The objectives of the EOC working group are to:
-advance the enabling technologies that support 2-D optical processing
-identify applications uniquely suited to optical information processing technology
-develop architectures and algorithms that efficiently exploit the technology and demonstrate its long-term potential.
There are four subgroups, covering:
-II-VI compounds, silicon and liquid-crystal devices
-III-V compound electro-optic devices
-free-space optical interconnect technology
-architectures, systems and algorithms.
The status of optical information technology is reviewed to identify collaboration opportunities and key areas where research is needed. A clear picture of the future direction of optical information processing is emerging and gives rise to a major research proposal.
Progress was reviewed in the 4 main topic areas. Investigations of thermal optical nonlinearities of II to VI semiconductors has led to the construction of a novel temperature sensor, based on the optical bistability of cadmium sulphide. Other cadmium sulphide optical properties are being investigated as a route to directly cascadable devices in the short time (picosecond) regime. Molecular beam deposition work has enabled the preparation of nonlinear interference filters of good uniformity that exhibit high resistance to laser beam damage. The performance of such filters in various optical processing circuits is being assessed.
Assembly of data has continued for a review of Perspectives for Parallel Optical Interconnects, a forthcoming publication which will include state of the art contributions.
Progress has been made on many issues, including optical pipeline processors, fast arithmetic using optical symbolic substitution logic, tolerancing studies, development of the optical cellular logic image processor and an edge detection cellular automation. A demonstration of optical most significant digit addition has also been staged.
Pump and probe beam and luminescence spectroscopy were used to study the nonlinear response of gallium arsenide/(aluminium gallium) arsenide (GaAs/(AlGa)As) heterostructures to quasistationary excitation conditions. The carrier induced energetic shift of the 1hh-exciton as a function of the quantum well width showed a dimensional dependence of the carrier screening properties. This shift gave a good criterion to decide whether a system behaved more like a 2-dimensional or 3-dimensional systems. The high excitation regime was dominated by electron-hole plasma features. Many particle effects led to a renormalization of the fundamental bandgap. This effect was essential for understanding the physics of III-V semiconductor lasers. The carrier density and the reduced bandgap were determined via systematic evaluation of both gain and luminescence spectra. The observed behaviour can be described by a strict 2-dimensional theory using effective exciton parameters in order to account for the finite well widths of the structures. The study of the higher sub-bands revealed that both exciton bleaching and sub-band renormalization were mainly due to direct occupation of the specific sub-band while intersub-band effects were considerably smaller. By coating the 2 sides of a 50 by 100 Angstron multiple quantum well with semitransparent chromium-gold (Cr-Au) electrodes it was possible to control the energetic position of the 1hh-exciton as a function of the applied electric field and of the incoming light power. Several structures to optimize this effect in order to build an electrooptical switch or modulator was outlined.
The way the quantum confined Stark effect for asymmetric step quantum wells depends on specific quantum well parameters (total well width, width ratio, step height) has been studied. The calculations were performed for wells consisting of one layer of gallium arsenide (GaAs) and one layer of aluminium gallium arsenide (AlxGa(1-x)As) and barrier layers of aluminium gallium arsenide (Al0.3Ga0.7As). The energy levels and wave functions were calculated using the transfer matrix method.
The largest field induced shifts of the effective bandgap were obtained for the widest quantum wells. For wells of a total thickness of 75 Angstroms and an aluminium concentration in the potential step of 6%, the largest shifts were obtained for wells where 20% of the width consisted of gallium arsenide. For wells having a 15 Angstrom thick gallium arsenide layer and a 60 Angstrom wide aluminium gallium arsenide (AlGaAs) potential step. The largest energy level shifts were obtained for aluminium concentrations between 12% and 18%.
Barrier reservoir and quantum well electron transfer (BRAQWET) structures were also studied. For BRAQWETs consisting of gallium arsenide reservoir layers, 3 strained indium gallium arsenide wells and aluminium gallium arsenide barrier layers. The product of the voltage to obtain a pi phase change times the length of the waveguide could be 0.25 Vmm, which is 10 times less than has been reported for any other semiconductor structure.
