An investigation into the feasibility of applying pattern recognition concepts to the classification of metallic objects by their electromagnetic response was performed. The effect on the response of various factors such as object shape and orientation was examined and a pattern recognition scheme was proposed based on these results. Implementation of the proposal involved the development of a novel extension to the nearest mean vector type of classifier in which the class "centroid" was generalized to be a curve in the feature space rather than a point. The resultant pattern recognition scheme was tested on a representative test set which included 815 signatures of objects, corresponding to 104 variations in object and orientation. A success rate of greater than 98 percent was achieved. It is noted that the classifier extension developed provides a viable approach to classification of response signatures that vary continuously with respect to any single parameter. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

A study was conducted to evaluate the ability of police dog teams to identify "guilty" subjects in a simulated crime situation and to compare their accuracy with that of a polygraph examination. Research on the olfactory acuity of dogs, and on the role of olfactory cues such as pheromones in social communication, implies that the detection of guilt by experienced police dogs could occur as reliably as police dog handlers believe. The literature on polygraph investigations shows high reliability in detecting guilt. This was one of the reasons for using the polygraph as the comparison technique. Three experienced dog teams from the Vancouver Police Dog Squad and two expert polygraph field examiners were used. The subjects were 64 male volunteers, all university or college students. Subjects randomly assigned to the "guilty" condition were instructed to "steal" and conceal a $10 bill that had been left in an empty office, and to deny throughout the rest of the experiment that they had done so. Volunteers in the "not guilty" condition were told nothing about the "crime" Both groups were told that police dog teams and polygraph operators would try to find out whether they were guilty. They were promised $5.00 for participating plus a bonus of $10 if they succeeded in establishing that they were innocent. Police dog team performance was about chance level, while the polygraph examination was significantly more accurate than chance and than the dog teams. No individual difference was found among the dog teams. The failure of the dog teams could be attributed either to the impossibility of detecting guilty by smell cues or to some aspect of the simulation procedure. Further research should be directed at developing more realistic field studies. / Arts, Faculty of / Psychology, Department of / Graduate

This dissertation presents two lines of research that are superficially at opposite ends of the computational neuroscience spectrum. While models of adaptive motion detection in fruit flies and simulations inspired by monkeys that learn to control brain machine interfaces might seem like they have little in common, these projects both attempt to address the broad question of how real neural circuits flexibly compute. Sensory systems flexibly adapt their processing properties across a wide range of environmental and behavioral conditions. Such variable processing complicates attempts to extract mechanistic understanding of sensory computations. This is evident in the highly constrained, canonical Drosophila motion detection circuit, where the core computation underlying direction selectivity is still debated despite extensive studies. The first part of this dissertation analyzes the filtering properties of four neural inputs to the OFF motion-detecting T5 cell in Drosophila. These four neurons, Tm1, Tm2, Tm4 and Tm9, exhibit state- and stimulus-dependent changes in the shape of their temporal responses, which become more biphasic under specific conditions. Summing these inputs within the framework of a connectomic-constrained model of the circuit demonstrates that these shapes are sufficient to explain T5 responses to various motion stimuli. Thus, the stimulus- and state-dependent measurements reconcile motion computation with the anatomy of the circuit. These findings provide a clear example of how a basic circuit supports flexible sensory computation. The most flexible neural circuits are circuits that can learn. Despite extensive theoretical work on biologically plausible learning rules, however, it has been difficult to obtain clear evidence about whether and how such rules are implemented in the brain. In the second part of this dissertation, I consider biologically plausible supervised- and reinforcement-learning rules and ask whether biased changes in network activity during learning can be used to determine which learning rule is being used. Supervised learning requires a credit-assignment model estimating the mapping from neural activity to behavior, and, in a biological organism, this model will inevitably be an imperfect approximation of the ideal mapping, leading to a bias in the direction of the weight updates relative to the true gradient. Reinforcement learning, on the other hand, requires no credit-assignment model and tends to make weight updates following the true gradient direction. I derive a metric to distinguish between learning rules by observing biased changes in the network activity during learning, given that the mapping from brain to behavior is known by the experimenter. Because brain-machine interface (BMI) experiments allow for perfect knowledge of this mapping, I focus on modeling a cursor-control BMI task using recurrent neural networks, and show that learning rules can be distinguished in simulated experiments using only observations that a neuroscience experimenter would plausibly have access to.

