High dielectric constant (Ba, Sr) TiO₃ [BST] films have been deposited by liquid source chemical vapor deposition (CVD) method. The system consisted of a single-wafer, low-pressure thermal CVD reactor, a vaporizer for liquid source materials, and a shower-type gas nozzle head, giving stable BST film deposition on a 6-in. -diam substrate with uniform thickness and uniform chemical composition ratio. The source materials employed were Ba(DPM)₂, Sr(DPM)₂, and TiO(DPM)₂ dissolved in tetrahydrofuran (THF), resulting in a conformal step coverage of BST films at lowered substrate temperatures, where DPM denotes dipivaloylmethanato. Moreover, the two-step deposition technique was developed to restrain protrusions formed on BST film surfaces at low temperatures, where BST films consisted of a buffer layer and a main layer; the buffer layer was an about 60-Å-thick layer of CVD-BST films annealed in N₂ ambience. Thus, the two-step CVD deposition of BST films on Pt and Ru electrodes achieved an equivalent SiO₂ thickness of t_eq_??_0.5nm, a leakage current of J_L_??_1.0×10^-8A/cm² (at+1.1V) and a dielectric loss of tan δ_??_0.01 at a total film thickness of 250Å, along with a comformal coverage of 80% for a trench of an aspect ratio of 0.65. Then, for BST films deposited on patterned electrodes 0.24μm wide, 0.60μm deep, and 0.15μm high (each spaced by 0.14μm), the capacitance was demonstrated to be increased without a significant deterioration of the leakage current: the capacitance was measured to be increased in comparison with that for films on flat electrodes, by a factor corresponding to an increase in surface area due to sidewalls of storage-node-like pattern features. This capacitance increase reflects the most characteristic advantage of CVD, or an excellent step coverage on microscopic pattern features. These electrical properties obtained satisfy specifications of a capacitor for Gb-scale dynamic random access memories (DRAMS), giving a storage capacitance of more than 25fF/cell for a stacked capacitor having a storage node 0.2_??_0.3μm high.

抄録全体を表示

A new low temperature processing method to prepare SrBi₂Ta₂O₉ thin films is proposed. These thin films were prepared on Pt/Ta/SiO₂/Si substrates by a sol-gel method, and their structural and electrical properties were investigated. Films were annealed before and after the top Pt electrode deposition, respectively. The 1st annealing was performed in a 760Torr-oxygen ambient at 600°C for 30min., the 2nd annealing was performed in a 5Torr-oxygen ambient at 600°C for 30min.. The films were well crystallized and fine-grained after the 2nd annealing. The electrical characteristics of the 200-nm-thich film obtained by this new process, i.e., remanent polarization (Pr), coercive field (Ec), and the leakage current density (I_L), were as follows; Pr=8.5μC/cm², Ec=36kV/cm, I_L =1×10^-7A/cm² (at 150kV/cm). This process is very attractive for highly integrated ferroelectric nonvolatile memory applications.

抄録全体を表示

We have developed a new low temperature growth technique of Bi₄Ti₃O₁₂ thin films using MOCVD method, in which an ultra-thin double buffer layer (5nm-Bi₄Ti₃O₁₂/5nm-TiO₂) was used to control the crystallization and fine grain structure of Bi₄Ti₃O₁₂ thin films. The 100nm-Bi₄Ti₃O₁₂ thin films fabricated at 400°C showed an excellent smooth surface morphology and good electrical properties: large remanent polarization of Pr=11μC/cm², coercive field of Ec=90kV/cm and low leakage current IL=7×10^-9A/cm² at 3V, respectively. In addition, for the first time the fatigue free property, which is very important for nonvolatile ferroelectric memories (NVFRAMs) applications, was confirmed up to 1×10¹² switching cycles.

