James L. Blue

A small-signal theory of avalanche noise in IMPATT diodes

A general small-signal theory of the avalanche noise in IMPATT diodes is presented. The theory is applicable to structures of arbitrary doping profile and uses realistic ( \alpha \neq \beta in Si) ionization coefficients. The theory accounts in a self-consistent manner for space-charge feedback effects in the avalanche and drift regions. Two single-diffused n-p diodes of identical doping profile, one of germanium and the other of silicon, are analyzed in detail. For description of the noise of the diodes as small-signal amplifiers the noise measure M is used. Values for M of 20 dB are obtained in germanium from effects in the depletion region only, i.e., when parasitic end region resistance is neglected. Inclusion of an assumed parasitic end resistance of one ohm for a diode of area 10-4cm2produces the following noise measure at an input power of 5×104W/cm2, and at optimum frequency: germanium 25 dB, silicon 31 dB. For comparison, a noise figure of 30 dB has been reported [1] for a germanium structure of the same doping profile as used in the calculations. Measurements of silicon diodes of the same doping profile are not available, but typically silicon diodes give 6-8 dB higher noise figures than germanium diodes of comparable doping profile.

Evaluation of Pattern Classifiers for Fingerprint and OCR Applications

Abstract The classification accuracy of four statistical and three neural network classifiers for two image based pattern classification problems is evaluated. These are optical character recognition (OCR) for isolated handprinted digits, and fingerprint classification. It is hoped that the evaluation results reported will be useful for designers of practical systems for these two important commercial applications. For the OCR problem, the Karhunen-Loeve (K-L) transform of the images is used to generate the input feature set. Similarly for the fingerprint problem, the K-L transform of the ridge directions is used to generate the input feature set. The statistical classifiers used are Euclidean minimum distance, quadratic minimum distance, normal, and k -nearest neighbor. The neural network classifiers used are multi-layer perceptron, radial basis function, and probabilistic neural network. The OCR data consist of 7480 digit images for training and 23,140 digit images for testing. The fingerprint data used consist of 2000 training and 2000 testing images. In addition to evaluation for accuracy, the multi-layer perceptron and radial basis function networks are evaluated for size and generalization capability. For the evaluated datasets the best accuracy obtained for either problem is provided by a probabilistic neural network. Minimum classification error is 2.5% for OCR and 7.2% for fingerprints.

NIST Form-Based Handprint Recognition System

1

Fast implementations of nearest neighbor classifiers

Abstract Standard implementations of non-parametric classifiers have large computational requirements. Parzen classifiers use the distances of an unknown vector to all N prototype samples, and consequently exhibit O( N ) behavior in both memory and time. We describe four techniques for expediting the nearest neighbor methods: replacing the linear search with a new kd tree method, exhibiting approximately O (N 1 2 ) behavior; employing an L ∞ instead of L 2 distance metric; using variance-ordered features; and rejecting prototypes by evaluating distances in low dimensionality subspaces. We demonstrate that variance-ordered features yield significant efficiency gains over the same features linearly transformed to have uniform variance. We give results for a large OCR problem, but note that the techniques expedite recognition for arbitrary applications. Three of four techniques preserve recognition accuracy.

Micromagnetic structure of domains in Co/Pt multilayers. I. Investigations of wall structure

An analysis of the micromagnetic structure of domains and domain walls in Co/Pt multilayer films is reported. Magneto‐optically written domains have been imaged in a scanning transmission electron microscope by using the modified differential phase contrast mode of Lorentz electron microscopy. These have been compared with computer‐simulated images based on a two‐dimensional model of a circular, perpendicular magnetized domain with a Bloch‐like wall structure. Agreement is found for the domain and stray field contrast, but the absence of wall contrast in the experimental images indicates a more complex wall structure in the multilayer than was assumed by the model. In a further series of calculations the magnetic microstructure of a Co/Pt multilayer was modeled by solving the Landau–Lifshitz–Gilbert equations. These suggest that the wall structure varies throughout the thickness of the multilayer, allowing significant saving of magnetostatic energy through the establishment of flux closure paths close to ...

