低压化学汽相淀积含氢氮化硼薄膜

本实验采用二硼烷(B_1H_6)和氨气(NH_(?)),在388～700℃下进行低压化学汽相淀积生成含有硼、氮、氢的膜(B_(1～3)NH),用于X光曝光掩模的基底.已测得膜的各种特性如淀积速率、电阻率、元素成份比、晶型、折射率、红外吸收光谱、应力变化、对光的透射率、化学不活泼性及等离子腐蚀速率等.     

Adhesion force change on multilayer EUVL mask due to laser induced plasma shock wave

Change in adhesion force between a borosilicate glass microsphere and 40 Al₂O₃/TaN/Ru/MoSi pairs on a silicon wafer used as a multilayer extreme ultraviolet lithography mask stack were characterized by force-distance spectroscopy after cleaning Al₂O₃ layers using a laser induced plasma (LIP) shock wave. The adhesion force of the Al₂O₃ surface decreased at a higher laser energy and a lower gap distance above a threshold gap distance without changes in surface roughness. Frictional electrostatic repulsion, triboelectricity, was identified as the cause of lower adhesion forces on Al₂O₃ surface due to the high velocity and pressure of the LIP shock waves. The adhesion force decreased by increasing the number of exposures of LIP shock waves to the substrate.     

Determination of mask opening size in creating a fluid hole on brittle material by double-side sand blasting

In creating a hole on brittle materials by double-side sand blasting, the rebounding sand particle flux during the process may result in underetching at the edge of the mask opening, and leads to a larger sized fluid hole than the desired one. Determination of the correct mask opening size was made mainly by trial and error in the past. In this paper, relationships between the mask opening size and desired size of a hole on both the front and the back sides of the substrate are derived. For the front side, by taking into account the underetching effect, an equation is derived based on the kinetic energy theory. For the back side, there is negligible rebounding sand particles, and the mask opening size is set to be equal to the desired size of the hole. Experiments were conducted to verify the derived relationships. It is found that the measured sizes of the eroded holes on both the front and the back sides of the wafer substrate are distributed normally. The desired hole sizes deviate slightly from the median of a normal distribution curve, and the maximum predicted errors are 2.4% and 3.0% for front side and the back side sand blasting, respectively. The very satisfactory result together with ANOVA and test of homogeneity of variance of the predicted errors for various hole size shows that the derived relationships is applicable for determination of the mask opening size in the sand blasting process.     

多刀切割技术在石英玻璃片生产中的应用

本文探讨了多刀切割技术在石英玻璃片大规模生产中遇到的重要问题,并提出解决方案,建立了石英玻璃的多刀切割工艺。     

Development of rapid mask fabrication technology for micro-abrasive jet machining

Micro-machining of a brittle material such as glass or silicon is important in micro fabrication. Particularly, micro-abrasive jet machining (μ-AJM) has become a useful technique for micro-machining of such materials. The μ-AJM process is mainly based on the erosion of a mask which protects brittle substrate against high velocity of micro-particles. Therefore, fabrication of an adequate mask is very important. Generally, for the fabrication of a mask in the μ-AJM process, a photomask based on the semi-conductor fabrication process was used. In this research a rapid mask fabrication technology has been developed for the μ-AJM. By scanning the focused UV laser beam, a micro-mask pattern was fabricated directly without photolithography process and photomask. Therefore, rapid and economic mask fabrication can be possible for the micro-abrasive jet machining. Two kinds of mask patterns were fabricated by using SU-8 and photopolymer (Watershed 11110). Using fabricated mask patterns, abrasive-jet machining of Si wafer was conducted successfully. This paper was recommended for publication in revised form by Associate Editor Dae-Eun Kim Seung Pyo Lee is a research engineer of GM Daewoo Auto & Technology. He receives his Ms degree in Precision Mechanical Engineering at the Chungbuk National University in 2007. Hyun-Wook Kang is currently a Ph.D. candidate in the Department of Mechanical Engineering at POSTECH, Korea. He received his B.S. degree in the Department of Mechanical Engineering from POSTECH. His current research interests are in the solid freeform fabrication technology for the engineered tissue construction. Seung-Jae Lee received the M.S. degree from the Dept. of Me-chanical Engineering from the POSTECH in 2002, and his Ph.D. degree in Dept. of Mechanical Engineering from POSTECH in 2007. His Ph.D research is the study of Microfabrication and Tissue engineering. In Hwan Lee is a professor of School of Mechanical Engineering at Chungbuk National University, Korea. He receives his Ph.D. degree in Mechanical Engineering from the POSTECH in 2003. His research is focused on micro-manufacturing and bio-system. Tae Jo Ko is a professor of mechanical engineering at Yeungnam University, Korea. Also, he is responsible for the Gyoungbuk Hybrid Technology Institute that is regional research innovation center and initiates the idea for hybrid manufacturing. He earned Ph.D in mechanical engineering from POSTECH, Korea, in 1994. He worked for Doosan Infracore Co. Ltd. (formerly Daewoo) from 1985 to 1995. His research interests include machine tools, metal cutting as well as nontraditional machining. Dong-Woo Cho is a professor in the Department of Mechanical Engineering at the POSTECH. He received his Ph.D. degree in Mechanical Engineering from the University of Wisconsin-Madison in 1986. His research focuses on the manufacturing system for Tissue Engineering.     

