Control of multiproduct bulk server diffusion/oxidation processes. Part 2: Multiple servers

We investigate how knowledge of future arrivals can be used to control bulk server diffusion and oxidation processes in semiconductor manufacturing to reduce the average waiting time of lots. While past research has dealt with the control of bulk server queueing systems, only a few studies have addressed the use of knowledge of future arrivals, and those studies were limited to a single server system. We extend prior strategies for the single product-single server case to a multiple product-multiple server case, and devise a control strategy that is tested through the use of simulation. The performance of the new policy is compared to that of the optimal control strategy ignoring future arrivals (i.e., a Minimum Batch Size strategy). Results indicate that the new strategy performs well under a wide variety of circumstances. To demonstrate the control strategy performance in a realistic setting, a detailed simulation model of the diffusion area of an existing wafer fab was developed. The model was run with several start rates and the results compared to those from a Minimum Batch Size strategy. Results indicate that the new strategy performs well over a wide range of start rates.