Scheduling and control of robotized semiconductor manufacturing equipment with reentrant job flows and time window constraints

Abstract

Modern semiconductor manufacturing equipment tend to integrate several processing modules and multiple wafer handling robots to improve wafer quality and reduce the flow time. Such robotized equipment include cluster tools and track equipment. Since such robotized integrated equipment have no buffer between the processing modules and in-progress wafers or cassettes cannot be returned into the loading buffers before all processes are completed, the scheduling problem is complicated. Advanced wafer processing processes require more scheduling constraints. First, some modern advanced processes such as atomic layer deposition require reentrant job flows that a wafer visits a single processing module more than once to form the specified deposition thickness. Such reentrant job flows may cause a deadlock for an improper robot task sequence. Second, some processes such as low pressure chemical vapor deposition have strict time window constraints that a wafer processed at a processing module should be unloaded from the processing module within a specified time limit. Otherwise, the wafer has severe quality problems due to residual gases and heat within the processing module``s chamber. Even if the wafer delay within a processing chamber does not have a strict upper limit, the longer or non-uniform wafer delays cause more quality troubles. Therefore, the wafer delays should be minimized or regulated. Third, integrated equipment for some processes such as wet cleaning have multiple robots and process multiple wafer jobs concurrently that have different flow patterns. Therefore, robot collisions and deadlocks might occur. We have a challenge for scheduling robotized, integrated semiconductor manufacturing equipment that have such constraints and complications. We examine scheduling problems for of robotized semiconductor manufacturing equipment with reentrant job flows and wafer delay constraints. We first examine a scheduling problem for a single-armed cluster tool with ...