Mask 3D model based on complex-valued convolution neural network for EUV lithography

The oblique incidence of extreme ultraviolet (EUV) light rays and the high aspect ratio of absorber layer on reflective mask induce pronounced mask three-dimensional (M3D) effects that significantly influence the imaging performance of EUV lithography. This paper proposes a fast learning-based M3D model dubbed complex-valued U-Net (CVU-Net) for EUV lithography simulation. The diffraction near-field (DNF) of EUV mask is calculated based on a set of complex-valued diffraction matrices, and each diffraction matrix can be rapidly synthesized using a well-trained CVU-Net. The network parameters are trained in a supervised manner. We set up a DNF dataset to train and test the proposed model. The comparison between the proposed method and some other M3D simulation methods is provided and discussed.