OCD系统中DSP和超分辨技术的研究与应用

发布时间：2018-01-14 18:25

本文关键词：OCD系统中DSP和超分辨技术的研究与应用　出处：《电子科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着半导体技术和半导体工艺制造技术的发展,工业上对半导体器件的关键尺寸的要求越来越高,这就为工艺过程中关键尺寸的正确检测提出了很高的要求。传统的尺寸测量方法包括扫描电子显微镜(SEM)、原子力显微镜(AFM)、扫描探针显微镜(SPM)等,而这些测量方法有原理简单的优点,但其缺点也很明显:测量精度不高、速度慢、对样品可能造成伤害等,这些缺点为其实现在线实时测量起到了极大的阻碍作用。相比较传统测量方法,光学关键尺寸(OCD)测量有着成本更低、精度更高、速度更快、对样品无损等的优点。OCD软件平台主要包括优化模块和模拟模块,优化模块负责对样本进行建模,模拟模块负责针对样本模型进行模拟计算分析其光学响应,并与实际得到的光学信息进行对比,然后在修改样本模型,重复上述步骤,直到模拟得出的光学响应与实验光学信息数据相吻合,这时候的模型的尺寸才是真实的结构尺寸。模拟模块根据精确的光学衍射理论,采用常用的严格耦合波分析法(RCWA)算法求解得到样本模型的光学响应,模拟模块的性能好坏直接决定了OCD软件平台的性能好坏。RCWA算法其实质是麦克斯韦方程组的求解,但是通常情况下RCWA都要计算进行大量的矩阵计算(具体的计算量视所取衍射次数),尤其是其中的特征值特征向量的求解非常耗费时间,这直接制约了模拟模块的效率。本文尝试使用TI公司生产的TMS320C6678 DSP芯片来进行对特征值特征向量的加速效果实现。DSP本身并不做通用计算用,但随着DSP技术的发展,其在乘法性能和存储器访问性能上得到了很大的提升,而这两项对矩阵运算非常有帮助,这是本文进行这个工作的重要原因;再加上DSP扩展容易、功耗低等的特点,使DSP运用到本文中的加速可行。随着微电子结构尺寸的进一步缩小,还有另外一种提高测量精度的方法即是使用超分辨技术。超分辨技术可以突破传统器件的衍射极限,实现较高的分辨效果。本文最后介绍了超分辨技术,并尝试运用于本课题中。
[Abstract]:With the development of semiconductor technology and semiconductor manufacturing technology, the key dimensions of semiconductor devices are becoming more and more important in industry. The traditional measurement methods include scanning electron microscope (SEM) and atomic force microscope (AFM). Scanning probe microscope (SPM) and so on, and these measuring methods have the advantages of simple principle, but their disadvantages are also obvious: the measurement accuracy is not high, the speed is slow, and the sample may be damaged. Compared with the traditional measurement methods, the optical critical dimension (OCD) measurement has lower cost, higher accuracy and faster speed. The OCD software platform mainly includes the optimization module and the simulation module, and the optimization module is responsible for modeling the sample. The simulation module is responsible for the simulation calculation and analysis of the optical response of the sample model, and compared with the actual optical information, then modify the sample model and repeat the above steps. Until the optical response obtained by the simulation is consistent with the experimental optical information data, the size of the model is the real size of the structure. The simulation module is based on the accurate optical diffraction theory. The optical response of the sample model is obtained by using the conventional strictly coupled wave analysis (RCWAA) algorithm. The performance of simulation module directly determines the performance of OCD software platform. The essence of RCWA algorithm is the solution of Maxwell equations. But usually RCWA has to calculate a large number of matrix calculations (the specific calculation depends on the number of diffraction), especially the eigenvalue of which the solution of eigenvector is very time-consuming. This directly restricts the efficiency of the simulation module. This paper attempts to use TI's TMS320C6678. DSP chip to implement the acceleration effect of eigenvalue eigenvector. DSP itself is not used for general calculation. However, with the development of DSP technology, its performance of multiplication and memory access has been greatly improved, and these two terms are very helpful to matrix operation, which is an important reason for this work. Combined with the easy expansion of DSP, low power consumption and other characteristics, DSP can be used in this paper to accelerate the feasibility, with the further reduction of the size of the microelectronic structure. Another way to improve the measurement accuracy is to use super-resolution technology, which can break through the diffraction limit of traditional devices and achieve higher resolution effect. Finally, this paper introduces the super-resolution technology. And try to apply in this subject.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN305

【相似文献】