Você está na página 1de 8

Methods for Optical Computing

Shannon Ghorbani Physics 325

Our current and ever increasing need for computing power has recently been pushing the limits of electrical components and current engineering designs. Since the inception of the laser in the latter half of the 20th century, fiber optics has utilized the advantages and efficiency of light to transmit large volumes of data at extremely high speeds. With modern design concepts, new and old ideas are being explored to utilize light in computing even more so than before. Research is ongoing and progressing to create all-optical processors as well as increasing data transmission capabilities. Perfecting optical computing is the next feasible step in increasing our processing power and data transmission rates. There are two main approaches that fall under the general term of optical computing; using signal processors for an all-optical analyzing processor, and using optical interconnects between processors and other analyzing components to transmit data quickly and efficiently. Expanding our abilities using known physical processes and popular, historical experiments as a guide has brought optical computing closer to a reality within these two approaches. An ambitious method of creating an all-optical processor involves the use of logic gates. Logic gates are currently used in series and parallel combinations in electrical circuits with todays transistors found in microprocessors. Logic gates are usually used to give Boolean operation outputs (AND, OR, NOT, XOR, and others are the most commonly used). Optical circuits using logic gates are designed to output these signals based on circuit setup and input signals. A method of creating an optical logic gate uses well-known interferometer techniques to discern various interference patterns which then allow the optical circuit to output the different operators.

One basic setup involves splitting an incoming beam of light (visible spectrum if desired) sent through fiber optic cable, and altering one of the beams path lengths. This will alter the phase difference between the two beams. As Farahi explains,1 when the two beams come together in a beam splitter on their way to two separate detectors, the interference pattern will be adjusted depending on the phase difference. General intensity functions that depend on sinusoidal phase differences at each detector can be expressed as
I A = I 0 1 + cos
b b c

I B = I 0 1 @ cos

Here, = 2L / , where L is the path length difference between the two light beams. This also allows calibration to occur and to discern the beam splitter and detectors positions as we can alter the phase between the two beams to be an even multiple of , so that the interference will cancel out transmitted light at one of the two detectors. This allows the assigning of matrices with simple binary elements of whether
1

light is present, a 1, or no light present, a 0.2 Using this setup, the detector is split up into different quadrants, where each quadrant is represented in each matrix by a single element. Another setup given by Farahi involving interference patterns utilizes partially opaque screens that limit the incoming light. By setting up the screens in series or parallel configurations, interference patterns will show up on certain quadrants of the detector. This quadrant configuration is identical to the setup used above. These
1

different setups allow using the same matrix transformations to allow the Boolean operators to be assigned to different intensity matrices, rather than just based on the setup. The signal processing using matrix transformations across different setups allow

the operation just like an electrical logic gate. Some examples of matrix assignments are as follows,
D

0 1 a AND 00
E

1 1 a OR 01
E

A closely related method of using interferometer techniques to create optical logic gates involves the use of interference patterns resulting from coherent light beams passing through multiple slits plates. This slit setup given by Wang involves detecting
2

intensities from the well-known, multi-slit intensity function,

I = I 0 sinc
` a b

f c sin N g2 fffff fffff fffff fffff

sin

and translating into matrix notation involving ones and zeros. These logic gates utilize simple, well-known optical effects to allow signal processing. The unfortunate part of using these optical gates is the combination of gates (in series or parallel) on a large quantity scale. Current electrical microprocessors have transistors by the millions on chips on the scale of just a few square centimeters. To have the density and efficiency of so many gates, these optical setups are just not small enough yet. It also requires that phase and path length differences are also very tiny. Visible wavelengths may not be ideal for such differences needed on this scale. A concern among using the slit setup is the diffraction effects from multiple slits, which could make reading intensity values harder with a large number of slits.
2

Another, more practical evolution to current computing is the use of optical interconnects. Optical interconnects speed up and increase the volume of data that is transmitted between various signal processors. Although the processors themselves might be electrical, efficient and fast transmission of data via current electrical circuits is

slow (~200Mhz) compared to ideal optical transmission methods of much higher bandwidth.3 Existing optical data transmission methods utilize fiber optics between computers and other processing components. This has allowed the increased data capacity and transmission speed of information. However, fiber optics of today usually send information in one-directional format, so an upstream/downstream cabling structure is needed to effectively transmit data back and forth. Modern laser optics and optically transparent materials allow the transmission of much more information bi-directionally (and possibly even multi-directional paths). One setup that has already been partially tested is the use of a multi-layered optical waveguide. Known as a slab waveguide,4 the construction is similar to silicon wafer boards used in modern electronics. Here, the construction would be from different optically transparent materials of different refraction indices layered upon each other. The different layers allow different sources of light to be transmitted with different wavelengths, so different detectors can be attached at each end to receive the different signals. The most important part of slab waveguides is the ability to send a larger quantity of signals (or pulses) through the same layer of the waveguide without interference. As the concept is explained by Feldman, by utilizing modes of wave transmission, and
4

