Investigation of Multiple Feedforward Modes for On Product Overlay Control

In semiconductor manufacturing process, large lot to lot (L2L) variation induced by multiple process steps is a big challenge for On-Product Overlay (OPO) control. Feedforward control has been proven to be an effective method to handle such a variation. However, different feedforward modes are needed depending on alignment and overlay schemes of the layers. In this paper, we report an investigation of using suitable feedforward modes for OPO control of a number of critical layers from a logic node. Different feedforward modes are defined according to differences in alignment and overlay tree to ensure the best results. Besides commonly used correction feedforward in semiconductor industry, we implement also de-correction fingerprint feedforward to some layers. We identify also that for some layers using partial feedforward is better than feedforwarding the whole correction or de-correction fingerprints. Our results show that combination of feedforward and feedback control is able to improve OPO by 40%~70% comparing to feedback-only one. With such a large improvement, the fab can reduce the number of pilot-run layers by 84%. This saves cost significantly. Integrating the aforementioned multiple feedforward control modes into the fab automation system has also been realized.