Eutectic bonding of Al-based high aspect ratio microscale structures

Metal-based microelectromechanical systems (MEMS) have important advantages over Si-based MEMS. To form any functional metal-based microdevice from metallic high-aspect-ratio microscale structures (HARMS), proper assembly and packaging are required. In this paper, we report successful eutectic bonding of Al-based HARMS using Al–Ge intermediate layers. A series of Al–Ge composite thin films were vapor deposited. Their composition and micro-/nano-scale structure were characterized. The morphology of bonded Al structures was examined, including the morphology of separated surfaces.     

Reanalysing structures with small modifications

A simple method developed for reanalysing slightly modified structures is outlined. The computational process is described and examples of application using two computer programs released by the Highway Engineering Computer Branch of the Department of Transport are presented.     

In-silico bonding schemes to encode chemical bonds involving sharing of electrons in molecular structures

Encoding of covalent and coordinate covalent bonds in molecular structures using ground state valence electronic configuration is achieved. The bonding due to electron sharing in the molecular structures is described with five fundamental bonding categories viz. uPair–uPair, lPair–uPair, uPair–lPair, vPair–lPair, and lPair–lPair. The involvement of lone pair electrons and the vacant electron orbitals in chemical bonding are explained with bonding schemes namely “target vacant promotion”, “source vacant promotion”, “target pairing promotion”, “source pairing promotion”, “source cation promotion”, “source pairing double bond”, “target vacant occupation”, and “double pairing promotion” schemes. The bonding schemes are verified with a chemical structure editor. The bonding in the structures like ylides, PCl₅, SF₆, IF₇, N-Oxides, BF₄⁻, AlCl₄⁻ etc. are explained and encoded unambiguously. The encoding of bonding in the structures of various organic compounds, transition metals compounds, coordination complexes and metal carbonyls is accomplished.     

Transfer learning of syntactic structures for building taxonomies for search engines

We apply a paradigm of transfer learning to build a taxonomy of entities intended to improve search engine relevance in a vertical domain. The taxonomy construction process starts from the seed entities and mines available source domains for new entities associated with these seed entities. New entities are formed by applying the machine learning of syntactic parse trees (their generalizations) to the search results for existing entities to form commonalities between them. These commonality expressions then form parameters of existing entities, and are turned into new entities at the next learning iteration. To match natural language expressions between source and target domains, we use syntactic generalization, an operation which finds a set of maximal common sub-trees of constituency parse trees of these expressions.Taxonomy and syntactic generalization are applied to relevance improvement in search and text similarity assessment. We conduct an evaluation of the search relevance improvement in vertical and horizontal domains and observe significant contribution of the learned taxonomy in the former, and a noticeable contribution of a hybrid system in the latter domain. We also perform industrial evaluation of taxonomy and syntactic generalization-based text relevance assessment and conclude that a proposed algorithm for automated taxonomy learning is suitable for integration into industrial systems. The proposed algorithm is implemented as a component of Apache OpenNLP project.     

面向少样本网状结构体的候选区域自适应检测方法

在仅有少量标签数据的图像网状结构检测任务中,需要大量训练数据的目标检测模型检测性能大幅下降。基于区域候选的目标检测模型在预测时,检测目标越多,检测时间越长。若基于区域候选的目标检测模型产生候选框的数量固定不变,而不同图像中网状结构目标数量不同,造成目标检测中额外的时间消耗。针对该问题,通过对训练样本中网状结构目标在图片中的密度分析以及根据网状结构体在图片中的特征分布,提出一种面向少样本网状结构体的候选区域自适应检测方法。该方法通过基于二值标签图标注方法得到大量训练样本,由候选区域自适应方法选取合理的候选框数量。与未改进的模型相比,在几乎不损失准确率的情况下,其加快了检测速度,尤其在目标数量稀少的数据中优势更为明显。     

基于多层结构的短信息应用系统研究与实现

为促进复杂网络计算环境信息交换的准确性和高效性,提出了基于多层网络结构模型的解决方法.应用该方法在面向银行短信业务系统方案中,提出了主动服务与被动服务概念,进而实现了短信业务及其系统管理功能设计,并基于XML规范交易报文的消息交换,实现了关键网络通信功能算法.通过该系统的可靠运行,表明了基于多层结构设计的短信应用系统已成为银行复杂信息系统集成的灵活、重要组成部分.     

A collapse result for constraint queries over structures of small degree

Collapse results, which are central for understanding constraint database queries, show that in terms of the expressive power, a large class of queries collapses to a much smaller one, typically involving only a restricted form of quantification. Most collapse results have been proved over constraints involving a linear order, and proofs are typically rather nontrivial. In this short note we give an easy proof of a powerful form of collapse for a large class of constraints without a linear order, namely those in which all basic relations are of small degree.     

