CAMD (Center for Advanced Microstructures and Devices) utilizes the radiation from a 1.2-1.5GeV electron storage ring with a critical-wavelength range 9.5-4.8Å. The storage ring at CAMD provides suitable radiation for X-ray micro-lithography and microfabrication work. High resolution, thin and thick resist processing techniques, including resist application, exposure and development were explored. New techniques such as transfer mask and equivalent filtration are introduced and demonstrated. Issues of latent image formation in exposed resist are addressed. Examples of structures fabricated using described techniques are presented.
This paper is devoted to the description of the LIGA process using a 600MeV superconducting compact synchrotron light source. The realization of deep-etch x-ray lithography is based on a new resist and mask. The resist is made of a copolymer of methyl methacrylate (MMA) and methacryl acid (MAA). The main benefit is its high sensitivity, which is one order of magnitude greater than that of polymethyl methacrylate (PMMA) used in the LIGA process. The mask is composed of a 2μm -thick silicon nitride membrane with high transparency supporting tungsten absorber which is a 5μm-thick. Experimental results about deep-etch x-ray lithography, electroforming and molding techniques are presented. Micro-ultrasonic transmitter obtained with these techniques is also shown. The purpose of this study is the realization of low cost micro-components for a variety of industrial applications.
Micromachining of polytetrafluoroethylen (PTFE) and formation of crystalline thin PTFE films were carried out by direct synchrotron irradiation of a PTFE target in vacuum. The microstructures made in PTFE had the smallest surface detail down to less than ten microns and the largest structural height up to thousand microns. The maximum aspect-ratio was more than 20. The PTFE films which can preserve the chemical compositions, crystalline features and surface morphologies of the starting material, could be produced at low substrate temperatures with high deposition rates (200 Å/s). In situ mass spectrometric analysis of gaseous species evolved during the irradiation shows that, different from the laser ablation of PTFE for the deposition, the SR-induced reactions should be ablative photochemical decomposition (APD) rather than photothermal unzipping, yielding saturated fluorocarbons rather than monomers as the main gaseous products. Detection of a trace of CF3 in the deposited films by XPS and FTIR was consistent with our APD mechanism.
It is known that infrared reflection absorption spectroscopy (IRAS) on semiconductor or insulator surfaces becomes practicable by using buried metal layer (BML) substrates, in which the metal thin film is buried under semiconductor or insulator films. In this work, IRAS has been measured for Langmuir-Blodgett films deposited on the BML substrate with SiO2/Al/Si(100) structure and the observed spectrum intensity has been quantitatively compared with the calculation assuming the ideal multilayer structure for the BML Substrate. The BML-IRAS using CoSi2 has been adopted to the detection of SiHn on the Si(100) substrate during synchrotron radiation (SR) stimulated Si2H6 gas source molecular beam epitaxy. It has been found that SiH2 and SiH3 on the Si(100) surface are easily decomposed by SR, but SiH can't be decomposed. From these experiments, it has been concluded that the BML-IRAS is an useful in situ observation technique for the photo-stimulated surface reactions.
As supply voltage is reduced, a speed superiority of BiCMOS to CMOS can be diminished but BiCMOS even has such an advantage as inducing relatively smaller characteristics degradation on nMOS transistor by supressing the drain voltage in the gate. Introducing a new quantitative methodology to evaluate hot-electron induced characteristics degradation of nMOS transistor, this report expects that both BiNMOS gate and CBiCMOS gate will have a considerable domination to CMOS gate within the voltage rangs from 2.5 volts to 3.3 volts even in the same nMOS transistor characteristics. BiNMOS has 30% faster speed than CMOS only by the effect of amplification of pMOS drain current by a npn transistor. CBiCMOS has 40% faster speed and seven oders of magnitude longer life time than CMOS. The speed improvement comes from amplification of both pMOS and nMOS drain current by npn and pnp transistor, respectivery, and the life time improvement is due to voltage drop effect through pnp transistors on the drain terminal of nMOS transistor. The analytical methodology was also utilized to choose an optimum drain structure. though CMOS necessitate LDD (Lightly Doped Drain), SD (Single Drain) structure, so that the speed of the gates can be further improved and the speed of CBiCMOS is expected to be 45% faster than the CMOS gate. Even in the case of SD structure, the life time of CBiCMOS was estimated to be two orders of magnitude longer than that of CMOS with LDD structure.
