Physics-based deep learning network for parallelized two-photon polymerization lithography using a spatial light modulator

Physics-based deep learning network for parallelized two-photon polymerization lithography using a spatial light modulator — Valeriia Sedova1, Thomas Le Deun2, Joёl Rovera2, Jonas Wiedenmann3, Kevin Heggarty2, Andreas Erdmann1 1Fraunhofer Institute for Integrated Systems and Device Technology (IISB), Erlangen, Germany 2IMT Atlantique, Brest, France 3Heidelberg Instruments Mikrotechnik GmbH, Würzburg, Germany ICPST-42 2025, Himeji, Hyogo, Japan ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Overview 25.11.2025Page 2 Motivation and introduction Forward model development Inverse Lithography Techniques (Optimization) Results Summary ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan FABulous project - https://fabulous3d.eu/ 25.11.2025Page 3 ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Two-photon lithography 25.11.2025Page 4 ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Limitations of the process 25.11.2025Page 5 Speed of printing Conventional point by point multiphoton lithography (MPL) oHow to make process faster? Parallelization of the process © Fraunhofer IISB Printed Metasurface Parallelized MPL for 3D metasurface writing ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Limitations of the process 25.11.2025Page 6 oThe objective field of view oProximity Effects (In-Plane Overlap) oProximity Effects (In-Plane Overlap) Modeling is important to overcome the limitations, to scale up the process, to unlock the full potential of metasurface technology and more! ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Limitations of the process 25.11.2025 © Fraunhofer IISBPage 7 Optimization of the problem (ILT) Target Intensity/power distribution Fabricated structure Missing: the right input Intensity/power distribution Optimization procedure algorithm is needed Target Target ? ? ? ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan How to solve inverse problem? 25.11.2025 © Fraunhofer IISBPage 8 1. Build the Forward Model: 2. Solve the Optimization Problem Use gradient-based methods (e.g., backpropagation) Employ a generator model to reconstruct the solution … Polymer Quenching Diffusion Quenching Termination Propagation Polymerization R + O2 R + O2 O2 R R R + M Mack model Resist Exposure ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Overview of the models 25.11.2025Page 9 Optical model Generation of point spread function (PSF) within resist Resist model Threshold describes polymerization Exposure kinetics Diffusion of a single species Development of the processed photopolymer Exposure kinetics Temperature profile Diffusion and kinetics of multiple species Presence of quencher Development of the processed polymer A full model of polymerization Compact modelThreshold model ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Neural network and differentiable 3D lithography model •Differentiable 3D lithography model •Includes key steps of the lithography process •Input: dose distribution Output: 3D resist profiles Differentiable 3D lithography model Predicted dose distribution Predicted 3D structure Physics-based part NNBimg NNInhibitor NNMackModel NNResist ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Neural network and differentiable 3D lithography model 25.11.2025Page 17 1. Freeze NNlitho3D and set UNet to train mode 2. Provide Resist 3D to UNet (Z axis is Channel axis) 3. Let UNet predict a mask 4. Provide predicted mask and precomputed PSF to NNLitho3D 5. Let NNLitho3D compute a Resist 3D 6. Calculate loss between predicted Resist 3D and target Resist 3D using MSE loss. 7. Backpropagate loss 8. Optimize UNet weights until convergence ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Neural network and differentiable 3D lithography model 18 Differentiable 3D lithography model U-Net Desired 3D structure Predicted dose distribution Predicted 3D structure Loss is defined by the desired and predicted resist pattern Data-driven part Physics-based part ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Current results A. B. ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Conclusion 25.11.2025 © Fraunhofer IISBPage 20 •U-Net + Differentiable 3D lithography model Differentiable 3D lithography model U-Net Target pattern (3D desired shape) Predicted dose distribution Simulated resist shape Data-driven part Physics-based part Update 1. Developed a time-dependent differentiable forward model using PyTorch 2. Optimization approach for inverse lithography were developed: 3. Applied to SLM designs for experimental validation Thank you! For more information, please contact: [email protected] This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement nº 101091644. UK participants in Horizon Europe Project FABulous are supported by UKRI grant nº 10062385 (MODUS). Backup slides ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Generalized compact model 25.11.2025Page 23 DNQ-type resist Erdmann, Andreas. SPIE. 2021 CAR Erdmann, Andreas. SPIE. 2021 ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Two-Photo Resist (TPR) model 25.11.2025Page 24 Termination Q = 0.01 Q = 0.05 Q = 0.10 Q = 0.20 Q = 0.01 Q = 0.1 Q = 0.2 Polymer P Active center A Monomer M Quencher (added) Q O2 ICPST-42, June 24-27, 2025, Himeji City, Hyogo, Japan Overview of the models 25.11.2025 © Fraunhofer IISB Page 25 Generalized compact model *Yuan Yu Q = 0.01 Q = 0.1 Q = 0.2 O2Added quencher Active center Polymer Final inhibitor DArT i. Quenching (O2) ii. Diffusion (A, O2) iii. Quenching (O2) iv. Termination (A, Q) v. Polymerization (A)