Optimization of diffusion furnace loading based on substrate and recipe selection

Traditional furnace batching in semiconductor manufacturing considers trade offs between wait-to-batch time and processing time to determine cycle reductions afforded by batching. An interesting aspect of this problem is the creation of potential batching opportunities through the appropriate insertion of substrates at the start of the line. In this paper we present a Mixed Integer Programming model as a decision support tool to provide substrate selection recommendations based on wafer start requirements with the objective of maximizing batching opportunities in the diffusion furnace, with consequent reductions in the furnace loading. Our MIP model takes into account the exclusivity requirements and aid the substrate material and recipe selection process for weekly starts in the Fab. We have achieved a 10% improvement in the overall furnace loading.