National Institute of Advanced Industrial Science and Technology (AIST) This page is a page of the former research institute. We stopped updating on March 31.2001.
E-mail to webmaster (Japanese) E-mail to webmaster (English)

　
　8-2 Implementation of R&D Works

　Research themes currently being implemented at ETL are listed in the Appendix 22. For further details, refer to the electronic text posted at a Web site to be linked from the ETL home page.

　
　8-3. Super-Labs and Project Teams
　
　　8-3-1 New Information Processing (Real-World Intelligence) Superlab
　The Real World Computing (RWC) Project -- a 10 year national project on new information processing begun by Japanﾕs Ministry of International Trade and Industry (MITI) in 1992 -- was initiated with the establishment of the RWC Partnership. The project's objective is to realize flexible, real-world information processing. The ETL has played an important role in compiling RWC Project master plans and the ETL itself is conducting research supporting RWC Partnership R&D.

　In the RWC Project's first half decade (1992-1996), some 60 researchers from 13 sections of 6 ETL Divisions have focused on 9 research themes including rather domain-specific, parallel research in theoretical foundation, novel functions, massively parallel systems, and optical computing. A transversal research group was set up in 1993 to unify individual section research and extend cooperation across research fields.

　Research activities include the following:

　

(1) Multimodal Interaction (Dialogue)

Multimodal interaction involves computer systems equipped with a form of eyes, ears, and mouth. A virtual computer graphics (CG) agent with an expressive face displayed on a screen recognizes the user, and provides information services and supports users' intellectual tasks through spoken dialogue (Photo 1). This "electronic secretary" prototype represents the friendly human interfaces of the near future. By developing functions understanding users' expressions and gestures and expanding vocabulary, research is working toward realizing a real-world intelligent system able to learn and to support users by integrating multimodal information.

　(2) Autonomous Learning

Autonomous learning involves system equipped with movement and autonomous motion, e.g., in an office environment. The office-operable mobile robot JIJO2 (Photo 2), is the prototype of an intelligent mobile robot that will gather and provide information such as reception and guidance and replacing human personnel in tasks such as postal pickup and delivery. Currently, necessary, elementary functions are being integrated into a commercially available mobile robot (NOMAD); such as, understanding of real environment via multi-sensors, obstacle avoidance, map learning, navigation planning, and learning via conversation with nearby personnel.

　　

Photo.1"Electronic secvetary"　　　　　　Photo.2 Mobile Robot"JIJO2"

　(3) Common-Use Theories and Methods

Superlab researchers are examining problems in learning and inference using incomplete information considered from the probabilistic and statistical standpoint. Methods developed include neural network computing, genetic algorithms, and probabilistic constraint programming.

　(4) Computational Support

To support computation, ETL researchers concerned developed the parallel computer EM-X based on a new architecture, and began research on a flexible operating systems (OS) for estimating the degree of parallelism, load balancing, and scheduling. An event-driven language was also developed. Adaptive, evolvable hardware was developed based on neural and genetic algorithms, and neural networks and fuzzy inference were implemented optically.

　(5) RWC Project in the Second Half Decade

The interim RWC Project evaluation and reorganization in fiscal 1996 determined that the second half decade would devote R&D resources to parallel and distributed computing promoted by the RWCP Tsukuba Research Center, and to real-world intelligence promoted by the ETL. Coincidentally, the ETL had reorganization to abolish the conventional sections and to start more mission-oriented Labs. One of these was the 8- laboratory New Information Processing Superlab, or Real-World Intelligence (RWI) Center.

　(6) Real-World Intelligence R&D

The RWI field involves intensive cooperation between theoretical bases and actual application-domain R&D centering on information integration and learning/self-organization. RWI application domains are classified into multimodal information abstraction/organization, interface interaction and behavior. R&D focuses on basic, actual RWI problems and on verifying the feasibility of the novel functions developed. To support, accelerate, and evaluate R&D from hardware and software sides, adaptable devices such as evolvable hardware and intellectual resources, real-world databases, and software library must be developed compiling common elements (Fig. 20).

