公开(公告)号：EP4418042A1

公开(公告)日：2024-08-21

申请号：EP23156643.1

申请日：2023-02-14

申请人： ASML Netherlands B.V.

发明人： SHENG, Zhifeng ,  BAJALAN, Nastaran ,  JACOBS, Jort, Adrianus, Thomas

IPC分类号： G03F7/00 ,  G06N3/0455 ,  G06N3/0895 ,  G06N20/00 ,  H01L21/66

CPC分类号： G03F7/706845 ,  G03F7/70525 ,  G03F7/70633 ,  G03F7/70616 ,  G06N3/0895 ,  G06N3/0455 ,  H01L22/20 ,  H01L22/12 ,  G06N3/047 ,  G06N3/096 ,  G06N3/09

摘要： A method for predicting process information characterizing a lithographic process from image data. The neural network is configured to receive image data of a patterned portion of a substrate and to generate predicted process information. The method comprises obtaining a plurality of labeled training items, each training item including training image data, a domain label and process information. The method further comprises processing the training image data to generate a domain latent representation, a process information latent representation and a residual latent representation. The method further comprises generating predicted image data, predicted domain label and predicted process information. The method further comprises generating a training domain latent representation and a training process information latent representation. The method further comprises updating weights of the neural network to reduce discrepancies between the labeled training items and the predicted image data, domain label and predicted process label.