This paper introduces pass-transistor logic design with dual-threshold voltages. A set of single-rail, fully restored, pass-transistor gates are presented. Logic transistors are implemented with low threshold voltages and signal restoration transistors with high threshold voltages. Simulation is used to characterize the leakage power consumption, switching energy, and propagation delay of the proposed gates. A method to reduce circuit power by selectively replacing CMOS gates with the proposed gates is discussed and applied to the ISCAS'85 benchmark circuits. Relative to circuits composed entirely of conventional CMOS gates, use of the proposed SDPL gates achieves up to 49% reduction in leakage power and up to 63% reduction in total power consumption.

We evaluate the effectiveness of dual-Vt design in the presence of both subthreshold leakage and leakage due to gate oxide tunneling. At the device level, we use detailed HSPICE simulation to investigate the total leakage impact of three methods of dual-Vt implementation: multiple channel doping, channel length, and oxide thickness. At the system level, we generate and characterize a standard cell library and apply three representative delay-constrained leakage minimization dual-Vt assignment algorithms to the ISCAS'85 combinational benchmark circuits. Results show that oxide thickness modulation effectively reduces total leakage power consumption, but channel doping and channel length modulation are less effective.

A spectral-domain method is described for solving Schrodinger's equation based on the multidomain pseudospectral method and boundary patching. The computational domain is first divided into nonoverlapping subdomains. Using the Chebyshev polynomials to represent the unknown wave function in each subdomain, the spatial derivatives are calculated with a spectral accuracy at the Chebyshev collocation points. Boundary conditions at the subdomain interfaces are then enforced to ensure the global accuracy. Numerical results demonstrate that this spectral-domain method has an exponential accuracy and is flexible, and thus is an attractive method for large-scale problems. With only about four cells per wavelength, the results have an error less than 1% in our typical examples. For a typical quantum well, the method is about 51 and 295 times faster than the second-order finite-difference method for 1% and 0.1% accuracy, respectively. The spectral grid method has also been validated by results obtained by the finite-element method, semianalytical (Airy function) method, and the Numerov's method.

A technique for the determination of two-dimensional impurity profiles in silicon using methods for emission computed tomography is presented. Several one-dimensional impurity profiles obtained for different directions through the sample are used to reconstruct the two-dimensional profile. A simulation study of the experiment is described, and effects of various experimental and reconstruction parameters are discussed. Reconstructions of an area of 4 mu m*4 mu m from thirteen one-dimensional measurements, with a resolution of 1000 AA, are numerically possible.>