Scheduling cluster tools for concurrent processing of multiple wafer types

Abstract

Modern semiconductor manufacturing industry has widely adopted cluster tools for improving wafer quality and productivity. A cluster tool consists of several processing modules (PMs) and a transport robot in a closed environment. Most existing scheduling works on cluster tools assume that the tool keeps processing identical wafers. However, in practice, such assumption might not be valid no longer due to concurrent processing of wafers to utilize PMs better. Whereas cluster tools have 4-8 PMs, wafer recipes or wafer flow patterns often use only 2-3 process steps in a tool for earlier inspection and metrology to satisfy strict wafer quality requirement. Therefore, nowadays, fabs confront this issue by considering the concurrent processing of multiple wafer types for production tools. Thus, in this thesis, we examine the possibilities of concurrently processing multiple wafer types in a cluster tool. In addition, for easy implementation of scheduling rules, we examine a scheduler architecture of cluster tools and propose an open standard for specifying and exchanging the scheduling rules, the wafer recipe, and the scheduling requirements. In Chapter 2, we examine the possibilities of concurrently processing two wafer types in a cluster tool. Due to restriction on the number of load ports, only two wafer types can be concurrently processed in practice to increase the utilization of the tool. We therefore examine a cyclic scheduling problem of cluster tools that concurrently process two wafer types. For a given cycle plan, we wish to determine the robot task sequence so as to minimize the tool cycle time. When a single wafer type is processed, the backward and swap sequences are optimal for single-armed and dual-armed tools, respectively, and are being prevalently used because of simplicity and robustness. To maintain such advantages in concurrent processing, we introduce and define concurrent backward and swap sequences (CBSs and CSSs, respectively). We then deve...