Training and application of physics-inspired neural networks for two-photon lithography

Training and application of physics-inspired neural networks for two-photon lithography Valeriia Sedova1, Andreas Erdmann1, Joёl Rovera2, Florie Ogor2, Kevin Heggarty2 (1Fraunhofer IISB, 2IMTA) -Workshop 6th-8th June - Motivation 09.06.2024 © Fraunhofer IISBPage 2 Example of 2D metasurfaces Source: https://vosdogs.com/harvard-scientists-monitor-points-of-darkness-for-remote-sensing-and-covert-sensing-applications Source: Princeton University Source: FABulous proposal Source: Edinburg Instruments -Workshop 6th-8th June - Outline 09.06.2024 © Fraunhofer IISBPage 3 Model Description Compact model Neural Network (Digital twin) Optimization Conclusion and outlook -Workshop 6th-8th June - Outline 09.06.2024 © Fraunhofer IISBPage 4 Model Description Compact model Neural Network (Digital twin) Optimization Conclusion and outlook -Workshop 6th-8th June - Exposure kinetics Diffusion of a single species Development of the processed photopolymer Threshold describes polymerization Overview of the models 09.06.2024 © Fraunhofer IISBPage 5 Optical model Generation of point spread function (PSF) within resist Resist model Exposure kinetics Temperature profile Diffusion and kinetics of multiple species Presence of quencher Development of the processed polymer Threshold model Compact model A full model of polymerization SPIE Optical Systems Design: Advances in modeling and optimization for two-photon lithography V. Sedova, F. Ogor, J. Rovera, O. Tsilipakos, L. Lemberg, K. Heggarty, A. Erdmann -Workshop 6th-8th June - Compact model 09.06.2024 © Fraunhofer IISBPage 6 -Workshop 6th-8th June - Statement of the problem 09.06.2024 © Fraunhofer IISBPage 7 We want to inverse the problem Intensity distribution Resulted structure First step –Training of NN -Workshop 6th-8th June - Outline 09.06.2024 © Fraunhofer IISBPage 8 Model Description Compact model Neural Network (Digital twin) Optimization Conclusion and outlook -Workshop 6th-8th June - Close the Design-to-Manufacturing Gap in Computational Optics with a ’Real2Sim’ Learned Photolithography Simulator 09.06.2024 © Fraunhofer IISBPage 9 -Workshop 6th-8th June - Results for forward pass of NN 09.06.2024 © Fraunhofer IISBPage 16 Difference Difference in cross-section px px µm µm -Workshop 6th-8th June - New loss function for the NN 09.06.2024 © Fraunhofer IISBPage 17 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 +𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 =𝐺𝐺 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 =𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 �𝑑𝑑 − 𝑑𝑑 ∗ 1−𝐺𝐺 max(𝐺𝐺) now desired Loss Epochs µm µm µm µm -Workshop 6th-8th June - Results with new loss function 09.06.2024 © Fraunhofer IISBPage 18 Difference Difference in cross-section px px µm Dr. LiTHO Learned NN µm -Workshop 6th-8th June - Outline 09.06.2024 © Fraunhofer IISBPage 19 Model Description Compact model Neural Network (Digital twin) Optimization Conclusion and outlook -Workshop 6th-8th June - Recap and optimization 09.06.2024 © Fraunhofer IISBPage 20 Intensity distribution Resulted structure Second step - Optimization -Workshop 6th-8th June - Optimization run for binary mask 09.06.2024 © Fraunhofer IISBPage 21 px px px px px px µm µm Intensity distribution target Intensity distribution predicted 2.5D structure target 2.5D structure predicted µm µm -Workshop 6th-8th June - Optimization run for grayscale mask 09.06.2024 © Fraunhofer IISB Page 22 •Results after the optimization 2.5D structures Difference between structures px px px px pxpx px px px px px µm µm Intensity distribution target Intensity distribution predicted 2.5D structure target 2.5D structure predicted µm px µm -Workshop 6th-8th June - Outline 09.06.2024 © Fraunhofer IISBPage 23 Model Description Compact model Neural Network (Digital twin) Optimization Conclusion and outlook -Workshop 6th-8th June - Conclusion / Outlook 09.06.2024 © Fraunhofer IISBPage 24 Conclusion: •Understanding Gained •Fabrication of Training Data •Customization of Neural Networks •Optimization technique was implemented and used for both binary and grayscale masks Outlook: Further Investigation Needed: There is a clear need for continued exploration and research, particularly in the areas of inverse lithography (ILT) and optical proximity correction (OPC) tasks Thank you!