A Mathematical Programming Approach for Determining Control Plans in Semiconductor manufacturing
Abstract
Worldwide competition, the move to ever smaller geometries in manufacturing processes, and the increasing of complexity in High-Mix semiconductor plants led to the introduction of numerous controls at different manufacturing stages. However, with the costs associated to metrology, i.e. non added value operations, it becomes increasingly important and challenging to reduce the number of controls for delivering products at high yield and at minimal costs. This has aroused a huge interest for dynamic sampling techniques. In this paper, an Integer Linear Programming model is proposed in order to determine key parameters for an optimal dynamic sampling plan. These key parameters correspond to the maximum risk that can be tolerated for each production tool while taking into account the metrology capacity and the current control plan. The key parameters are chosen to optimize the number of controls, reduce risks on production tools, and use metrology tools efficiently. The Integer Linear Programming model has been tested on real instances and validated through a smart sampling simulator prototype.