应用碳化硅表面改性技术降低全息-离子束刻蚀光栅刻槽的粗糙度

采用具有良好比刚度和热稳定性的碳化硅材料作为基底,使用全息-离子束刻蚀技术制作了光栅。碳化硅材料表面固有缺陷导致制作的光栅刻槽表面粗糙度高,槽底和槽顶粗糙度分别达到了29.6 nm和65.3 nm (R_q)。通过等离子辅助沉积技术在碳化硅表面镀制一层均匀的硅改性层,经过抛光可以获得无缺陷的超光滑表面。XRD测试表明制备的硅改性层为无定形结构。原子力显微镜的测试结果表明:经过抛光后,表面粗糙度为0.64 nm(R_q)。在此表面上制作的光栅刻槽表面粗糙度明显降低,槽底和槽顶粗糙度分别为2.96 nm和7.21 nm,相当于改性前的1/10和1/9。     

Fabrication of Micro-Grooves in Silicon Carbide Using Femtosecond Laser Irradiation and Acid Etching

A simple method using an 800-nm femtosecond laser and chemical selective etching is developed for fabrication of high-aspect-ratio grooves in silicon carbide. Micro grooves with an aspect ratio of approximately 40 are obtained. The morphology and chemical compositions of the grooves are analyzed using a scanning electronic microscope equipped with an energy dispersive x-ray spectroscopy. The formation mechanism of SiC grooves is attributed to the chemical reactions of the laser induced structural changes with a mixed solution of hydrofluoric acid and nitric acid. In addition, the effects of laser irradiation parameters on the aspect ratio of the grooves are investigated.     

Modern Applications of Plasma Etching and Patterning in Silicon Process Technology

Besides plasma etching of through-wafer interconnects in wafer stacks for vertical integration of chips, fabrication of platinum (Pt) electrodes with non-tapered sidewalls for the storage node in modern memories (DRAMs and FeRAMs) is one of the most challenging tasks of plasma process technology today. This paper describes the achievement of vertical integration of chips by plasma etching of high aspect ratio interchip vias. The etching processes for dielectrics, single crystal silicon, and the organic glue layer were all optimized for minimum reactive ion etching (RIE) lag i.e. for minimum decrease of etch rate with increasing etch depth. Furthermore the fabrication of perfect Pt electrodes for modern DRAMs and FeRAMs is reported. Vertical Pt profiles were achieved by plasma processing with resist mask. In this novel approach, the build-up of thin redepositions of Pt onto the sidewalls of the resist, obtained as a result of processing in pure Ar plasmas, is utilized to achieve a sidewall steepness of the patterned Pt film which is determined by the steepness of the pre-etch resist profile. After pattern transfer and resist stripping, the portion of the redepositions protruding above the fabricated storage node was completely removed by chemical mechanical polishing.     

混沌光学系统之前向神经网络混沌加速的系统辨识研究

研究了利用前向神经网络对混沌光学系统进行混沌加速系统辨识的可能性，计算机数值仿真发现，利用三层前向神经网络混沌光学系统辨识器。在基于混沌动力学角度的修正ＢＰ算法（混沌加速ＢＰ算法）支持下可克服由常规ＢＰ算法导致的辨识时间长的缺点，在较少的训练次数内即可对布拉格声双稳混沌系统进行良好的系统辨识，此研究结果表明，在混沌加速ＢＰ算法支持下，三层前向神经网络可用来快速处理混沌光学时间序列以进行相应的动力学     

