Deep Learning Mask Shape Recovery From Signed Distance Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mask manufacturing methods using inverse lithography technology (ILT) face challenges in quickly and accurately restoring a precise curvilinear mask shape from a mask image, leading to decreased optimization performance, increased optimization difficulty, and reduced restoration speed due to irreversible transformation processes and the need for additional iterative optimizations.
Innovation Solution
The method involves obtaining reversible signed distance images through transformation algorithms like Euclidean or signed distance computation, and performing deep learning to infer a transformation model between rasterized and signed distance images, enabling quick and accurate restoration of mask shapes without additional iterative optimizations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If inverse lithography technology (ILT) is used to restore mask shape from mask image, then manufacturing precision is improved, but device complexity and optimization difficulty increase
Solution Approach 1:
The patent introduces signed distance images as an intermediary representation between the mask image and the final mask shape. This intermediary form simplifies the optimization process by providing a mathematically convenient representation that preserves geometric information while enabling more efficient optimization algorithms, thereby reducing optimization difficulty while maintaining manufacturing precision
Solution Approach 2:
The patent transforms the mask shape representation parameters by using signed distance functions instead of traditional binary or grayscale image representations. This parameter change enables the optimization process to work with continuous distance values, improving convergence properties and reducing optimization complexity while maintaining the ability to achieve high manufacturing precision
2Manufacturing precision
If iterative optimization is performed to restore mask shape, then manufacturing precision is improved, but productivity decreases
Solution Approach 1:
The patent performs preliminary transformation of the mask image into signed distance images before the main optimization process. This preliminary action pre-processes the data into a form that is more amenable to optimization, reducing the number of iterative steps required and thereby improving productivity while maintaining the precision gains from iterative optimization
Solution Approach 2:
The patent replaces traditional mechanical iterative optimization approaches with a deep learning-based transformation model. By training a neural network to learn the mapping from mask images to mask shapes, the system substitutes multiple iterative optimization cycles with a single forward pass through the trained model, dramatically improving restoration speed while maintaining high accuracy
3Manufacturing precision
If transformation from mask image to mask shape is performed, then manufacturing precision is improved, but loss of time increases
Solution Approach 1:
The patent creates a copied representation of the mask image in the form of signed distance images, which preserves all necessary geometric information in a transformed domain. This copying approach allows the optimization to work with the copied representation rather than repeatedly accessing and processing the original image, reducing computation time while maintaining manufacturing precision
Data Source
AI summary
Provided are a method of forming a mask, the method accurately and quickly restoring an image on the mask to the shape on the mask, and a mask manufacturing method using the method of forming the mask. The method of forming a mask includes obtaining first images by performing rasterization and image correction on shapes on the mask corresponding to first patterns on a wafer, obtaining second images by applying a transformation to the shapes on the mask, performing deep learning based on a transformation relationship between ones of the first images and ones of the second images corresponding to the first images, and forming a target shape on the mask corresponding to a target pattern on the wafer, based on the deep learning. The mask is manufactured based on the target shape on the mask.