geometrical layouts of the sources and detectors, multiple channels or signals of information can be sent via a single layer of waveguide. The number of modes given by Feldman of an optical, rectangular waveguide is given by

wwwwwc wwwww wwwww wwwww wwwww wwwww wwwww 2fff b 2 d f fr ff f m = n @1

So as the wavelength(s) used decreases and the thickness of the waveguide d increases, so do the number of modes. If light is sent in the plane of the waveguide (i.e. propagation direction of wave is
4

parallel to waveguide), then the signal will output at some angle (either negative or positive to initial angle depending on wavelength relative to length of waveguide) that can be estimated to some approximation using Snells law. The advantage of this
4

transmission setup allows different sources to use the same waveguide and emit signals over different input angles along the polar axis to light propagation. Assuming the waveguide is somewhat larger on the same order in height compared to source separation, you should be able to input several different propagating waves and be able to discern the signal with clarity on the other end. The same can be done with out-of-plane of propagation waves. These waves are
4

sent initially perpendicular to plane of propagation, but the refractive nature of the waveguide sends the signal to the other end, allowing placement of multiple sources and detectors to be placed horizontally to waveguide width. Another advantage is the use of phase differences or time delays in the signal pulses.5 This allows the same source to alternate between channels and still send information at a very fast speed. In order to utilize the layered effect of the slab, so that the different refractive layers can effectively utilize the geometrical wave propagation properties described above, Feldman describes how partially transparent mirrors can be used, collimating lenses, as well as sources of varying wavelengths. The mirrors can filter or redirect light of certain wavelengths (e.g. blazed diffraction gratings) to more densely packed detectors on the detector side. Lenses can take largely spaced sources or detectors and allow the

use of smaller scale waveguides. This allows each layer to either send or receive large number of multiple signals and have each side have a finite number of sources and detectors of nearly any desired size scale. Since these the waveguides are layered and made of optical transmitting material of finite dimensions, we have to be mindful of certain limitations. Foremost is loss of signal or signal intensity. Even with the ability to send multiple signal pulses through a waveguide, if the signal is degraded or altered enough, detectors may not be able to differentiate signals amongst others. The refractive nature of all optical material leads to dispersive light waves of varying frequencies that will reduce optical intensities as light may leak from boundaries of the physical waveguides. So choosing frequencies that have maximum transmittance within different refractive indices is a limiting factor on designing light sources. Being in the vicinity of electrical components that radiate heat could also have an affect on the waveguide material than if the material was thermally isolated. Just as frequency dispersion ranges depend on refractive index (and conversely), temperature gradients may alter the indices of some materials thus forcing the source light to alter its generating frequency for optimal transmission. Another shortcoming is the ability to efficiently convert optical signals to electrical signals without noticeable signal loss. This is currently one of the major concerns, as too much loss is hampering the development and implementation of optical interconnects. However, some experiments have shown that even with a very low estimate of 10% efficiency between source and detector of optical waveguides, it is still
4

possible to transmit thousands of modes of signals or channels of information and still detect the individual signals compared to current fiber optic technologies. These two main areas of research in optical computing offer different paths to increasing our processor and analyzing abilities. Using logic gates to fabricate all-optical processors may be in its infancy with the research shown here, the motivation to move to faster processors still drives the idea. The other method of waveguide optical interconnects is proving to become a good step for feasible advances in data transmission capabilities, which could drastically bring the total combined speed of processing and transmission much higher than current speeds. Rapid inclusions of these optical processes in modern computing will hopefully occur within a short time frame, so that we all can benefit more from lights advantages.

References
1 2 3 4

F. Farahi and T.W. Mayes, Am. J. Phys. 62 (5), 467-469 (1994). L.A. Wang, S.H. Chang, and Y.F. Lin, Opt. Eng. 37 (3) 1011-1018 (1998). Argawal et al, IEEE J. Select. Topics Quantum Electron. 9 (2) 410-418 (2003). M. Feldman, R. Vaidyanathan, and A. El-Amawy, IPPS/SPDP Workshops, High Speed, High Capacity A. Louri and A.K. Kodi, Appl. Opt. 42 (17) 3407-3417 (2003).

Bused Interconnects Using Optical Slab Waveguides, (1999), 924-937.


5

Você também pode gostar