　The results of ETL R&D are complied in annual reports and distributed to interested institutes.

　

　　8-3-2 Thin-Film Silicon Solar Cell Superlab
　The Thin Film Silicon Solar Cell Superlab was set up to develop the high-quality, highly stable, high-growth-rate thin-film amorphous and microcrystalline silicon (a-Si:H and ｵc-Si:H) materials essential to fabricating thin-film silicon solar cells.

The Superlab's flexible research activities currently involve high-rate, high-quality film preparation; ｵc-Si:H growth and characterization; photoinduced degradation mechanism research; photo-induced degradation control; reaction diagnosis; and solar cell fabrication. Staffing includes six ETL members, eleven private-company researchers, eleven professors and doctoral university students, and two post-doctoral fellows.

Recent activities include the following:
　(1) Schottky-Cell I-V Characteristics Fill Factor and a-Si:H Thin-Film Growth Rate

A guiding principle for obtaining highly stable a-Si:H under high growth rates is to be proposed based on an examination of plasma reactive species for determining micro- and mesoscopic network structures in films (Fig. 21).

　(2) Plasma CVD ｵc-Si:H Growth Mechanism

A variety of surface reaction diagnostic techniques was used to obtain a one-to-one correspondence between the surface hydrogen-bonding configuration during growth and the crystallite lattice orientation in the resulting ｵc-Si:H. Based on this, Superlab researchers fabricated epitaxial-like film on a (100) single-crystal Si substrate at a mere 200 ｡C (Photo 3). This suggests that the important factor in controlling structural properties in ｵc-Si:H is selective, preferential growth of orientation-arranged nuclei and followed by epitaxial growth of crystallites.

　

(3) Photoinduced Degradation Mechanism

Superlab researchers recently measure the very small change in film volume after light soaking using a cantilever in a trial to clarify the mechanism underlying photoinduced defect generation. The film volume was found to be expanded slightly but strictly in prolonged light soaking and is restored by thermal annealing (Fig. 22).

Division researchers have found that deuterated amorphous silicon (a-Si:D) shows photoinduced degradation differing from conventional a-Si:H. Somewhat different degradation behavior was also observed in a-Si:H film (Fig. 23) prepared using newly developed method of defect reduction by energized precursors (DREP) to produce a situation similar to a-Si:D growth by accounting for the mass difference in surface-covering atoms and surface-diffusing reactive species. Exaggerating this growth condition is expected to eventually lead to stable a-Si:H fabrication.

　

(4) A-SiGe:H Surface Hydrogen Thermal Removal

Surface hydrogen bonded to Ge is completely removed within 30 min after a-SiGe:H growth, while surface hydrogen bonded to Si remains, suggesting that surface-bonded hydrogen does not hop to other sites on the film growth surface (Fig. 24).

　

Superlab researchers will continue studying the control of silicon thin-film micro- and mesoscopic structures based on a detailed understanding of reactive-plasma film-growth processes, developing a new concept in materials research -- process control through process diagnosis

　
　8-3-3 Cooperative Architecture Project Team

　The Cooperative Architecture project team was set up in June 1991 to study basic mechanisms for the New Models for Software Architecture Project under the Industrial Science and Technology Frontier Program, AIST, MITI. The project ends in 1998. Its purpose is to yield a new software methodology.

Recent activities include the following:

　

(1) New software Methodology

The Project Team is focusing on architecture that combines units, rather than on the intelligence of individual units. "Cooperation" is used to refer generally to mechanisms through which large systems are composed from smaller units, in contrast to broader connotations.

In conventional software architecture, programmers must list all possible situations to which the program may be exposed and specify all necessary instructions -- an assumption practically impossible to satisfy in the real world. The Project Team is developing a methodology in which (a) the programmer can prepare elemental descriptions without having to "re"-present all information, particularly that available in the environment, and (b) the system can automatically combine elements provided by programmers. This enables a complex, flexible system to be constructed out of elements.

