Abstract : Photonic Integrated Circuits (PIC) are of tremendous interest for photonics system in order to reduce their power consumption, size, fabrication cost and improve their reliability of fiber optics linked discrete component architecture. However, unlike for microelectronics, in photonics different heterostructures are required depending on the type of device (laser sources, detectors, modulators, passive waveguides…). Therefore photonics integration needs a technology able to produce multiple bandgap energy wafers with a suitable final material quality in a reproducible manner and at a competitive cost: a technological challenge that has not been completely solved yet. Quantum Well Intermixing (QWI) is a post growth bandgap tuning process based on the localized and controlled modification of quantum well composition profile that aims to address these matters. UV laser induced QWI (UV-Laser-QWI) relies on high power excimer laser to introduce point defects near the heterostructure surface. By adjusting the laser beam shape, position, fluence and the number of pulse delivered, the different regions to be intermixed can be defined prior to a rapid thermal annealing step that will activate the point defects diffusion across the heterostructure and generate quantum well intermixing. UV-LaserQWI presents the consequent advantage of allowing the patterning of multiple bandgap regions without relying on photolithographic means, thus offering potentially larger versatility and time efficiency than other QWI processes. UV-Laser-QWI reproducibility was studied by processing samples from an InGaAs/InGaAsP/InP 5 quantum well heterostructure emitting at 1.55 µm. 217 different sites on 12 samples were processed with various laser doses. The quantum well intermixing generated was then characterized by room temperature photoluminescence (PL) mapping. Under those experimental conditions, UV-Laser-QWI was able to deliver heterostructures with a PL peak wavelength blue shift controlled within a +/- 15 % range up to 101.5nm. The annealing temperature proved to be the most critical parameter as the PL peak wavelength in the laser irradiated areas varied at the rate of 1.8 nm per degree Celsius. When processing a single wafer, thus limiting the annealing temperature variations, the bandgap tuned regions proved to be deliverable within ± 7.9%, hence establishing the potential of UV-Laser-QWI as a reproducible bandgap tuning solution. The UV-Laser-QWI was used to produce multiple bandgap wafers for the fabrication of broad spectrum superluminescent diodes (SLD). Multiple bandgap energy profiles were tested and their influence on the SLDs’ performances was measured. The most favorable bandgap modifications for the delivery of a very broadband emitting SLD were analyzed, as well as the ones to be considered for producing devices with a flat top shaped spectrum. The intermixed SLDs spectra reached full width at half maximum values of 100 nm for a relatively flattop spectrum which compare favorably with the ≈ 40nm of reference devices at equal power. The light-intensity characteristics of intermixed material made devices were very close to the ones of reference SLD made from as-grown material which let us think that the alteration of material quality by the intermixing process was extremely limited. These results demonstrated that the suitability of UV-Laser-QWI for concrete application to photonic devices fabrication. Finally, an alternative laser QWI technique was evaluated for SLD fabrication and compared to the UV laser based one. IR-RTA relies on the simultaneous use of two IR laser to anneal local region of a wafer: a 980 nm laser diode coupled to a pigtailed fiber for the wafer background heating and a 500 µm large beam TEM 00 Nd:YAG laser emitting at 1064 nm to anneal up to intermixing temperature a localized region of the wafer. The processed samples exhibited a 33 % spectral width increase of the spectrum compare to reference device at equal power of 1.5 mW. However, the PL intensity was decreased by up to 60 % in the intermixed regions and the experiments proved the difficulty to avoid these material degradations of material quality with IR-RTA. / Résumé : L’intégration de circuit photonique vise à réduire la consommation énergétique, la taille, le coût et les risques de panne des systèmes photoniques traditionnels faits de composants