InAs/ln(0.52)Al(0.48)As quantum wire structure with the specific layer-ordering orientation on InP(001)

Quantum wires were formed in the 6-period InAs/In0.52Al0.48As structure on InP(0 0 1) grown by molecular beam epitaxy. The structure was characterized with transmission electron microscopy. It was found that the lateral periodic compositional modulation in the QWR array was in the [1 (1) over bar 0] direction and layer-ordered along the specific orientation deviating from the [0 0 1] growth direction by about 30 degrees. This deviating angle is consistent with the calculation of the distribution of elastic distortion around quantum wires in the structure using the finite element technique. (C) 1999 Elsevier Science B.V. All rights reserved.

p-type co-doping study of GaN by photoluminescence

Photoluminescence measurements were performed on p-type co-doping effects of C, As, and Mg in GaN. The dopants were incorporated into GaN by ion implantation performed at 77 K. We find that the 3.42 eV luminescence line is sensitive to hole concentration, and propose that after cartful calibration the 3.42 eV line may be used as a probe to measure hole concentration in GaN. Simply doping one kind of accepters will not result in holes, while co-doping can substantially improve p-type doping efficiency. As + C and As + Mg co-doping induce an acceptor level of 180 meV above the valence band. Mg + C co-doping is the most promising method for p-type doping, the related acceptor level is determined to be as shallow as 130 meV. The improvement of the doping efficiency by co-doping is probably due to the decrease of the acceptor ionization energy. (C) 1999 Elsevier Science B.V. All rights reserved.

Uniformity enhancement of the self-organized InAs quantum dots

Growth interruption was introduced after the deposition of GaAs cap layer, which is thinner than the height of quantum dots. Uniformity of quantum dots has been enhanced because the full-width of half-maximum of photoluminescence decrease from 80 to 27 meV in these samples as the interruption time is increased. Meanwhile, we have observed that the peak position of photoluminescence is a function of interruption time, which can be used to modulate energy level of quantum dots. All of the phenomenon mentioned above can be attributed to the diffusion of In atoms from the tops of InAs islands to the top of GaAs cap layer caused by the difference between the surface energies of InAs and GaAs. (C) 1999 Elsevier Science B.V. All rights reserved.

MOVPE growth of GaN and LED on (111) MgAl2O4

The growth of wurtzite GaN by low-pressure metalorganic vapor-phase epitaxy on (1 1 1) magnesium aluminate (MgAl2O4) substrates have been studied. The morphological, crystalline, electrical and optical properties are investigated. A p-n junction GaN LED was fabricated on the MgAl2O4 substrate. (C) 1998 Elsevier Science B.V. All rights reserved.

Dependence of photoluminescence induced by carbon contamination on GeSi structure

We studied the dependence of photoluminescence induced by carbon contamination on the Ge/GeSi structure. It is found that a carbon and silicon defect complex may be formed in a special structure by opening the in situ high-energy electron diffraction test during growth. There is an important difference in the dependence of photoluminescence on the temperature between the defect complex in our samples and in bulk Si. where the impurity-active center is generated by high-energy electron (about several MeV) irradiation. The quenching temperature of the photoluminescence from the impurity-active center is higher in our Ge/GeSi structure than in bulk Si. The defect complex may serve as an impurity-active center for a possible application in making Si-based light-emitting diodes whose wavelength is around 1.3 mu m in the window of optical communication. (C) 1998 Elsevier Science B.V. All rights reserved.