The elements described above are not fixed, and may have to be reprogrammed due to changes in specifications or the environment. The target methodology of the project limits the scope of change locally and still enables reconstruction of a new combination using already existing elements whenever possible.

　

(2) Organic Programming

The Project Team has proposed an "organic programming" concept based on the dynamism -- dynamic structure changes -- and interaction -- among parts or between parts and a whole -- characterizing organic systems. In organic programming, programs are divided and stored in large numbers of units called cells that are then dynamically combined to form a large flexible system structure.

The above dynamic combination is expected to enable a system to be constructed whose capabilities are more than the sum of its parts due to their situation-dependent, flexible combination. In other words, components adapt to changes in their environment by interacting with each other and with the environment.

Program units must thus be constructed in a situation-oriented way and be combined in a situation-dependent manner.

　

(3) Gaea Programming Language

The Project Team developed a programming language, Gaea, based on organic programming and made the system available to other researchers through a web page (http'//cape.etl.go.jp/gaea) Gaea extends the logic programming language Prolog with its cell structure, and with parallel programming based on multithreads. It is implemented on Euslisp, also developed at the ETL. Gaea runs on Solaris and Linux.

　

(4) Soccer-Game Benchmark Problem

The Project Team proposed a soccer game as a benchmark problem for multiagent systems and provided a soccer server used in RoboCup. The problem is acknowledged as a new standard problem in artificial intelligence. This organic programming design was also tested in programming soccer players, with a "dynamic subsumption architecture" proposed to describe complex action. (For details of the achievements, see 8-4-9.)

　
　8-3-4. Femtosecond Technology Project Team

　The MITI-proposed 1995 Femtosecond Technology Project was to develop fundamental technology required for realizing ultrafast optoelectronics that overcome speed limitations on conventional electronics.

　The Project's purpose was to extend the electromagnetic wave frequency range handled to the visible-light frequency. Using the physical properties of light as an information medium is expected to provide an innovative technological base that handles symbolic information in communication and computing as well as direct information from the environment and physical and chemical materials properties.

　The Femtosecond Technology Project Team is in partial charge of the central project involving researchers from the Optoelectronics, Electron Devices, and Materials Science Divisions. Laser technology, optoelectronics, and materials technology were combined to set up femtosecond superelectronics in which photon-electron phenomena in an ultrafast time region can be controlled intentionally by clarifying photon-electron interaction and finding new device principles in the femtosecond region. The project is conducted in close cooperation with the Femtosecond Technology Association (FESTA).

Recent activities include the following:

　

(1) Laser Technology

　The Project's purpose is to develop technology for pulse generation in the time region from hundreds of femtoseconds to a few femtoseconds (monocycles), precise pulse control, and to evaluate pulse characteristics. A synchronized system of ultrashort pulse lasers and an ultrashort pulse electron-beam accelerator is used for producing hard X-ray pulses to monitor power plant facilities. Ultrashort optical and electron pulses must be generated, stabilized and synchronized.

　Femtosecond self-start mode-locked ultrashort pulse lasers have been developed in the ~1.3 ｵm wavelength region using a Cr-doped forsterite crystal and a newly-developed semiconductor mirror that includes quantum-well-saturable surface absorber layers. The semiconductor saturable observer mirror (SESAM) produces a robust femtosecond pulse laser. Intracavity dispersion compensation was improved to generate 20 fs pulses -- the shortest in the wavelength region exceeding 1.0 ｵm.

　A procedure has been proposed to design dispersion-controlled mirrors (spectrum-inverse method), realizing a wideband dispersion compensation mirror whose group dispersion is -500 fs2. Compensation bandwidths are 40 nm for AlGaAs multilayers and 80 nm for Ta2O5/SiO2 multilayers. It is anticipated that compact femtosecond lasers will be able to be fabricated using these mirrors.