distincts connectés par fibre optique. Cependant, contrairement à la microélectronique, des hétérostructures spécifiques sont requises pour chaque composant : lasers, détecteurs, modulateurs, guides d’ondes… De cette constatation découle le besoin d’une technologie capable de produire des gaufres d’hétérostructures III/V de qualité à plusieurs énergies de gap, et ce de façon reproductible pour un coût compétitif. Aucune des techniques actuelles ne répond pour l’instant pleinement à tous ces impératifs. L’interdiffusion de puits quantique (IPQ) est un procédé post épitaxie basé sur la modification locale de la composition des puits quantiques. L’IPQ induite par laser UV (IPQ-UV) est basée sur l’utilisation de laser excimer (Argon-Fluor émettant à 193 nm ou Krypton-Fluor à 248 nm) pour introduire des défauts ponctuels à la surface de l’hétérostructure. En ajustant la taille du faisceau, sa position, son énergie ainsi que le nombre d’impulsions laser délivrées à la surface du matériau, on peut définir les régions à interdiffuser ainsi que leur futur degré d’interdiffusion. Un recuit de la gaufre active ensuite la diffusion des défauts et par conséquent l’interdiffusion du puits. L’IPQ-UV présente l’avantage considérable de se passer de photolithographie pour définir les zones de différentes énergies de gap, diminuant ainsi la durée et potentiellement le coût du procédé. La reproductibilité de l’IPQ-UV a été étudiée pour l’interdiffusion d’une structure à 5 puits quantiques d’InGaAs/InGaAsP/InP émettant à 1.55 µm. 217 régions sur 12 échantillons ont été irradiés par un laser KrF avec des nombres d’impulsion variables selon les sites et avec une densité d’énergie constante de 155 mJ/cm². Les modifications de la structure générée par ce traitement furent ensuite mesurées par cartographie en photoluminescence (PL) à température ambiante. L’analyse des données montra que l’IPQ-UV permet un contrôle du décalage vers le bleu du pic de PL à +/- 15 % jusqu’à 101.5nm. La température du recuit est apparue comme le paramètre crucial du procédé, puisque la longueur d’onde du pic de PL des zones interdiffusées varie de 1.8 nm par degré Celsius. En considérant les sites irradiés sur une seule gaufre, c’est à dire en s’affranchissant des variations de température entre deux recuits de notre système, la variation du pic de PL est contrôlable dans une plage de ± 7.9%. Ces résultats démontrent le potentiel de l’IPQ-UV en tant que procédé reproductible de production de gaufre à plusieurs énergies de gap. L’IPQ-UV a été utilisé pour la fabrication de diodes superluminescentes (DSLs). Différents type de structure à énergie de gap multiple ont été testés et leurs influences sur les performances spectrales des diodes évalués. Les spectres des DSLs faites de matériau interdiffusé ont atteint des largeurs à mi-hauteur dépassant les 100 nm (jusqu’à 132 nm), ce qui est une amélioration conséquente des ≈ 40nm des DSLs de référence à puissance égale. Les caractéristiques intensité–courant des DSLs interdiffusés furent mesurées comme étant très proches de celle des dispositifs de référence faits de matériau brut, ce qui suggère que l’IPQ-UV n’a pas ou très peu altéré la qualité du matériau initial. Ces résultats prouvent la capacité de l’IPQ-UV à être utilisé pour la fabrication de dispositifs photoniques. Une technique alternative d’IPQ par laser a été évaluée et comparée à l’IPQ-UV pour la fabrication de DSL. Le recuit rapide par laser IR est basé sur l’utilisation simultanée de deux lasers IR pour chauffer localement l’hétérostructure jusqu’à une température suffisante pour provoquer l’interdiffusion: une diode laser haute puissante émettant à 980 nanomètre couplée dans une fibre chauffe la face arrière de la gaufre sur une large surface à une température restant inférieure à celle requise pour provoquer l’interdiffusion et un laser Nd:YAG TEM 00 émettant à 1064 nm un faisceau de 500 µm de large provoque une élévation de température additionnelle localisée à la surface de l’échantillon, permettant ainsi l’interdiffusion de l’hétérostructure. Les dispositifs fabriqués ont montré une augmentation de 33 % de la largeur à mi-hauteur du spectre émis à puissance égale de 1.5 mW. Cependant, l’intensité du pic de PL dans les zones interdiffusées est diminuée de 60 %, suggérant une dégradation du matériau et la difficulté à produire un matériau de qualité satisfaisante.