抄録全体を表示

Bismuth layer-structured ferroelectric SrBi₂Ta₂O₉ and Bi₄Ti₃O₁₂ thin films have been prepared on Pt and Si at low temperature of 400-550°C by laser ablation method using an ArF excimer laser. Crystallographic properties of these films deposited at various substrate temperatures, laser frequencies and oxygen gas pressures are characterized. Depth profile of X-ray photoelectron spectra reveales a homogeneous composition but Bi oxidation is not satisfactory. Electrical properties of Bi₄Ti₃O₁₂ films on SiO₂/Si such as J-V and C-T characteristics are good. C-V hysteresis loop for SrBi₂Ta₂O₉ and Bi₄Ti₃O₁₂ films on SiO₂/Si and D-E hysteresis loop of Bi₄Ti₃O₁₂ film on Pt and Si are obtained

抄録全体を表示

A one-mask-patterned ferroelectric capacitor test structures designed with a 0.5μtm feature size were fabricated. Oxygen plasma treatment after dry etching decreased the leakage current to as low as as-deposited film. The one-mask-patterned ferroelectric capacitors with switching charge almost equal to as-deposited film were achieved. Ferroelectric memories as dense as dynamic random access memories will become possible with this technology. However, the leakage current density for micron-size capacitors is as 2-10 times high as large capacitor. This current increase was found to be caused by perimeter leakage current through damaged layer on the PZT side-wall.

抄録全体を表示

In the first section of this paper, the author reports the measurment of a tensile strain distribution on field-installed OPGW (optical fiber composited over-head ground wire) using the method of Brillouin optical fiber time domain analysis (BOTDA). The strain differences between strung sections and non-strung sections of OPGW are clealy differenciated and observed. The obtained differences are about 0.1% in the winter and about 0.2% in the summer. Those values are less than the allowable strain on the OPGW. An analysis of a relationship between measured strain and temperature is also described. It is found that the temperature dependency of strain data agrees quite well with a thermal expansion coefficient of OPGW.
In the second section, it is introduced a new strain measurement technique named “Ring Amplification Method”, which enables measurements of rapidly varying strains such as the one caused by galloping. A principle and a measurable distance by this method are studied, and it is clarified that this method is applicable to a measurement on a long-extended fiber up to 40km. Furthermore, the ability of measuring the time-varying strain and its distribution has been demonstrated in the basic experiments.

抄録全体を表示

The authors already proposed Chaos Control Method on Universal Learning Network (U. L. N.). The method can control chaotic phenomena on such U. L. N. that have bounded functional nodes. In this paper, the chaos control is extended to multi-branch U. L. N. for the purpose of investigating the influence of increasing the number of branches between nodes and changing delay times. Generation and die-out of chaotic phenomena can be realized by changing Lyapunov Number of U. L. N. to a positive and negative value respectively by U. L. N. parameter learning. The parameter learning is carried out by using a gradient method with a newly devised calculation algorithm of second order derivatives of U. L. N. and a random search method which stabilizes the learning procedure. In the simulations of multi-branch U. L. N., the results show that, in the case of generating chaotic phenomena, the Lyapunov Number can be more quickly changed to the positive desired value on the multi-branch U. L. N. than on the single-branch U. L. N., that is, the ordinary recurrent neural network. The results also show that the larger the time delay is, the more difficult changing the Lyapunov Number to the desired positive value is, in other words, generating chaotic phenomena becomes more difficult. Based on the simulation results of the single-branch U. L. N., namely the ordinary recurrent neural network and the multi-branch U. L. N., suitable U. L. N. parameters and structures have been made clear that can realize generation and die-out of chaotic phenomena more efficiently.

抄録全体を表示

One of the serious problems of neural networks is that the number of weights increases explosively as size of the network becomes larger. A lot of studies have been done for the multilayer perceptron type networks. However, there is no such a study for neural network associative memories. This paper proposes two effective algorithms to reduce the number of weights in BAM (Bidirectional Associative Memory): one is a simple algorithm which does not require too much additional computations; the other is designed to reduce the weights as many as possible. Both of the algorithms can guarantee the recall of the stored data. Computer simulations are carried out to confirm the validity of the proposed algorithms and to examine the performance.