密码Hash函数相关问题研究

密码Hash函数是信息安全密码学的一个重要研究内容，是一类广泛应用的密码算法，用于把任意长度的字符串压缩成特定长度的字符串，同时需要在各种应用环境下满足一定的安全要求如抗碰撞，抗原象等。Hash函数广泛应用于数字签名、可证明安全、密码算法的构造以及重要的安全协议中。对Hash函数进行研究、分析Hash函数的安全性、构造安全高效的Hash算法有着重要意义。 本文研究了Hash函数的安全性质、设计结构以及常用分析方法，研究了Hash函数扩散层部件的设计，并且对MAME压缩函数算法进行了分析，取得了如下研究结果： (1) 研究了密码Hash函数的安全性质、设计结构、设计原理和常用分析方法，归纳总结了51个SHA-3候选算法的设计特点、设计原理和实现效率，研究了最新的分析进展，总结了新的攻击方法如REBOUND攻击等。NIST仿照AES的征集过程的SHA-3竞赛，目标是选出新的Hash函数标准SHA-3。进入第一轮的候选算法有51个，经过筛选选出其中的14个作为当前第二轮的候选算法。这些新Hash算法是由世界各国密码学家精心设计，是Hash函数领域最新设计思想的集体展示，当中涌现出很多新的设计结构和设计方法，同时激励密码学家发展新的分析方法。 (2) 设计并实现了了有限域上的扩散层构造算法以及扩散层分支数测试的算法，并针对多元域上的扩散层矩阵，本文使用编码理论，利用GRS码和柯西矩阵等设计了多元域扩散层矩阵的构造算法；使用有限域上的高斯消元法和线性码的性质设计了多元域扩散层矩阵的分支数的检测；设计了高效的二元域扩散层矩阵分支数测试算法。 (3) 针对MAME压缩函数算法进行差分分析，MAME算法是SHA-3候选算法Lesamnta的前身，于CHES 2007上提出的面向硬件有效实现的Hash算法。本文利用差分攻击对MAME算法进行分析，首先针对MAME的结构性质利用对通用Feistel结构的攻击方法构造了22轮差分攻击，碰撞攻击的复杂度为2^97，（第二）原象攻击的复杂度为2^197；对23轮的差分攻击需要的预计算是2^64张表，每张表的大小为2^64；对24轮的差分攻击需要的预计算是2^128张表，每张表的大小为2^64。针对24轮差分攻击很大的内存复杂度，我们利用了算法的细节特性，改进了差分攻击，新的差分不需要预计算的辅助内存，（第二）原象的复杂度为2^224。

encoding-based method for pattern matching

A new algorithm for compiling pattern matching is presented. Different from the traditional traversal-based approaches, it can represent a sequence of patterns as an integer by an encoding method and translate equations into case-expressions. The algorithm is simple to implement, and efficient for a kind of patterns, i.e. simple and dense patterns. This method can be complementary to traditional approaches.

基于局部信息和领域知识的聚类算法研究

近些年来，各种信息处理系统和互联网快速发展，大量的数据随之产生。如何从这些数据中发现有用的知识就变得非常重要，因此，数据挖掘在近些年是热点研究领域。 而在数据挖掘研究中，聚类分析是一种关键的非监督学习技术。本文首先就基于局部信息的聚类算法进行了深入的研究，包括基于局部回归的聚类算法研究和基于局部重构的聚类算法研究。其次，由于领域知识的应用在聚类分析的过程中很重要，所以本文就基于领域知识的聚类算法进行了研究创新。具体说来，本文的研究工作和贡献包含以下几个方面。 第一，提出一种新的基于局部回归的聚类分析方法。该聚类方法基于局部学习的策略，以绝对值误差和作为簇标签错误度量，比以往文献中使用的平方误差和要更加合理和健壮。当与核回归相结合时，可以最终推导出一个谱聚类算法。在多个数据集上进行的实验结果表明，我们提出的聚类算法一般情况下效果较好。 第二，提出一种新的基于正则化局部重构的聚类分析方法。该聚类方法基于局部重构的策略，此策略在维数约减领域使用较多，可以用于处理高维空间的低维流形上的数据。根据这种策略，我们提出了一种基于正则化局部重构的聚类模型，并由此推导出一种新的谱聚类算法。在多个数据集上的实验结果表明，该聚类算法和之前提出的基于局部回归的聚类算法一般情况下效果较好。 第三，我们提出一种新颖的用于聚类分析的领域知识形式：特征顺序偏好。不同于聚类领域以前经常使用的实例层面的领域知识，这是一种特征层面的领域知识。然后， 我们提出一个合理且自然的基于特征顺序偏好的聚类模型。 此模型可以使用 Bregman 散度作为失真度量，这是一大类失真函数。该聚类模型易于扩展，可以使用更多的失真函数和权重惩罚函数；其正则化项可以采用泛化熵；还能够被扩展以处理离散数据。此外，我们给出了该模型的概率解释。在一些数据集上进行的实验结果表明，我们提出的基于特征顺序偏好的聚类模型确实能够提升聚类算法的性能。