Plasmons can be generated in the two dimensional electron gas (2DEG) of grating-gated high electron mobility transistors (HEMTs). The grating-gate serves dual purposes, namely to provide the required wavevector to compensate for the momentum mismatch between the free-space radiation and 2D-plasmons, and to tune the 2DEG sheet charge density. Since the plasmon frequency at a given wavevector depends on the sheet charge density, a gate bias can shift the plasmon resonance. In some cases, plasmon generation results in a resonant change in channel conductance which allows a properly designed grating-gated HEMT to be used as a voltage-tunable resonant detector or filter. Such devices may find applications as chip-scale tunable detectors in airborne multispectral detection and target tracking. Reported here are investigations of InGaAs/InP-based HEMT devices for potential tunable resonant sub-THz and THz detectors. The HEMTs were fabricated from a commercial double-quantum well HEMT wafer by depositing source, drain, and semi-transparent gate contacts using standard photolithography processes. Devices were fabricated with metalized transmission gratings with multiple periods and duty cycles. For sub-THz devices, grating period and duty cycle were chosen to be 9 ?m and 22%, respectively; while they were chosen to be 0.5 ?m and 80% for the THz device. The gratings were fabricated on top of the gate region with dimensions of 250 ?m x 195 ?m. The resonant photoresponse of the larger grating-period HEMT was investigated in the sub-THz frequency range of around 100 GHz. The free space radiation was generated by an ultra-stable Backward Wave Oscillator (BWO) and utilized in either frequency modulation (FM), or amplitude modulation (AM) experiments. The photoresponse was measured at 4K sample temperature as the voltage drop across a load resistor connected to the drain while constant source-drain voltages of different values, VSD, were applied. The dependence of such optoelectrical effect to polarization of the incident light, and applied VSD is studied. The results of AM and FM measurements are compared and found to be in agreement with the calculations of the 2D-plasmon absorption theory, however, a nonlinear behavior is observed in the amplitude and the line-shape of the photoresponse for AM experiments. For detection application, the minimum noise-equivalent-power (NEP) of the detector was determined to be 235 and 113 pW/Hz1/2 for FM and AM experiments, respectively. The maximum responsivity of the detector was also estimated to be ~ 200 V/W for the two experiments. The far-IR transmission spectra of the device with nanometer scale period was measured at 4 K sample temperature for different applied gate voltages to investigate the excitation of 2D-plasmon modes. Such plasmon resonances were observed, but their gate bias dependence agreed poorly with expectations.

A fiber optic fluid level sensor based on light transmission attenuation due to bending losses is designed, built and tested. Fibers formed with reverse curvatures of decreasing radii will induce and increasing amount of lower order mode light loss to the cladding as the light propagates along the step index multimode fiber. The sensor is arranged in the fluid in a vertical position such that the light travels along the fiber from the bottom or low fluid point to the top or full point. As the fluid covers increasing lengths of the exposed fiber, it strips even more power from the cladding (assuming the fluid refractive index is greater than the cladding refractive index). Data taken with a sensor of this configuration show a monotonic decrease of the output intensity as a function of increasing fluid level. As much as a 14 dB change occurs over a one-foot fluid level change. A mathematical model, based on both field theory and geometrical optics, is developed to evaluate and predict the performance of this fiber optic fluid level sensor. Comparisons of the theoretical predictions and the experimental results under laboratory conditions show very good agreement.

