n型砷化镓微区光电化学腐蚀过程

在n-GaAs电解液界面,用聚焦He-Ne激光照射, 使n-GaAs表面发生微区光电化学腐蚀, 用计算机控制步进马达, 使试样在X-Y二维方向扫描移动, 能在晶片上得到刻蚀点直径2 μm的刻蚀图案. 研究了激光相对光强, KOH、H_2SO_4、KCl等刻蚀剡的浓度, 光腐蚀的时间, 电极电位等因素对腐蚀点的直径和深度的影响, 通过实验数据找出腐蚀过程的规律, 并用光电化学原理进行解释.     

安培检测芯片毛细管电泳及其初步应用

An end-channel amperometric detector with a guide tube for working electrode was designed and integrated on a home-made glass microchip. The guide tube was directly patterned and fabricated at the end of the detection reservoir, which made the fixation and alignment of working electrode relatively easy. The fabrication was carried out in a two-step etching process. A 30 μm carbon fiber microdisk electrode and Pt cathode were also integrated onto the amperometric detector. The baseline separation of dopamine (DA), catechol (CA) and epinephrine (EP) was achieved within 80 s. Relative standard deviations of not more than 5.2% were obtained for both peak currents and migration times of DA and CA (n=5). Using standard adding method, DA in tLrine and plasma samples was detected. The recoveries were in the range of 83%—103%.     

Temperature-dependent dry etching characteristics of III–V semiconductors in HBr- and HI-based discharges

Microwave discharges of HBr/H₂/Ar and H/H₂/Ar with additional do biasing of the sample were used to etch InP, GaAs, and AlGaAs at temperatures between 50–250°C. The etch rates increase by factors of 3–50 and 5–9, respectively, for HBr-and HI-based discharges over this temperature range, but display non-Arrhenius behavior. The etched surfaces became very rough above 100°C for InP with either discharge chemistry due to preferential loss of P, while GaAs and AlGaAs are more tolerant of the elevated temperature etching. The near-surface electrical properties of InP are severely degraded by etch temperatures above 100°C, while extensive hydrogen in-diffusion occurs in GaAs and AlGaAs under these conditions, leading to dopant passivation which can be reversed by annealing at 400°C.     

Characteristics of vinyl iodide microwave plasma etching of GaAs/AlGaAs and InP/InGaAs heterostructures

Vinyl iodide (C₂H₃I) microwave discharges with additions of H₂ and Ar are found to provide faster etch rates than conventional CH₄/H₂/Ar discharges for InP, InGaAs, GaAs, and AlGaAs. This is a result of the relatively high volatilities of indium, gallium, and aluminum iodide species. The etched features are smooth and anisotropic over a wide range of do self-biases (–150 to –350 V), process pressures (1–20mTorr), and microwave powers (150–500 W). The polymer that forms on the mask during the plasma exposure can be readily removed in O₂ discharges. Electron spectroscopy for chemical analysis (ESCA) showed that the etched surfaces are slightly deficient in the group V elements under most conditions, but changes to the optical properties of the semiconductors are minimal. No defects are visible by transmission electron microscopy (TEM) in GaAs or InP samples etched at dc biases –250 V.     

A model for the etching of Si in CF4 plasmas: Comparison with experimental measurements

A model has been developed to describe the chemistry which occurs in CF₄ plasmas and the etching of Si both in the plasma and downstream. One very important feature of this model is that for discharge residence times which vary by more than an order of magnitude, the amount of CF₄ consumed is low and relatively constant. This is because the gas-phase combination reactions between F and both CF₃ and CF₂ lead to the rapid reforming of CF₄. The model predicts that CF₂ is a major species in the gas phase and that the [F] detected as a sample point downstream is a very sensitive function of [CF₂]/[F] in the discharge. Even though the calculations show that [F] in the discharge varies only slightly over the wide range of experimental conditions considered, large variations in [F] at the sample point occur because the [CF₂]/[F] ratio in the discharge changes. The concentrations of C₂F₆ and SiF₄ are predicted to within a factor of 2 over a very wide range of experimental conditions. This confirms the importance of gas-phase free radical reactions in the etching of Si.     

A model for the etching of silicon in SF6/O2 plasmas

A model has been developed to describe the chemistry which occurs in SF₆/O₂ plasmas and the etching of silicon in these plasmas. Emphasis is placed nn the gas-phase free radical reactions, and the predictions n( the model are compared with experimental results. Forty-seven reactions are included, although a subset of 18 reactions describes the chemistry equally well. Agreement between the calculated and measured concentrations of stable products downstream of the plasma is better than a factor of 2. The need for additional kinetic data and fàr well-characterized diagnostic studies of SF₆/O₂ plasmas is discussed.     

The use of “actinometer” gases in optical diagnostics of plasma etching mixtures: SF6-O2

The spectroscopic emission intensities from excited F atoms in SF₆-O₂ discharges at 1 torr have been correlated to the densities of atoms in their ground electronic state by measuring the excitation efficiencies of the electrons in the energy range 11 to 17 eV with a method which essentially consists in the analysis of the emission of Ar or N₂, added as actinometer gases to the discharge mixtures. The general applicability of the method has been tested by a direct titration of F atoms with chlorine. The spectroscopic analysis has allowed the determination of useful information on the trends of both the electron densities and their energies as a function of the oxygen percent in the feed.     

基于自适应对比度增强的红外小目标检测网络

红外探测系统需要尽早发现目标以便及时拦截,但是红外图像上的小目标检测是一个挑战十足的任务。为了提高检测准确率,提出一种基于自适应对比度增强的红外小目标检测方法。为了利用自注意力机制和卷积各自的优势,设计了一个高效的特征提取网络和一个面向小目标的检测头。同时为了解决实际应用中出现的弱目标,在检测子网络前添加了一个图像预处理子网络,该模块可以自适应地调节图像对比度。在红外空中小目标数据集上的实验表明,提出的方法能达到93.76%的检测精度,与经典的检测方法相比,能够更好地平衡检测精度和召回率,证明了方法的巨大应用潜力。     

道路交通网络节点分配优化策略研究进展

为了进一步提高城市道路交通网络的通行效率,粒子群优化和神经网络等多种智能优化算法受到越来越多的关注。近年来,深度学习技术的普及与应用大幅提升了城市交通网络的节点识别效率,而交通网络的节点调度又扩展了深度学习技术的应用。文中详细分析了交通节点调度所面临的关键问题,归纳并总结了相关网络节点分配的研究现状。在此基础上,深入研讨了城市交通网络节点调度与深度学习的应用前景,并对交通网络节点分配优化策略的未来研究方向进行了展望。     

基于拓扑链路识别的光网络流量数据合成算法

基于深度学习的光网络流量诊断与预测等场景中,由于保密等原因,光链路的流量数据采集和存储工作受限。针对数据量少而无法支撑深度学习的问题,文章提出了一种基于拓扑链路识别的光网络流量数据合成算法,其核心思想是在生成对抗网络框架下,联合基于光网络拓扑的条件生成模型和基于光网络流量的数据合成模型,以自监督的方式合成指定光链路的流量数据。仿真结果表明,所提算法合成的光网络流量数据在自相关系数指标上与真实数据接近且使得基于全连接神经网络的流量预测模型准确率达到95%以上。