subject.

A useful division of the field of optical information processing is into:

(i) Techniques, (ii) Algorithms, (iii) Architectures and (iv) Applications. This division immediately emphasises the interdisciplinary nature of the subject. Techniques are traditionally the realm of physicists and electronics engineers, involving optical phenomena, the interaction mechanisms of light and matter, and the construction of devices that exploit these mechanisms in a controllable and efficient manner. The fabrication processes and the materials advancement involve also the expertise of materials scientists and chemists. Algorithm development is essentially a mathematics topic, it covers the representation of numeric or physical data and the arithmetic or logic-based methods by which data can be manipulated to solve a given problem. The computer scientist is concerned with the Architecture by which the components developed under the techniques heading are put together in a system that efficiently implements the algorithms on the data presented to it. Finally the construction of any processing system must be aimed towards one or more Applications. Therefore the existing and perceived needs of research, consumer or military markets must be borne in mind, also the achievements and potential of alternative information processing systems.

A textbook on optical computing could usefully be subdivided as described above. An attempt has been made to structure this proceedings along these lines. It must be borne in mind, however, that the invited experts are each involved with more than one aspect of the subject, often being concerned both with the fundamental science and optimisation of particular devices and with the use of these devices in optical processing circuitry. I believe that it is fair to say however that the research described in Chapters 3, 10 is predominantly ‘techniques-led’, that of Chapters 11, 14 concerns mainly algorithms and architectures, whilst the final two chapters present the relevant achievements of electronic computing. Applications of optical computing are addressed variously throughout.

Any information processing system requires (i) optical sources, (ii) a method for modulating the sources in order to carry the information, (ii) an information processing stage in which decisions are made and the information may be altered, (iv) optical communication & interconnection, (v) detection and (vi) display and/or storage. A techniques section of a definitive optical computing text could contain chapters on each of these components. There are of course entire texts devoted to laser sources, optical fibre communication, detectors, displays and optical storage [4]. These topics are still subject to advancement but are not at the early stages of development that properly form the subject of a Summer School Proceedings. We concentrate here on techniques for Information Processing and for Data Interconnection either between the processing elements themselves or between the elements and the input/output stages. Chapters 3, 7 are concerned primarly, but by no means exclusively, with processing techniques, in Chapters 8, 10 optical information communication is foremost.

At the fundamental level the striking difference between optics and electronics is that there is effectively no interaction between photons travelling in vacuum whereas the Coulombic interaction between electrons is very strong. Consequently optical communication without crosstalk of streams of data and 1-D or 2-D images is easy, whereas electronic communications are restricted in zero or 1-D and impossible in highly parallel 2-D format. It is the combination of massive parallelism and cross-talk and interference free interconnection that gives optics its biggest advantage over electronics. These Proceedings concentrate on areas where this advantage is likely to have its greatest potential, namely in the manipulation of two-dimensional images.

Given the absence of direct photon-photon interactions it is essential that indirect interactions are employed if optical image processing is to be achieved, it is then a matter of the degree to which material excitation (charge displacement) is involved. One option is to detect the light, the electronic signals are processed electronically, and the resulting information interfaced back to optics via a modulator. In this way the interconnect freedom of optics can be used in combination with the processing power of electronic circuitry; limitations, at present lie with the interfacing input and output. Viewed as a black box the above device is simply a transmitter (or reflector) with a nonlinear optical transfer function, the function itself being controlled by the address to the electronic circuitry. There are several alternative methods to catalyse the interaction of photons with photons.

In Chapter 3 optically addressed spatial light modulators are described; here for example an optical field incident on one medium generates a static electric field that in turn is used to alter the refractive index of a second medium and hence the transfer properties of the electro-optic medium to a second optical field. In Chapters 4 and 7, bistable devices that rely on either the photo-generation of carriers in semiconductors or the subsequent material heating, followed by a refractive index change (of the same material), are described. Similar devices in which the absorption rather than the refraction dominates the transfer nonlinearity are discussed in Chapter 5; static electric fields induced optically are used to produce relatively large absorption changes – the self-electroabsorptive effect. None of the above devices relies on coherent input light. Therefore, if sensitive enough they can be used as incoherent-to-coherent converters. This property is exploited in the use of spatial light modulators in Fourier optics, Chapter 3. The photorefractive devices described in Chapter 7 also rely on the generation of static fields. In this case the radiation absorption generates carriers that diffuse to neighbouring trap sites. The resulting space-charge fields create refractive index changes via the electrooptic effect and hence again produce a nonlinear transfer.

In the 1980’s one of the most active areas of optical computing has been the development of the above techniques for the optical control of optical information and of the exploitation of relatively crude devices in prototype optical processing circuits. Such circuits are described in Chapter 6. In each case the processing device is an optically addressed 2-D spatial light modulator. Other components in processing circuits are of course also 2-D light modulators, but with quite different modes of address. Thus a single lens is a 2-D phase modulator, the address being the mechanically controlled fabrication of the thickness variation. A hologram has structurally controlled index modulation, a computer generated hologram has a controlled attenuation modulation. These fixed, rather than real-time variable, devices are key to many optical interconnection schemes; example uses are presented in Chapter 8.