Direct silicon-silicon bonding by electromagnetic induction heating

A novel heating technique, electromagnetic induction heating (EMIH), uses electromagnetic radiation, ranging in frequency from a few megahertz to tens of gigahertz, to volumetrically heat silicon above 1000/spl deg/C in only a few seconds. Typical power requirements fall between 900 to 1300 W for silicon wafers 75 to 100 mm in diameter. This technique has successfully produced direct silicon wafer-to-wafer bonds without the use of an intermediate glue layer. Infrared images indicate void free bonds that could not be delaminated with knife-edge tests. In addition, four pairs of stacked wafers were bonded simultaneously in 5 min, demonstrating the potential for multiwafer bonds and high-throughput batch processing.     

Device-level hermetic packaging of microresonators by RTP aluminum-to-nitride bonding

This paper presents a device-level microelectromechanical systems (MEMS) packaging process with accelerated tests and reliability analysis. Surface-micromachined microresonators are sealed inside microcavities by a rapid thermal processing (RTP) aluminum-to-silicon nitride bonding and packaging technique. Chip-to-chip bonding is used to form packages both under atmospheric pressure and in vacuum. The hermeticity of the package seals are evaluated by IPA (isopropyl alcohol) leak tests. The vacuum seal is evaluated by measuring the Q-factor (quality factor) of the packaged microresonators. The measured Q-factor of a vacuum-packaged comb-resonator is 1800/spl plusmn/200, corresponding to a 200 motor vacuum inside the micro cavity, and has not degraded over 37 weeks of shelf-life. The reliability information is evaluated by combining accelerated testing of the packages in a harsh environment (an autoclave chamber, 130/spl deg/C, 2.7 atm and 100% RH) and statistical analysis. The mean-time-to-failure (MTTF) of the packaged device is estimated as 29.7 weeks in an autoclave chamber, and tests on vacuum-packaged devices have confirmed the estimation.     

一种圆片级硅三层键合的三明治加速度传感器

提出了一种利用体微机械加工技术制作的硅三层键合电容式加速度传感器.采用硅各向异性腐蚀和深反应离子刻蚀技术实现中间梁一质量块结构的制作,通过玻璃软化键合方法完成上、下电极的键合.在完成整体结构圆片级真空封装的同时通过引线腔结构方便地实现了中间电极的引线.传感器芯片大小为6.8 mm×5.6 mm×l.26 ITUTI,其中敏感质量块尺寸为3.2 mm×3.2 mm×0.42 mm.对封装的传感器性能进行了初步测试,结果表明制作的传感器灵敏度约4.15 pF/g,品质因子为56,谐振频率为774 Hz.     

MEMS微结构变形行为尺度效应的研究进展

微米尺寸驱动结构广泛应用于微机电系统（MEMS）和集成电路等微／纳米系统中。由于这些微米尺寸驱动结构的几何尺寸和其微结构尺寸均在微米至纳米范围,它们的弹性、塑性性能及其变形行为具有明显的尺度效应。简要概述了近年来国内外有关微米尺寸驱动结构的弹性性能和变形特性尺度效应的研究情况,介绍了高阶理论的发展和塑性应变理论与微极理论在尺度效应分析中的应用,并对MEMS今后需要重点研究的方向进行了展望。     

芯片焊接中焊点自动定位技术的研究

为实现自动焊接,在已定位芯片的基础上,对自动定位芯片和框架上的焊点进行了研究,提出了一个高效精确的方法.在系统运行初始,人工引导测定保存每对焊点相对于芯片中心的坐标,使得在自动焊接过程中,定位了芯片中心,就得到了每对焊点坐标.芯片在手工粘贴时存在旋转和偏移误差,还需修正每对焊点坐标.该方法实现了高速精确的焊点自动定位.     

控向金字塔颜色传递

提出了一个基于控向金字塔的颜色传递算法,其基本思想是在目标图像和源图像对应的子带信息中分别实现统计信息的传递．统计信息包含各子带信息的均值、方差、偏度系数和峰度系数．为了避免通道交叉,目标图像和源图像的RGB像素信息首先被变换至与分量无关的laβ颜色空间．然后对变换后的各分量分别进行控向金字塔分解,并在倒塌金字塔时实现统计信息的传递．作为灵活的图像变换结构,控向金字塔允许对需传递的子带进行选择,从而可以获得多种不同的结果。     

Flip chip electrical interconnection by selective electroplating and bonding

This work presents a parallel electrical interconnection process by means of flip-chip, selective electroplating and bonding. The electrical interconnection lines are built on a glass substrate made of 500/2000 Å of Cr/Au with 3150 m in length and 10 m in width. Two silicon chips are processed as the device chips to be electrically interconnected. It has been demonstrated that 98 out of 102 interconnects are established in parallel with a successful rate of 96% and the average resistance of the electroplating bond is 12 . This process has potential applications in replacing the conventional, serial wire bonding or tape automated bonding (TAB) process for massive interconnection requirements in IC or MEMS devices. Reliability test is also performed by putting the interconnects into boiling liquid nitrogen (–195 °C) repeatedly. It is found that 100% of the interconnects survive after 2 cycles of the quenching process.The author would like to thank Mr. Ron Wilson for taking SEM pictures. This project is supported in part by a NSF award (DMI-0096220) and a Sandia National Laboratory grant. This paper was presented at the Fourth International Workshop on High Aspect Ratio Microstructure Technology HARMST 2001 in June 2001.     

