Fast mask near-field calculation using fully convolution network

Near-field calculation for thick mask is a fundamental task in lithography simulations. This paper proposes a fully convolution network (FCN) method to improve the efficiency of three-dimensional (3D) mask near-field calculation compared to the rigorous electromagnetic field (EMF) simulation methods. Taking into account the 3D mask effects, the network is trained based on a set of mask samples and the corresponding near-field data obtained by the EMF simulator. During the testing stage, the trained FCN is used to rapidly predict the diffraction near-field of the testing mask patterns. The performance of the proposed FCN approach is evaluated by simulations based on EUV lithography.