A novel, efficient and fast vertical nin aluminium gallium arsenide/gallium arsenide (AlGaAs/GaAs) multiple quantum well (MQW) microresonator grown by molecular beam epitaxy (MBE) has been demonstrated. The electrical current of this hybrid device is kept low by using undoped aluminium gallium arsenide (AlGaAs) barrier layers cladding the active MQW material. For high speed operation, a novel travelling wave modulator with a special coplanar metallic waveguide structure has been proposed. Electrooptical modulation was studied using a titanium:sapphire (Ti:sapphire) laser pumped by an argon ion laser.
Nonlinear Fabry-Perot (NLFP) devices have been fabricated which operate at a threshold power 1 mW with a contrast ratio of about 10:1 and less than 1 dB attenuation. The absorbed switching energy is about 4 pJ in the NLFP. Due to critical slowing down this figure is independent of the transition time, which can vary from a few ns to us, depending on the magnitude of the switching signal used. The threshold power, together with the dissipated switching energy, corresponds to an estimated GBWP of 100 MHz. The suitability of these gallium arsenide (GaAs) etalons as integrating/threshold units has been investigated theoretically. Due to the wavelength selectivity of the resonant structure, the device should be used with stabilized laser systems, until adequate frequency stability and tunability is achieved with semiconductor lasers. These devices have minimal spatial fan-in capacity. The trend in NLFP devices is towards shorter, higher finesse cavities, to reduce the switching intensity. However, because of the limited optical confinement provided by the integrated dielectric mirrors, diffraction losses become increasingly important for shorter cavities. Calculations for devices based on aluminium gallium arsenide (AlGaAs) material indicate that the switching energy tends to saturate at an asymptotic value of about 0.1 pJ for finesse values larger than Fc about 100. To eliminate diffraction losses and to improve the switching energy, it is necessary to use waveguiding. Such devices will accept a single spatial mode, and their temporal fan-in will decrease proportionally with finesse, beyond Fe. At present, arrays containing about 1000 devices on an area of less than 1 square millimetre are conceivable.
In synthetic diffractive optics, the sensitivity of the resonance domain designs to the fabrication errors can be estimated by introducing random perturbations to the optimized parameters. The deviations from the optimized groove structure increase the reconstruction error. Numerical simulations show that the inaccuracy of the transition points should not exceed one twentieth of the wavelength for satisfactory performance. Therefore, the present day microlithographic fabrication methods restrict the demonstrations to infrared wavelengths.
Two reflection type resonance domain fan-out gratings were fabricated for wavelength 10.6 um. The fabrication method guarantees almost perfect reflectivity at this wavelength. The reconstruction errors of the fabricated beamsplitters with fan-out to six and seven were 11% and 15%, compared to the design values 2.3% and 3.5%, agreeing roughly with the numerical analysis of fabrication errors, assuming that the transverse fabrication accuracy is 0.2 um. For normal incidence (and odd fan-out), the zeroth order is particularly sensitive to the fabrication errors of the groove depth. The relief depth error for the one to seven fan-out grating was about 1%, and the zeroth order was consequently weaker than the others. Excluding the zeroth order, a reconstruction error of 8% was measured for the remaining orders.
A technique for the reduction of the peak intensity of noise which arises from Fourier plane array generators has been developed. This is achieved by designing the array generator as a quasiperiodic computer generated hologram. In this way an even distribution of noise is obtained while the required diffraction orders are left as tightly sampled spots. Using the method of generalized error diffusion, several different quasiperiodic 4 x 4 array generators were designed. A 14.5 dB reduction in peak noise was obtained for a hologram with 8 x 8 quasiperiods. Diffraction efficiency and array uniformity were not adversely affected by this technique.
A microlens fabrication process has been developed, based on the fact that an irradiation of polymethyl methacrylate (PMMA) with a high energy ion beam reduces the molecular weight by splitting the polymer chains and thus changes the diffusion properties of the material. Thus there are 2 processing steps, an irradiation and a subsequent diffusion.
