Dual-armed cluster tools for semiconductor manufacturing typically have had two arms fixed in opposite directions. Recently, new cluster tool robot systems with two independent robot arms have been introduced with the expectation that the arms' flexibility will improve the throughput. However, the productivity gain has yet to be examined. Accordingly, we examine under which circumstances and the extent to which productivity gains can be achieved and how the robot task sequences should be scheduled to maximize the throughput. For this purpose, we develop a Petri net model that represents the tool behavior. We show that the well-known swap sequence, which is known to be optimal for conventional dual-armed tools, is not always optimal. Instead, we identify two other sequences that are optimal under certain conditions. We define the workloads for each process step and the transport module to derive conditions for optimality of the sequences, based on the Petri net model and the workload. We also develop a mixed integer programming (MIP) model to determine optimal sequences among one-cyclic schedules for the cases in which the proposed sequences are not optimal. Furthermore, we analyze and demonstrate how the two independent arms can increase the throughput in comparison to a conventional dual-armed robot.