Nonlinear interference filters have offered the prospects of 2-D arrays of all-optical memory and logic functions. Image processing and numeric computation architectures that suit the properties of nonlinear filters, and prototype demonstrators, are described.

Optical processing techniques are presented that provide the necessary input data for multi-target trackers. We detail optical processing techniques for the detection of sub-pixel targets. We also detail optical processing techniques for the processing of composite frames of data from the first stage processor to detect target tracks and to estimate the position/velocity parameters of all targets and the covariance of all target state parameters. The first stage processor employs optical correlation, shift, interpolation and subtraction techniques. This second stage processor employs optical feature extraction techniques. Both processors utilize hybrid optical/digital processing techniques. Simulated data for both processor stages is provided. A new thresholded Hough transform processor is shown to provide reduced false track detections.

There is at present a growing interest in developing photonic switching and signal-processing devices for future applications in communications and computing. In this paper we will briefly outline some of the reasons for this interest, and then describe some recent work on ultrafast all-optical switching devices using glass as the nonlinear material. We will conclude with some speculations regarding the future prospects for photonic switching.

We have investigated the switching dynamics of thermo-optical devices, working at room temperature. Slowing down and transverse effects are studied for ZnSe interference filters and platelets of CdS.

Monolithic optical logic devices 1.5-5 μm across are defined by ion-beam assisted etching through a GaAs/AlAs Fabry-Perot structure grown by molecular beam epitaxy. They show reduced energy requirements (factor 30 smaller than the unetched heterostructure), uniform response over small arrays, negligible crosstalk at 3-μm center-center spacing, <200 ps recovery time and thermal stability at 82 MHz operating frequency. All experiments were performed at room temperature.

Time-dependent behavior of GaAs nonlinear Fabry-Perot etalons has been investigated with a computer model using the Banyai-Koch plasma theory for the GaAs nonlinear optical properties. Investigation of single-wavelength transient operation of GaAs etalons as optical logic gates has shown that useful differential energy gain is not achievable for optical pulses shorter than about ten times the charge carrier lifetime in the semiconductor.

Holography can be used for arbitrary, parallel, weighted mappings between planes with up to to 106 pixels each at I/O limited rates. This allows precalculated mappings to occur at very high speed. The applicarttions for one-to-one, one-to-many, many-to-one, and many-to-many maps are explored here.

An associative memory for fingerprint identification has been constructed using a Van der Lugt correlator and an interference filter as a reflective thresholding device.

Optical memories consisting of arrays of bistable pixels are considered. Attainable pixel density is limited by cross-talk, so that there is a minimum pixel separation for which all pixels may be independently held "on" or "off". This problem is analysed for diffusive cross-talk between members of a linear array of pixels. A nonlinear dynamics analogy links pixel independence to spatial chaos, and enables estimates of minimum pixel density. The stability of pixel patterns is governed by a form of Schrodinger equation. More realistic models should show similar features.

We discuss novel detectors and modulator structures fabricated from GaAs/AtxGai_xAs multiquantum well superlattices, which are compatible with high speed optical computing at the CO2 laser wavelength A = 10.6 μm. These components can be integrated with transparent GaAs optical waveguides and also with GaAs FETs or bipolar transistors for high speed electronic signal processing.

The physics of low energy switching in both optical and electronic devices is briefly introduced. The properties and applications of bistable diode amplifiers are reviewed. It is stressed that bistable dode laser amplifiers are the most energy efficient optical switching devices so far demonstrated. The concept of colored optical interconnects is introduced for applications in closely spaced wavelength division multiplexing transmission systems. The implementation of a generalized, optically updatable, optical crossbar switch is discussed.

Changes in the reflection and transmission spectra of GaAs doping superlattices due to optical excitation of carriers are measured. Modulated transmission and reflection spectra show that both the real and imaginary parts of the complex refractive index are affected in the spectral region near and below the GaAs band gap. Modulated transmission measurements of 1064-nm light through a 1-mm-long waveguide sample show significant modulation at room temperature. In the sample studied, large below-gap absorption was observed which could not be reduced by optical excitation. We associate this absorption with an impurity or trap level located approximately 200 meV below the GaAs band gap.