By irradiation of PMMA with a high energy proton beam through a structured metal mask, domains with reduced molecular weight are produced. For the fabrication of microlenses metal masks with circular apertures have to be used.
After irradiation the structured substrate is placed into an atmosphere of monomer vapour. The diffusion of monomer vapour causes the volume of the irradiated domains to expand. By surface tension these volumes form lenslike shapes.
3-dimensional microoptic systems can be fabricated by a thermal imprinting processes in polymethyl methacrylate (PMMA). Each microoptic component is constructed by thermally imprinting a metal master into the substrate. This process, though simple, provides high quality components and allows mass production.
The possibility of free design of the components and their locations opens a wide range of optical applications. In particular, optical interconnections and microoptic elements for digital optical data processing systems can be made cost efficiently. Due to the refractive or reflective nature of the imprints, the angular sensitivity and the spectral sensitivity for transmitted or reflected light are significantly smaller as compared to diffractive interconnections.
A mass production of imprinted structures is also possible by electroforming. To this end, the substrates can be coated with silver (Ag) or nickel (Ni) by chemical vapour deposition (CVD), resulting in a conducting surface. The thickness and stability of the metal layers is increased by electroforming of nickel. The PMMA substrates are then dissolved, eg by acetone. Consequently, a negative copy of the imprinted component arrangement can be achieved. This copy can be used later for thermally imprinting the whole scheme is one step. A reactive of thermal injection moulding may cause problems due to the shrinking during polymerization. With deep thermal imprinting, however, the distances can be maintained sufficiently.
A hybrid processor architecture for an associative memory with complex images has been developed, using double correlator architecture, where a reflective hologram on a BGO crystal gives the desired squared correlation which is back convolved with the memory images recorded on the first hologram.
First experiments utilized invariant filters as fixed masks with a video input. A grating with the same spatial carrier as the filter was placed in front of the liquid crystal light valve (LCLV). Using a joint transform filter set up, image recording was demonstrated, but there were difficulties in aligning the LCLV and the photographic filter. With the matched filter configuration, implementation was easier, and the initial poor signal to noise ratio requires that the complex filters also be recorded on the LCLV.
Holographic and diffractive optical elements have been finding widespread use in parallel optical computing demonstration experiments. Three basic functions can be identified:
high level fan-out (one to hundreds) of a single beam to provide the multiple inputs required to power logic planes containing arrays of optical logic elements;
low level fan-out (one to a few) to provide regular interconnection patterns between elements within processing arrays (eg between nearest neighbours);
irregular (space variant) interconnects across processing arrays.
Holographic/diffractive techniques can satisfy a wide range of interconnect requirements within experimental optical computing systems. This remains a rapidly developing research area, with many new advances in sight, particularly in the context of surface relief diffractive optical elements. It can be anticipated that the optical processing systems of the future will rely on holographic fan-out/fan-in components of this sort, closely integrated into the compact optomechanical assemblies that are also currently under development.
Since the recognition of the importance of optical array illuminators in the construction of parallel digital optical processors, a wide variety of optical components have been designed and demonstrated that convert a single laser beam into a regularly spaced array of M x N equal intensity light spots. Space variant array illuminators, which form the spot array in a Fresnel plane or an image plane of the aperture, are in general easy to design and fabricate, but they require uniform plane wave illumination that is difficult to provide. Space invariant array illuminators generate the spot array in the Fourier plane and are therefore rather immune to the exact shape of the incident beam, but they are more difficult to design and require tight fabrication tolerances.
Three techniques of achieving space invariant optical array illuminators by means of diffraction gratings with a computer synthesized periodic structure have been examined. The spot separation and the compression ratio in the Fourier plane can then be controlled straightforwardly by choosing a suitable grating period and the size of the illuminating beam, respectively. The relative intensities of the diffraction orders are, however, highly nonlinear functions of the grating structure, even in the domain where the paraxial scalar diffraction theory is valid.