A Kinetic Model for Plasma Etching Silicon in a SF6/O2 RF Discharge

Time-dependent Boltzmann electron distribution calculations have been made at constant power and pressure in a SF6/O2 plasma with a varying oxygen mole fraction. The results show that as the oxygen fraction increases in a SF6/O2 plasma, the number of high-energy electrons in the tail of the electron distribution and the mean electron energy both increase significantly while the plasma is kept at the same reduced electric field E/N. Rate coefficients have been computed for the electron kinetic processes of these plasmas and merged within a kinetic equilibrium model for the plasma etch process, including neutral gas-phase chemistry, ion chemistry, and surface reactions. Model simulations show good agreement with experimental results for SF6/O2 etching of polysilicon and demonstrate that the anisotropic character of dilute SF6 plasma etching is related to the shift in the electron distribution with increasing oxygen fraction. Competition between F and O species for adsorption to silicon etching sites is also shown to be a factor in determining etch rates, but this competition is not significant until very large (> 80 percent) oxygen concentrations are present. Ionization rates and ion transport to the surface are shown to be much more important. The model simulations provide a rationale for explaining the very high etch rates observed at low-SF6 partial pressures and the increasing anisotropic etch character with greater oxygen dilution of SF6.     

Millimeter-wave Micromachined Filter Design by Artificial Neural Network Modeling Technique

This paper presents a design approach for a 34 GHz λ/2 resonator micormachined bandpass filter by using the artificial neural network (ANN) modeling technique. Three important dimensions of the filter layout are used to capture critical input-output relationships in the ANN model. Once fully developed, the ANN model has been shown to be as accurate as an EM simulator and much more efficient computationally in the design optimization of the filter.     

Size Control of Porous Silicon Nanoparticles by Electrochemical Perforation Etching

Size‐controlled porous silicon‐based nanoparticles are prepared by pulsed electrochemical etching of single crystal silicon wafers, followed by ultrasonic fracture of the freestanding porous layer. When high‐current density pulses are applied periodically during the porous layer etching process, a porous multilayer results in which porous layers are separated by thin layers of much higher porosity. Ultrasonic fracture selectively cleaves the porous film along these high‐porosity perforations, providing greater size control and improved yields (by 5x) of the resulting porous nanoparticles. The effect of pulse width and repetition rate is systematically studied: tunability of the average nanoparticle size in the range 160–350 nm is demonstrated.     

金刚石线锯切割多晶硅片的气相刻蚀制绒研究

采用气相刻蚀制绒法研究金刚石线锯切割多晶硅片制绒.加热体积比1∶3、总体积400 mL的HF-HNO3酸混合溶液到90℃,使酸混合溶液受热产生气相,利用气相对金刚石线锯切割多晶硅片表面进行制绒.结果表明,制绒15 min之后,硅片表面的切割纹被完全去除;小腐蚀坑密布硅片表面,尺寸小于1μm,而传统湿法酸制绒所形成的腐蚀坑尺寸大于10μm.气相刻蚀后的金刚石线锯切割多晶硅片表面的微观粗糙度比传统酸混液制绒后的金刚石线锯切割多晶硅片表面的微观粗糙度高3倍多.气相制绒效果明显,并仅有12.11％的低反射率.     

The Effect of Substrate Temperature on the Properties of Nanostructured Silicon Carbide Films Deposited by Hypersonic Plasma Particle Deposition

Nanostructured silicon carbide films have been deposited on molybdenum substrates by hypersonic plasma particle deposition. In this process a thermal plasma with injected reactants (SiCl₄ and CH₄) is expanded through a nozzle leading to the nucleation of ultrafine particles. Particles entrained in the supersonic flow are then inertially deposited in vacuum onto a temperature-controlled substrate, leading to the formation of a consolidated film. In the experiments reported, the deposition substrate temperature T_s has ranged from 250°C to 700°C, and the effect of T_s on film morphology, composition, and mechanical properties has been studied. Examination of the films by scanning electron microscopy has shown that the grain sizes in the films did not vary significantly with T_s. Micro-X-ray diffraction analysis of the deposits has shown that amorphous films are deposited at low T_s, while crystalline films are formed at high T_s. Rutherford backscattering spectrometry has indicated that the films are largely stoichiometric silicon carbide with small amounts of chlorine. The chlorine content decreases from 8% to 1.5% when the deposition temperature is raised from 450°C to 700°C. Nanoindentation and microindentation tests have been performed on as-deposited films to measure hardness, Young's modulus and to evaluate adhesion strength. The tests show that film adhesion, hardness and Young's modulus increase with increasing T_s. These results taken together demonstrate that in HPPD, as in vapor deposition processes, the substrate temperature may be used to control film properties, and that better quality films are obtained at higher substrate temperatures, i.e. T_s700°C.     

