Real-time scheduling of semiconductor integrated single-wafer processing tools

Abstract

Clustering technology is usually used in recent semiconductor industry to prevent several sequential process steps from exposing to air and to achieve better productivity of equipment. The equipment that performs two or more sequential processes in a single unit is called integrated single-wafer processing tools. The typical examples of the integrated single-wafer processing tools are cluster tools and photolithography processing tools. The trend toward the integrated single-wafer processing tools in wafer fabs accelerates the necessity of a sophisticated and dedicated scheduler that can manage them more autonomously and flexibly. Thus, this dissertation addresses crucial problems for real-time scheduling of the semiconductor integrated single-wafer processing tools. The first part of this dissertation proposes the deadlock-free scheduling approach that can perform real-time scheduling and deadlock management in efficient way for the integrated single-wafer processing tools in spite of dynamic contingencies. First, a novel framework is introduced where deadlock management problem is decomposed into sub-problems for computational efficiency, and then they are integrated with real-time scheduling algorithms. In other words, deadlock-prone modules and deadlock-safe modules are identified in the context of the deadlock management, and then only deadlock-prone modules are focused to guarantee deadlock-freeness when scheduling algorithms are applied to the entire modules. To realize the proposed framework, a Resource Request Matrix is newly introduced to represent operational states. The Resource Request Matrix is a representation useful in identifying deadlock-prone modules and applying deadlock management algorithms, because deadlock conditions are expressed explicitly in this representation. This dissertation also presents algorithms with polynomial complexity to identify deadlock-prone modules, and algorithms to manage deadlock considering machine failures. The...