Performance evaluation of photolithography cluster tools

Abstract

The photolithography cluster tool is typically the most expensive tool set utilized in the production of semiconductor wafers and is often selected as a fabricator bottleneck. Modeling such a tool as a serial processing cluster tool, we deduce measures of tool performance. Queueing models reveal that the mean cycle time in the presence of a Poisson arrival process is related to the parallelism inherent in the system configuration. As a consequence, the normalized mean cycle time behavior has a different form than that of the standard single server queue. The process time of a lot and the throughput are evaluated in the presence of disruptions common in practical manufacturing environments. For multiple products with different process rates, it is shown that the throughput is not influenced by the order in which the lots are sequenced.