The origine of the nonlinear optical properties of semiconductors are discussed. We treat in detail 3 examples for electronic mechanisms: a) induced exciton absorption due to dynamical line-broadening; b) the band-edge nonlinearities due to plasma band-filling, screening, and renormalization; and c) the coherent band-edge nonlinearities for ultra-short pulse excitation.

Excitonic optical nonlinearities in epitaxial ZnSe thin films have been measured at 150 and 300K. The measured nonlinear response, An/N = 1.8X10-19 cm3 at 300K, is comparable to that observed in bulk GaAs and GaAs/AlGaAs multiple quantum well systems. Analysis of these results using our partly-phenomenological plasma theory indicates that the mechanisms responsible for this nonlinearity at room temperature are exciton screening and density-dependent broadening. At 150K, exciton screening makes the major contribution to the nonlinearity. Using femtosecond laser techniques, we have determined that the nonlinearity turns off in times on the order of 50-100 ps. These excitonic nonlinearities are, therefore, both large in magnitude and extremely rapid.

Semiconductor structures in lower dimensions, dubbed quantum dots, exhibit novel properties which result from size quantization of their charge carriers, as well as from their large surface-to-volume ratio. Optical measurements, combined with scanning tunneling microscopy, can provide the detailed information required to model the nonlinear optical response of these clusters.

We discuss the saturation intensities and saturation densities as a function of well width for the room temperature excitonic absorption resonance of GaAs/A1GaAs multiple quantum well structures grown by metalorganic chemical vapor deposition. The optimum growth temperature is considered. We discuss techniques for calculating absorption coefficients that remove Fabry-Perot effects as well as the effects of carrier diffusion. The difference between pulsed and CW excitation of the mutliple quantum wells is addressed. We discuss the calculation of the nonlinear index of refraction from a discrete set of absorption data. We introduce a new structure (a hetero n-i-p-i) which combines the advantages of multiple quantum wells and doping superlattices.

We present a study of the dependence of the magnitude of optical nonlinearities of GaAs/AlGaAs multiple quantum wells on quantum well thickness. Using four-wave mixing and nonlinear absorption measurements the refractive nonlinearity was determined in 17 samples grown by MOCVD and MBE. We find a small variation (less tban a factor of three) in the change in refractive index per photoexcited carrier for well sizes between 50 A and 300 A. or bulk GaAs.

Optical nonlinearities of bulk GaAs and GaAs/A1GaAs multiple quantum well samples are measured and compared over the wavelength region in the vicinity of bandgap at room temperature. The maximum index variation (An) is found to increase by a maximum of a factor of 3 as the well size is decreased from bulk to 76 A. The nonlinear absorption spectra are analyzed using a semi-empirical fit.

Photorefractive materials have been used as the nonlinear optical media in various optical computing and image processing applications. Devices such as passive phase conjugate mirrors, image subtractors, optical limiters and thresholders can be constructed with materials whose index of refraction is a function of light intensity. It is the purpose of this paper, then to describe some of these materials on the basis of their nonlinear optical response.

Photorefractive gratings have been written with picosecond pulses in GaAs, BaTiO3, LiNb03 and Bi12S1020. These observations are reviewed and compared to each other and to the theoretical limit on photorefractive sensitivity.

Measurements carried out on semiconductor-doped glass channel waveguides are reported. The waveguides exhibit a fast resonantly enhanced saturable nonlinearity.