碳化硅晶须流延浆料的制备

通过分析浆料分散性和粘度等因素,研究分散剂和粘结剂等对碳化硅晶须浆料流变行为的影响,确定合适的浆料制备工艺参数。结果表明:选择合适剂量的分散剂(2 wt%)能够有效减弱晶须之间的吸引力,减少晶须团簇,提高晶须分散性。同时选择3 wt%的粘结剂能够得到粘度适宜的浆料。优化工艺参数,最终获得分散性及稳定性较好且适于流延的晶须浆料。     

单频干涉仪瞬时相位计算的神经网络模型

用反三角函数表述的单频干涉仪瞬时相位的解析计算模型通常是分段或不连续的 ,不利于系统性能的综合分析。这里提出了基于神经网络的干涉仪瞬时相位的连续型计算模型 ,给出了网络学习方法。仿真研究结果表明 ,该模型对干涉仪瞬时相位的辨识精度优于 0 5° ,同时对干涉仪的光强波动有良好的鲁棒性 ;实验结果验证了这一点 ,为进一步提高单频干涉仪信号处理精度奠定了基础。此外 ,简要述及了该模型在其他测量领域 ,特别是速度 加速度测量领域的应用前景。     

改进BP神经网络算法对煤矿水源的分类研究

煤矿安全对煤炭工业的健康持续发展至关重要,而煤矿水灾又是煤矿事故的重大隐患,因此煤矿水源数据的处理对于预防矿井突水事故具有重要意义.实验在激光器的辅助下利用激光诱导荧光技术获取7种水源的数据信息,设定激光发射功率为100 mW,向被测水体发射波长405 nm激光,获取实验水样210组的荧光光谱数据,为了剔除光谱在采集过...     

一种基于遗传优化的BP神经网络的测光红移估计算法

除了星系的光谱红移之外,星系测光红移的估计也对研究宇宙大尺度结构及演变有着重要的研究意义。利用斯隆巡天项目最新发布的SDSS DR13的150 000个星系的测光及光谱数据,在红移值Z<0.8范围内,先使用SOM自组织神经网络对星系样本进行早型星系和晚型星系的聚类,然后用遗传算法优化后的BP神经网络对星系的测光红移进行估算。估算结果与作为标准的已知星系光谱红移进行比对,早型星系的红移估计最小均方误差约为0.001 3,晚型星系最小均方误差约为0.001 7。实验结果表明,遗传优化的BP算法在精度上优于BP神经网络算法,且效率上优于K近邻、核回归等传统测光红移估计算法。     

等离子体刻蚀工艺中的光学发射光谱仪数据的模型研究(英文)

原位光学发射光谱仪(OES)已经成为等离子体刻蚀工艺控制过程中的一种非常有潜力的在线传感器系统。采用光谱仪实时采集高密度等离子体刻蚀机中的OES光谱数据,利用BP神经网络算法对特定波长的OES数据进行分析及建模,以便对等离子体刻蚀工艺过程进行反馈控制。由刻蚀产物SiCl发出的405nm谱线被选为特征谱线来确定刻蚀工艺过程的终点。     

Automation of a Mass Flow Controller for Application in Time‐Multiplex SF6+CH4 Plasma Etching of Silicon

