应用化学 ›› 2021, Vol. 38 ›› Issue (9): 1199-1208.DOI: 10.19894/j.issn.1000-0518.210121

• 研究论文 • 上一篇    

基于卷积运算的嵌段共聚物自组装图像缺陷分析统计方法

田昕1,2, 赖翰文1,2, 刘亚栋1*, 季生象1,2*   

  1. 1中国科学院长春应用化学研究所,中国科学院生态环境高分子材料重点实验室,长春 130022
    2中国科学技术大学, 合肥 230026
  • 收稿日期:2021-03-17 接受日期:2021-04-12 出版日期:2021-09-01 发布日期:2021-09-06
  • 通讯作者: *E-mail:ydliu26@ciac.ac.cn; sji@ciac.ac.cn
  • 基金资助:
    国家自然科学基金(Nos.51973212, 51773201)、吉林省科技发展计划中青年科技创新领军人才及团队项目(No.20200301017RQ)、长春市科技计划科技创新“双十工程”重大科技攻关项目(No.19SS005)、吉林省与中国科学院科技合作高技术产业化专项资金项目(No.2019SYHZ0002)资助

Analysis of Defects in Block Copolymer Films by a Convolution Algorithm

TIAN Xin1,2, LAI Han-Wen1,2, LIU Ya-Dong1*, JI Sheng-Xiang1,2*   

  1. 1Key Laboratory of Polymer Ecomaterials, Changchun Institute of Applied Chemistry Chinese Academy of Sciences, Changchun 130022, China
    2University of Science and Technology of China, Hefei 230026, China
  • Received:2021-03-17 Accepted:2021-04-12 Published:2021-09-01 Online:2021-09-06
  • Supported by:
    National Natural Science Foundation of China (Nos.51973212, 51773201), the Department of Science and Technology of Jilin Province (No.20200301017RQ), the Bureau of Science and Technology of Changchun (No.19SS005) and the Joint Program of CAS-Jilin Province (No.2019SYHZ0002)

摘要: 嵌段共聚物引导自组装是制备sub-10 nm尺寸半导体器件的潜在技术方案之一。 然而由于缺少高效准确的实验室分析工具,难以对嵌段共聚物自组装/引导自组装的结构进行深入定量分析,特别是对其组装结构的缺陷密度及线边粗糙度等进行定量研究。 本文参考图形分析中的图像增强算法和模板匹配算法,采用二值化和骨架化算法对嵌段共聚物自组装/引导自组装图像进行处理,构建相应的卷积核并利用卷积运算分析和定位组装结构中的缺陷,并对缺陷密度进行统计。 经验证,该算法能够减少噪点对统计结果的影响,区分图像边缘截断所产生的断点和组装缺陷,快速准确地自动统计自组装/引导自组装图像中的端点、分叉和岛缺陷结构。 同时对分析所需的图片放大倍率和幅面尺寸极限进行了探索,结果表明嵌段共聚物本征相分离周期与像素尺寸的比值至少为6.69, 有效统计嵌段共聚物自组装缺陷所需的视场面积至少为1.5 μm2。 缺陷分析速度测试表明,本文算法相较于邻近像素判断法快136~147倍。

关键词: 嵌段共聚物, 自组装, 引导自组装, 卷积, 缺陷统计

Abstract: Directed self-assembly (DSA) of block copolymers (BCP) is one of the potential methods to manufacture sub-10 nm structures for semiconductors. However, the lack of tools in labs makes it difficult to quantitatively analyze defects and defect densities in BCP films prepared by self-assembly and DSA. Inspired by the general image processing method, image enhancement and template matching, a convolution algorithm was developed to detect defects in the scanning electron microscopy (SEM) images of BCP films and count the densities of each types of defects. Defects and artificial noises are accurately distinguished as verified by hand. The dot, terminal point and junction defects are automatically counted without any error in most SEM images by this algorithm. In order to get the reliable and reproducible results, the ratio of the BCP period (L0) to nanometers per pixel (NPP) needs to be >6.69 and the smallest image area is 1.5 μm2. Finally, direct comparison of the two algorithms on our workstation shows that the compute speed of the convolution algorithm is about 136~147 times faster than that of the adjacent pixel determination algorithm.

Key words: Block copolymer, Self-assembly, Directed self-assembly, Convolution, Defect statistics

中图分类号: