This paper dicusses the emitter-resistance (R_E) effect on the cutoff frequency (f_T) of silicon heteroemitter bipolar transistors (Si HBTs) by using two-dimensional device simulation, circuit simulation and small-signal analysis. It is shown that an approximate formula f_T=f_T0/(1+2πf_T0R_EC_C) agrees well with the device and circuit simulation results. The emitter resistance affects the cutoff frequency greatly. The conductance modulation in lightly-doped heteroemitter layers with a heavily-doped backup layer is also analyzed. It is found that the carrier injection from the heavily-doped backup layer into the lightly-doped heteroemitter layer reduces the emitter resistance to one-fourth the value at best. The permissible lower limit of electron mobility in heteroemitter is estimated roughly to be 2 cm²v^-1s^-1.

This paper describes the dynamic behavior of a trench capacitor dRAM cell, named the SCC (Surrounded high-Capacitor Cell). A transient three-dimensional simulation reveals the raising of the substrate potential and the leakage current in the low-to-high state operation. The simulation results are verified by the experimental data using a test device. The characterization of this phenomenon allows design consideration of the scaled SCC.

This paper describes a new simulation-based design methodology for process and device development of submicron MOS VLSIs. The main purpose of this work is to bridge the gap between simulations and actual experimental data through a transformation of the RSF (Response Surface Function) which determines a quadratic relationship between measured device characteristics and process conditions. To achieve the reliable RSF, we have developed two key techniques: (1) Transformation of variable and response before application of the RSM (Response Surface Method) design by simulations. (2) Improvement of the RSF which is determined by process and device simulations, by using the measured data of MOSFET characteristics. The new design methodology is applied to obtain RSFs having good device threshold voltage and maximum drain current. MOS devices fabricated with an experimental 0.8 µm technology are utilized to verify the results. The device parameters are 11-20 nm for the gate oxide thickness, and 0.8-4.0 µm for the gate length. The averaged RMS errors between the obtained RSF and the experimental data are 0.02 V for the threshold voltage and 1.46 % for the maximum drain current. A quantitative explanation on the effect of the transformation technique is given.

Efficient device simulation methods to realize circuit level analyses have been proposed. Highly conductive wire regions, the current source, and lumped elements such as resistances and capacitances can be handled within the framework of a present device simulation software. Additionally, the simulation area can be reduced and computational efforts can be saved by adopting the inner Neumann boundary condition between devices which affect each other very little except through the wire region. The regularity of the coefficient matrix can be preserved in all cases, so it is possible to get stability of matrix solver and save computer memory. The advantage of matrix regularity is enhanced to speed up when using a vector-concurrent computer. Hence, the computational cost can be saved considerably. These numerical techniques have been implemented in the authors' device simulation software.

Low-temperature MOSFET is a promising device for future high-speed VLSI. We have developed a three-dimensional device simulator which can be used for the analysis of low-temperature deep-submicron MOSFET's. In order to improve the convergence property, the method of physical limiting on increment (PLI) was suggested. Two types of PLI, i.e., the limiting on potential increment (LPI) and the limiting on carrie-concentration increment (LCI) were showed to be very simple and effective methods for both 300 K and 77 K. Using the simulated results of COLD3, we showed the threshold variation in a low-temperature MOSFET due to the narrow channel effect can be suppressed if the device is designed according to the temperature scaling law.

The nonstationary electron transports in the BP-SAINT GaAs MESFETs with submicron gate lengths have been studied under the various conditions at 300 K by an ensemble Monte Carlo simulation. It is shown that the calculated drain currents fairly agree with the experiments. It has been found that Wang's effective saturation velocity of electrons, which is defined as 0.8 V_max, depends not only on the gate length, but alsos on the gate voltage, the drain voltage, and the doping concentration of the channel, where V_max is the maximum velocity in the channel. The dependence of the effective saturation velocity upon a gradient of the electric field is discussed. The spatial distributions of the valley population ratio (or the effective mass for the effective-single valley model) and the kinetic energy of electrons under the gate are studied by the ensemble Monte Carlo method to evaluate the validity of the relaxation time approximation for these device simulations. It is shown that the effective mass cannot be always specified as a fuction of the mean energy only and that the kinetic energy is not negligibly small around the middle of the channel compared with the thermal energy.

