一种基于自适应差分进化算法的薄膜参数表征方法研究（特邀）

依据穆勒椭偏测量方法中偏振光的传输方式,文中提出了一种基于自适应差分进化算法（SADE）的各向同性纳米薄膜厚度与光学常数的表征方法。通过建立出射光强关于待测标准样片穆勒矩阵的最小二乘模型,用SADE算法对穆勒矩阵元素进行求解,并将拟合得到的穆勒光谱曲线与用双旋转补偿器穆勒矩阵椭偏仪(DRC-MME)测量得到的穆勒光谱图进行了比较,利用传输矩阵求解薄膜厚度。对标定值分别为(104.2±0.4) nm和(398.4±0.4) nm的SiO₂/Si标准样片进行仿真计算,实验表明:当分别迭代到80次和87次时,目标函数光强的残差平方和收敛到最小值0.97和1.01,得到的膜厚计算值分别是(103.8±0.6) nm和(397.8±0.6) nm,相对误差均小于1%。同时用计量型椭偏仪根据得到的折射率进行计算,得到膜厚的计算值分别为(104.1±0.6) nm和(398.2±0.6) nm,验证了SADE在相近收敛速度下对各向同性纳米薄膜参数求解过程中具有计算简单和可以准确的找到全局最优解的特点。     

金属反射镜对外差干涉椭偏测量精度的影响

采用声光调制器设计了一种透射式外差干涉椭偏测量系统.实验测量了单层透明氧化铟锡(ITO)膜,膜厚和折射率测量误差分别达8nm和7%.采用琼斯矢量法分析了金属反射镜引起的光束椭偏化对测量结果的影响.从光学系统中移出被测样品得到的标定数据,可以消除金属反射镜本身退偏效应的影响,但无法消除退偏效应和方位角误差共同作用所引入的椭偏参数测量误差.计算结果表明,退偏效应和方位角误差共同作用引入的膜厚测量误差可达4 nm左右,该误差与薄膜参数无关,与方位角误差近似成线性关系.     

光学元件磁流变加工驻留时间粒子群优化算法

为实现光学元件磁流变高精度加工,基于脉冲迭代原理,提出基于粒子群算法的驻留时间优化方法。该方法在脉冲迭代法的基础上引入粒子群算法对整体面型残差进行优化,通过对整体驻留时间的判定,从而实现每个驻留时间点的最优选择,达到高精度面形加工。通过对Φ156mm光学表面仿真加工,均方根(RMS)值和峰谷(PV)值从初始的169.164和1161.69nm收敛到23.4925和807.2156nm。仿真结果表明,该算法能在保证面形收敛精度的同时快速获得稳定可靠的驻留时间分布,能有效降低中频误差,其算法性能优于常用的脉冲迭代法。该算法为磁流变抛光光学元件过程中的驻留时间计算提供了一种解决方案。     

基于粒子群-神经网络算法的纳米薄膜参数表征

提出了一种利用椭圆偏振测量进行纳米薄膜参数数据处理的混合优化算法。以人工神经网络模型为基础,利用改进粒子群算法选择人工神经网络中的权值和阈值,建立改进粒子群-神经网络(IPSO-NN)算法模型对纳米薄膜参数进行数据处理,以获得更高精度的纳米薄膜参数。利用IPSO-NN算法模型计算标称厚度值为50和997.7 nm的硅上二氧化硅(SiO_2/Si)纳米薄膜厚度标准样片的薄膜参数,结果表明:两种尺寸的薄膜厚度计算结果相对误差均小于3%,说明了混合优化算法具有高精度的薄膜厚度和复折射率等薄膜参数的计算能力。同时通过实验证明了IPSO-NN算法模型能有效地优化迭代次数,具有收敛速度快、测量效率高等优势。     

干涉椭偏测量系统中分光镜引入的非线性误差

研究了一种采用声光调制器实现的透射式外差干涉椭偏(IE)测量系统。实验测量了单层透明ITO膜, 膜厚和折射率测量误差分别达到4 nm和6％。除了激光源和偏振器件之外, 分光镜也是重要的非线性误差源。研究了分光镜(BSs)退偏效应和方位角对椭偏测量误差的影响。采用琼斯矢量法推导出误差理论模型, 并数值计算了误差随分光镜光学参数和方位角的变化规律。计算结果表明, 由此引入的膜厚测量误差可达数纳米量级, 且与方位角误差近似成线性关系。退偏效应和方位角误差引入的非线性测量误差是互相关的, 不能通过移出被测样品的标定过程来完全消除。为了达到亚纳米级测量精度, 需要控制分光镜方位角误差在0.01°以内。根据分光镜退偏参数与非线性误差的关系, 可以设计或选择合适的分光镜。     