Binary and multilevel surface relief gratings fabricated directly on fused silica by microlithographic methods involving electron beam written masks and reactive ion etching have been examined. Additionally, the method of hybrid holography was applied to record (in dichromated gelatin) the object wave produced by spatially filtering the output of an electron beam written binary amplitude master grating. The potential and the limitations of these techniques have been compared quantitatively, using 32 x 1 array illuminators as an illustration.
The binary gratings were the easiest to fabricate and, along with the multilevel elements, the easiest to use since they reconstruct the array on-axis. Of these 2, the multilevel approach offers higher efficiency but large scale fabrication requires low cost replication techniques. Hybrid holograms can be replicated by optical holography but the present recording material (dichromated gelatin) requires careful processing.
The prospect of wide bandgap light emitting diodes (LEDs) and lasers has been transformed by recent advances in p-doping in materials such as zinc selenide (ZnSe) grown by molecular beam epitaxy. Junctions of p-n ZnSe on gallium arsenide (GaAs) substrates using iodine as the n-type dopant and nitrogen from a plasma discharge source as the p-type dopant has been grown. LEDs have been fabricated using a gold contact to the p-type layer and blue continuous wave (CW) emission has been observed under forward bias. Stripe geometry laser structures have been fabricated and blue stimulated emission has been observed for the first time at low temperatures.
Electronic dispersive optical nonlinearities of cadmium sulphide (CdS) platelets have been investigated by wave mixing and pump probe experiments. These nonlinearities enabled the realization of a picosecond single wavelength. Fabry-Perot type logical gate that exploits polarization dichroism.
Coherent nonlinear resonances due to extended and localized excitons in cadmium selenide (CdSe) and cadmium selenide sulphide (CdSexS(1-x)) have been investigated by picosecond time resolved degenerate 4 wave mixing and differential transmission experiments. Large nonlinear coefficients chi(3) greater than 10 square centimetres per square volt were found, with coherence times in the picosecond range.
The nonlinear spectra of absorption near the band edge have been calculated for quantum well wires with up to 3 subbands. The calculations take into account phase space filling, plasma screening and band gap renormalization due to an optically excited electron-hole plasma. Large optical nonlinearities are obtained around the exciton ground state mainly due to state filling by the optically excited thermal electron-hole plasma, while the plasma screening effects are found to have relatively little influence. For all plasma densities n (including n = 0) the free carrier transition spectra differ strongly from those calculated with Coulomb interaction.
The plasma density dependence of the spectra of absorption and dispersion of gallium arsenide/gallium aluminium arsenide (GaAs/GaAlAs) quantum well wires (QWWs) with up to 3 subbands has been calculated. These calculations are an extension of corresponding calculations for bulk and quantum will semiconductors which are generally in good agreement with corresponding measurements.
In pump and probe beam experiments on II-VI semiconductor quantum dots at room and helium temperature no holeburning was found, but a strong bleaching of both maxima in the absorption spectrum, up to 75% of the optical density under a pump intensity up to 60 MW per square centimetre. This bleaching was nearly independent of the pump energy and the resulting small energetic shift was identified with the inhomogeneous broadening due to the size distribution of the crystallite. For both temperatures the halfwidth of the bleaching peak was comparable, which suggests a strong coupling of the excited states to the lattice. Corresponding results were obtained also from linear luminescence measurements. From both a Huang-Rhys factor from 1 to 2 was calculated fro the system. The samples were investigated in a self diffraction experiment at room temperature. The efficiency spectrum was a very broad band, corresponding to the broad structures in the absorption spectra of the sample with a halfwidth of about 2 nm. The effective value to chi(3) was calculated, about 1e-9 esu for the samples.
The implementation of a parallel optoelectronic automaton for a specific algorithm, the 'Lattice-Gas' has been studied. Three components were implemented for this automaton:
a microelectronic circuit for the treatment;
a Talbot effect hologram for instructions broadcast;
optoelectronic modulators for parallel readout.
