Chemical information extraction from scanning electron microscopy images on the basis of image recognition

Traditional resist materials have faced challenges as the extreme ultraviolet (EUV) light source with a wavelength of 13.5 nm brought the evolution of lithography to the semiconductor industry. A significant issue in the development of resist materials or the discovery of new type resists is that numerous parameters involved in the resist pattern printing process cause the generation of defects. Meanwhile, the inherent chemical variation in resist materials and processes causes the stochastic defects. In addition, the stochastic defects caused by the inherent chemical variation in resist materials and processes become increasingly significant as feature scales continue to shrink. Consequently, the number of pattern data with failures is much greater than those without defects. However, by utilizing the information contained in pattern failures, chemical parameters can be adjusted to improve resist resolution. In this study, a new method is proposed for evaluating resist patterns with defects by fitting the experimental scanning electronic microscopy (SEM) images of line-and-space patterns with defects to simulated images.

Automated summary

Traditional resist materials face challenges with the evolution of lithography in the semiconductor industry due to the introduction of extreme ultraviolet (EUV) light source. The development of resist materials and the discovery of new types are hindered by the generation of defects caused by numerous parameters involved in the resist pattern printing process. Inherent chemical variation in resist materials and processes result in stochastic defects, which become more significant as feature scales decrease. This study proposes a new method for evaluating resist patterns with defects by fitting experimental scanning electron microscopy (SEM) images of line-and-space patterns to simulated images.