Performance models of systems of multiple cluster tools

Abstract

Over the last 15 years, many semiconductor processes have migrated to single-wafer vacuum cluster tools. Although performance models of these tools are found in the literature, they are usually restricted to the analysis of single tools, even if, inevitably, several such tools may be configured into systems, and operated as such. The objective of this work is to determine, using an equation-based model, the optimal configuration and operation of systems of cluster tools in the presence of scheduled maintenance. The two extremes in the spectrum of possible cluster configurations are the serial configuration, in which the modules in a cluster tool are all different (each representing a step in process sequence), and the more popular parallel configuration, in which the modules in a tool are identical. This paper predicts that systems of parallel-configured tools can offer higher throughputs than their serial-configured counterparts. However, this advantage may be slight when equipment downtime is relatively schedulable and infrequent, in which case the serial configuration may be preferable because of its superior cycle times. We also derive the optimal lot size and lot release policy for systems of cluster tools. We conclude that cluster tools will gradually migrate from parallel configurations to serial as cluster tools become more reliable and/or cycle time becomes more important.