Scheduling cluster tools for concurrent processing of multiple wafer types with identical job flows

Abstract

This paper discusses the cyclic scheduling problem of multiple wafer types processing, a problem where several types of wafers are concurrently produced with chambers sharing in a cluster tool. With the widespread small-lot production of a variety wafer types, modern fabs require concurrent processing of various types of wafers. The optimality of robot task sequence and the parallel chamber utilization are examined. We present robot task sequence based on the backward sequence and swap sequence for concurrent processing of multiple wafer types. Such sequences widely used in identical wafer types processing for single-armed cluster tool and dual-armed cluster tool, respectively, due to its simple implementation and robustness. We show that the backward and swap sequences are always optimal for a given cycle plan and configuration. Then, we show that a proper cycle plan increases the utilization rates of processing chambers, and the tool productivity can be maximized accordingly. More specifically, we identify that the tool cycle time is minimized when the utilization rates of parallel chambers are evenly balanced. Finally, we suggest a novel scheduling strategy that determines an efficient cycle plan, and identify the conditions for which the proposed cycle plan achieves the optimal tool performance. Through extensive experimental experiences, it is verified that the suggested scheduling method achieves the optimal tool throughput for most of the practical tool conditions. This research also examines wafer delay issues of dual-armed cluster tools with concurrent processing of multiple wafer types. Recently, circuit widths of wafers have been significantly reduced and their designs have become extremely complicated. As a result, wafer quality control has become one of the most important issues in wafer manufacturing processes. When a wafer resides within a processing chamber for a long time after the processing is finished, the wafer is prone to have quality problems due to residual gases and heats. When multiple wafer types are concurrently processed in a tool, such quality problems become much more critical as each wafer type should competitively use the robot and chambers. To address and resolve such wafer delay issues in cluster tools with multiple wafer types, we first identify that such wafer delays are caused by two different factors. From these, we suggest a new insight that a wafer delay can occur even for the chambers of the bottleneck process step. Then, we show that the wafer delays as well tool cycle time can be reduced by balancing the workloads of parallel chambers. Once the workloads of parallel chambers are balanced, we also confirm that the wafer delays can be further reduced by imposing intentional postponements to the start times of robot tasks. This study provides practical and fundamental insights on how the workload balancing in cluster tools can improve the tool throughput as well as the wafer quality.