抄録全体を表示

The extended Kalman filter (EKF) is an approximate filter for nonlinear systems. The EKF is well used for the adaptive filtering problems which identify both the system states and parameters for linear systems. For nonlinear systems, the adaptive filer could not work well because the linear filter could not identify the nonliear characteristics. In this paper, we propose a new adaptive filter for nonlinear systems using a neural network based on the EKF. First, the linear filter with EKF is extended to an approximately linearized adaptive filter for the nonlinear systems based on first-order linearization. Next, a multi-layered neural network is applied to the linearized adaptive filter. A transition matrix and an ovservation matrix of the system are required in order to solve the filtering algorithms. We derive the elements of the transition matrix and the ovservation matirix using the functions of the neural network. Finally, simulation results show that the proposed filter has good performances for the state estimation problems of the nonlinear systems.

抄録全体を表示

In thermal power plants, it is an important operation to improve the control accuracy of main steam pressure and temperature etc. during load up/down. This paper focused on temperature control that is most difficult to control due to the nonlinearity and long dead time of power plants. We applied some control methods such as MRAC, neural network and long range predictive control to the power plant main control system. Each method was evaluated by a simulator using detailed physical models that represents accurately the dynamics of power plants. We confirmed that each method is possible to control properly, but long range predictive control is better than other two methods. It is also stated that the thermal power plant is so complex that some consideration (persistently exiting condition, learning method of neural networks etc.) is necessary for the application of theoretical algorithm.

抄録全体を表示

In this paper, we study the relation between the optimal deadbeat control signal and pulse transfor function. And we also proposed a design method of optimal deadbeat control system with stste feedback. The sequence of the optimal deadbeat control signal which makes the deadbeat response for step. ramp, or constant acceleration variation of the desired value can be easily obtained from the matrix computation by the step response of the system. Then, the pulse transfor function derived by using sequence of the deadbeat control signal. The method proposed here is very simple comparing with other methods. A numerical example is given to illustrate the effectiveness of the proposed method. Finally, mutual relation to the optimal deadbeat control signal, pulse transfor function, and compensation parameter of the optimal deadbeat servo-system are indicated to the appendix.

抄録全体を表示

In the control of actual plant, there commonly exits a constraint on the control input caused by mechanical constraint on the actuator or to protect the plant. This constraint usually exerts adverse effect on control performance called windup phenomenon. Anti-windup technique is known as one of the most effective techniques to prevent such windup phenomenon.
In this paper, we propose a new design of anti-windup controller based on state feedback H^∞ control theory. This design consists of representation of anti-windup controller that includes a free parameter and decision of the free parameter by using state feedback H^∞ control theory. Then, we also discuss the problem on discretization of anti-windup controller in implementation. In numerical case study, we demonstrate the performance of proposed anti-windup controller by computer simulation.

抄録全体を表示

This paper presents a new method for model reference adaptive control for SISO nonminimum phase continuous-time systems. The usual model reference adaptive techniques cannot be applied to nonminimum phase systems, because the boundedness of all signals is not kept in the model reference adaptive control system. This paper uses an approximate inverse system applicable to the nonminimum phase continuous-time system which assures that the system input and output remain bounded at any time. The least squares approximation is used to find these approximate inverse systems. Finally, the results of computer simulation are presented to illustrate the effectiveness of the proposed method.

抄録全体を表示

It is well known that infrastructure networks like telecommunications, freight and person trip play a very important role in the social activities and that the facilities of these networks constantly change in both quality and quantity. There have been developed some models to analyze their structures, they treated only the static data. This paper proposes an extension of the structure to deal with the dynamic interrelationship changes model, combining the spatial interaction model and multidimensional scaling model. This model employs the polar coordinate to clarify the dynamic convergence or divergence behavior among regions. In this paper, we applied it to the analyze telecommunications, freight, and person trip OD data. The mainfinding are as follows: 1. telecommunication shows single ring pattern, converging around Tokyo, 2. freight transportation network gives some clusters converging around Tokyo and Osaka, 3. person trip network indicates intermediate pattern between those of telecommunication and freight.

抄録全体を表示