Due to their inherent BW capacity, optical interconnect (OI) offers a means of replacement to BW limited copper as bottlenecks begin to appear within the various interconnect levels of electronics systems. Low-cost optically enabled printed circuit boards are a key milestone on many electronics roadmaps, e.g. iNEMI. Current OI solutions found in industry are based upon optical fibres and are capable of providing a suitable platform for inter-board applications especially on the backplane. However, to allow component assembly onto high BW interconnects, an integral requirement for intra-board applications, optically enabled printed circuit boards containing waveguides are essential. Major barriers to the deployment of optical printed circuit boards include the compatibility of the technique, the cost of acquiring OI and the optical power budget. The purpose of this PhD research programme is to explore suitable techniques to address these barriers, primarily by means of laser material processing using UV and IR source lasers namely 248 nm KrF Excimer, 355 nm UV Nd:YAG and 10.6 μm IR CO2. The use of these three main lasers, the trio of which dominates most PCB production assembly, provides underpinning drive for the deployment of this technology into the industry at a very low cost without the need for any additional system or system modification. It further provides trade-offs among the suitable candidates in terms of processing speed, cost and quality of waveguides that could be achieved. This thesis presents the context of the research and the underlying governing science, i.e. theoretical analysis, involving laser-matter interactions. Experimental investigation of thermal (or pyrolitic) and bond-breaking (or photolytic) nature of laser ablation was studied in relation to each of the chosen lasers with regression analysis used to explain the experimental results. Optimal parameters necessary for achieving minimum Heat Affected Zone (HAZ) and surface/wall roughness were explored, both of which are key to achieving low loss waveguides. While photochemical dominance - a function of wavelength and pulse duration - is desired in laser ablation of photopolymers, the author has been able to find out that photothermallyprocessed materials, for example at 10.6 μm, can also provide desirable waveguides. Although there are literature information detailing the effect of certain parameters such as fluence, pulse repetition rate, pulse duration and wavelength among others, in relation to the etch rate of different materials, the machining of new materials requires new data to be obtained. In fact various models are available to try to explain the laser-matter interaction in a mathematical way, but these cannot be taken universally as they are deficient to general applications. For this reason, experimental optimisation appears to be the logical way forward at this stage of the research and thus requiring material-system characterisation to be conducted for each case thereby forming an integral achievement of this research. In this work, laser ablation of a single-layer optical polymer (Truemode™) multimode waveguides were successfully demonstrated using the aforementioned chosen lasers, thus providing opportunities for rapid deployment of OI to the PCB manufacturing industry. Truemode™ was chosen as it provides a very low absorption loss value < 0.04 dB/cm at 850 nm datacom wavelength used for VSR interconnections - a key to optical power budget - and its compatibility with current PCB fabrication processes. A wet-Truemode™ formulation was used which required that optical polymer layer on an FR4 substrate be formed using spin coating and then UV-cured in a nitrogen oxygen-free chamber. Layer thickness, chiefly influenced by spinning speed and duration, was studied in order to meet the optical layer thickness requirement for multimode (typically > 9 μm) waveguides. Two alternative polymers, namely polysiloxane-based photopolymer (OE4140 and OE 4141) from Dow Corning and PMMA, were sparingly utilized at some point in the research, mainly during laser machining using UV Nd:YAG and CO2 lasers. While Excimer laser was widely considered for polymer waveguide due to its high quality potential, the successful fabrication at 10.6 μm IR and 355 nm UV wavelengths and at relatively low propagation loss at datacom wavelength of 850 nm (estimated to be < 1.5 dB/cm) were unprecedented. The author considered further reduction in the optical loss by looking at the effect of fluence, power, pulse repetition rate, speed and optical density on the achievable propagation but found no direct relationship between these parameters; it is therefore concluded that process optimisation is the best practice. In addition, a novel in-plane 45-degree coupling mirror fabrication using Excimer laser ablation was demonstrated for the first time, which was considered to be vital for communication between chips (or other suitable components) at board-level.

No description available.

No description available.

No description available.

