Search papers, labs, and topics across Lattice.
The paper introduces Gen-Fab, a conditional GAN (cGAN) based on Pix2Pix, to predict fabrication variations in nanophotonic devices from design layouts. Gen-Fab incorporates a latent noise vector to enable one-to-many mapping, generating diverse, high-resolution predictions of fabrication outcomes that resemble SEM images. Experiments on out-of-distribution datasets demonstrate that Gen-Fab outperforms deterministic U-Nets, Monte Carlo Dropout U-Nets, and ensembles of U-Nets in both accuracy (IoU of 89.8%) and uncertainty modeling, showing better alignment with real fabrication outcomes.
Nanophotonic device design just got a digital twin: Gen-Fab accurately predicts fabrication variations, outperforming U-Net baselines and capturing uncertainty with impressive fidelity.
Silicon photonic devices often exhibit fabrication-induced variations such as over-etching, underetching, and corner rounding, which can significantly alter device performance. These variations are non-uniform and are influenced by feature size and shape. Accurate digital twins are therefore needed to predict the range of possible fabricated outcomes for a given design. In this paper, we introduce Gen-Fab, a conditional generative adversarial network (cGAN) based on Pix2Pix to predict and model uncertainty in photonic fabrication outcomes. The proposed method takes a design layout (in GDS format) as input and produces diverse high-resolution predictions similar to scanning electron microscope (SEM) images of fabricated devices, capturing the range of process variations at the nanometer scale. To enable one-to-many mapping, we inject a latent noise vector at the model bottleneck. We compare Gen-Fab against three baselines: (1) a deterministic U-Net predictor, (2) an inference-time Monte Carlo Dropout U-Net, and (3) an ensemble of varied U-Nets. Evaluations on an out-of-distribution dataset of fabricated photonic test structures demonstrate that Gen-Fab outperforms all baselines in both accuracy and uncertainty modeling. An additional distribution shift analysis further confirms its strong generalization to unseen fabrication geometries. Gen-Fab achieves the highest intersection-over-union (IoU) score of 89.8%, outperforming the deterministic U-Net (85.3%), the MC-Dropout U-Net (83.4%), and varying U-Nets (85.8%). It also better aligns with the distribution of real fabrication outcomes, achieving lower Kullback-Leibler divergence and Wasserstein distance.