Major advances in fiber optic transmissions have brought about a need for highly sensitive photodetectors. In order to detect this type of transmission, the photodetector must be able to detect one of the two low loss windows of the fiber optics transmission. The photodetector must also be characterized by a high gain and fast speed of response without generating excessive noise power. This report compares different types of high speed phodetectors with emphasis on the merits of using an avalanche photodiode. The report studies the avalanche multiplication and the signal-to-noise power ratio in the ternary InGaAs. The effects of the absorption coefficient, the depletion width, and the impurity concentration are studied. Finally, an optimization of the signal-to-noise power ratio is achieved by selecting the proper impurity concentration profile at suitable values of absorption coefficient and epitaxial width.

[Truncated abstract] HgCdTe is the leading semiconductor material for the fabrication of high performance infrared photon detectors, in particular, for detection of radiation beyond the near infrared. State-of-the-art infrared detection and imaging systems are currently based around high density focal plane arrays consisting of HgCdTe photodiodes as detector elements. Despite the high performance of HgCdTe infrared detectors, and the many benefits they can offer to industry and society, their utilisation remains limited due to the high cost of production. The chemical composition and narrow bandgap of the HgCdTe material used for infrared detection means that the material is inherently very susceptible to defect formation caused by the processing procedures required for device fabrication. Consequently, fabrication of HgCdTe photodiode arrays have traditionally been characterised by low yields and high costs for arrays that meet required operability specifications. In this thesis a new photodiode fabrication technology with the potential to improve device yields over traditional fabrication technologies is presented. This new fabrication technology is distinguished from others by the use of plasma-induced p-to-n type conversion of HgCdTe for junction formation. This allows great simplification of the fabrication process and avoids high temperature processing during and after junction formation, and keeps the junction protected from the atmosphere at all stages of fabrication. The development of the photodiode fabrication technology using plasma-induced junction formation has involved characterising the electrical transport properties of the type-converted layers, fabrication and characterisation of photodiodes, and photodiode dark current modelling

Made available in DSpace on 2014-10-09T12:40:44Z (GMT). No. of bitstreams: 0 / Made available in DSpace on 2014-10-09T14:06:14Z (GMT). No. of bitstreams: 1 02811.pdf: 5186762 bytes, checksum: f178f608c95e9fecbaf736562e08dae5 (MD5) / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Tese (Doutoramento) / IPEN/T / Instituto de Pesquisas Energeticas e Nucleares - IPEN/CNEN-SP / FAPESP:94/03409-4

Made available in DSpace on 2014-10-09T12:40:44Z (GMT). No. of bitstreams: 0 / Made available in DSpace on 2014-10-09T14:06:14Z (GMT). No. of bitstreams: 1 02811.pdf: 5186762 bytes, checksum: f178f608c95e9fecbaf736562e08dae5 (MD5) / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Tese (Doutoramento) / IPEN/T / Instituto de Pesquisas Energeticas e Nucleares - IPEN/CNEN-SP / FAPESP:94/03409-4