This automaton exploits the performances of the electronics for nonlinear operation and the optics for the input and output interconnection.
The performances of a complete setup for 2 mm complementary metal oxide semiconductor (CMOS) technology (French MCP process) have been estimated as:
15000 iterations per second;
0.2 W power dissipation for 100 EPs;
high connectivity about 1.68 x 1e10 bits/square centimetre.
Some generalization of the use of such an automaton for other algorithms using symbolic substitution is possible. It can be made by simply changing the contents of the lookup memory.
A fully functional optical adder based on systolic arrays and symbolic substitution (SSL) has been demonstrated. This application shows the feasibility of optical design concepts based on incoherent data processing systems with optoelectronic threshold amplifier arrays. The optical setup applies 4 SSL rules in parallel to a 16 by 16 pixel data plane. Low hardware effort with integrable modules has been achieved.
The results of the comparison of different addition algorithms for parallel digital optical computing showed that the design of high performance parallel arithmetic units with self electrooptic devices (SEEDs) based on the known design methods is not primarily a question of having the right algorithm. It appears that for efficient optical arithmetic units the hardware side must be developed considerably. The assumptions taken for SEEDs, eg limited fan-in and fan-out of 2, application of regular interconnects only and the programmable logic array (PLA) design method based on universal AND stages for generation of all possible minterms, seriously limit the performance of the computing systems, independently of the algorithm used.
Hence the investigation of other design methods for SEED modules, the application of irregular (holographic for instance) interconnections, and/or the development of more powerful components like 'smart pixels' seem to be necessary in order to benefit from the advantages of digital optical parallel processing. Furthermore, the application of symbolic substitution in parallel adder design using SEEDs should be taken into account.
A board to board optical interconnect system has been developed. The interconnect distance is int he range of some centimetres to some metres. It comprises a 2-dimensional emitter array, a fibre bundle and a monolithic custom designed receiver array for application as a link within a multiprocessor system. Each channel has a data rate of 10 MBit/s.
Nonlocal algorithms that could be implemented on designs of extended cellular logic image processor (CLIP) like machines that incorporate one or more shuffle interconnects have been developed. A first indication of the power of single cycle nonlocal reconfiguration of data lies in the sort procedure that plays a major role in computation. Time savings of up to 1000 compared to dedicated software have been demonstrated for a 32 x 32 8-bit sort with and without a perfect shuffle connection. Such interconnects have been demonstrated optically, using lens combinations or holographic techniques.
With regard to the system components, the ongoing development of both optically and electrically addressed spatial light modulators is seen as key to fast parallel input, along with parallel data acquisition from disc-like optical storage. The use of SEEDs, and recent advances in pnpn structure optical thyristor components, leads to the question of smart pixels. How much logic should be incorporated into each optical cell? The answer would appear to be the maximum electronics, for both the cell itself and for local connections, that does not compromize the optical parallelism and hence the power of optics to carry out the nonlocal interconnection.
Finally, the issue of packaging is to be addressed. Our own demonstrators have advanced from independent optical mounts and 2 metre long gas lasers, through rail mounts and solid state lasers, to backplates and diode lasers. The overall size of the latter circuit is of order 50 cm x 50 cm, still considerably too large.
A cellular logic image processor was designed, constructed and successfully operated by interconnecting 2 symmetric self electrooptic effect device (S-SEED) arrays. Some of the design issues associated with the implementation of a free space digital optical system have been addressed. It is argued that pixellated devices are more desirable than other devices and that if differential data representation is used, high contrast is not required.
