FWave: a fast 3D resist model with dual-branch architecture for EUV lithography simulation.

Rigorous physical resist models are highly accurate but computationally expensive. Specifically, solving reaction-diffusion equations to predict inhibitor concentration severely limits simulation speed. Furthermore, compared to deep ultraviolet (DUV), extreme ultraviolet (EUV) lithography introduces stochastic effects, resulting in irregular pattern edges. To overcome these challenges, we propose FWave, a fast 3D resist model for inhibitor concentration prediction during PEB. FWave utilizes the dual-branch architecture: the Fourier neural operator (FNO) branch efficiently captures global features, while the discrete wavelet transform (DWT) branch extracts edge variations caused by EUV stochastic effects. Additionally, a spatially weighted mean squared error (WMSE) loss function is employed to accurately fit contact hole regions. Applied to EUV contact hole simulations, FWave demonstrates sub-nanometer critical dimension (CD) errors and achieves a approximately 2000x speedup compared to rigorous physical model.