光学表面粒子污染物散射的单层薄膜调控特性

为了降低超精密低损耗光学元件表面粒子污染物的光散射损耗,文中提出通过在光学表面沉积单层薄膜来调控表面场强分布,从而降低散射损耗的方法。理论分析了K9玻璃超光滑光学表面不同厚度单层二氧化硅（SiO₂）和单层二氧化钛（TiO₂）薄膜表面上方半径为100 nm粒子污染物所在处的电场强度,理论分析结果发现,当SiO₂薄膜厚度为137.4 nm,TiO₂薄膜厚度为12.3 nm时,表面粒子污染物所在处的电场强度最小。在此基础上分别计算了光学元件表面沉积厚度为137.4 nm的单层SiO₂薄膜以及厚度为12.3 nm的单层TiO₂薄膜,表面粒子污染物的总散射损耗(S)和双向反射分布函数(BRDF),计算结果表明,在波长为632.8 nm的光垂直入射时,单层SiO₂薄膜和单层TiO₂薄膜可有效降低其表面粒子的BRDF,且可将K9玻璃表面的总散射分别降低12.40%和25.04%。实验验证了单层SiO₂薄膜对于表面粒子污染物散射降低的有效性。     

掺氮氧化钒薄膜的椭偏光谱表征

利用反应溅射法,在Si(100)衬底上沉积了掺N氧化钒薄膜.借用SE850椭偏仪对薄膜进行了近红外波段(1300-2300 nm)的光谱测量,运用Tauc-Lorentz模型对薄膜的椭偏光谱数据进行了拟合,并计算了薄膜的光学参数(n,k)和厚度.结果表明:随着掺N量的增加,氧化钒薄膜的光学参数(n,k)随之增加,而其沉...     

基于混合优化算法的纳米薄膜参数表征

为了在椭圆偏振测量过程中得到精确的纳米薄膜参数,提出了一种求解纳米薄膜参数的混合优化算法。结合人工神经网络算法反向传播和粒子群算法快速寻优的特点,建立了改进粒子群-神经网络(Improved Particle Swarm Optimization-Neural Network,IPSO-NN)混合优化算法。该算法在较少的迭代次数下具有快速跳出局部最优解的能力,从而快速寻找椭偏方程最优解。文中使用该算法对标称值为(26.7±0.4)nm的硅上二氧化硅纳米薄膜厚度标准样片进行薄膜参数计算。结果表明:采用IPSO-NN混合优化算法计算薄膜厚度时相对误差小于2%,折射率误差小于0.1。同时,文中通过实验对比了传统粒子群算法与IPSO-NN算法,验证了IPSO-NN算法计算薄膜参数时能有效优化迭代次数和寻找最优解的过程,实现快速收敛,提高计算效率。     

基于模式理论光栅椭偏参数反演的数值模拟

将一种广泛用于求解系统优化问题的方法——正单纯形法，求解光栅的椭偏方程。首先，利用求解光栅的傅立叶模式理论对TE和TM波的复反射系数进行求解。然后计算出其相应的椭偏参数(△，Ψ)，并在该值的基础上加入不同偏差的随机高斯噪声，将加入噪声后的值(△m，Ψm)作为模拟测量值。最后使用优化算法进行反演。通过对几种常用面形光栅椭偏参数的数值模拟，一方面表明傅立叶模式理论计算光栅的椭偏参数不仅精度高。而且速度快；另一方面表明利用正单纯形法得到的光栅参数值很接近于正演时假设的参数值，从而从理论上证明了利用椭偏法测量光栅各种光学参数的可行性。     

多步旋涂CsPbBr3薄膜复光学常数的椭偏光谱研究

为了研究多步旋涂法制备的CsPbBr₃薄膜的光学常数，以溴化铅和溴化铯为原料，采用多步旋涂法在硅和FTO衬底上制得CsPbBr₃薄膜。利用光弹调制式椭偏光谱仪对硅衬底上的薄膜进行了椭偏光谱分析，使用Tanguy和Tauc-Lorentz 3组合模型对变角度的椭偏光谱进行参数拟合，得到了薄膜光学常数在1.00 eV～5.00 eV范围内的色散关系，并利用荧光发射光谱、吸收谱验证椭偏拟合结果。结果表明，多步旋涂法制备的CsPbBr₃薄膜的光学常数与其它方法相比具有一定的差异性，其中折射率可能与薄膜表面粗糙度呈负相关；椭偏拟合所得带隙为2.3 eV,验证了荧光光谱、吸收谱的计算结果。该研究为多步旋涂法制备的CsPbBr₃薄膜椭偏光谱拟合分析提供了参考。     