Objetivo: Investigar as alterações biomecânicas da córnea de suínos induzidas pela confecção de lamelas pediculadas de diferentes espessuras pelo laser de femtossegundo. Métodos: Para a formação dos dois grupos, 12 olhos de porcos foram usados: lamelas pediculadas de 100 e de 300 micrômetros confeccionadas pelo laser de femtossegundo. Cada olho foi submetido aos seguintes exames, antes da criação das lamelas: topografia por rasterstereography, Ocular Response Analyzer (ORA), tomografia do segmento anterior por coerência óptica para a avaliação paquimétrica corneal e das lamelas criadas e sistema de velocidade de onda (SVO), que mede a velocidade de propagação de ondas acústicas entre dois transdutores posicionados na superfície corneal antes e imediatamente, após a feitura da lamela. O primerio passo foi desenhado para o estudo das diferenças em relação à histerese corneal, fator de resistência corneal, mudanças na curvatura e velocidade de propagação de onda acústica entre córneas com lamelas finas e espessas. Posteriormente, as lamelas foram amputadas, e as medidas do sistema de velocidade de onda foram repetidas. Resultados: A média de espessura das lamelas ± desviopadrão (DP) foi de 108,5±6,9 (8,5% da espessura total) e 307,8±11,5 m (22,9% da espessura total), para os grupos de lamelas finas e espessas, respectivamente (p< 0,001). Histerese corneal e o fator de resistência corneal não apresentaram diferença estatística, após a criação de lamelas finas (p = 0,81 e p = 0,62, respectivamente). Histerese corneal foi significantemente mais baixa, depois da confecção de lamelas mais espessas (8,0±1,0 para 5,1±1,5 mmHg para medidas pré e pós-operatórias, respectivamente, p = 0,003, diminuição de 36,25%) e fator de resistência corneal também mostrou significante diminuição nesse grupo, após o procedimento cirúrgico; valores médios pré e pós-operatórios de 8,2±1,6 e 4,1±2,5 mmHg respectivamente (p= 0,007), diminuição de 50%. A ceratometria média simulada apresentou maiores valores, após a confecção das lamelas mais espessas em relação ao pré-operatório (ceratometria pré e pós-operatória de 39,5±1 D e 45,9±1,2 D, respectivamente, p= 0,003). Para o grupo de lamelas finas, não houve diferença estatisticamente significante (ceratometria pré e pós-operatória de 40,6±0,6 D e 41,4±1,0 D, respectivamente, p=0,55). Em relação ao Sistema de Velocidade de Onda, após a criação das lamelas e sua amputação, houve diminuição da velocidade de propagação acústica, embora na maior parte das posições não fosse estatisticamente significante. Conclusão: Nas condições experimentais estabelecidas por este estudo, a criação de lamelas de maior espessura pareceu exercer efeito mais relevante sobre a biomecânica da córnea de suínos / Purpose: To study the impact of programmed flaps at two different thicknesses on the biomechanical properties of the swine corneas. Methods: Twelve pig eyes were enrolled in this study and were formed two groups: 100m and 300 m flaps performed with the femtosecond laser. Each eye had the following procedure before the flap creation: raster photograph topographic maps, Ocular Response Analyzer (ORA), Optical Coherence Tomography to measure the pachymetry and flap thickness and Surface Wave Velocity system which is a prototype system that measures sonic wave propagation time between two transducers positioned on the corneal surface before and after flap creation. This first step was designed to investigate the differences in respect to corneal hysteresis, corneal resistance factor, curvature change and ultrasonic wave propagation between the groups with thinner and thicker flaps. After this initial procedure, flap amputation was performed and new measurements with the surface wave velocity system were taken again. Results: Measured flap thicknesses averaged 108.5±6.9 (8.5% of the total cornea) and 307.8±11.5 m (22.9% of the total cornea) for thin and thick flap groups, respectively (p< 0.001). Hysteresis and corneal resistance factor did not change significantly after flap creation in the thin flap group (p = 0.81 and p = 0.62, respectively). With thicker flaps, both parameters decreased significantly from 8.0±1.0 to 5.1±1.5 mmHg (p=0.003, reduction of 36.25%) and from 8.2±1.6 to 4.1±2.5 mmHg, respectively (p = 0.007), reduction of 50%. Simulated keratometry values increased in the thick flap group (from 39.5±1 D to 45.9±1.2 D, p=0.003) after flap creation and not in the thin flap group (from 40.6±0.6D to 41.4±1.0D, p= 0.55). Regarding surface wave velocity analysis, the surgical procedures induced lower values in some positions although most of them did not present statistically different results. Conclusion: In this experimental model, thicker flaps seemed to have a greater effect on the biomechanics of the swinish cornea

