【作者】 李宁；

【导师】 赵学增；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2010， 博士

【摘要】 随着集成电路制造业的迅速发展,芯片内的刻线宽度作为其工艺特征尺寸已下降到100nm以下量级。对纳米尺度刻线尺寸越来越严格的控制要求使得半导体刻线边缘粗糙度(Line edge roughness,LER)的测量与控制成为光刻工艺和集成电路制造业关注的热点问题之一。原子力显微镜(Atomic force microscopy,AFM)以其良好的三维成像精度以及对样本和测量环境的广泛适用性而成为纳米尺度线边缘粗糙度测量研究的重要工具,本文针对目前AFM测量LER所存在的一些问题,进行了纳米尺度线边缘粗糙度测量与表征方法的研究。由于目前LER仍没有公认的明确定义,无法进行测量结果的有效比较,这对加工工艺的监控及优化、器件性能仿真模型的建立与测试、线宽测量精度的控制和提高极为不利。为解决这一问题,本文从半导体制造工艺与LER的形成、LER对线宽测量和器件电气性能的影响三个方面对测量需求进行分析,从测量学角度提出了一个新的线边缘粗糙度定义,并结合线边缘粗糙度的自身特点与应用要求给出了具体的表征参数,包括幅值参数均方根粗糙度RMS、频谱分析参数功率谱密度函数PSD和空间评定参数自相关函数R(τ)。根据所提出的LER定义并考虑到AFM扫描图像的特点,建立了一个合理的LER测量模型,其主要任务包括确定线边缘位置和评定基准以提取LER特征。首先研究了基于图像处理技术的线边缘确定方法,该方法综合考虑了侧墙形貌及其形位信息,确定出由边缘检测算子和高度阈值所决定的局部高度范围内的线边缘。由于该方法中人为选取参数会对LER测量结果产生影响,因此又提出了一种基于关键点的线边缘确定方法,利用相邻数据点之间的变化率确定出每一条扫描线上的6个关键点,并通过所有扫描线上的关键点构成不同局部高度区域的线边缘。采用最小二乘中线作为评定基准来提取LER特征,并计算出表征参数,包括左、右单线边缘及线宽均方根粗糙度值σL、σR和σLWR,以及左、右线边缘相关长度LcL、LcR。用上述两种方法对同一样本的AFM测量数据进行处理,对测量结果进行了比较和分析。针对现有LER表征参数不能完全满足测量要求这一问题,提出了一种基于冗余第二代小波变换的LER多尺度表征方法,该方法将LER特征分解到各尺度包含不同空间频率成分的独立频带中,对各尺度重构信号作进一步分析,从而确定出粗糙度指数R、各尺度上重构信号的空间频率分布特性及均方根值,并用最小尺度空间的细节信号估计测量噪声的影响。采用随机轮廓仿真数据对LER多尺度表征方法进行了验证,结果表明,对具有相近幅值的线边缘,粗糙度指数R可以有效反映出刻线边缘的不规则程度,并能较直观地提供LER空间频率的分布信息。单晶硅刻线样本实验数据的分析结果进一步表明,LER多尺度表征方法可以有效分析对IC器件性能影响较大的LER高频成分,并可以对具体空间频率范围的LER幅值进行量化表征,从而为半导体制造工艺的监控与优化提供更完善的测量基础。根据AFM扫描方式和成像特点,对AFM测量LER的影响因素,如探针针尖尺寸与形状、AFM扫描图像的噪声、压电晶体驱动精度、扫描采样间隔等进行了理论研究和实验分析。对各因素的影响程度进行了探讨并分别提出了抑制及修正方法,从而减小测量误差,提高测量结果的可信度。更多还原

【Abstract】 With the rapid development of semiconductor integrated circuit (IC) manufacturing, the line width of chip has been shrinked into sub-100nm as the critical dimension of process technologies. The more and more stringent control request on the nano-scale dimension of line structure makes the measurement and control of line edge roughness (LER) to be one of the serious issues in the field of lithography processes and IC industry. Atomic force microscopy (AFM) has become an important technique for its good 3D imaging capability, its insensitivity to sample materials, and its capability for working at ambient conditions. Aimed on some problems existing in LER measurement using AFM, this paper studied nano-scale LER measurement and charactrization method.Because there is no unambiguous definition of LER acceptable for the research field at present, the measurement results can not be compared effectively, which is not be propitious to monitoring and optimization of lithographical process technologies, establishing and testing the device performance simulation model as well as controlling and improving the precision of line width measurement. For the purpose of resolving this problem, in this paper based on the demand analysis of three aspects such as the semiconductor manufacturing technologies and LER formation, the impact on the linewidth measurement and electric performance of device, a new LER definition is proposed from the perspective of measure. Additionally combined with the intrinsical characteristics and application request of LER, the specific characterization parameters, including the amplitude parameters root mean square roughness value (RMS), spectral analysis parameter power spectral density (PSD) and spatial evaluation parameter autocorrelation function R(τ), are also provided.According to the proposed LER definition and considering the feature of AFM scanning image, a reasonable measurement model of LER is established. The main task of the model contains the determination of line edge location and evaluation benchmark in order to extract the LER feature. Firstly the edge detection method based on image processing technique is studied. In this method the sidewall topography and its location information are considered synthetically, and the line edge within the local height range is determined by the edge detection operator and height threshold value. Considering there exist the artificial chosen parameters in this method, which have an influence on LER measurement results. So the edge detection method based on essential points is presented. In this method 6 essential points are determined at each scan line profile using the change rate of adjacent data points, and thus the line edge points of different height region composed of essential points of all scan lines can be obtained. Each single line edge is fitted by a least squares line as the evaluation benchmark to extract the LER feature, and the characterization parameters are calculated, including the RMS value of left and right single line edge and line width roughnessσL,σR andσLWR, and the correlation length LcL and LcR. The AFM measurement data of the same sample is processed by these two methods, and analysis and comparison of measurement result are made.Aimed at the existing LER characterization parameters can not completely satisfy the measure requests, the multi-scale characterization method of LER based on redundant second generation wavelet transform (RSGWT) is proposed. The LER feature is decomposed into the separate spatial frequency range which contains different spatial frequency components, then the resconstructed signals at the different scale are analyzed furthermore, accordingly determining the roughness exponent R, the spectral property and RMS value of reconstructed signal at each scale, moreover the influence of high frequency noise is evaluated by means of the detail signal at the smallest scale.The LER multi-scale characterization method is testified using the random profile simulation data. The results show that for the line edge with approximately amplitude, the roughness exponent R can reflect the irregularity of line edge effectively, and provide the intuitionistic information of LER spatial frequency distribution. The results of experimental data of single crystal silicon show that LER multi-scale characterization method can effectively analyze high frequency LER which have greater effect on the IC device performance, and quantatively characterize the LER amplitude in the spectific spatial frequency range, accordingly provide the more perfect measurement base for the monitoring and optimizing the semiconductor manufacturing technologies.Based on the feature of AFM scanning mode and imaging, the influence factors of LER measurement by AFM are studied. The theortical and experimental analysis of the influence of various factors on LER measurement, including the size and shape of probe tips, the signal noises of AFM scanning images, the driving precision of piezo crystal, scan and sampling interval and so on. The error sources of nano-scale LER measurement and their effect degree are analyzed, and the methods for restraining and amending the errors are resepectively proposed to reducing the measurement error and improving the reliability of measure result.更多还原