Real-time modulators that are used for the transfer of information between one discipline and another, or from one dimensionality to another, include acousto-optic Bragg cell devices, electrically addressed 2-D spatial light modulators – index changes brought about by lattice vibrations or electric fields are again the mechanism by which the light-matter interaction is controlled. The former are mentioned in Chapter 2; the techniques of acousto-optic information processing are now well established (although not widely applied), details can be found in a number of monograph references [5]. Externally addressed 2-D light modulators are discussed in Chapter 3.

From the optics viewpoint the algorithms of data processing can be categorised by the amount of fan-out demanded of each signal. Thus in an electronic processing scheme it might be necessary only to achieve optical Data Reordering, the output from one device being fed to only one detector. The ‘perfect shuffle’ for example is the basis for fast sorting algorithms and for the fast discrete Fourier transform. Clock-distribution (Chapter 10), however, would demand one-to-many fan-out, with likely restrictions on uniformity and registration of the generated beam arrays.

For processing itself the output of any one active device must usually be fed to two or more further devices, and equally each device must be able to accept at least two information signals. If the fan-in/out is small, less than 1-to-10 say, one is usually concerned with Digital Optics. Here the values of ‘zero’ and ‘one’ levels representing binary numeric or image data are to be monitored accurately throughout calculations. The number representation, algorithm and architectures for parallel processing of digital optics are presented particularly in Chapters 11, 13. There is presently a clear division between proposed cellular 2-D schemes for digital optics processing and schemes with multiple fan-out/in presently proposed for optical associative memory and neural networks. The latter are discussed in Chapter 14. With fanning of perhaps 1-to-100 up to 1-to-10⁴ it is not possible to maintain the accuracy of digital arithmetic; one is concerned with decision making by thresholding the combined input to a given processor at some approximate level. I term this approach Threshold Optics. Finally the global fan-in/out achieved by lens and diffraction optics, is the basis for the established field of Fourier Optics [6] and is represented herein in the use of photorefractive devices for real-time holography and reconfigurable optical interconnects (Chapters 7 and 10).

The final two Chapters (15 and 16) are included in order to place the achievements of optics in perspective and to indicate those areas of ‘computing’ in which optics is likely to best complement (rather than supercede) electronic machines. These Chapters cover briefly the hardware developments of electronics, parallel architectures, and the use of parallel architectures for specific applications including neural networks.

There is no way that the Proceedings of a single, two-week Summer School could hope to encompass the full range of topics under the general heading of Optical Computing. It is hoped that those topics selected will give the readers a feel for the subject as a whole however as well as details of some of the most active areas of development that are currently being pursued.

References

1. Proc. Symp. on Optical & Electrooptical Information Processing, Boston 1964, Ed. J. Tippett et al, MIT Press 1968.

2. cf. the following Conference Proceedings & Journal Special Issues. IEEE Trans. Comp. C-24, April 1975. Special Issue on Optical Computing. Proc. IEEE 65, Jan. 1977. Special Issue on Optical Computing. SPIE 231/2, Proc. Int. Optical Computing Conf., April 1980. SPIE 388, Adv. in Optical Information Processing, Jan. 1983. ‘ SPIE 422, Proc. 10th Int. Optical Computing Conf., April 1983. Opt. Eng., 23 Jan/Feb 1984. Special Issue on Optical Computing. SPIE 456, Optical Computing, Jan. 1984. Proc. IEEE 72, July 1984. Special Issue on Optical Computing. Technical Digest of the 13th congress of the International Commission for Optics, Sapporo, Aug. 1984. Opt. Eng. 24, Jan/Feb 1985. Special Issue on Optical Computing & Optical Information Processing Components. Technical Digest of OSA topical meeting on Optical Computing, Incline Village, Nevada, March 1985. Opt. Eng. 25, Jan. 1986. Special Issue On Digital Optical Computing. SPIE 625, Optical Computing, Jan. 1986 SPIE 634, Institute for Advanced Optical Technologies, Optical & Hybrid Computing, March 1986. Applied Optics 25, May 1986. Special Issue on Optical Computing. Opt. Eng. 26, Jan. 1987, Special Issue on Optical Computing & Nonlinear Optical Signal Processing. Technical Digest of OSA topical meeting on Optical Computing, Incline Village, Nevada, March 1987. Opt. Eng. 26, March 1987. Special Issue on Optical Information Processing. ICO Meeting on Optical Computing, Toulon, Aug. 1988, to be published as SPIE 963.

3. “Optical Computing, A survey for Computer Scientists”, D. G. Feitelson, MIT Press 1988.

4. Example text in areas relevant to Optical Computing. Lasers: physics, systems & techniques. Proc. 23rd Scottish Universities Summer School in Physics. Ed. W.J. Firth & R.G. Harrison, SUSSP Edinburgh, 1983. Optical Fibre Communications. T. Senior, Prentice-Hall International Ltd., London, 1985. Detection of Optical & Infrared Radiation. R.H. Kingston Springer Series in Optical Sciences. Springer-Verlag NY, 1978. Principles of Optical Disc Systems. Ed. G. Bouwhuis et al. Adam Hilger Ltd., Bristol UK, 1985.

5. Acousto-optic Signal Processing. Ed. N.J. Berg & J.N. Lee, Marcel Dekker Inc. NY. Proc. IEEE 69, Jan. 1981. Special Issue on Acousto-Optic Processing.

6....