Temperature-dependent Sellmeier coefficients and chromaticdispersions for some optical fiber glasses

Temperature-dependent Sellmeier coefficients are necessary to optimize optical design parameters of the optical fiber transmission system. These coefficients are calculated for fused silica (SiO₂ ), aluminosilicate, and Vycor glasses for the first time to find the temperature dependence of chromatic dispersion at any wavelength from UV to 1.7 μm. The zero dispersion wavelength λ₀ (1.273 μm for SiO₂, 1.393 μm for aluminosilicate, and 1.265 μm for Vycor glasses at 26°C) varies linearly with temperature, and dλ₀/dT is 0.03 nm/K for aluminosilicate and Vycor glasses, whereas for SiO₂ it is 0.025 nm/K. This study interprets the recently observed experimental value of dλ₀/dT for two dispersion shifted optical fibers; and the dominantly material origin of dλ₀/dT is confirmed here as a fundamental property of the optical fiber glasses     

不同退火氛围对TiN/HfO2/SiO2/Si结构电荷分布的影响

热退火技术是集成电路制造过程中用来改善材料性能的重要手段。系统分析了两种不同的退火条件(氨气氛围和氧气氛围)对TiN/HfO₂/SiO₂/Si结构中电荷分布的影响,给出了不同退火条件下SiO₂/Si和HfO₂/SiO₂界面的界面电荷密度、HfO₂的体电荷密度以及HfO₂/SiO₂界面的界面偶极子的数值。研究结果表明,在氨气和氧气氛围中退火会使HfO₂/SiO₂界面的界面电荷密度减小、界面偶极子增加,而SiO₂/Si界面的界面电荷密度几乎不受退火影响。最后研究了不同退火氛围对电容平带电压的影响,发现两种不同的退火条件都会导致TiN/HfO₂/SiO₂/Si电容结构平带电压的正向漂移,基于退火对其电荷分布的影响研究,此正向漂移主要来源于退火导致的HfO₂/SiO₂界面的界面偶极子的增加。     

Thin-gate SiO2 films formed by in situ multiple rapidthermal processing

A reliable method of forming very thin SiO₂ films (<10 nm) has been developed by rapid thermal processing (RTP) in which in situ multiple RTP sequences have been employed. Sub-10-nm-thick SiO₂ films formed by single-step RTP oxidation (RTO) are superior to conventional furnace-grown SiO₂ on the SiO₂ /Si interface characteristics, dielectric strength, and time-dependent dielectric-breakdown (TDDB) characteristics. It has been confirmed that the reliability of SiO₂ film can be improved by pre-oxidation RTP cleaning (RTC) operated at 700-900°C for 20-60 s in a 1%HCl/Ar or H₂ ambient. The authors discuss the dielectric reliability of the SiO₂ films formed by single-step RTO in comparison with conventional furnace-grown SiO₂ films. The effects and optimum conditions of RTC prior to RTO on the TDDB characteristics are demonstrated. The dielectric properties of nitrided SiO₂ films formed via the N₂O-oxynitridation process are described     

Visible light emitting Si/SiO2 superlattices

Six-period superlattices of Si/SiO₂ have been grown at room temperature using molecular beam epitaxy. With this mature technology, the ultra-thin (1–3 nm) Si layers were grown to atomic layer precision. These layers were separated by 1 nm thick SiO₂ layers whose thickness was also well controlled by using a rate-limited oxidation process. The chemical and physical structures of the multilayers were characterized by cross-sectional TEM, X-ray diffraction, Raman spectroscopy, Auger sputter-profile, and X-ray photoelectron spectroscopy. The analysis showed that the Si layer is free of impurities and is amorphous, and that the SiO₂/Si interface is sharp (0.5 nm). Photoluminescence (PL) measurements were made at room temperature using 457.9 nm excitation. The PL peak occurred at wavelengths across the visible range for these multilayers. The peak energy position E was found to be related to the Si layer thickness d by E (eV) = 1.60+0.72d⁻² in accordance with a quantum confinement mechanism and the bulk amorphous-Si band gap.     