Optimization of resolution by FDTD analysis in aligner lithography

A mask aligner can transcribe a pattern from a photomask to an exposure substrate by Fresnel diffraction. A diffraction fringe, specific to Fresnel diffraction, appears on a light intensity distribution of the pattern (a diffraction pattern image), and the formation of the pseudo-pattern restricts the resolution performance. The diffraction fringe can be smoothened by expanding the spread of the illumination source, and thus, the pseudo diffraction can be attenuated. However, this also causes a change in light intensity at the pattern edge to be attenuated, and the error of pattern line width to process change becomes large. Since edge diffraction patterns can be calculated by finite difference time domain (FDTD) analysis, the size of light source providing optimum resolution can be predicted by calculating and comparing pattern images corresponding to the size of the light source using this analysis. Therefore, by introducing an illumination optical system that can arbitrarily set the size of a light source to a predicted value, optimum resolution can be obtained without prior trial exposure. This study shows that the resolution of an aligner can be optimized by prior prediction, by introducing a multiple smoothing optical system that has been developed as an illumination system. This system realizes uniform illumination distribution on both a pupil plane and photomask plane and has an adjustable aperture mechanism that, by combining it with FDTD analysis, can arbitrarily set the size of the light source.     

基于Mask R-CNN的铁谱磨粒智能分割与识别

针对铁谱图像因背景复杂、尺寸分布广、颗粒重叠等导致难以精确分割与识别的问题,以相似度高的疲劳剥块、严重滑动磨粒、层状磨粒共3种异常磨粒作为研究对象,提出基于深度神经网络模型Mask R-CNN的对多目标铁谱磨粒进行智能分割与识别的方法,并对特征提取层分别选用深度不同的残差网络ResNet50和ResNet101进行对比试验。实验结果表明,基于迁移学习方法的Mask R-CNN+ResNet101模型能够在复杂背景下对多目标、多类型、多尺寸的相似磨粒进行有效分割与识别,测试集的平均精度高达76.2%,模型具有较好的泛化能力。     

Si局域分子束外延技术

利用局域分子束外延技术生长了含有SiGe量子阱的岛状结构。实现了生长的岛状结构侧面不与掩模材料相接触。这种含有量子阱的岛状结构具有比相同结构量子阱强20多倍的光致发光强度。     

基于Mask RCNN的矿仓入料口

针对矿仓入料口堵塞矿石识别过程中现场工况环境复杂、矿石识别检测难度大等问题,采用深度学习和图像处理技术开展矿石智能识别检测的研究,提出基于Mask RCNN的矿石识别检测方法。该方法可以实现对矿石识别的同时进行实例分割,并提出利用矿石轮廓的形心坐标取代Mask RCNN中的外接矩形框定位方法,有效解决矿石定位不精确的问题。实验结果表明:基于Mask RCNN网络的矿石识别模型可以实现对多种数量、不同位姿以及堆叠的矿石精准识别,综合准确率达到97.6%,采用矿石轮廓形心坐标的定位方式可以有效避免因矿石形状和位姿而带来的定位误差,为智能清堵机械手提供精确的视觉引导。      

基于深度学习的白菜田杂草分割

田间除草在农业生产中具有重要意义,传统杂草识别方式具有效率低或者局限性大的缺点。为此提出将Mask R-CNN算法应用到自然光照下杂草幼苗和白菜幼苗的图像识别。实验选取常见杂草幼苗和白菜幼苗图像作为训练集训练网络,经测试集测试得到81%的合格率。与传统阈值分割算法对比,Mask R-CNN在不同环境下都能精确地识别出杂草幼苗,解决了传统图像算法在复杂光照和叶片遮挡环境下图像难以分割的问题,并且避免了分类器设置工作。