　A novel method proposed for measuring the mode-locked laser pulses timing fluctuation by phase-demodulating the pulse intensity has a larger dynamic range and a wider frequency span than conventional methods. The effectiveness of the method was demonstrated by measuring the timing fluctuation of a laser-diode-pumped Cr:LiSAF laser in the Fourier frequency from 50 mHz to 1 MHz. The all-solid-state self-mode-locked laser generated 26-fs (6 cycle) pulses at 1.3 ｵm as shown by the fringe-resolved autocorrelation trace in Photo 4. Photo 5 shows the autorecovery behavior of a mode-locked pulse train in the laser oscillator with an intracavity chopper. Our semiconductor saturable absorber realizes mode-locking self-starting.

(2) Optoelectronics

　The fusion of femtosecond laser technology and ultrafast electronics is expected to enable the fabrication of ultrafast optoelectronic devices. In ultrafast electronics, ultrafast electronic pulses are generated over a coplanar transmission line using ultrafast photoconductive switches (Fig. 25).

　Project Team researchers recently generated 380-fs-wide 4-V-high electrical pulses (Fig. 26) on a transmission line by using a narrow-gap (0.1 ｵm) photoconductive switch fabricated using nanometer-scale anodization of a thin Ti film with an AFM probe. This is, to our knowledge, one of the shortest electrical pulses generated in a circuit. The pulse height is sufficient for many applications.

　An electro-optical sampling system with a dye-laser light source was developed to measure such short pulses. A new electro-optical sampling technique -- electro-optic vector sampling -- also developed by Project Team researchers, independently and accurately measures longitudinal and transverse electric fields.

　Ultrafast photon-electron interaction in carrier transport and other semiconductor phenomena are being studied to eventually realize ultrafast optoelectronic devices with ultrashort optical pulse, ultrafast generation, transmission, control, and measurement.

　Two-dimensional electron flow in GaAs/AlGaAs InGaAs/GaAs QWs and wires has been visualized with 90-ps time resolution. The measurement system, developed by combining a microphotoluminescence measurement system and an ultrafast electronic shutter camera, is a useful tool for understanding carrier transport in electronic materials, for clarifying femtosecond optoelectronic device operation, and for designing and evaluating such devices. (For the details of the achievement, see 8-4-4.)

　(3) Fundamental Femtosecond Optoelectronics Materials

　To make breakthroughs in innovative ultrafast optoelectronic devices, new types of ultrafast phenomena at microscopic levels in materials must be found and their controllability determined. This is necessary for long-range projects to be successful. In materials technology, optoelectronic materials with mesoscopic structures are being used as present prototypes and ultrafast electron and photon phenomena in femtosecond time domains are being studied and evaluated. They are expected to contribute to the development of ultrafast optoelectronic devices and to open new ways toward "state-to-state electronics," which we propose as a leading concept in realizing our objectives. Coherence and its control have a significant role in this development.

　Project Team researchers have developed a way to fabricate MQW structures using II-VI diluted semimagnetic semiconductors, i.e., CdTe/Cd1-xMnxTe. The use of these structures has revealed many new properties for ultrafast relaxation phenomena of localized spin and photo-excited carrier interaction. The new ways to measure these phenomena at low temperatures and/or in high magnetic fields are also being developed. The ultimate purpose of these activities is to develop ultrafast new functional optoelectronic spin devices.

　Signal propagation properties are important to future ultrafast devices. Project Team researchers are studying these properties, and have demonstrated the generation and propagation of THz radiation -- phonon-polaritons -- in electro-optical crystal LiTaO3 -- by using transient method with asymmetric optical pulse arms of ~100 fs. They have also developed the basics of novel and direct ways to evaluate the dielectric response of electro-optical materials in the THz region.

　To detect and evaluate ultrafast properties of mesoscopic structural materials, which include functional molecules, polymers, their clusters, and quantum wires and/or dots, they are also developing a "femtosecond" scanning near-field microscope (fs-SNOM). Fabrication techniques are now being developed and optimized for SNOM tips. Extension of the observation wavelength and temperature region toward cryogenic temperatures is now in progress. Ways to measure ultrashort, ultraweak light pulses are also needed in developing fs-SNOM.