Roughness-enhanced reliability of MOS tunneling diodes

Both electrical and optical reliabilities of PMOS and NMOS tunneling diodes are enhanced by oxide roughness, prepared by very high vacuum prebake technology. For rough PMOS devices, as compared to flat PMOS devices, the Weibull plot of T_BD shows a 2.5-fold enhancement at 63% failure rate, while both the D₂ and H₂-treated flat PMOS devices show similar inferior reliability. For rough NMOS devices, as compared to flat NMOS devices, the Weibull plot of T_BD shows a 4.9-fold enhancement at 63% failure rate. The time evolutions of the light emission from rough PMOS and NMOS diodes degrade much less than those of flat PMOS and NMOS diodes. The momentum reduction perpendicular to the Si/SiO₂ interface by roughness scattering could possibly make it difficult to form defects in the bulk oxide and at the Si/SiO₂ interface by the impact of the energetic electrons and holes     

溶胶-凝胶膜光学表面激光清洗工艺研究

为了解决光学元件表面的颗粒污染问题，在单发次激光干式清洗的基础上，提出了气流置换系统辅助的激光清洗方法，使用波长为355nm的Nd:YAG激光器，针对镀溶胶-凝胶SiO₂薄膜熔石英光学表面粒径为1μm~50μm的典型SiO₂颗粒污染物，进行了理论分析和清洗实验，取得了可用于激光清洗的工艺参量。结果表明，对于镀溶胶-凝胶膜熔石英样品的单发激光干式清洗，最佳激光能量密度为2.29J/cm2，与未镀膜石英的激光清洗工艺参量存在一定差异；在最佳工艺参量下，单发次激光清洗对于粒径1μm以上的SiO₂颗粒清洗效果明显，移除率可达82.96%；当污染密度过高时会导致清洗效果的减弱及对基底的损伤，而气流置换系统辅助的激光清洗方法可进一步增强对光学表面颗粒污染的去除效果。该研究对大型高功率固体激光装置中的光学元件在线清洗及清洗装备的设计具有重要的研究意义与实用价值。     

Silicon–O–M–O–silicon superlattice

The success of heterojunction quantum wells and quantum dots in III–Vs has not been extended to silicon because the ideal barrier, SiO₂, is amorphous, preventing the formation of quantum structures with silicon. The possibility of a few monolayers of oxide inserted between adjacent silicon layers was proposed and realized with a superlattice (SL) structure consisting of Si–Si–O–Si–Si–Si, having a monolayer of oxygen in each period introduced by adsorption onto the 2×1 reconstructed surface along the Si(1 0 0). Reduction of the period leads to a slight up-shift of the energy of the emitted light, indicating that the essential objective of boosting the optical transition by promoting direct transitions has not been realized. Annealing in H₂+O₂ results in significant improvement in PL and EL, showing that specific defects, e.g., Si–O complexes may be responsible for the observed light emission. The role of Si–O complex being the origin of emission is further supported by the observation that the emission of visible light from polycrystalline Si and SiO₂ structure is similar to the epitaxial superlattice with oxygen. The computed strain in a new type of superlattices consisting of SiO₂, and GeO₂ is much lower than the Si–O SL. The EL in Si–O superlattice with the use of a Schottky barrier to provide electron–hole accumulation allows double injection into states higher in energy than the bandgap of Si, a prerequisite for injection laser without the need to use a wide-band pn-junction.     

Ge-laminated polarizer for 790-850-nm applications with extinctionratios larger than 50 dB and insertion losses below 0.4 dB

A laminated structure composed of alternating layers of Ge and SiO ₂ (Ge-LAMIPOL) is usable as a miniaturized polarizer at 790-850 nm in fiber optic gyroscopes, for instance. However, peeling of the sputter deposited layers, due to the weak binding strength between Ge and SiO₂, is a serious problem during the slicing process in preparation for assembly in the fiber. In order to improve the adhesive strength, the 1-nm-thick Si layers are inserted between Ge and SiO₂ layers. The Si layer functions as the adhesive layer via the Si-O bonding formation in place of the weaker Ge-O bond. The new Ge-LAMIPOL, including 125 layers of (1-nm Si)/(4.5-nm Ge)/(1-nm Si)/ (800-nm SiO₂) is successfully prepared without any fracture. The extinction ratio and the insertion loss were 51 and 0.33 dB at 850 nm, respectively, corresponding to 59.6 and 0.36 dB as the designed values