Silica-waveguide based optical interconnections for both LSI chip-to-chip data transmissions and optical clock signal distributions are described. The optical interconnection circuit is composed of waveguide paths, a waveguide mixer, laser diodes and photodiodes. The interconnection network is based on LAN star coupler network to perform both broadcast data transmissions and clock distributions. In an experimental 4-chip optical interconnection circuit, a possible signal bit rate estimated from measured receiving optical level was 500 Mbps. In the data transmission and clock distribution experiments, 170 Mbps data rate and 250 Mbps clock rate were confirmed.

Electrical interconnect technology is becoming the major limitation in the realization of high-performance computing machines. The use of optical interconnections promises to alleviate key interconnect bottlenecks such as pinout, fanout, wiring density, and communication bandwidth. We present results of work on guided-wave optical interconnect circuits for eventual use within and between multichip packages. Low-loss integrated optical waveguides (<0.5 dB/cm), right-angle waveguide corner bends (<0.4 dB/cm), right-angle waveguide crossovers (<0.2 dB), and right-angle 1:2 waveguide branches (<0.5 dB) are presented.

A hybrid approach to optical computing is currently studied at the University of California, San Diego for performing parallel processing. Figure 1 shows a promising parallel processing architecture, in which an optical processing array is connected massively in parallel to an optical memory array. An optical processing array consists of an array of electronic logic gates; each cell in the array has at least one photodetector (or phototransistor) and one modulator for optical input/output. Similarly, an optical memory array consists of an array of electronic memory circuits; each cell in the array has also optical input/output. Within the optical processing array or the optical memory array, local interconnections among neighboring cells can be established electronically. Between the optical processing array and the optical memory array, global interconnections among distant cells can be established optically using the optical input/output in each cell. One important application of such a massively interconnected architecture would be to perform matrix-tensor multiplications (Ref. 1) not only for numeric computing, but also for artificial intelligence and neural computing (Ref. 2). There are three unique aspects associated with the parallel architecture of Figure 1, which may serve to justify this hybrid approach toward optical computing: optical interconnection, 3-D computational topology and optical memory.

A matched filter based architecture for associative memories ( MB'AM ) has been proposed by many researchers[1-12]. The correlation from a leg of a matched filter bank, after being altered nonlinearly, weights its corresponding library vector. The weighted vectors are summed and clipped to give an estimate of the library vector closest to the input. We analyze the performance of such architectures for binary and/or bipolar inputs and libraries. Sufficient conditions are derived for the correlation nonlinearity so that the MFAM outputs the correct result. If, for example, N bipolar library vectors are stored, then the correlation nonlinearity Z(x) = Nx/2, will always result in that library vector closest to the input in the Hamming sense.

Direct free-space optical interconnection techniques are described for the Hypercube concurrent processor machine using a holographic optical element. Computational requirements and optical constraints on implementation are briefly summarized with regard to topology, power consumption, and available technologies. A hybrid lens/HOE approach is described that can support an 8-dimensional cube of 256 nodes.

A novel switch geometry using photorefractive crystals has been demonstrated which has nondestructive readout and reduces the order of complexity of a non-blocking switching network. It is a routing full duplex design capable of handling the full bandwidth of the input signals. It is fully reconfigurable in times which depend principally upon the photorefractive material used and the intensity of the writing beams. The concept is applicable for use with a variety of photorefractive materials. The capacity of this switch has been calculated to be on the order of 100, based upon the sum of inputs and outputs, but is primarily limited by its simultaneous interconnectability. This limit is set by the number of gratings which can be stored in a single photorefractive crystal at one time. A demonstration of a 2 x 3 switch using BGO with 514nm writing beams and 633nm signal (readout) beams is descibed.

Photodetectors fabricated in GaAs have been vertically coupled to optical fibers using a cavity anisotropically etched in the backside of the GaAs substrate. The detector used was a Schottky photodiode amplified by an integrated MESFET. The cavity was etched from the backside using reactive ion etching (RIE). AlGaAs epitaxially grown on the substrate was used as an etch stop layer. A tapered optical fiber inserted in the backside cavity is accurately aligned to the detector and is mechanically stable. The vertical coupling approach is real-estate efficient and is particularly well suited for spatially parallel optical computing.