In this work, we have designed, simulated, fabricated and characterized homojunction Hg1-xCdxTe detector for high operating temperature in the MWIR region. The IR photon detectors need cryogenic cooling to suppress thermal generation. The temperature of operation in narrow gap semiconductor devices is limited by the noise due to statistical nature of thermal generation-recombination in narrow gap semiconductors. To make IR systems affordable they have to be operated without cooling or with minimal cooling compatible with low cost, low power and long life. Several fundamental and technological limitations to uncooled operation of photon detectors have been discussed in Chapter-1 of this thesis. Way and means adopted to increase the operating temperature, such as non-equilibrium operation, use of multilayer stacked hetero¬structures, optical immersion etc. have also been discussed. Key to improving the detector performance at any temperature is reduction of dark currents to level below the photocurrent and ultimately to the level where detector noise is determined by the fluctuations in photon flux from the scene (BLIP limit). In addition, design of present generation uncooled Hg1-xCdxTe infrared photon detectors relies on complex hetero-structures with a basic unit cell of type n+/π/p+. Theoretical modeling and numerical simulations on TLHJ device consisting of backside illuminated n+/π/p+ photodiodes have been performed. A numerical model for solving carrier transport equations for Hg1-xCdxTe infrared photodiodes was developed in MATLAB. Finite difference discretization of carrier transport equations and successive over relaxation method have been adopted. Numerical models are more appropriate than analytical models when analyzing multi-layer hetero-structures because we can account for realistic doping profiles, compositional grading and hetero-structures using this model. The model can be suitably modified to accommodate different device architectures, designs, material properties and operating temperature. Such a generalized model is useful to a device designer to customize the detector performance as per the availability of the material to suit the application specific requirements. The present work therefore proposes a more flexible, accurate and generalized methodology to accommodate the user needs by simulating the position dependence of carrier concentration, electrostatic potential and g-r rates and their effect on detector performance vis-à¬vis contact doping, absorber doping and absorber width on device performance. We detail aspects of our simulation model by developing a library of Hg1-xCdxTe properties using analytical and empirical expressions for material parameters (energy band gap, electron affinity, intrinsic carrier concentration, carrier effective mass, carrier mobility, dielectric constant and absorption coefficient). The PDEs were solved using the FDM coupled with SOR method. Behavior of Hg1-xCdxTe diodes (homo/hetero-junction) under different biasing, illumination and non equilibrium situations were modeled. Model has been validated for experimental measured data on n on p Hg1-xCdxTe photodiodes. The numerical computations are next applied to simulation/modeling of MWIR (λc=4.5 μm) n+/π/p+ TLHJ device for operation at T=250K. Several recombination processes occur in Hg1¬-xCdxTe depending on material quality, operating temperature, device design and processing conditions. Detailed mathematical models of radiative, Auger, Shockley Read Hall (SRH), surface recombination and optical g-r are analyzed and their effect on carrier lifetime have been evaluated. Analytical models for dark currents affecting the performance of Hg1-xCdxTe diodes at different temperatures are discussed. The mechanisms contributing to dark current are: (i) the thermal diffusion of minority carriers from the neutral regions (IDiff); (ii) generation-recombination from the space charge region of diode (IG-R) (iii) trap assisted tunneling currents, wherein the traps in the depletion region or the traps in the quasi neutral p region close to the depletion edge participate in the tunneling process(ITAT); (iii) band-to-band tunneling currents (IBTB) and (iv) surface leakage currents due to shunt resistance. Total current of a photodiode is ITOT=IDiff+IG-R+ITAT+IBTB+ISH-IP, where IP is the photocurrent. We evaluate the variation of electrostatic potential, carrier concentration, and electric field and g-r profiles as a function of position. The effect of variation in absorber width, doping and contact doping on D* is also analyzed. The mathematical models of different g-r processes (Auger, SRH, radiative, surface recombination and optical generation) affecting the device performance analyzed and their affect on carrier lifetimes are investigated. Responsivity ~3.25Amp-Watt-1, noise current~2pA/Hz1/2 and D* ~8x109 cmHz1/2watt-1 at 0.1V reverse bias have been calculated using optimized values of doping concentration, absorber width and carrier lifetime. The suitability of the method has been illustrated by demonstrating the feasibility of achieving the optimum device performance by carefully selecting the device design and other parameters. The