Lifetime of solid-state lasers is limited by the Kr arc lamps for optical pumping. The causes to shorten the lifetime of lamp have been investigated. Consequently, a new lamp with long lifetime was developed by adopting a cathode with tapered structure which functioned as a thermionic emission type. The lifetime of new lamp is about 3-10 times compared with commercially available lamps.
The huge number of neurons in vivo form the complex neural networks. The complexity prevents us from understanding the functions of nerve cells. So, our approach is to take the solitary neurons in vitro as a fundamental unit. And we are groping for interface substrates to communicate the neurons. The object neurons of our study are sensory nerve cells dissociated from vertebrate tissue. Using the proteolytic enzyme, we isolated and identifie_??_ visual cells, bipolar cells and horizontal cells from a carp retina, olfactory receptor cells from a newt olfactory epithelium. Their reactions were recorded by whole-cell mode of patch clamp methods. It is one of techniques used in the field of electrophysiology. A pair of ITO thin-film electrodes was fabricated by using a conventional photolithography to stimulate the neurons. The reacted cell signals were successfully recorded after the extracellular stimulation by the thin film electrodes. The effectiveness of newly fabricated measurement system was confirmed experimentally in the field of Functional Electrical Stimulation (FES).
A simple neuron circuit which has a chaotic firing property similar to that of a neuron is proposed. The design of the circuit is based on a mathematical neuron model which we previously discussed. The model was developed by modifying the McCulloch-Pitts and Nagumo-Sato models. In the modifications, nonlinear functions of sigmoidal and notchlike shapes, which are made up of exponential functions, were introduced. The shapes of these functions are crucial for the model to have the neuronal chaotic firing property. In the present study, the functions are formed by nonlinear function generators with nonlinear properties of the diode and transistor. Since the circuit is simple, its design may enable the integration of a large number of neuron circuits on an IC. The use of the circuit will have the advantages of high-speed and parallel processing which are not provided when using a mathematical model which requires the use of a computer.
This paper considers the Linear-Quadratic-Gaussian (LQG) problem for nonstandard singularly perturbed systems making use of the recursive technique. In order to obtain the optimal control law, we must solve the generalized algebraic Riccati equation. Using the recursive technique, we show that the solution of the generalized algebraic Riccati equation converges with the rate of convergence of O(ε). The existence of a bounded solution of error term can be proved by the implicit function theorem. As a result, the solution of LQG problem for nonstandard singularly perturbed systems can be obtained with an accuracy of O(εk).
Impedance control is one of the most effective control method for a manipulator in contact with its environment. The method, however, cannot control the end-effector impedance perfectly without an accurate model of the manipulator. Also, unexpected external disturbances are often applied to the manipulator, so that errors between a target impedance and a realized one may arise. The present paper proposes an impedance control of a manipulator in the presence of modeling error, and validity of the proposed method is verified through experiments using a direct-drive robot.
This paper proposes an intuition-based agent model in an Intelligent Industrial System(IIS). This is a model of an agent that cooperates human beings or the outside world. It has three-layered structure of knowledge. The upper layer has abstract knowledge as macro level. The middle one has concrete knowledge as micro level. The lower one has a local feed back, which consists of a feature extractor and action generator, by which communicates the outside world. The model has three functions since it is easy to recognize all knowledge in the model and the model processes knowledge top-down and bottom-up between the upper and the middle layer. (1) An interface function that communicates human beings or the outside world using images and sounds. (2) A learning function using function (1). (3) An adaptive function that corresponds to environmental changes using creativity. Then, this paper shows the following two experimental systems to achieve the above functions. (a) A gesture instruction learning system in which a mobile robot learns a trajectory using the qualitative sense inherent in a human macro qualitative instruction. (b) A self-organizing system based on chaotic dynamics in which a mobile robot creates a new knowledge adapted to an environmental change and learn one autonomously. Some experimental results show the effectiveness of the agent model.
This paper presents implementation of Artificial Neural Network Controller (ANNC) for induction motor starting, to track the reference condition scheme. To achieve mapping of the desired input/output we combine generally known method for control, the model reference adaptive system and the learning algoritlmm. The reference data as tracking trajectory computed from the model reference adaptive of induction motor starting (MRAIMS). A supervised learning neural network algorithm is used to return the induction motor starting speed (rotation starting) disturbance condition caused by the varies of load, back to the starting speed reference condition. The ANNC continuously monitors the status of the system parameter and the performance of motor starting, and the ANNC output is used as the current control signal of induction motor starting, because control of starting current, will improve the behavior of starting speed.