Internally funded project
Start date : 01.02.2021
This project is dedicated to the exploration of the capabilities of deep learning models for EUV lithography simulations and utilize them to speed-up a variety of computationally intensive applications. A wide range of techniques to optimize the accuracy and data efficency of deep learning models for lithography are also investigated. The developed accurate models and the frameworks for training data optimizations are applied to practical EUV use-cases in addition to experimental SEM images of wafer prints.
The purpose of this project is to explore the capabilities of deep learning models for EUV lithography simulations and utilize them to speed-up a variety of computationally intensive applications. An objective of this project is the development of accurate and efficient data driven deep learning models for EUV lithographic imaging. The developed deep learning models are applied to EUV lithography settings including demonstration of potential advantages and comparison compared to rigorous physical simulation models. Furthermore, optimizations of the deep learning model’s data efficiency to minimize the training data requirement for EUV data using techniques such transfer learning and data selection are investigated. The developed frameworks are also applied for experimental data, including SEM images of wafers. This project also involves a demonstration of perspectives of deep learning models for computationally intensive optimization techniques such Source Mask Optimization (SMO), mask biasing or Optical Proximity Corrections (OPC).