Disappearance of impurity levels in silicon and germanium due to screening

We have studied the disappearance of impurity levels in silicon and germanium due to free‐carrier screening of the Coulomb field of the impurity ions. The ground‐state eigenfunctions and eigenvalues have been calculated for electrons described by an ellipsoidal effective‐mass Hamiltonian. A two‐dimensional finite‐element analysis was used to obtain the solutions. Only moderate carrier densities (1019 cm−3 for silicon and 1018 cm−3 for germanium) are needed to cause the impurity levels to disappear into the conduction band, the result at high doping densities being simply a degenerate semiconductor.

Training dynamics and neural network performance

We use an analysis of a simple model of recurrent network dynamics to gain qualitative insights into the training dynamics of feedforward multilayer perceptrons (MLPs) used for classification. These insights suggest changes to the training methods used for MLPs that improve network performance significantly. In previous work, the probabilistic neural network (PNN) was shown to provide better zero-reject error performance on character and fingerprint classification problems than radial basis function and MLP-based neural network methods. We will show that performance equal to or better than PNN can be achieved with a single three-layer MLP by making fundamental changes in the network optimization strategy. These changes are: 1) use of neuron activation functions, which reduce the probability of singular Jacobians; 2) use of successive regularization to constrain the volume of the minimized weight space; 3) use of Boltzmann pruning to constrain the dimension of the weight space; 4) use of Prior class probabilities to normalize all error calculations, so that statistically significant samples of rare but important classes can be included without distorting the error surface. All four of these changes are made in the inner loop of a conjugate gradient optimization iteration and are intended to simplify the training dynamics of the optimization. On handprinted digits and fingerprint classification problems these modifications improve error-reject performance by factors between 2 and 4, and reduce network size by 40-60%. Copyright 1997 Elsevier Science Ltd.

Two-dimensional analysis of semiconductor devices using general-purpose interactive PDE software

Analyzing currents and fields in VLSI devices requires solving three coupled nonlinear elliptic partial differential equations in two dimensions. Historically, these equations have been solved using a special-purpose program and batch runs on a large fast computer. We use a general-purpose program and interactive runs on a large minicomputer. We discuss the physical formulation of the semiconductor equations and give three example solutions: a short-channel MOSFET near punchthrough, a DMOS power transistor in the ON state, and a beveled p-n junction. These examples demonstrate that solutions to a very general class of semiconductor-device problems can be obtained using these methods.

High-accuracy physical modeling of submicrometer MOSFET's

When short-channel MOSFET transistor models are compared to experimental data, the uncertainty in some of the physical input variables often requires that some of the input variables be adjusted to fit the data. This uncertainty is increased by a lack of knowledge of process sensitivity information on critical parameters. These uncertainties have been eliminated using a two-dimensional finite-element model of a MOSFET with no free parameters. The model is compared to four self-aligned silicon-gate n-channel MOSFETs with channel lengths of 0.80, 1.83, 2.19, and 8.17 µm. The 0.80, 1.83, and 8.17-µm devices have phosphorus sources and drains. The 2.19-µm device has an arsenic source and drain. These devices span the range of channel lengths from a short-channel device, totally dominated by velocity saturation and source-drain profile shape, to a long-channel device, well characterized by a long-channel model. Using the data obtained from the measurements described in this work, it is possible to model the drain current for all of the transistors studied without adjustable parameters. Transistors with 0.80-µm channel length differ in model input from those with 8.17-µm channel length only in the length of the polysilicon gate. If sufficiently accurate parameters are available, these methods allow the characteristics of submicrometer transistors to be predicted with ±5-percent accuracy. These simulations show that the observed short-channel effects can be accounted for by existing mobility data and a simple empirical model of these data. Triode and saturation effects are dominated by two-dimensional drain field penetration of the channel region. Subthreshold effects are caused by distortion of fields in the entire channel region by the drain field.

Defect induced lowering of activation energies at step bands in Co/Cu(100)

Complex topological features such as rectangular voids and step inclusions that were seen in secondary electron micrographs of Co films grown on Cu(100) at room temperature were reproduced in Monte Carlo simulations in the presence of step bands. Lowered activation energies at defects such as steps, kinks, and vacancies enhance step edge restructuring during growth and upon annealing. This results in features such as faceted step edges, rectangular pits, incorporation of Co into terraces, surface alloying, and surface segregation. Simulated growth structures are directly compared with those observed in an ultrahigh vacuum scanning transmission electron microscope.