Silicon on sapphire (SOS) has been studied a nonlinear waveguide at lambda = 1.06 um. Resonant excitation of guided waves in submicron undoped SOS films can be obtained with a neodymium: yttrium aluminium garnet (Nd:YAG) laser beam for precise values of the incidence angle by means of a submicron period grating coupler etched in silicon. Sharp and contrasted angular resonances can be observed near the fundamental TEo mode excitation on either the reflected or transmitted beam. The wavelength being slightly smaller than the silicon absorption edge, fast optical and electrical switchings have been observed in the nonlinear pulsed regime (20 ns and 200 ps pulse durations). These switchings result, at first from electronic nonlinearities induced by a carrier density in excess of the order of 1e18 per cubic centimetre due to high optical excitation and later (with 20 ns pulses) from the competition between these effects and thermal nonlinearities of opposite sign due to the fast relaxation of the electron/hole pairs in excess.
When the devices operate under continuous wave illumination, the predominant nonlinearity has a thermal origin; it can be greatly enhanced by Joule effect, thanks to a voltage applied between electrodes deposited on the silicon film. Studies carried Joule enhanced sensitivity devices using thick silicon substrate as Fabry-Perot resonator and Schottky photodiode structures have shown interesting results in optical bistability with individual devices, although the nonuniformity of the silicon plate thickness has not allowed realization of 2-dimensional arrays. The good spatial uniformity of the TEo resonance on the SOS plate characterized during the studies carried out on an individual device led to the realization of a 6 x 3 gate array. The transfer characteristic in transmission allows NOR logic gate operation in an off axis geometry.
An optical oscillator has been demonstrated by introducing the sample into a hybrid ring resonator with long round trip time. In the case of a bistable input-output characteristic (IOC) of the nonlinear element (using an interference filter in reflection consisting of a glass matrix doped with cadmium sulphur selenide (CdSSe) strong mode locking to the resonator round trip time was obtained with a complex mode structure including Farey tree transitions and mode coexistence. For a barely bistable IOC (realized with a cadmium sulphide (CdS) single crystal) a completely different kind of dynamics was found. Relaxation oscillations became important and stabilized the system to prevent a transition to chaos. Further investigations are concerned with the electrooptic bistability in zinc selenide (ZnSe) single crystals. By applying an electric field perpendicular to the light beam changes of the optical properties due to the temperature change and to the Franz-Keldysh effect were observed.
With the aim of constructing a temperature sensor based on thermally induced optical bistability that can be operated by a conventional diode laser, the absorptive behaviour of gallium arsenide/(aluminium gallium) arsenide (GaAs/(AlGa)As) multiple quantum well structures were investigated. Even at room temperature, these structures exhibit a steep excitonic absorption edge in the near infrared spectral region. With rising temperature, this absorption edge shifts strongly enough to the red to lead to a sharp increase in the absorption of suitable photon energies below the absorption edge of the quantum wells. This opens the possibility of observing thermally induced optical bistability in the sample with moderate pumping intensities. The bistability is observed by focusing an infrared laser beam on the multiple quantum well structure. The transmitted laser intensity shows the switching of the sample between the 2 possible absorptive states. The incident laser intensities at which this switching occurs are extremely sensitive to the temperature of the material surrounding the sample. Using this effect, an optical temperature sensor can be achieved in a very practical design using an optical fibre to guide the incident and reflected laser beam, than could be provided by a diode laser.
Fibre optical sensor systems offer great potential for the application of nonlinear optical switching elements. For practical reasons, a working wavelength in the near infrared wavelength region, where solid state light sources can be used, would be very important. In addition, compatibility of the required light intensities for switching with available solid state light sources as well as high stability with environmental conditions are required. Fast switching speed and high parallelism are of lower importance for these applications. As shown with the bistable etalon with absorbed transmission (BEAT) element, an application with a solid state laser source is, in principle, possible. The concept of the optical reference switch gives an example that even with a rather simple use of such elements, the functionality of an optical sensor system can be enhanced. On the other hand, there is certainly a great effort necessary to improve the properties of optical switching elements, even for relatively simple technical applications.
Shaping of 1.9 ns laser pulses and optical switching behaviour with about 100 ps switching on/off times are measured in a bulk n-gallium arsenide (n-GaAs) Fabry-Perot etalon. The etalon structure is not optimized, having a finesse of about 4 at low incident power. The impurity related fast nonlinearity causes optical switching at wavelengths slightly lower than that of the fundamental absorption edge with incident intensities of about 1e6 W/square centimetre, corresponding to a switch energy of about 1 pJ/square micrometre.