Task Transfer by a Developmental Robot

Scaffolding is a process of transferring learned skills to new and more complex tasks through arranged experience in open-ended development. In this paper, we propose a developmental learning architecture that enables a robot to transfer skills acquired in early learning settings to later more complex task settings. We show that a basic mechanism that enables this transfer is sequential priming combined with attention, which is also the driving mechanism for classical conditioning, secondary conditioning, and instrumental conditioning in animal learning. A major challenge of this work is that training and testing must be conducted in the same program operational mode through online, real-time interactions between the agent and the trainers. In contrast with former modeling studies, the proposed architecture does not require the programmer to know the tasks to be learned and the environment is uncontrolled. All possible perceptions and actions, including the actual number of classes, are not available until the programming is finished and the robot starts to learn in the real world. Thus, a predesigned task-specific symbolic representation is not suited for such an open-ended developmental process. Experimental results on a robot are reported in which the trainer shaped the behaviors of the agent interactively, continuously, and incrementally through verbal commands and other sensory signals so that the robot learns new and more complex sensorimotor tasks by transferring sensorimotor skills learned in earlier periods of open-ended development     

Adhesive bonding by SU-8 transfer for assembling microfluidic devices

SU-8 is largely used to make microfluidic molds or components, but mainly for producing high-precision and thermally stable structures. We present a versatile method that employs SU-8 as glue to perform an adhesive bonding between micro-patterned structures. More in general, this technique enables an easy assembly of microfluidic devices, which can also be made by different materials, where selective bonding is required. The adhesive bonding is achieved by transferring a thin layer of SU-8 5 (thickness ≤15?μm) on a substrate by means of a polyimide foil. The method is described in detail and an example of its application is given. Finally, a shear test is carried out to prove sufficient adhesion strength for microfluidic applications.     

Transfer active learning by querying committee简

In real applications of inductive learning for classifi cation, labeled instances are often defi cient, and labeling them by an oracle is often expensive and time-consuming. Active learning on a single task aims to select only informative unlabeled instances for querying to improve the classifi cation accuracy while decreasing the querying cost. However, an inevitable problem in active learning is that the informative measures for selecting queries are commonly based on the initial hypotheses sampled from only a few labeled instances. In such a circumstance, the initial hypotheses are not reliable and may deviate from the true distribution underlying the target task. Consequently, the informative measures will possibly select irrelevant instances. A promising way to compensate this problem is to borrow useful knowledge from other sources with abundant labeled information, which is called transfer learning. However, a signifi cant challenge in transfer learning is how to measure the similarity between the source and the target tasks. One needs to be aware of different distributions or label assignments from unrelated source tasks;otherwise, they will lead to degenerated performance while transferring. Also, how to design an effective strategy to avoid selecting irrelevant samples to query is still an open question. To tackle these issues, we propose a hybrid algorithm for active learning with the help of transfer learning by adopting a divergence measure to alleviate the negative transfer caused by distribution differences. To avoid querying irrelevant instances, we also present an adaptive strategy which could eliminate unnecessary instances in the input space and models in the model space. Extensive experiments on both the synthetic and the real data sets show that the proposed algorithm is able to query fewer instances with a higher accuracy and that it converges faster than the state-of-the-art methods.     

Non-adhesive direct bonding of tiny parts by means of ultraprecision trapezoid microgrooves

 The study proposes a new method of bonding two small parts without the use of any adhesive by making use of ultraprecision trapezoid microgrooves on the surfaces of each part. These trapezoid microgrooves allow small parts to be bonded together with high positioning accuracy due to a wedge effect and an adhesion phenomenon between the mating surfaces. Ultraprecision trapezoid microgrooves are created by using an ultraprecision machining center having 1 nm positioning resolution and a diamond milling cutter. From experimental and simulated results, it is found that the bonding strength in general increases with decreasing wedge angle. The effect of the wedge length on bonding strength varies with the wedge angle where the strength increases with increasing length for wedge angles greater than 8° but increases with decreasing length for wedge angles less than 8°. Received: 5 May 2000/Accepted: 30 May 2000     

Water glass bonding

The author developed a low-temperature (80°C), low-external loads (electric field, magnetic field, load, etc.) bonding technology using water glass which is used in making molds (silica sand). The adhesive area ratio attained by this bonding technology was more than 95% and the bonding strength was about 290 kgf/cm². As this water glass bonding technology is applicable at comparatively low-temperatures, the residual stress of the bond is so small as to be able to bond eight 4-inch wafers together in the 3rd dimensional direction. In addition, as the thickness of a bonding layer is as small as several nanometers, the precision bonding of a tolerance of ±3 μm was also possible through alignment, and we could successfully fixed a cap with having a bonding seal of 0.32 mm wide in the vacuum. The leak rate was less than the detection limit of the He leak detector (1×10⁻¹⁰ Pa m³/s), showing excellent air-tightness. Using this bonding technology, the author made a self-package type IR microsensor on an experimental basis, and carried out an accelerated environmental test. As a result, its MTTF (mean time to failure) was estimated to be 6 years.