Scheduling And Simulation Of Single-arm Cluster Tools With Wafer Reentrant Processes

Nowadays, the Internet and informationization in the world play a very important role in the economic and social life, which is greatly dependent on semiconductor manufacturing Thus, development level of semiconductor manufacturing is an important symbol of national economy and technology. With the increase of wafer size in semiconductor manufacturing, there is an increasing use of cluster tools for most wafer factories. A cluster tool is composed of a number of processing modules and a wafer handling robot. With the effective integration of wafer processing and transportation, it improves productivity, makes the best use of work spaces, and reduces contamination. Cluster tool is automated and tightly coupled equipment that operates in a vacuum environment. In a cluster tool, there is no intermediate buffer between the process modules, which makes the scheduling of a cluster tool very complicated. Semiconductor fabrication needs to take many factors into consideration such as wafer residency constraints, reentrant process, and large number of processing steps, etc. Thus, it is more challenging to schedule a cluster tool with reentrant process being taken into consideration. This paper addresses the scheduling problem of single-arm cluster tools with wafer reentrant process.Atomic layer deposition (ALD) is a typical revisiting process in wafer fabrication. For such processes, the wafers need to visit some processing modules a number of times. For such a process, it is deadlock-prone, which makes the scheduling of the system much complicated. Petri net is widely used not only in simulation and control of discrete event system, but also it is a math and graphic modeling tool. In this work, A Petri net model is developed for single-arm cluster tools to describe wafer reentrant process. With the Petri net model and the optimal schedule for the revisiting process, study is conducted to schedule the overall system. Often the cycle time of the revisiting process is longer than the workload of other steps. Thus, it is desired to schedule the system such that cycle time for the overall system equals to that for the revisiting process. Conditions are presented to schedule the system in this way. When the conditions are not satisfied, strategies are given to schedule the system such that the cycle time is minimized. Because Petri net is hard to understand for practitioners and cluster tools are costly in purchasing and operation, an on eM-Plant-based system is developed to simulate the scheduling process. With this system, optimal cycle time of the overall system can be obtained by simulation. Through the embedded Simtalk language to change the number of wafer reentrant times, it provides an effective tool for engineers to verify the correctness. Illustrative examples are given to show the effectiveness of the proposed scheduling method.