Objetivo: Investigar as alterações biomecânicas da córnea de suínos induzidas pela confecção de lamelas pediculadas de diferentes espessuras pelo laser de femtossegundo. Métodos: Para a formação dos dois grupos, 12 olhos de porcos foram usados: lamelas pediculadas de 100 e de 300 micrômetros confeccionadas pelo laser de femtossegundo. Cada olho foi submetido aos seguintes exames, antes da criação das lamelas: topografia por rasterstereography, Ocular Response Analyzer (ORA), tomografia do segmento anterior por coerência óptica para a avaliação paquimétrica corneal e das lamelas criadas e sistema de velocidade de onda (SVO), que mede a velocidade de propagação de ondas acústicas entre dois transdutores posicionados na superfície corneal antes e imediatamente, após a feitura da lamela. O primerio passo foi desenhado para o estudo das diferenças em relação à histerese corneal, fator de resistência corneal, mudanças na curvatura e velocidade de propagação de onda acústica entre córneas com lamelas finas e espessas. Posteriormente, as lamelas foram amputadas, e as medidas do sistema de velocidade de onda foram repetidas. Resultados: A média de espessura das lamelas ± desviopadrão (DP) foi de 108,5±6,9 (8,5% da espessura total) e 307,8±11,5 m (22,9% da espessura total), para os grupos de lamelas finas e espessas, respectivamente (p< 0,001). Histerese corneal e o fator de resistência corneal não apresentaram diferença estatística, após a criação de lamelas finas (p = 0,81 e p = 0,62, respectivamente). Histerese corneal foi significantemente mais baixa, depois da confecção de lamelas mais espessas (8,0±1,0 para 5,1±1,5 mmHg para medidas pré e pós-operatórias, respectivamente, p = 0,003, diminuição de 36,25%) e fator de resistência corneal também mostrou significante diminuição nesse grupo, após o procedimento cirúrgico; valores médios pré e pós-operatórios de 8,2±1,6 e 4,1±2,5 mmHg respectivamente (p= 0,007), diminuição de 50%. A ceratometria média simulada apresentou maiores valores, após a confecção das lamelas mais espessas em relação ao pré-operatório (ceratometria pré e pós-operatória de 39,5±1 D e 45,9±1,2 D, respectivamente, p= 0,003). Para o grupo de lamelas finas, não houve diferença estatisticamente significante (ceratometria pré e pós-operatória de 40,6±0,6 D e 41,4±1,0 D, respectivamente, p=0,55). Em relação ao Sistema de Velocidade de Onda, após a criação das lamelas e sua amputação, houve diminuição da velocidade de propagação acústica, embora na maior parte das posições não fosse estatisticamente significante. Conclusão: Nas condições experimentais estabelecidas por este estudo, a criação de lamelas de maior espessura pareceu exercer efeito mais relevante sobre a biomecânica da córnea de suínos / Purpose: To study the impact of programmed flaps at two different thicknesses on the biomechanical properties of the swine corneas. Methods: Twelve pig eyes were enrolled in this study and were formed two groups: 100m and 300 m flaps performed with the femtosecond laser. Each eye had the following procedure before the flap creation: raster photograph topographic maps, Ocular Response Analyzer (ORA), Optical Coherence Tomography to measure the pachymetry and flap thickness and Surface Wave Velocity system which is a prototype system that measures sonic wave propagation time between two transducers positioned on the corneal surface before and after flap creation. This first step was designed to investigate the differences in respect to corneal hysteresis, corneal resistance factor, curvature change and ultrasonic wave propagation between the groups with thinner and thicker flaps. After this initial procedure, flap amputation was performed and new measurements with the surface wave velocity system were taken again. Results: Measured flap thicknesses averaged 108.5±6.9 (8.5% of the total cornea) and 307.8±11.5 m (22.9% of the total cornea) for thin and thick flap groups, respectively (p< 0.001). Hysteresis and corneal resistance factor did not change significantly after flap creation in the thin flap group (p = 0.81 and p = 0.62, respectively). With thicker flaps, both parameters decreased significantly from 8.0±1.0 to 5.1±1.5 mmHg (p=0.003, reduction of 36.25%) and from 8.2±1.6 to 4.1±2.5 mmHg, respectively (p = 0.007), reduction of 50%. Simulated keratometry values increased in the thick flap group (from 39.5±1 D to 45.9±1.2 D, p=0.003) after flap creation and not in the thin flap group (from 40.6±0.6D to 41.4±1.0D, p= 0.55). Regarding surface wave velocity analysis, the surgical procedures induced lower values in some positions although most of them did not present statistically different results. Conclusion: In this experimental model, thicker flaps seemed to have a greater effect on the biomechanics of the swinish cornea