numerical models provided insight about the operation and performance of Hg1-xCdxTe Auger-suppressed infrared photodiodes. Hetero-junction configuration increases the dynamic resistance, while the heavily doped contacts reduce the contact resistance. Wide gap/heavily doped contacts present a barrier to injection of minority carries into the absorber layer. At the same time they allow collection of minority carriers generated in the absorber region at the contacts. Hg1-xCdxTe hetero-diodes are grown by MOCVD and MBE with precise doping and compositional gradient control to reduce g-r contributions from defects and dislocations to the dark current in order to reap advantages of Auger suppression. Measured dark currents in hetero-junction photodiodes continue to be larger than expected in spite of the advancements in MBE technique. Delineation of an array on hetero-structures involves mesa separation of the diodes thus creating additional surface requiring passivation. Overall, the whole effort of fabricating a hetero Hg1-xCdxTe detector array is disproportionate to the overall gain in the performance. Therefore, we employ a much simpler fabrication process of homo-junction Hg1-xCdxTe detectors. It involves a planar device fabrication approach thus minimizing the surface passivation problem. We have deliberated upon the specific growth, characterization techniques and processing steps employed in our study. We discuss some of the experimental issues. We also presented results on the novel processing techniques developed that are potentially applicable to HOT technology and Hg1-xCdxTe technology in general. Hg1-xCdxTe (x=0.27-0.31) layer of ~ 15×15mm2 area and 15-20µm thickness is grown on CdZnTe substrate by Liquid Phase Epitaxy (LPE) in-house. As grown wafer is vacancy doped p-type with a carrier concentration of ~5×1015-1x1016 cm-3 and hole mobility of ~400cm2V-1s-1@80K. Planar n+/ν/p junction ~2-3µm deep is formed by B+ ion implantation and subsequent annealing; details are outlined in Chapter-4. Hall measurements and differential Hall measurements were used to find the carrier concentration, carrier mobility, resistivity of the wafer. The diodes are formed in the form of a 2D array along with various PEV’s for process characterization. Composition of Hg1-xCdxTe wafers used for the work is in the range of 0.27¬ 0.31 as determined by FTIR, corresponding to cutoff wavelength of 4.5-6.5µm. Junction depth and doping profile of the diodes after ion implantation was characterized by differential Hall technique. Transient minority carrier lifetime in fabricated MWIR n+/ν/p Hg1-xCdxTe (x=0.27) diodes were characterized using diode reverse-recovery technique. We prefer this method because it is a direct indicator of device as well as material quality post processing. By this time the device has undergone all the chemical/mechanical treatments and the measured lifetime is the cumulative of g-r mechanisms operative in bulk, space charge region and surface of diode. The value of lifetime extracted from the measured data lies in the range of 80-160ns. Variable temperature lifetime data was also extracted to determine the prevalent g-r process operative in the device. Diode dark I-V and junction C-V measurements were also made to correlate the observed behavior of the measured lifetime with g-r processes. Evidence of Auger suppression at room temperature is seen in the dark I-V characteristics via observation of negative differential resistance in the homo-junction Hg1-xCdxTe diodes. The experimental data is fitted using the numerical and analytical models developed. Based on this fitting, the current mechanisms limiting the dark current in these photodiodes are extracted. An improved analytical I-V model is reported by incorporating TAT and electric field enhanced Shockley-Read-Hall generation recombination process due to dislocations. Tunneling currents are fitted before and after the Auger suppression of carriers with energy level of trap (Et), trap density (Nt) and the doping concentrations of n+ and νregions as fitting parameters. Values of Et and Nt were determined as 0.78-0.80Eg and ~7-9×1014 cm-3 respectively in all cases. Doping concentration of νregion was found to exhibit non-equilibrium depletion from a value of 2×1016 to 4×1015 cm-3. Quantum efficiency of the diodes was found to ~25-30%. Note, that these are wafer level measurements on unpackaged device without backside AR coating. In addition to junction diodes, we present results on several PEV's such as VADA, MIS/MIM capacitors and TLM structures both at room and low temperature. Variable temperature measurements for a VADA tile and subsequent analysis provide evidence of g-r processes originating from defects, dislocations and dislocation loops, which are non-uniformly distributed across the Hg1-xCdxTe wafer and contributes to TAT current at high temperatures. MIS analysis yielded surface charge density lying between 3×1010-1×1011 cm-2 for ZnS/CdTe surface corresponding to a near flat band condition. Results of low and variable temperature measurements on the devices have also been shown to correlate it with the possibility of operating the device at mid temperatures such as 180-250K.