Dose control strategy using random logic device patterns and massive metrology in a foundry high volume manufacturing environment

One of the most critical challenges for lithography process is to effectively control all critical patterns over one field, across wafer, and from lot to lot consistently. With design rules shrink, the lithography process has become more vulnerable, and the critical patterns control becomes more and more challenging. ASML Imaging optimizer (IMO) dose control has been widely adopted to control CDU of critical patterns, with designed marks or patterns in most of the situations [1–3]. However, this traditional method has a few weaknesses at logic foundry environment. The first, the designed marks are often not representative of random logic critical device patterns, thus leads to a situation that marks are controlled well, while device patterns are not. The designed marks normally put more constrains on device design layout, which is not wanted in some cases. In large die case, where there is limited space to place marks, the traditional dose control method can become even more limited. Moreover, the traditional method using limited data point is due to limited CD-SEM tool throughput. Weak point (WP) patterns are typically 2D patterns in critical BEOL layers. To measure just a single WP, metrology noise is normally high. Thus, to reduce the metrology noise, to measure more 2D weak point patterns and make an average are preferred for dose control. This requires massive amount of measurement. A high throughput metrology SEM tool to make an averaging to reduce noise is thus required. To address the problems, we have developed a method to use yield limiting device patterns to directly control dose thus improve CDU. To reduce the metrology noise of 2D patterns, a lot weak points per die have been selected which can well cover the full die. A high speed eBeam metrology tool, EP5, is used to measure all these identified weak points. To make this method to work effectively, a CDU budget breakdown (BB) has analyzed to identify and quantify CDU contributor[4,5], such as Local CDU (LCDU), metrology noise, reticle fingerprint, etc.. This can lead to optimized number of WP measurements to achieve optimal CDU correction balancing metrology tool time at HVM application.