Recent advances in the miniaturization, functionality, and integration of integrated circuits and packages, such as the system-on-package (SOP) methodology, require increasing use of microvias that generates vertical signal paths in a high-density multilayer substrate. A scanning projection excimer laser system has been utilized to fabricate the microvias. In this thesis, a novel technique implementing statistical experimental design and neural networks (NNs) is used to characterize and model the excimer laser ablation process for microvia formation. Vias with diameters from 10 50 micrometer have been ablated in DuPont Kapton(r) E polyimide using an Anvik HexScan(tm) 2150 SXE pulsed excimer laser operating at 308 nm. Accurate NN models, developed from experimental data, are obtained for microvia responses, including ablated thickness, via diameter, wall angle, and resistance. Subsequent to modeling, NNs and genetic algorithms (GAs) are utilized to generate optimal process recipes for the laser tool. Such recipes can be used to produce desired microvia responses, including open vias, specific diameter, steep wall angle, and low resistance. With continuing advancement in the use of excimer laser systems in microsystems packaging has come an increasing need to offset capital equipment investment and lower equipment downtime. In this thesis, an automated in-line failure diagnosis system using NNs and Dempster-Shafer (D-S) theory is implemented. For the sake of comparison, an adaptive neuro-fuzzy approach is applied to achieve the same objective. Both the D-S theory and neuro-fuzzy logic are used to develop an automated inference system to specifically identify failures. Successful results in failure detection and diagnosis are obtained from the two approaches. The result of this investigation will benefit both engineering and management. Engineers will benefit from high yield, reliable production, and low equipment down-time. Business people, on the other hand, will benefit from cost-savings resulting from more production-worthy (i.e., lower maintenance) laser ablation equipment.

Microelectromechanical systems (MEMS) based artificial sensory hairs for flow sensing have been widely explored, but the processes involved in their fabrication are lithography intensive, making the process quite expensive and cumbersome. Most of these devices are also based on silicon MEMS, which makes the fabrication of out-of plane 3D flow sensors very challenging. This thesis aims to develop new fabrication technologies based on Polymer MEMS, with minimum dependence on lithography for the fabrication of piezoresistive 3D out-of-plane artificial sensory hairs for sensing of air flow. Moreover, the fabrication of a flexible sensor array is proposed and new materials are also explored for the sensing application. Soft lithography based approaches are first investigated for the fabrication of an all elastomer device that is tested in a bench top wind tunnel. Micromolding technologies allow for the mass fabrication of microstructures using a single, reusable mold master that is fabricated by SU-8 photolithography, reducing the need for repetitive processing. Polydimethylsiloxane (PDMS) is used as the device material and sputter deposited gold is used as both the piezoresistive as well as the electrode material for collection of device response. The fabrication results of PDMS to PDMS metal transfer micromolding (MTM) are shown and the limitations of the process are also discussed. A dissolving mold metal transfer micromolding process is then proposed and developed, which overcomes the limitations of the conventional MTM process pertinent to the present application. Testing results of devices fabricated using the dissolving mold process are discussed with emphasis on the role of micro-cr  acking as one failure mode in elastomeric devices with thin film metal electrodes. Finally, a laser microfabrication based approach using thin film Kapton as the device material and an electrically conductive carbon-black elastomer composite as the piezoresistor is proposed and demonstrated. Laminated sheets of thick and thin Kapton form the flexible substrate on which the conductive elastomer piezoresistors are stencil printed. Excimer laser ablation is used to make the micro-stencil as well as to release the Kapton cantilevers. The fluid-structure interaction is improved by the deposition of a thin film of silicon dioxide, which produces a stress-gradient induced curvature, strongly enhancing the device sensitivity. This new approach also enables the fabrication of backside interconnects, thereby addressing the commonly observed problem of flow intrusion while using conventional interconnection technologies like wire-bonding. Devices with varying dimensions of the sensing element are fabricated and the results presented, with smallest devices having a width of 400 microns and a length of 1.5 mm with flow sensitivities as high as 60 Ohms/m/s. Recommendations are also proposed for further optimization of the device.

科学研究費補助金 研究種目:一般研究(A) 課題番号:63420036 研究代表者:後藤 俊夫 研究期間:1988-1990年度