Development of a subsystem and algorithm for alignment and overlay improvement for microlithography

Microlithography is the process of transfer of minute electronic circuit patterns from a template (also called a mask) onto the surface of wafers for the manufacture of integrated circuits (ICs). A typical IC consists of seven to twenty layers of patterns. In this process it is essential to align patterns to the previous layer with an accuracy of better than the minimum feature size on the patterns. Failing to meet alignment targets leads to failure of ICs and hence a reduced process yield. Manufacturing processes in mask making, wafer processing and the process of pattern transfer itself cause the distortion of patterns. Technologies that work with smaller pattern sizes (below 65nm) like imprint lithography hence have more stringent requirements. Hence, overlay alignment is a huge challenge and a very important process of lithography. This thesis discusses the development of a sub-module concept that can be added to commercial lithography machines to reduce alignment and overlay errors and hence improve overlay. The concept makes use of the fact that the masks used in optical and imprint lithography are 250 mil thick plates of fused silica, a mechanically robust substrate as compared to X-ray masks which are made of thin silicon films. Distortion correction methods have been disclosed for X-ray lithography in literature claiming to control magnification and distortions by applying forces on the ends of the thin silicon film. A similar idea is much easier to implement on optical lithography masks. This thesis discusses a method to correct alignment and overlay errors by the application of a number of forces on the sides of the optical mask. The mask was modeled using Finite Element Methods. The development of optimization routines to minimize distortions is discussed. Overlay improvement results with actual pattern placement data are presented. The results show that significant reduction in distortions is achievable through this method.