In this work is proposed the automation of a gas injection (mass flow) system in order to generate timemultiplex SF₆/CH₄ radiofrequency plasma applied for silicon (Si) etching process. The control of the gas injection system is important in order to better control the process anisotropy, i.e., the high‐aspect‐ratio of mask pattern transfer to substrate surface. In other words, this control allows the attainment of deep Si etching process. Here, the automation of the gas injection system was realized through the interface between a computer and a data acquisition board. The automation software developed allows controlling the gas flow rate switching it on and off during whole process through the use of a square waveform routine, intermittent flow, beyond the conventional condition of a fixed value for gas flow rate, continuous flow. In order to investigate the time‐multiplex SF₆/CH₄ plasma etching of Si, the residual gas analysis was performed. The investigations were made keeping the following process parameters: flow of SF₆: 10 sccm, flow of CH₄: 6 sccm, 100 W rf power, wave period: 20 sec. It were monitored the partial pressure of SF⁺ ₅ (parent neutral specie: SF₆), CH⁺₄ (CH4) and SiF+ ₃ (SiF4) species as a function of time for different gas flow switching and duty cycle. The results showed that with the generation of plasma occurs a drastic change in behavior of partial pressures of SF⁺ ₅ and CH⁺₄ species. Moreover, it is evidenced that the interactions between the SF₆ and CH₄ fragments promotes a high production rate of HF molecule and consequently a decrease of atomic fluorine, mainly when plasma is on. Finally, the behavior of partial pressure of SiF⁺ ₃ specie for alternatively intermittent SF₆ and CH₄ flow operation shows us that both the etching processes and the deposition of a polymer passivation layer are occurring alternatively, a desirable feature for multi‐step etching process (© 2012 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

等离子体刻蚀工艺中的光学发射光谱仪数据的模型研究

原位光学发射光谱仪（OES）已经成为等离子体刻蚀工艺控制过程中的一种非常有潜力的在线传感器系统。采用光谱仪实时采集高密度等离子体刻蚀机中的OES光谱数据，利用BP神经网络算法对特定波长的OES数据进行分析及建模，以便对等离子体刻蚀工艺过程进行反馈控制。由刻蚀产物SiCl发出的405nm谱线被选为特征谱线来确定刻蚀工艺过程的终点。     

硅衬底上纳米晶金刚石生长

研究了蚀刻气体对生长在硅衬底上纳米晶金刚石合成的影响.合成方法为热丝化学气相沉积法,衬底温度为550 oC,反应压力为4 kPa. 其中甲烷和氢气分别作为源气体和稀释气体. 氮气、氢气和氨气用作蚀刻气体. 结果表明,仅氢气作为蚀刻气体可获得最佳工艺条件.     

基于神经网络的叶绿素含量精细测量建模方法研究

活体植物叶片叶绿素含量SPAD值易受叶片厚度、水分等影响,提出了基于多参数神经网络建模的叶绿素含量精细反演方法。通过测量叶片在中心波长分别为650,940和1 450 nm光照射下的透过率,获得叶片的SPAD值和水分指数WI(water index),同时用数字螺旋测微仪测量相应的叶片厚度并用分光光度法测得其叶绿素含量。利用建模集样本分别建立SPAD值与实测叶绿素含量之间的单参数模型和基于BP神经网络的WI、厚度及SPAD值与实测叶绿素含量之间的非线性模型。利用这两种模型分别计算获得验证集样本的叶绿素含量预测值,对预测值和实测值进行了相关分析和相对误差的分析。实验以340个三种不同植物叶片为样本,用以上方法进行了分析。结果表明,利用BP神经网络建模后,每种植物样本的叶绿素含量预测精度都有不同程度的提高,尤其对于叶片厚度值较大的样本,效果更为明显。数据显示所有混合样本平均相对误差绝对值由单参数模型的7.55%降低到5.22%,实测值与预测值的拟合决定系数由0.83提高到0.93。验证了利用多参数BP神经网络模型可以有效地提高活体植物叶绿素含量预测精度的可行性。     

基于改进RBF神经网络的变形场插值算法研究

传统的物体表面力学变形场计算方法存在计算量大,无法计算边缘点变形等问题;提出一种改进的萤火虫算法优化RBF神经网络的变形插值方法,利用阈值约束RBF神经网络隐含层结点数,运用可变步长萤火虫算法优化RBF神经网络隐含层节点的中心和宽度,采用递推最小二乘法计算隐含层到输出层之间的权值,建立物体表面位移神经网络插值模型;为提高位移插值精度,在训练和测试的输入中增加坐标组合数据;应用于混凝土梁三点弯实验,仿真结果表明,该算法比常用的神经网络算法有更快的仿真速度和更高的预测精度,可用于土工材料表面变形场的快速、准确的计算。