We investigated the limit of applicability of the impact ionization model as a function of the local electric field by using of the Monte Carlo technique. We found out that the deviation between the Monte Carlo result and estimation of the model of the local electric field becomes significant where the field-gradient is high and for our electric field profile it is about 1.0 10¹⁰ V/cm². We also calculated the substrate current by the regional Monte Carlo calculation. We proposed an absorbent boundary at the end of the calculation region to reduce the computational time. Computational time is greatly reduced and the result is the same as that calculated by the conventional boundary.

A three-dimensional simulation program of oblique rotating ion implantation using the Monte Carlo method has been newly developed to simulate the N^- and N⁺ drain formation of the gate/N^- overlapped LDD MOSFET's. The binary scattering approximation is used for nuclear scattering, and the Lindhard-Scharff formula and the Bethe-Bloch formula are used for electronic scattering. The azimuth of the ion is initialized by a random number uniformly distributed between 0 and 2π to express the wafer rotation. The topography of the MOSFET's is approximately expressed in algebraic form to obtain effectively the touchdown points of ion particles on the target surface. The vectorized Monte Carlo method is used to reduce the CPU time. The simulation provides the two-dimensional distribution of the dopant and the Frenkel pairs (vacancy-interstitial), using the Kinchin-Pease equation. From the results of the calculation, it appears that the overlap length, which is defined as the distance between the polysilicon gate edge and the intersection of the 10⁴/cm (equi-concentration/doseline) on the silicon surface, increases in accordance with the increase of the incidence angle of the ion beam, and it extends to 0.1 µm when 40-keV of phosphorus is implanted with an incidence angle of 60. It also appears that the concentration of the Frenkel pairs becomes lower in accordance with an increase in the incidence angle of the ion beam. The simulation also reveals that the effect of a shadowed drain region caused by the polysilicon gate is enlarged in accordance with the increase in the incidence angle, especially in the case of an incidence angle of 60, when the shadowed N⁺ drain region extends to the point 0.6 µm from the edge of the sidewall which is 0.35 µm in height.

A two-dimensional process simulator using the coordinate transformation method has been developed. Four transformation methods are compared by solving a diffusion equation; that is (1) boundary-fitted coordinate transformation solving the Poisson's equations with an orthogonal boundary condition, (2) boundary-fitted coordinate transformation solving the Poisson's equations with a fixed boundary condition, (3) vertical coordinate transformation, and (4) parameter mapping coordinate transformation. It has been found that the total calculation time to transform the coordinates and to solve the diffusion equation is very small by the last two methods (3) and (4).

A new simulation system, which is composed of two-dimensional process simulator and inference engine, has been developed as one of the prototype systems for analysing process tolerance and fluctuation in individual process parameters. A general concept of the system, algorithm of process fluctuation analysis and inference engine of AI software package are presented. Typical evidences of results obtained are also demonstrated.

Based on the loss behavior of optical fibers coated with high and low thermal expansion coefficient (α) secondary jackets in the -180 to 20 temperature range, the mechanism of loss increase has been discussed. The low α liquid-crystalline-polyester(LCP)-jacketed fiber shows the microbending loss of 5 dB/km at -180, while the high α nylon-jacketed fiber exhibits the maximum microbending loss of about 30 dB/km at -120 and then shows the decrease in loss with decreasing temperature. The relatively low loss increase in the LCP-jacketed fiber arises from crystallization of the primary silicone and slight eccentricity of the LCP jacket. The unusual loss behavior in the nylon-jacketed fiber can be explained in terms of the arrangement of randomly distributed fiber curvature at -120, which is predicted by Yabuta's fiber-buckling model and supported by the occurrence of fiber-strain relaxation.

A wavelength selective filter using a vertical micro-cavity surface emitting laser is proposed. A 5 µm long GaInAsP/InP micro-cavity with 95% dielectric multilayer reflectors was fabricated. A wavelength bandwidth of less than 10 was obtained at 1.52 µm of operating wavelength. Also, polarization independent filtering characteristics are demonstrated. By introducing an amplification/wavelength-tuning section, a tunable filter with bandwidth of less than 1 is expected.

The first successful imaging by laser absorption computed tomography of in vitro specimens has been achieved by means of the Coherent Detection Imaging (CDI) method realized with the optical heterodyne detection technique and image reconstruction from back projection of the data obtained via optical absorption measurements in a parallel beam geometry.

All the elements of the direct-sum representation circuit matrix for E-plane symmetrical T-junctions are shown to be given by the stationary values in a variational principle. It is proved that the matrix is approximately singular.