Various solutions have been proposed in order to reduce the individual power consumption of integrated microresonators, and optimize their nonlinear response. In particular, at least one order of magnitude is expected to be gained in the threshold intensity by simply optimizing the structure geometry. Up to now power density thresholds of 500 W/square centimetre have actually been observed. The experimental analysis of the mechanism of the nonlinear saturation has shown that it originates from the saturation of the band-tail absorption, leading to refined criteria in the structure design. The possibility of fabricating nonlinear microresonator arrays whose continuous wave (CW) nonlinear characteristics improve after pixellation has been demonstrated. The alloy mixing technique which has been used allows a good confinement of both carriers and light.
All optical bistability has been observed between 2 different charge distributions in a type II heterostructure. In the OFF state, most charge accumulation occurs in the heart of the structure, while in the ON state holes accumulate at the extremities of the structure. This charge transfer bistability is, in this actual structure, rather slow (10 ns for OFF/ON commutation and about 1 us for ON/OFF commutation). However, a key feature is that it needs no external polarization and then no electrical addressing. Moreover, threshold values are at least 2 orders lower than in current devices based on more classical optical nonlinearities, and an important point is that the structure parameters and experimental conditions are not critical.
The structure and fabrication of a vertical cavity surface emitting laser diode have been explored. The lowest threshold currents are 16 mA. The light emission is single longitudinal mode at 915 nm wavelength. Under pulsed excitation the maximum output power is 3.5 mW. A wavelength selective photodetector of related structure shows a quantum efficiency of 10% at resonance with 1.5 nm spectral width at half maximum.
The double heterostructure optical thyristor is a promising device for optical information processing. It has the important advantage of combining the receiving, transmitting, and memory functions in a single device, and simultaneously achieving a considerable optical gain. The dynamics of the switching of a single device and of several devices in a winner-ta
APPROACH AND METHODS
Participants meet annually to disseminate information and to plan a potential European research effort. The Working Group also:
-holds regular sub-group meetings
-fosters working visits and longer-term exchanges
-invites other specialists in optical technology to address the annual meeting
-establishes technical objectives for device performance characteristics
-generates state-of-the-art reports.
PROGRESS AND RESULTS
A full Workshop meeting was convened at the Vrije Universiteit, Brussels in September 1991 and attended by 65 delegates from the full and associate participating centres. Progress was reviewed in the four main topic areas. Investigations of thermal optical nonlinearities of II-VI semiconductors has led to the construction of a novel temperature sensor based on the optical bistability of CdS. Other CdS optical properties are being investigated as a route to directly cascadable devices in the short-time (pi cosecond) rgime. Molecular Beam Deposition work at Edinburgh has enabled the preparation of nonlinear interference filters of good uniformity that exhibit high resistance to laser beam damage. The performance of such filters in our various optical proces sing circuits is being assessed.
Assembly of data has continued for a review of Perspectives for Parallel Optical Interconnects, a forthcoming publication which will include state-of-the-art contributions from authorities within the WOIT consortium.
The architecture group has made good progress on many issues, including optical pipeline processors, fast arithmetic using optical symbolic substitution logic (Erlangen), tolerancing studies and further develoment of the optical cellular logic image processor (Edinburgh), the Institute of Optics, Paris is developing an edge detection cellular automation and the University of Toulon has staged a demonstration of optical most significant digit addition.
Optical information technology is continuing to expand in importance. The exploitation of massive parallelism and interconnect complexity offered by optics could permit significant advances in the computational power available for tasks such as image processing and finite element analysis. Although considerable expertise exists around Europe, the creation of a world-competitive research programme requires its coherent coordination within a well targeted programme. Milestone demonstrators have been identified as essential components of such a project. This will permit assessment of current technology and processing architectural concepts and will allow the ultimate physical limits of both logic and interconnect devices to be fully explored.