A comparison of the performance of optoelectronic (laser diode-based) and electronic residue position-coded look-up tables is presented. The comparison is made in terms of speed and power consumption. Experimental results for the electronic look-up tables are also presented.

The use of optical phase conjugation (OPC) process for parallel digital and symbolic optical computing is described. Using spatially encoded logic and sym-bolic variables, various OPC-based parallel ultrafast optical logic, symbolic as well as interconnect processors are detailed. The proposed devices are experimen-tally verified using picosecond pulses from a mode-locked Nd3+:YAG laser. Based on these processors, an OPC-based ultrafast optical computing architecture is proposed.

The Prolog language provides the flexibility to implement many expert system approaches and exhibits high levels of parallelism. However, it is too slow to have wide applicability to real-world problems or for real-time problems. The most computationally demanding steps in Prolog are the search and substitution required in unification. An all-digital opto-electronic architecture is described in which optical symbolic substitution is used to significantly speed unification by using high levels of parallelism. A simulator is being developed for the system to verify operation and assist performance evaluation. Progress in spatial light modulators suggests that such architectures will soon be feasible and superior to all-electronic counterparts.

The role of fiber optics as an interconnection mechanism for optoelectronic programmable logic arrays (OPLA's) is critically examined. The effect of fundamental physical limitations on the computational capacity of OPLA's is considered. A comparison of OPLA technology to that of existing electronic logic arrays is presented, and scaling relations are derived. The role of higher order decoders in minimizing the requirements on computational capacity is delineated. A novel technique is proposed for effecting a reconfigurable equivalent to a spatial light modulator based logic system.

The optical transfer characteristics of Microchannel Spatial Light Modulator are presented. The addressability and the readability of the device and their relationships are discussed from the viewpoint of the device structure, performance and potential applications.

The concept of using color-multiplexed four-wave mixing to achieve optical matrix-matrix multiplication with N3 parallelism is described. Experimental results, for the case of 2 x 2 matrices, using two-color four-wave mixing in a photorefractive crystal are presented.

A two-level neural network is proposed for the implementation of general deterministic logic (switching) functions. The network is potentially capable of implementing any set of binary switching functions of n variables. A cascade of two neural-like processor levels gives rise to a high-performance nonlinear functional memory. The first neural layer implements a linearly separable psuedorandom mapping that maps n dimensional binary input vectors into a higher m dimensional space of randomly scattered vectors, while the second neural layer implements a one-pass associative neural memory (ANM) that maps the output of the first layer into prerecorded target vectors. The interconnection weights of this layer are synthesized using a new and highly efficient recording technique[l]. The high fan-out of the first layer mapping and the highly distributed parallel architecture of the proposed network are ideal for optical implementation.

An optical pattern recognizer which classifies An input sequence into one of two classes is described. The language of regular expressions is used to identify a set of input sequences to be recognized. The system is realized without memory elements by using an optical iterative array it with the method of symbolic substitution. The hardware must be able to implement arbitrary and multiple substitution rules for the general design of a pattern recognizer.

An optical bidirectional associative memory (BAM) offering a potential of operating up to 106 neurons with 1012 interconnections is described. Except possibly, for input and output, all operations are optical and parallel.

An important step in multitarget tracking (MTT) is the assignment of measurement/target pairing probabilities, i.e., the probability that a certain measurement is associated with a particular target. This is one of the more computationally intensive steps of the Joint Probabilistic Data Association (JPDA) and other MTT algorithms. Determining such assignment probabilities in the conventional JPDA algorithm involves the generation and analysis of all possible binary feasibility matrices, where each feasibility matrix represents the feasible event that each measurement is due to only one target (or clutter) and that each target causes at most one measurement. We discuss a new and most efficient technique for obtaining the measurement/target pairing probabilities directly (without first producing the feasibility matrices) and its implementation on a new analog optical computer.